]>
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 | ||
1da177e4 LT |
34 | #include <linux/kernel.h> |
35 | #include <linux/module.h> | |
1da177e4 LT |
36 | #include <linux/types.h> |
37 | #include <linux/fcntl.h> | |
38 | #include <linux/interrupt.h> | |
39 | #include <linux/ptrace.h> | |
40 | #include <linux/ioport.h> | |
41 | #include <linux/in.h> | |
169a3e66 | 42 | #include <net/ip.h> |
1da177e4 | 43 | #include <linux/ip.h> |
df98be06 MC |
44 | #include <linux/icmp.h> |
45 | #include <linux/icmpv6.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> |
278b2083 | 80 | #include <linux/rculist.h> |
1bd758eb | 81 | #include <net/flow_dissector.h> |
18cb261a | 82 | #include <net/xfrm.h> |
1ef8019b DM |
83 | #include <net/bonding.h> |
84 | #include <net/bond_3ad.h> | |
85 | #include <net/bond_alb.h> | |
89df6a81 TT |
86 | #if IS_ENABLED(CONFIG_TLS_DEVICE) |
87 | #include <net/tls.h> | |
88 | #endif | |
1da177e4 | 89 | |
73b5a6f2 MB |
90 | #include "bonding_priv.h" |
91 | ||
1da177e4 LT |
92 | /*---------------------------- Module parameters ----------------------------*/ |
93 | ||
94 | /* monitor all links that often (in milliseconds). <=0 disables monitoring */ | |
1da177e4 LT |
95 | |
96 | static int max_bonds = BOND_DEFAULT_MAX_BONDS; | |
bb1d9123 | 97 | static int tx_queues = BOND_DEFAULT_TX_QUEUES; |
ad246c99 | 98 | static int num_peer_notif = 1; |
b98d9c66 | 99 | static int miimon; |
3d632c3f SH |
100 | static int updelay; |
101 | static int downdelay; | |
1da177e4 | 102 | static int use_carrier = 1; |
3d632c3f SH |
103 | static char *mode; |
104 | static char *primary; | |
a549952a | 105 | static char *primary_reselect; |
3d632c3f | 106 | static char *lacp_rate; |
655f8919 | 107 | static int min_links; |
3d632c3f SH |
108 | static char *ad_select; |
109 | static char *xmit_hash_policy; | |
7bdb04ed | 110 | static int arp_interval; |
3d632c3f SH |
111 | static char *arp_ip_target[BOND_MAX_ARP_TARGETS]; |
112 | static char *arp_validate; | |
8599b52e | 113 | static char *arp_all_targets; |
3d632c3f | 114 | static char *fail_over_mac; |
b07ea07b | 115 | static int all_slaves_active; |
d2991f75 | 116 | static struct bond_params bonding_defaults; |
c2952c31 | 117 | static int resend_igmp = BOND_DEFAULT_RESEND_IGMP; |
73958329 | 118 | static int packets_per_slave = 1; |
3a7129e5 | 119 | static int lp_interval = BOND_ALB_DEFAULT_LP_INTERVAL; |
1da177e4 LT |
120 | |
121 | module_param(max_bonds, int, 0); | |
122 | MODULE_PARM_DESC(max_bonds, "Max number of bonded devices"); | |
bb1d9123 AG |
123 | module_param(tx_queues, int, 0); |
124 | MODULE_PARM_DESC(tx_queues, "Max number of transmit queues (default = 16)"); | |
ad246c99 | 125 | module_param_named(num_grat_arp, num_peer_notif, int, 0644); |
90e62474 AG |
126 | MODULE_PARM_DESC(num_grat_arp, "Number of peer notifications to send on " |
127 | "failover event (alias of num_unsol_na)"); | |
ad246c99 | 128 | module_param_named(num_unsol_na, num_peer_notif, int, 0644); |
90e62474 AG |
129 | MODULE_PARM_DESC(num_unsol_na, "Number of peer notifications to send on " |
130 | "failover event (alias of num_grat_arp)"); | |
1da177e4 LT |
131 | module_param(miimon, int, 0); |
132 | MODULE_PARM_DESC(miimon, "Link check interval in milliseconds"); | |
133 | module_param(updelay, int, 0); | |
134 | MODULE_PARM_DESC(updelay, "Delay before considering link up, in milliseconds"); | |
135 | module_param(downdelay, int, 0); | |
2ac47660 MW |
136 | MODULE_PARM_DESC(downdelay, "Delay before considering link down, " |
137 | "in milliseconds"); | |
1da177e4 | 138 | module_param(use_carrier, int, 0); |
2ac47660 | 139 | MODULE_PARM_DESC(use_carrier, "Use netif_carrier_ok (vs MII ioctls) in miimon; " |
b3c898e2 | 140 | "0 for off, 1 for on (default)"); |
1da177e4 | 141 | module_param(mode, charp, 0); |
90e62474 | 142 | MODULE_PARM_DESC(mode, "Mode of operation; 0 for balance-rr, " |
2ac47660 MW |
143 | "1 for active-backup, 2 for balance-xor, " |
144 | "3 for broadcast, 4 for 802.3ad, 5 for balance-tlb, " | |
145 | "6 for balance-alb"); | |
1da177e4 LT |
146 | module_param(primary, charp, 0); |
147 | MODULE_PARM_DESC(primary, "Primary network device to use"); | |
a549952a JP |
148 | module_param(primary_reselect, charp, 0); |
149 | MODULE_PARM_DESC(primary_reselect, "Reselect primary slave " | |
150 | "once it comes up; " | |
151 | "0 for always (default), " | |
152 | "1 for only if speed of primary is " | |
153 | "better, " | |
154 | "2 for only on active slave " | |
155 | "failure"); | |
1da177e4 | 156 | module_param(lacp_rate, charp, 0); |
90e62474 AG |
157 | MODULE_PARM_DESC(lacp_rate, "LACPDU tx rate to request from 802.3ad partner; " |
158 | "0 for slow, 1 for fast"); | |
fd989c83 | 159 | module_param(ad_select, charp, 0); |
0d039f33 | 160 | MODULE_PARM_DESC(ad_select, "802.3ad aggregation selection logic; " |
90e62474 AG |
161 | "0 for stable (default), 1 for bandwidth, " |
162 | "2 for count"); | |
655f8919 | 163 | module_param(min_links, int, 0); |
164 | MODULE_PARM_DESC(min_links, "Minimum number of available links before turning on carrier"); | |
165 | ||
169a3e66 | 166 | module_param(xmit_hash_policy, charp, 0); |
e79c1055 | 167 | MODULE_PARM_DESC(xmit_hash_policy, "balance-alb, balance-tlb, balance-xor, 802.3ad hashing method; " |
90e62474 | 168 | "0 for layer 2 (default), 1 for layer 3+4, " |
32819dc1 | 169 | "2 for layer 2+3, 3 for encap layer 2+3, " |
7b8fc010 | 170 | "4 for encap layer 3+4, 5 for vlan+srcmac"); |
1da177e4 LT |
171 | module_param(arp_interval, int, 0); |
172 | MODULE_PARM_DESC(arp_interval, "arp interval in milliseconds"); | |
173 | module_param_array(arp_ip_target, charp, NULL, 0); | |
174 | MODULE_PARM_DESC(arp_ip_target, "arp targets in n.n.n.n form"); | |
f5b2b966 | 175 | module_param(arp_validate, charp, 0); |
90e62474 AG |
176 | MODULE_PARM_DESC(arp_validate, "validate src/dst of ARP probes; " |
177 | "0 for none (default), 1 for active, " | |
178 | "2 for backup, 3 for all"); | |
8599b52e VF |
179 | module_param(arp_all_targets, charp, 0); |
180 | MODULE_PARM_DESC(arp_all_targets, "fail on any/all arp targets timeout; 0 for any (default), 1 for all"); | |
3915c1e8 | 181 | module_param(fail_over_mac, charp, 0); |
90e62474 AG |
182 | MODULE_PARM_DESC(fail_over_mac, "For active-backup, do not set all slaves to " |
183 | "the same MAC; 0 for none (default), " | |
184 | "1 for active, 2 for follow"); | |
ebd8e497 | 185 | module_param(all_slaves_active, int, 0); |
37b7021d | 186 | MODULE_PARM_DESC(all_slaves_active, "Keep all frames received on an interface " |
90e62474 | 187 | "by setting active flag for all slaves; " |
ebd8e497 | 188 | "0 for never (default), 1 for always."); |
c2952c31 | 189 | module_param(resend_igmp, int, 0); |
90e62474 AG |
190 | MODULE_PARM_DESC(resend_igmp, "Number of IGMP membership reports to send on " |
191 | "link failure"); | |
73958329 NA |
192 | module_param(packets_per_slave, int, 0); |
193 | MODULE_PARM_DESC(packets_per_slave, "Packets to send per slave in balance-rr " | |
194 | "mode; 0 for a random slave, 1 packet per " | |
195 | "slave (default), >1 packets per slave."); | |
3a7129e5 | 196 | module_param(lp_interval, uint, 0); |
197 | MODULE_PARM_DESC(lp_interval, "The number of seconds between instances where " | |
198 | "the bonding driver sends learning packets to " | |
199 | "each slaves peer switch. The default is 1."); | |
1da177e4 LT |
200 | |
201 | /*----------------------------- Global variables ----------------------------*/ | |
202 | ||
e843fa50 | 203 | #ifdef CONFIG_NET_POLL_CONTROLLER |
fb4fa76a | 204 | atomic_t netpoll_block_tx = ATOMIC_INIT(0); |
e843fa50 NH |
205 | #endif |
206 | ||
c7d03a00 | 207 | unsigned int bond_net_id __read_mostly; |
1da177e4 | 208 | |
58deb77c MC |
209 | static const struct flow_dissector_key flow_keys_bonding_keys[] = { |
210 | { | |
211 | .key_id = FLOW_DISSECTOR_KEY_CONTROL, | |
212 | .offset = offsetof(struct flow_keys, control), | |
213 | }, | |
214 | { | |
215 | .key_id = FLOW_DISSECTOR_KEY_BASIC, | |
216 | .offset = offsetof(struct flow_keys, basic), | |
217 | }, | |
218 | { | |
219 | .key_id = FLOW_DISSECTOR_KEY_IPV4_ADDRS, | |
220 | .offset = offsetof(struct flow_keys, addrs.v4addrs), | |
221 | }, | |
222 | { | |
223 | .key_id = FLOW_DISSECTOR_KEY_IPV6_ADDRS, | |
224 | .offset = offsetof(struct flow_keys, addrs.v6addrs), | |
225 | }, | |
226 | { | |
227 | .key_id = FLOW_DISSECTOR_KEY_TIPC, | |
228 | .offset = offsetof(struct flow_keys, addrs.tipckey), | |
229 | }, | |
230 | { | |
231 | .key_id = FLOW_DISSECTOR_KEY_PORTS, | |
232 | .offset = offsetof(struct flow_keys, ports), | |
233 | }, | |
234 | { | |
235 | .key_id = FLOW_DISSECTOR_KEY_ICMP, | |
236 | .offset = offsetof(struct flow_keys, icmp), | |
237 | }, | |
238 | { | |
239 | .key_id = FLOW_DISSECTOR_KEY_VLAN, | |
240 | .offset = offsetof(struct flow_keys, vlan), | |
241 | }, | |
242 | { | |
243 | .key_id = FLOW_DISSECTOR_KEY_FLOW_LABEL, | |
244 | .offset = offsetof(struct flow_keys, tags), | |
245 | }, | |
246 | { | |
247 | .key_id = FLOW_DISSECTOR_KEY_GRE_KEYID, | |
248 | .offset = offsetof(struct flow_keys, keyid), | |
249 | }, | |
250 | }; | |
251 | ||
252 | static struct flow_dissector flow_keys_bonding __read_mostly; | |
253 | ||
1da177e4 LT |
254 | /*-------------------------- Forward declarations ---------------------------*/ |
255 | ||
181470fc | 256 | static int bond_init(struct net_device *bond_dev); |
c67dfb29 | 257 | static void bond_uninit(struct net_device *bond_dev); |
bc1f4470 | 258 | static void bond_get_stats(struct net_device *bond_dev, |
259 | struct rtnl_link_stats64 *stats); | |
ee637714 | 260 | static void bond_slave_arr_handler(struct work_struct *work); |
21a75f09 JV |
261 | static bool bond_time_in_interval(struct bonding *bond, unsigned long last_act, |
262 | int mod); | |
d4859d74 | 263 | static void bond_netdev_notify_work(struct work_struct *work); |
1da177e4 LT |
264 | |
265 | /*---------------------------- General routines -----------------------------*/ | |
266 | ||
bd33acc3 | 267 | const char *bond_mode_name(int mode) |
1da177e4 | 268 | { |
77afc92b HE |
269 | static const char *names[] = { |
270 | [BOND_MODE_ROUNDROBIN] = "load balancing (round-robin)", | |
271 | [BOND_MODE_ACTIVEBACKUP] = "fault-tolerance (active-backup)", | |
272 | [BOND_MODE_XOR] = "load balancing (xor)", | |
273 | [BOND_MODE_BROADCAST] = "fault-tolerance (broadcast)", | |
3d632c3f | 274 | [BOND_MODE_8023AD] = "IEEE 802.3ad Dynamic link aggregation", |
77afc92b HE |
275 | [BOND_MODE_TLB] = "transmit load balancing", |
276 | [BOND_MODE_ALB] = "adaptive load balancing", | |
277 | }; | |
278 | ||
f5280948 | 279 | if (mode < BOND_MODE_ROUNDROBIN || mode > BOND_MODE_ALB) |
1da177e4 | 280 | return "unknown"; |
77afc92b HE |
281 | |
282 | return names[mode]; | |
1da177e4 LT |
283 | } |
284 | ||
1da177e4 LT |
285 | /** |
286 | * bond_dev_queue_xmit - Prepare skb for xmit. | |
3d632c3f | 287 | * |
1da177e4 LT |
288 | * @bond: bond device that got this skb for tx. |
289 | * @skb: hw accel VLAN tagged skb to transmit | |
290 | * @slave_dev: slave that is supposed to xmit this skbuff | |
1da177e4 | 291 | */ |
ae46f184 | 292 | netdev_tx_t bond_dev_queue_xmit(struct bonding *bond, struct sk_buff *skb, |
3d632c3f | 293 | struct net_device *slave_dev) |
1da177e4 | 294 | { |
8387451e | 295 | skb->dev = slave_dev; |
374eeb5a | 296 | |
5ee31c68 | 297 | BUILD_BUG_ON(sizeof(skb->queue_mapping) != |
df4ab5b3 | 298 | sizeof(qdisc_skb_cb(skb)->slave_dev_queue_mapping)); |
ae35c6f7 | 299 | skb_set_queue_mapping(skb, qdisc_skb_cb(skb)->slave_dev_queue_mapping); |
374eeb5a | 300 | |
e15c3c22 | 301 | if (unlikely(netpoll_tx_running(bond->dev))) |
ae46f184 ED |
302 | return bond_netpoll_send_skb(bond_get_slave_by_dev(bond, slave_dev), skb); |
303 | ||
304 | return dev_queue_xmit(skb); | |
1da177e4 LT |
305 | } |
306 | ||
007feb87 TT |
307 | bool bond_sk_check(struct bonding *bond) |
308 | { | |
309 | switch (BOND_MODE(bond)) { | |
310 | case BOND_MODE_8023AD: | |
311 | case BOND_MODE_XOR: | |
312 | if (bond->params.xmit_policy == BOND_XMIT_POLICY_LAYER34) | |
313 | return true; | |
314 | fallthrough; | |
315 | default: | |
316 | return false; | |
317 | } | |
318 | } | |
319 | ||
9e2ee5c7 JM |
320 | static bool bond_xdp_check(struct bonding *bond) |
321 | { | |
322 | switch (BOND_MODE(bond)) { | |
323 | case BOND_MODE_ROUNDROBIN: | |
324 | case BOND_MODE_ACTIVEBACKUP: | |
39a0876d | 325 | return true; |
9e2ee5c7 JM |
326 | case BOND_MODE_8023AD: |
327 | case BOND_MODE_XOR: | |
39a0876d JM |
328 | /* vlan+srcmac is not supported with XDP as in most cases the 802.1q |
329 | * payload is not in the packet due to hardware offload. | |
330 | */ | |
331 | if (bond->params.xmit_policy != BOND_XMIT_POLICY_VLAN_SRCMAC) | |
332 | return true; | |
333 | fallthrough; | |
9e2ee5c7 JM |
334 | default: |
335 | return false; | |
336 | } | |
337 | } | |
338 | ||
18cb261a JW |
339 | /*---------------------------------- VLAN -----------------------------------*/ |
340 | ||
547942ca | 341 | /* In the following 2 functions, bond_vlan_rx_add_vid and bond_vlan_rx_kill_vid, |
cc0e4070 | 342 | * We don't protect the slave list iteration with a lock because: |
1da177e4 LT |
343 | * a. This operation is performed in IOCTL context, |
344 | * b. The operation is protected by the RTNL semaphore in the 8021q code, | |
345 | * c. Holding a lock with BH disabled while directly calling a base driver | |
346 | * entry point is generally a BAD idea. | |
3d632c3f | 347 | * |
1da177e4 LT |
348 | * The design of synchronization/protection for this operation in the 8021q |
349 | * module is good for one or more VLAN devices over a single physical device | |
350 | * and cannot be extended for a teaming solution like bonding, so there is a | |
351 | * potential race condition here where a net device from the vlan group might | |
352 | * be referenced (either by a base driver or the 8021q code) while it is being | |
353 | * removed from the system. However, it turns out we're not making matters | |
354 | * worse, and if it works for regular VLAN usage it will work here too. | |
355 | */ | |
356 | ||
1da177e4 LT |
357 | /** |
358 | * bond_vlan_rx_add_vid - Propagates adding an id to slaves | |
359 | * @bond_dev: bonding net device that got called | |
45a1553b | 360 | * @proto: network protocol ID |
1da177e4 LT |
361 | * @vid: vlan id being added |
362 | */ | |
80d5c368 PM |
363 | static int bond_vlan_rx_add_vid(struct net_device *bond_dev, |
364 | __be16 proto, u16 vid) | |
1da177e4 | 365 | { |
454d7c9b | 366 | struct bonding *bond = netdev_priv(bond_dev); |
81f23b13 | 367 | struct slave *slave, *rollback_slave; |
9caff1e7 | 368 | struct list_head *iter; |
dec1e90e | 369 | int res; |
1da177e4 | 370 | |
9caff1e7 | 371 | bond_for_each_slave(bond, slave, iter) { |
80d5c368 | 372 | res = vlan_vid_add(slave->dev, proto, vid); |
87002b03 JP |
373 | if (res) |
374 | goto unwind; | |
1da177e4 LT |
375 | } |
376 | ||
8e586137 | 377 | return 0; |
87002b03 JP |
378 | |
379 | unwind: | |
81f23b13 | 380 | /* unwind to the slave that failed */ |
9caff1e7 | 381 | bond_for_each_slave(bond, rollback_slave, iter) { |
81f23b13 VF |
382 | if (rollback_slave == slave) |
383 | break; | |
384 | ||
385 | vlan_vid_del(rollback_slave->dev, proto, vid); | |
386 | } | |
87002b03 JP |
387 | |
388 | return res; | |
1da177e4 LT |
389 | } |
390 | ||
391 | /** | |
392 | * bond_vlan_rx_kill_vid - Propagates deleting an id to slaves | |
393 | * @bond_dev: bonding net device that got called | |
45a1553b | 394 | * @proto: network protocol ID |
1da177e4 LT |
395 | * @vid: vlan id being removed |
396 | */ | |
80d5c368 PM |
397 | static int bond_vlan_rx_kill_vid(struct net_device *bond_dev, |
398 | __be16 proto, u16 vid) | |
1da177e4 | 399 | { |
454d7c9b | 400 | struct bonding *bond = netdev_priv(bond_dev); |
9caff1e7 | 401 | struct list_head *iter; |
1da177e4 | 402 | struct slave *slave; |
1da177e4 | 403 | |
9caff1e7 | 404 | bond_for_each_slave(bond, slave, iter) |
80d5c368 | 405 | vlan_vid_del(slave->dev, proto, vid); |
1da177e4 | 406 | |
e868b0c9 VF |
407 | if (bond_is_lb(bond)) |
408 | bond_alb_clear_vlan(bond, vid); | |
8e586137 JP |
409 | |
410 | return 0; | |
1da177e4 LT |
411 | } |
412 | ||
18cb261a JW |
413 | /*---------------------------------- XFRM -----------------------------------*/ |
414 | ||
415 | #ifdef CONFIG_XFRM_OFFLOAD | |
416 | /** | |
417 | * bond_ipsec_add_sa - program device with a security association | |
418 | * @xs: pointer to transformer state struct | |
419 | **/ | |
420 | static int bond_ipsec_add_sa(struct xfrm_state *xs) | |
421 | { | |
422 | struct net_device *bond_dev = xs->xso.dev; | |
9a560550 | 423 | struct bond_ipsec *ipsec; |
5cd24cbe JW |
424 | struct bonding *bond; |
425 | struct slave *slave; | |
b648eba4 | 426 | int err; |
18cb261a | 427 | |
5cd24cbe JW |
428 | if (!bond_dev) |
429 | return -EINVAL; | |
430 | ||
b648eba4 | 431 | rcu_read_lock(); |
5cd24cbe | 432 | bond = netdev_priv(bond_dev); |
f548a476 | 433 | slave = rcu_dereference(bond->curr_active_slave); |
105cd17a TY |
434 | if (!slave) { |
435 | rcu_read_unlock(); | |
436 | return -ENODEV; | |
437 | } | |
438 | ||
b1216933 TY |
439 | if (!slave->dev->xfrmdev_ops || |
440 | !slave->dev->xfrmdev_ops->xdo_dev_state_add || | |
441 | netif_is_bond_master(slave->dev)) { | |
18cb261a | 442 | slave_warn(bond_dev, slave->dev, "Slave does not support ipsec offload\n"); |
b648eba4 | 443 | rcu_read_unlock(); |
18cb261a JW |
444 | return -EINVAL; |
445 | } | |
446 | ||
9a560550 TY |
447 | ipsec = kmalloc(sizeof(*ipsec), GFP_ATOMIC); |
448 | if (!ipsec) { | |
449 | rcu_read_unlock(); | |
450 | return -ENOMEM; | |
451 | } | |
452 | xs->xso.real_dev = slave->dev; | |
453 | ||
b648eba4 | 454 | err = slave->dev->xfrmdev_ops->xdo_dev_state_add(xs); |
9a560550 TY |
455 | if (!err) { |
456 | ipsec->xs = xs; | |
457 | INIT_LIST_HEAD(&ipsec->list); | |
458 | spin_lock_bh(&bond->ipsec_lock); | |
459 | list_add(&ipsec->list, &bond->ipsec_list); | |
460 | spin_unlock_bh(&bond->ipsec_lock); | |
461 | } else { | |
462 | kfree(ipsec); | |
463 | } | |
b648eba4 TY |
464 | rcu_read_unlock(); |
465 | return err; | |
18cb261a JW |
466 | } |
467 | ||
9a560550 TY |
468 | static void bond_ipsec_add_sa_all(struct bonding *bond) |
469 | { | |
470 | struct net_device *bond_dev = bond->dev; | |
471 | struct bond_ipsec *ipsec; | |
472 | struct slave *slave; | |
473 | ||
474 | rcu_read_lock(); | |
475 | slave = rcu_dereference(bond->curr_active_slave); | |
476 | if (!slave) | |
477 | goto out; | |
478 | ||
479 | if (!slave->dev->xfrmdev_ops || | |
480 | !slave->dev->xfrmdev_ops->xdo_dev_state_add || | |
481 | netif_is_bond_master(slave->dev)) { | |
482 | spin_lock_bh(&bond->ipsec_lock); | |
483 | if (!list_empty(&bond->ipsec_list)) | |
484 | slave_warn(bond_dev, slave->dev, | |
485 | "%s: no slave xdo_dev_state_add\n", | |
486 | __func__); | |
487 | spin_unlock_bh(&bond->ipsec_lock); | |
488 | goto out; | |
489 | } | |
490 | ||
491 | spin_lock_bh(&bond->ipsec_lock); | |
492 | list_for_each_entry(ipsec, &bond->ipsec_list, list) { | |
493 | ipsec->xs->xso.real_dev = slave->dev; | |
494 | if (slave->dev->xfrmdev_ops->xdo_dev_state_add(ipsec->xs)) { | |
495 | slave_warn(bond_dev, slave->dev, "%s: failed to add SA\n", __func__); | |
496 | ipsec->xs->xso.real_dev = NULL; | |
497 | } | |
498 | } | |
499 | spin_unlock_bh(&bond->ipsec_lock); | |
500 | out: | |
501 | rcu_read_unlock(); | |
502 | } | |
503 | ||
18cb261a JW |
504 | /** |
505 | * bond_ipsec_del_sa - clear out this specific SA | |
506 | * @xs: pointer to transformer state struct | |
507 | **/ | |
508 | static void bond_ipsec_del_sa(struct xfrm_state *xs) | |
509 | { | |
510 | struct net_device *bond_dev = xs->xso.dev; | |
9a560550 | 511 | struct bond_ipsec *ipsec; |
5cd24cbe JW |
512 | struct bonding *bond; |
513 | struct slave *slave; | |
514 | ||
515 | if (!bond_dev) | |
516 | return; | |
517 | ||
a22c39b8 | 518 | rcu_read_lock(); |
5cd24cbe | 519 | bond = netdev_priv(bond_dev); |
f548a476 | 520 | slave = rcu_dereference(bond->curr_active_slave); |
18cb261a JW |
521 | |
522 | if (!slave) | |
a22c39b8 | 523 | goto out; |
18cb261a | 524 | |
9a560550 TY |
525 | if (!xs->xso.real_dev) |
526 | goto out; | |
527 | ||
528 | WARN_ON(xs->xso.real_dev != slave->dev); | |
18cb261a | 529 | |
b1216933 TY |
530 | if (!slave->dev->xfrmdev_ops || |
531 | !slave->dev->xfrmdev_ops->xdo_dev_state_delete || | |
532 | netif_is_bond_master(slave->dev)) { | |
18cb261a | 533 | slave_warn(bond_dev, slave->dev, "%s: no slave xdo_dev_state_delete\n", __func__); |
a22c39b8 | 534 | goto out; |
18cb261a JW |
535 | } |
536 | ||
537 | slave->dev->xfrmdev_ops->xdo_dev_state_delete(xs); | |
a22c39b8 | 538 | out: |
9a560550 TY |
539 | spin_lock_bh(&bond->ipsec_lock); |
540 | list_for_each_entry(ipsec, &bond->ipsec_list, list) { | |
541 | if (ipsec->xs == xs) { | |
542 | list_del(&ipsec->list); | |
543 | kfree(ipsec); | |
544 | break; | |
545 | } | |
546 | } | |
547 | spin_unlock_bh(&bond->ipsec_lock); | |
548 | rcu_read_unlock(); | |
549 | } | |
550 | ||
551 | static void bond_ipsec_del_sa_all(struct bonding *bond) | |
552 | { | |
553 | struct net_device *bond_dev = bond->dev; | |
554 | struct bond_ipsec *ipsec; | |
555 | struct slave *slave; | |
556 | ||
557 | rcu_read_lock(); | |
558 | slave = rcu_dereference(bond->curr_active_slave); | |
559 | if (!slave) { | |
560 | rcu_read_unlock(); | |
561 | return; | |
562 | } | |
563 | ||
564 | spin_lock_bh(&bond->ipsec_lock); | |
565 | list_for_each_entry(ipsec, &bond->ipsec_list, list) { | |
566 | if (!ipsec->xs->xso.real_dev) | |
567 | continue; | |
568 | ||
569 | if (!slave->dev->xfrmdev_ops || | |
570 | !slave->dev->xfrmdev_ops->xdo_dev_state_delete || | |
571 | netif_is_bond_master(slave->dev)) { | |
572 | slave_warn(bond_dev, slave->dev, | |
573 | "%s: no slave xdo_dev_state_delete\n", | |
574 | __func__); | |
575 | } else { | |
576 | slave->dev->xfrmdev_ops->xdo_dev_state_delete(ipsec->xs); | |
577 | } | |
578 | ipsec->xs->xso.real_dev = NULL; | |
579 | } | |
580 | spin_unlock_bh(&bond->ipsec_lock); | |
a22c39b8 | 581 | rcu_read_unlock(); |
18cb261a JW |
582 | } |
583 | ||
584 | /** | |
585 | * bond_ipsec_offload_ok - can this packet use the xfrm hw offload | |
586 | * @skb: current data packet | |
587 | * @xs: pointer to transformer state struct | |
588 | **/ | |
589 | static bool bond_ipsec_offload_ok(struct sk_buff *skb, struct xfrm_state *xs) | |
590 | { | |
591 | struct net_device *bond_dev = xs->xso.dev; | |
9a560550 TY |
592 | struct net_device *real_dev; |
593 | struct slave *curr_active; | |
594 | struct bonding *bond; | |
955b785e | 595 | int err; |
9a560550 TY |
596 | |
597 | bond = netdev_priv(bond_dev); | |
955b785e | 598 | rcu_read_lock(); |
9a560550 TY |
599 | curr_active = rcu_dereference(bond->curr_active_slave); |
600 | real_dev = curr_active->dev; | |
18cb261a | 601 | |
955b785e | 602 | if (BOND_MODE(bond) != BOND_MODE_ACTIVEBACKUP) { |
168e696a | 603 | err = false; |
955b785e TY |
604 | goto out; |
605 | } | |
a3b658cf | 606 | |
955b785e TY |
607 | if (!xs->xso.real_dev) { |
608 | err = false; | |
609 | goto out; | |
610 | } | |
9a560550 TY |
611 | |
612 | if (!real_dev->xfrmdev_ops || | |
613 | !real_dev->xfrmdev_ops->xdo_dev_offload_ok || | |
614 | netif_is_bond_master(real_dev)) { | |
955b785e TY |
615 | err = false; |
616 | goto out; | |
18cb261a JW |
617 | } |
618 | ||
955b785e TY |
619 | err = real_dev->xfrmdev_ops->xdo_dev_offload_ok(skb, xs); |
620 | out: | |
621 | rcu_read_unlock(); | |
622 | return err; | |
18cb261a JW |
623 | } |
624 | ||
625 | static const struct xfrmdev_ops bond_xfrmdev_ops = { | |
626 | .xdo_dev_state_add = bond_ipsec_add_sa, | |
627 | .xdo_dev_state_delete = bond_ipsec_del_sa, | |
628 | .xdo_dev_offload_ok = bond_ipsec_offload_ok, | |
629 | }; | |
630 | #endif /* CONFIG_XFRM_OFFLOAD */ | |
631 | ||
1da177e4 LT |
632 | /*------------------------------- Link status -------------------------------*/ |
633 | ||
547942ca | 634 | /* Set the carrier state for the master according to the state of its |
ff59c456 JV |
635 | * slaves. If any slaves are up, the master is up. In 802.3ad mode, |
636 | * do special 802.3ad magic. | |
637 | * | |
638 | * Returns zero if carrier state does not change, nonzero if it does. | |
639 | */ | |
2477bc9a | 640 | int bond_set_carrier(struct bonding *bond) |
ff59c456 | 641 | { |
9caff1e7 | 642 | struct list_head *iter; |
ff59c456 | 643 | struct slave *slave; |
ff59c456 | 644 | |
0965a1f3 | 645 | if (!bond_has_slaves(bond)) |
ff59c456 JV |
646 | goto down; |
647 | ||
01844098 | 648 | if (BOND_MODE(bond) == BOND_MODE_8023AD) |
ff59c456 JV |
649 | return bond_3ad_set_carrier(bond); |
650 | ||
9caff1e7 | 651 | bond_for_each_slave(bond, slave, iter) { |
ff59c456 JV |
652 | if (slave->link == BOND_LINK_UP) { |
653 | if (!netif_carrier_ok(bond->dev)) { | |
654 | netif_carrier_on(bond->dev); | |
655 | return 1; | |
656 | } | |
657 | return 0; | |
658 | } | |
659 | } | |
660 | ||
661 | down: | |
662 | if (netif_carrier_ok(bond->dev)) { | |
663 | netif_carrier_off(bond->dev); | |
664 | return 1; | |
665 | } | |
666 | return 0; | |
667 | } | |
668 | ||
547942ca | 669 | /* Get link speed and duplex from the slave's base driver |
1da177e4 | 670 | * using ethtool. If for some reason the call fails or the |
98f41f69 | 671 | * values are invalid, set speed and duplex to -1, |
c4adfc82 MB |
672 | * and return. Return 1 if speed or duplex settings are |
673 | * UNKNOWN; 0 otherwise. | |
1da177e4 | 674 | */ |
c4adfc82 | 675 | static int bond_update_speed_duplex(struct slave *slave) |
1da177e4 LT |
676 | { |
677 | struct net_device *slave_dev = slave->dev; | |
9856909c | 678 | struct ethtool_link_ksettings ecmd; |
61a44b9c | 679 | int res; |
1da177e4 | 680 | |
589665f5 DC |
681 | slave->speed = SPEED_UNKNOWN; |
682 | slave->duplex = DUPLEX_UNKNOWN; | |
1da177e4 | 683 | |
9856909c | 684 | res = __ethtool_get_link_ksettings(slave_dev, &ecmd); |
61a44b9c | 685 | if (res < 0) |
c4adfc82 | 686 | return 1; |
9856909c | 687 | if (ecmd.base.speed == 0 || ecmd.base.speed == ((__u32)-1)) |
c4adfc82 | 688 | return 1; |
9856909c | 689 | switch (ecmd.base.duplex) { |
1da177e4 LT |
690 | case DUPLEX_FULL: |
691 | case DUPLEX_HALF: | |
692 | break; | |
693 | default: | |
c4adfc82 | 694 | return 1; |
1da177e4 LT |
695 | } |
696 | ||
9856909c DD |
697 | slave->speed = ecmd.base.speed; |
698 | slave->duplex = ecmd.base.duplex; | |
1da177e4 | 699 | |
c4adfc82 | 700 | return 0; |
1da177e4 LT |
701 | } |
702 | ||
07699f9a | 703 | const char *bond_slave_link_status(s8 link) |
704 | { | |
705 | switch (link) { | |
706 | case BOND_LINK_UP: | |
707 | return "up"; | |
708 | case BOND_LINK_FAIL: | |
709 | return "going down"; | |
710 | case BOND_LINK_DOWN: | |
711 | return "down"; | |
712 | case BOND_LINK_BACK: | |
713 | return "going back"; | |
714 | default: | |
715 | return "unknown"; | |
716 | } | |
717 | } | |
718 | ||
547942ca | 719 | /* if <dev> supports MII link status reporting, check its link status. |
1da177e4 LT |
720 | * |
721 | * We either do MII/ETHTOOL ioctls, or check netif_carrier_ok(), | |
3d632c3f | 722 | * depending upon the setting of the use_carrier parameter. |
1da177e4 LT |
723 | * |
724 | * Return either BMSR_LSTATUS, meaning that the link is up (or we | |
725 | * can't tell and just pretend it is), or 0, meaning that the link is | |
726 | * down. | |
727 | * | |
728 | * If reporting is non-zero, instead of faking link up, return -1 if | |
729 | * both ETHTOOL and MII ioctls fail (meaning the device does not | |
730 | * support them). If use_carrier is set, return whatever it says. | |
731 | * It'd be nice if there was a good way to tell if a driver supports | |
732 | * netif_carrier, but there really isn't. | |
733 | */ | |
3d632c3f SH |
734 | static int bond_check_dev_link(struct bonding *bond, |
735 | struct net_device *slave_dev, int reporting) | |
1da177e4 | 736 | { |
eb7cc59a | 737 | const struct net_device_ops *slave_ops = slave_dev->netdev_ops; |
d9d52832 | 738 | int (*ioctl)(struct net_device *, struct ifreq *, int); |
1da177e4 LT |
739 | struct ifreq ifr; |
740 | struct mii_ioctl_data *mii; | |
1da177e4 | 741 | |
6c988853 PG |
742 | if (!reporting && !netif_running(slave_dev)) |
743 | return 0; | |
744 | ||
eb7cc59a | 745 | if (bond->params.use_carrier) |
b3c898e2 | 746 | return netif_carrier_ok(slave_dev) ? BMSR_LSTATUS : 0; |
1da177e4 | 747 | |
29112f4e | 748 | /* Try to get link status using Ethtool first. */ |
c772dde3 BH |
749 | if (slave_dev->ethtool_ops->get_link) |
750 | return slave_dev->ethtool_ops->get_link(slave_dev) ? | |
751 | BMSR_LSTATUS : 0; | |
29112f4e | 752 | |
3d632c3f | 753 | /* Ethtool can't be used, fallback to MII ioctls. */ |
a7605370 | 754 | ioctl = slave_ops->ndo_eth_ioctl; |
1da177e4 | 755 | if (ioctl) { |
547942ca NA |
756 | /* TODO: set pointer to correct ioctl on a per team member |
757 | * bases to make this more efficient. that is, once | |
758 | * we determine the correct ioctl, we will always | |
759 | * call it and not the others for that team | |
760 | * member. | |
761 | */ | |
762 | ||
763 | /* We cannot assume that SIOCGMIIPHY will also read a | |
1da177e4 LT |
764 | * register; not all network drivers (e.g., e100) |
765 | * support that. | |
766 | */ | |
767 | ||
768 | /* Yes, the mii is overlaid on the ifreq.ifr_ifru */ | |
43902070 | 769 | strscpy_pad(ifr.ifr_name, slave_dev->name, IFNAMSIZ); |
1da177e4 | 770 | mii = if_mii(&ifr); |
4ad41c1e | 771 | if (ioctl(slave_dev, &ifr, SIOCGMIIPHY) == 0) { |
1da177e4 | 772 | mii->reg_num = MII_BMSR; |
4ad41c1e | 773 | if (ioctl(slave_dev, &ifr, SIOCGMIIREG) == 0) |
3d632c3f | 774 | return mii->val_out & BMSR_LSTATUS; |
1da177e4 LT |
775 | } |
776 | } | |
777 | ||
a7605370 | 778 | /* If reporting, report that either there's no ndo_eth_ioctl, |
61a44b9c | 779 | * or both SIOCGMIIREG and get_link failed (meaning that we |
1da177e4 LT |
780 | * cannot report link status). If not reporting, pretend |
781 | * we're ok. | |
782 | */ | |
3d632c3f | 783 | return reporting ? -1 : BMSR_LSTATUS; |
1da177e4 LT |
784 | } |
785 | ||
786 | /*----------------------------- Multicast list ------------------------------*/ | |
787 | ||
547942ca | 788 | /* Push the promiscuity flag down to appropriate slaves */ |
7e1a1ac1 | 789 | static int bond_set_promiscuity(struct bonding *bond, int inc) |
1da177e4 | 790 | { |
9caff1e7 | 791 | struct list_head *iter; |
7e1a1ac1 | 792 | int err = 0; |
9caff1e7 | 793 | |
ec0865a9 | 794 | if (bond_uses_primary(bond)) { |
14056e79 | 795 | struct slave *curr_active = rtnl_dereference(bond->curr_active_slave); |
4740d638 | 796 | |
4740d638 ED |
797 | if (curr_active) |
798 | err = dev_set_promiscuity(curr_active->dev, inc); | |
1da177e4 LT |
799 | } else { |
800 | struct slave *slave; | |
dec1e90e | 801 | |
9caff1e7 | 802 | bond_for_each_slave(bond, slave, iter) { |
7e1a1ac1 WC |
803 | err = dev_set_promiscuity(slave->dev, inc); |
804 | if (err) | |
805 | return err; | |
1da177e4 LT |
806 | } |
807 | } | |
7e1a1ac1 | 808 | return err; |
1da177e4 LT |
809 | } |
810 | ||
547942ca | 811 | /* Push the allmulti flag down to all slaves */ |
7e1a1ac1 | 812 | static int bond_set_allmulti(struct bonding *bond, int inc) |
1da177e4 | 813 | { |
9caff1e7 | 814 | struct list_head *iter; |
7e1a1ac1 | 815 | int err = 0; |
9caff1e7 | 816 | |
ec0865a9 | 817 | if (bond_uses_primary(bond)) { |
14056e79 | 818 | struct slave *curr_active = rtnl_dereference(bond->curr_active_slave); |
4740d638 | 819 | |
4740d638 ED |
820 | if (curr_active) |
821 | err = dev_set_allmulti(curr_active->dev, inc); | |
1da177e4 LT |
822 | } else { |
823 | struct slave *slave; | |
dec1e90e | 824 | |
9caff1e7 | 825 | bond_for_each_slave(bond, slave, iter) { |
7e1a1ac1 WC |
826 | err = dev_set_allmulti(slave->dev, inc); |
827 | if (err) | |
828 | return err; | |
1da177e4 LT |
829 | } |
830 | } | |
7e1a1ac1 | 831 | return err; |
1da177e4 LT |
832 | } |
833 | ||
547942ca | 834 | /* Retrieve the list of registered multicast addresses for the bonding |
5a37e8ca FL |
835 | * device and retransmit an IGMP JOIN request to the current active |
836 | * slave. | |
837 | */ | |
f2369109 | 838 | static void bond_resend_igmp_join_requests_delayed(struct work_struct *work) |
5a37e8ca | 839 | { |
f2369109 | 840 | struct bonding *bond = container_of(work, struct bonding, |
841 | mcast_work.work); | |
842 | ||
4aa5dee4 | 843 | if (!rtnl_trylock()) { |
4beac029 | 844 | queue_delayed_work(bond->wq, &bond->mcast_work, 1); |
4aa5dee4 | 845 | return; |
5a37e8ca | 846 | } |
4aa5dee4 | 847 | call_netdevice_notifiers(NETDEV_RESEND_IGMP, bond->dev); |
5a37e8ca | 848 | |
4f5474e7 NA |
849 | if (bond->igmp_retrans > 1) { |
850 | bond->igmp_retrans--; | |
c2952c31 | 851 | queue_delayed_work(bond->wq, &bond->mcast_work, HZ/5); |
4f5474e7 | 852 | } |
f2369109 | 853 | rtnl_unlock(); |
5a37e8ca FL |
854 | } |
855 | ||
547942ca | 856 | /* Flush bond's hardware addresses from slave */ |
303d1cbf | 857 | static void bond_hw_addr_flush(struct net_device *bond_dev, |
3d632c3f | 858 | struct net_device *slave_dev) |
1da177e4 | 859 | { |
454d7c9b | 860 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 | 861 | |
303d1cbf JV |
862 | dev_uc_unsync(slave_dev, bond_dev); |
863 | dev_mc_unsync(slave_dev, bond_dev); | |
1da177e4 | 864 | |
01844098 | 865 | if (BOND_MODE(bond) == BOND_MODE_8023AD) { |
1da177e4 LT |
866 | /* del lacpdu mc addr from mc list */ |
867 | u8 lacpdu_multicast[ETH_ALEN] = MULTICAST_LACPDU_ADDR; | |
868 | ||
22bedad3 | 869 | dev_mc_del(slave_dev, lacpdu_multicast); |
1da177e4 LT |
870 | } |
871 | } | |
872 | ||
873 | /*--------------------------- Active slave change ---------------------------*/ | |
874 | ||
303d1cbf | 875 | /* Update the hardware address list and promisc/allmulti for the new and |
ec0865a9 VF |
876 | * old active slaves (if any). Modes that are not using primary keep all |
877 | * slaves up date at all times; only the modes that use primary need to call | |
303d1cbf | 878 | * this function to swap these settings during a failover. |
1da177e4 | 879 | */ |
303d1cbf JV |
880 | static void bond_hw_addr_swap(struct bonding *bond, struct slave *new_active, |
881 | struct slave *old_active) | |
1da177e4 | 882 | { |
1da177e4 | 883 | if (old_active) { |
3d632c3f | 884 | if (bond->dev->flags & IFF_PROMISC) |
1da177e4 | 885 | dev_set_promiscuity(old_active->dev, -1); |
1da177e4 | 886 | |
3d632c3f | 887 | if (bond->dev->flags & IFF_ALLMULTI) |
1da177e4 | 888 | dev_set_allmulti(old_active->dev, -1); |
1da177e4 | 889 | |
303d1cbf | 890 | bond_hw_addr_flush(bond->dev, old_active->dev); |
1da177e4 LT |
891 | } |
892 | ||
893 | if (new_active) { | |
7e1a1ac1 | 894 | /* FIXME: Signal errors upstream. */ |
3d632c3f | 895 | if (bond->dev->flags & IFF_PROMISC) |
1da177e4 | 896 | dev_set_promiscuity(new_active->dev, 1); |
1da177e4 | 897 | |
3d632c3f | 898 | if (bond->dev->flags & IFF_ALLMULTI) |
1da177e4 | 899 | dev_set_allmulti(new_active->dev, 1); |
1da177e4 | 900 | |
d632ce98 | 901 | netif_addr_lock_bh(bond->dev); |
303d1cbf JV |
902 | dev_uc_sync(new_active->dev, bond->dev); |
903 | dev_mc_sync(new_active->dev, bond->dev); | |
d632ce98 | 904 | netif_addr_unlock_bh(bond->dev); |
1da177e4 LT |
905 | } |
906 | } | |
907 | ||
ae0d6750 | 908 | /** |
909 | * bond_set_dev_addr - clone slave's address to bond | |
910 | * @bond_dev: bond net device | |
911 | * @slave_dev: slave net device | |
912 | * | |
913 | * Should be called with RTNL held. | |
914 | */ | |
b9245914 PM |
915 | static int bond_set_dev_addr(struct net_device *bond_dev, |
916 | struct net_device *slave_dev) | |
ae0d6750 | 917 | { |
1caf40de PM |
918 | int err; |
919 | ||
e2a7420d JW |
920 | slave_dbg(bond_dev, slave_dev, "bond_dev=%p slave_dev=%p slave_dev->addr_len=%d\n", |
921 | bond_dev, slave_dev, slave_dev->addr_len); | |
1caf40de PM |
922 | err = dev_pre_changeaddr_notify(bond_dev, slave_dev->dev_addr, NULL); |
923 | if (err) | |
924 | return err; | |
925 | ||
ae0d6750 | 926 | memcpy(bond_dev->dev_addr, slave_dev->dev_addr, slave_dev->addr_len); |
927 | bond_dev->addr_assign_type = NET_ADDR_STOLEN; | |
928 | call_netdevice_notifiers(NETDEV_CHANGEADDR, bond_dev); | |
b9245914 | 929 | return 0; |
ae0d6750 | 930 | } |
931 | ||
a951bc1e | 932 | static struct slave *bond_get_old_active(struct bonding *bond, |
933 | struct slave *new_active) | |
934 | { | |
935 | struct slave *slave; | |
936 | struct list_head *iter; | |
937 | ||
938 | bond_for_each_slave(bond, slave, iter) { | |
939 | if (slave == new_active) | |
940 | continue; | |
941 | ||
942 | if (ether_addr_equal(bond->dev->dev_addr, slave->dev->dev_addr)) | |
943 | return slave; | |
944 | } | |
945 | ||
946 | return NULL; | |
947 | } | |
948 | ||
547942ca | 949 | /* bond_do_fail_over_mac |
3915c1e8 JV |
950 | * |
951 | * Perform special MAC address swapping for fail_over_mac settings | |
952 | * | |
1c72cfdc | 953 | * Called with RTNL |
3915c1e8 JV |
954 | */ |
955 | static void bond_do_fail_over_mac(struct bonding *bond, | |
956 | struct slave *new_active, | |
957 | struct slave *old_active) | |
958 | { | |
faeeb317 JW |
959 | u8 tmp_mac[MAX_ADDR_LEN]; |
960 | struct sockaddr_storage ss; | |
3915c1e8 JV |
961 | int rv; |
962 | ||
963 | switch (bond->params.fail_over_mac) { | |
964 | case BOND_FOM_ACTIVE: | |
b9245914 PM |
965 | if (new_active) { |
966 | rv = bond_set_dev_addr(bond->dev, new_active->dev); | |
967 | if (rv) | |
e2a7420d JW |
968 | slave_err(bond->dev, new_active->dev, "Error %d setting bond MAC from slave\n", |
969 | -rv); | |
b9245914 | 970 | } |
3915c1e8 JV |
971 | break; |
972 | case BOND_FOM_FOLLOW: | |
547942ca | 973 | /* if new_active && old_active, swap them |
3915c1e8 JV |
974 | * if just old_active, do nothing (going to no active slave) |
975 | * if just new_active, set new_active to bond's MAC | |
976 | */ | |
977 | if (!new_active) | |
978 | return; | |
979 | ||
a951bc1e | 980 | if (!old_active) |
981 | old_active = bond_get_old_active(bond, new_active); | |
982 | ||
3915c1e8 | 983 | if (old_active) { |
faeeb317 JW |
984 | bond_hw_addr_copy(tmp_mac, new_active->dev->dev_addr, |
985 | new_active->dev->addr_len); | |
986 | bond_hw_addr_copy(ss.__data, | |
987 | old_active->dev->dev_addr, | |
988 | old_active->dev->addr_len); | |
989 | ss.ss_family = new_active->dev->type; | |
3915c1e8 | 990 | } else { |
faeeb317 JW |
991 | bond_hw_addr_copy(ss.__data, bond->dev->dev_addr, |
992 | bond->dev->addr_len); | |
993 | ss.ss_family = bond->dev->type; | |
3915c1e8 JV |
994 | } |
995 | ||
faeeb317 | 996 | rv = dev_set_mac_address(new_active->dev, |
3a37a963 | 997 | (struct sockaddr *)&ss, NULL); |
3915c1e8 | 998 | if (rv) { |
e2a7420d JW |
999 | slave_err(bond->dev, new_active->dev, "Error %d setting MAC of new active slave\n", |
1000 | -rv); | |
3915c1e8 JV |
1001 | goto out; |
1002 | } | |
1003 | ||
1004 | if (!old_active) | |
1005 | goto out; | |
1006 | ||
faeeb317 JW |
1007 | bond_hw_addr_copy(ss.__data, tmp_mac, |
1008 | new_active->dev->addr_len); | |
1009 | ss.ss_family = old_active->dev->type; | |
3915c1e8 | 1010 | |
faeeb317 | 1011 | rv = dev_set_mac_address(old_active->dev, |
3a37a963 | 1012 | (struct sockaddr *)&ss, NULL); |
3915c1e8 | 1013 | if (rv) |
e2a7420d JW |
1014 | slave_err(bond->dev, old_active->dev, "Error %d setting MAC of old active slave\n", |
1015 | -rv); | |
3915c1e8 | 1016 | out: |
3915c1e8 JV |
1017 | break; |
1018 | default: | |
76444f50 VF |
1019 | netdev_err(bond->dev, "bond_do_fail_over_mac impossible: bad policy %d\n", |
1020 | bond->params.fail_over_mac); | |
3915c1e8 JV |
1021 | break; |
1022 | } | |
1023 | ||
1024 | } | |
1025 | ||
b5a983f3 | 1026 | static struct slave *bond_choose_primary_or_current(struct bonding *bond) |
a549952a | 1027 | { |
059b47e8 | 1028 | struct slave *prim = rtnl_dereference(bond->primary_slave); |
1c72cfdc | 1029 | struct slave *curr = rtnl_dereference(bond->curr_active_slave); |
a549952a | 1030 | |
b5a983f3 MR |
1031 | if (!prim || prim->link != BOND_LINK_UP) { |
1032 | if (!curr || curr->link != BOND_LINK_UP) | |
1033 | return NULL; | |
1034 | return curr; | |
1035 | } | |
1036 | ||
a549952a JP |
1037 | if (bond->force_primary) { |
1038 | bond->force_primary = false; | |
b5a983f3 MR |
1039 | return prim; |
1040 | } | |
1041 | ||
1042 | if (!curr || curr->link != BOND_LINK_UP) | |
1043 | return prim; | |
1044 | ||
1045 | /* At this point, prim and curr are both up */ | |
1046 | switch (bond->params.primary_reselect) { | |
1047 | case BOND_PRI_RESELECT_ALWAYS: | |
1048 | return prim; | |
1049 | case BOND_PRI_RESELECT_BETTER: | |
1050 | if (prim->speed < curr->speed) | |
1051 | return curr; | |
1052 | if (prim->speed == curr->speed && prim->duplex <= curr->duplex) | |
1053 | return curr; | |
1054 | return prim; | |
1055 | case BOND_PRI_RESELECT_FAILURE: | |
1056 | return curr; | |
1057 | default: | |
1058 | netdev_err(bond->dev, "impossible primary_reselect %d\n", | |
1059 | bond->params.primary_reselect); | |
1060 | return curr; | |
a549952a | 1061 | } |
a549952a | 1062 | } |
3915c1e8 | 1063 | |
1da177e4 | 1064 | /** |
b5a983f3 | 1065 | * bond_find_best_slave - select the best available slave to be the active one |
1da177e4 | 1066 | * @bond: our bonding struct |
1da177e4 LT |
1067 | */ |
1068 | static struct slave *bond_find_best_slave(struct bonding *bond) | |
1069 | { | |
b5a983f3 | 1070 | struct slave *slave, *bestslave = NULL; |
77140d29 | 1071 | struct list_head *iter; |
1da177e4 | 1072 | int mintime = bond->params.updelay; |
1da177e4 | 1073 | |
b5a983f3 MR |
1074 | slave = bond_choose_primary_or_current(bond); |
1075 | if (slave) | |
1076 | return slave; | |
1da177e4 | 1077 | |
77140d29 VF |
1078 | bond_for_each_slave(bond, slave, iter) { |
1079 | if (slave->link == BOND_LINK_UP) | |
1080 | return slave; | |
b6adc610 | 1081 | if (slave->link == BOND_LINK_BACK && bond_slave_is_up(slave) && |
77140d29 VF |
1082 | slave->delay < mintime) { |
1083 | mintime = slave->delay; | |
1084 | bestslave = slave; | |
1da177e4 LT |
1085 | } |
1086 | } | |
1087 | ||
1088 | return bestslave; | |
1089 | } | |
1090 | ||
ad246c99 BH |
1091 | static bool bond_should_notify_peers(struct bonding *bond) |
1092 | { | |
4cb4f97b | 1093 | struct slave *slave; |
1094 | ||
1095 | rcu_read_lock(); | |
1096 | slave = rcu_dereference(bond->curr_active_slave); | |
1097 | rcu_read_unlock(); | |
ad246c99 | 1098 | |
ad246c99 | 1099 | if (!slave || !bond->send_peer_notif || |
07a4ddec VB |
1100 | bond->send_peer_notif % |
1101 | max(1, bond->params.peer_notif_delay) != 0 || | |
b02e3e94 | 1102 | !netif_carrier_ok(bond->dev) || |
ad246c99 BH |
1103 | test_bit(__LINK_STATE_LINKWATCH_PENDING, &slave->dev->state)) |
1104 | return false; | |
1105 | ||
313e2377 SK |
1106 | netdev_dbg(bond->dev, "bond_should_notify_peers: slave %s\n", |
1107 | slave ? slave->dev->name : "NULL"); | |
1108 | ||
ad246c99 BH |
1109 | return true; |
1110 | } | |
1111 | ||
1da177e4 | 1112 | /** |
acf61b3d | 1113 | * bond_change_active_slave - change the active slave into the specified one |
1da177e4 | 1114 | * @bond: our bonding struct |
45a1553b | 1115 | * @new_active: the new slave to make the active one |
1da177e4 LT |
1116 | * |
1117 | * Set the new slave to the bond's settings and unset them on the old | |
1118 | * curr_active_slave. | |
1119 | * Setting include flags, mc-list, promiscuity, allmulti, etc. | |
1120 | * | |
1121 | * If @new's link state is %BOND_LINK_BACK we'll set it to %BOND_LINK_UP, | |
1122 | * because it is apparently the best available slave we have, even though its | |
1123 | * updelay hasn't timed out yet. | |
1124 | * | |
1c72cfdc | 1125 | * Caller must hold RTNL. |
1da177e4 | 1126 | */ |
a77b5325 | 1127 | void bond_change_active_slave(struct bonding *bond, struct slave *new_active) |
1da177e4 | 1128 | { |
4740d638 ED |
1129 | struct slave *old_active; |
1130 | ||
1c72cfdc NA |
1131 | ASSERT_RTNL(); |
1132 | ||
1133 | old_active = rtnl_dereference(bond->curr_active_slave); | |
1da177e4 | 1134 | |
3d632c3f | 1135 | if (old_active == new_active) |
1da177e4 | 1136 | return; |
1da177e4 | 1137 | |
18cb261a | 1138 | #ifdef CONFIG_XFRM_OFFLOAD |
9a560550 | 1139 | bond_ipsec_del_sa_all(bond); |
18cb261a JW |
1140 | #endif /* CONFIG_XFRM_OFFLOAD */ |
1141 | ||
5cd24cbe | 1142 | if (new_active) { |
8e603460 | 1143 | new_active->last_link_up = jiffies; |
b2220cad | 1144 | |
1da177e4 | 1145 | if (new_active->link == BOND_LINK_BACK) { |
ec0865a9 | 1146 | if (bond_uses_primary(bond)) { |
e2a7420d JW |
1147 | slave_info(bond->dev, new_active->dev, "making interface the new active one %d ms earlier\n", |
1148 | (bond->params.updelay - new_active->delay) * bond->params.miimon); | |
1da177e4 LT |
1149 | } |
1150 | ||
1151 | new_active->delay = 0; | |
5d397061 JP |
1152 | bond_set_slave_link_state(new_active, BOND_LINK_UP, |
1153 | BOND_SLAVE_NOTIFY_NOW); | |
1da177e4 | 1154 | |
01844098 | 1155 | if (BOND_MODE(bond) == BOND_MODE_8023AD) |
1da177e4 | 1156 | bond_3ad_handle_link_change(new_active, BOND_LINK_UP); |
1da177e4 | 1157 | |
58402054 | 1158 | if (bond_is_lb(bond)) |
1da177e4 | 1159 | bond_alb_handle_link_change(bond, new_active, BOND_LINK_UP); |
1da177e4 | 1160 | } else { |
52333512 | 1161 | if (bond_uses_primary(bond)) |
e2a7420d | 1162 | slave_info(bond->dev, new_active->dev, "making interface the new active one\n"); |
1da177e4 LT |
1163 | } |
1164 | } | |
1165 | ||
ec0865a9 | 1166 | if (bond_uses_primary(bond)) |
303d1cbf | 1167 | bond_hw_addr_swap(bond, new_active, old_active); |
1da177e4 | 1168 | |
58402054 | 1169 | if (bond_is_lb(bond)) { |
1da177e4 | 1170 | bond_alb_handle_active_change(bond, new_active); |
8f903c70 | 1171 | if (old_active) |
5e5b0665 | 1172 | bond_set_slave_inactive_flags(old_active, |
1173 | BOND_SLAVE_NOTIFY_NOW); | |
8f903c70 | 1174 | if (new_active) |
5e5b0665 | 1175 | bond_set_slave_active_flags(new_active, |
1176 | BOND_SLAVE_NOTIFY_NOW); | |
1da177e4 | 1177 | } else { |
278b2083 | 1178 | rcu_assign_pointer(bond->curr_active_slave, new_active); |
1da177e4 | 1179 | } |
c3ade5ca | 1180 | |
01844098 | 1181 | if (BOND_MODE(bond) == BOND_MODE_ACTIVEBACKUP) { |
3d632c3f | 1182 | if (old_active) |
5e5b0665 | 1183 | bond_set_slave_inactive_flags(old_active, |
1184 | BOND_SLAVE_NOTIFY_NOW); | |
c3ade5ca JV |
1185 | |
1186 | if (new_active) { | |
ad246c99 BH |
1187 | bool should_notify_peers = false; |
1188 | ||
5e5b0665 | 1189 | bond_set_slave_active_flags(new_active, |
1190 | BOND_SLAVE_NOTIFY_NOW); | |
2ab82852 | 1191 | |
709f8a45 OG |
1192 | if (bond->params.fail_over_mac) |
1193 | bond_do_fail_over_mac(bond, new_active, | |
1194 | old_active); | |
3915c1e8 | 1195 | |
ad246c99 BH |
1196 | if (netif_running(bond->dev)) { |
1197 | bond->send_peer_notif = | |
07a4ddec VB |
1198 | bond->params.num_peer_notif * |
1199 | max(1, bond->params.peer_notif_delay); | |
ad246c99 BH |
1200 | should_notify_peers = |
1201 | bond_should_notify_peers(bond); | |
1202 | } | |
1203 | ||
b7bc2a5b | 1204 | call_netdevice_notifiers(NETDEV_BONDING_FAILOVER, bond->dev); |
07a4ddec VB |
1205 | if (should_notify_peers) { |
1206 | bond->send_peer_notif--; | |
b7bc2a5b AW |
1207 | call_netdevice_notifiers(NETDEV_NOTIFY_PEERS, |
1208 | bond->dev); | |
07a4ddec | 1209 | } |
7893b249 | 1210 | } |
c3ade5ca | 1211 | } |
a2fd940f | 1212 | |
5cd24cbe | 1213 | #ifdef CONFIG_XFRM_OFFLOAD |
9a560550 | 1214 | bond_ipsec_add_sa_all(bond); |
5cd24cbe JW |
1215 | #endif /* CONFIG_XFRM_OFFLOAD */ |
1216 | ||
5a37e8ca | 1217 | /* resend IGMP joins since active slave has changed or |
94265cf5 FL |
1218 | * all were sent on curr_active_slave. |
1219 | * resend only if bond is brought up with the affected | |
547942ca NA |
1220 | * bonding modes and the retransmission is enabled |
1221 | */ | |
94265cf5 | 1222 | if (netif_running(bond->dev) && (bond->params.resend_igmp > 0) && |
ec0865a9 | 1223 | ((bond_uses_primary(bond) && new_active) || |
01844098 | 1224 | BOND_MODE(bond) == BOND_MODE_ROUNDROBIN)) { |
c2952c31 | 1225 | bond->igmp_retrans = bond->params.resend_igmp; |
4beac029 | 1226 | queue_delayed_work(bond->wq, &bond->mcast_work, 1); |
a2fd940f | 1227 | } |
1da177e4 LT |
1228 | } |
1229 | ||
1230 | /** | |
1231 | * bond_select_active_slave - select a new active slave, if needed | |
1232 | * @bond: our bonding struct | |
1233 | * | |
3d632c3f | 1234 | * This functions should be called when one of the following occurs: |
1da177e4 LT |
1235 | * - The old curr_active_slave has been released or lost its link. |
1236 | * - The primary_slave has got its link back. | |
1237 | * - A slave has got its link back and there's no old curr_active_slave. | |
1238 | * | |
1c72cfdc | 1239 | * Caller must hold RTNL. |
1da177e4 | 1240 | */ |
a77b5325 | 1241 | void bond_select_active_slave(struct bonding *bond) |
1da177e4 LT |
1242 | { |
1243 | struct slave *best_slave; | |
ff59c456 | 1244 | int rv; |
1da177e4 | 1245 | |
e0974585 NA |
1246 | ASSERT_RTNL(); |
1247 | ||
1da177e4 | 1248 | best_slave = bond_find_best_slave(bond); |
1c72cfdc | 1249 | if (best_slave != rtnl_dereference(bond->curr_active_slave)) { |
1da177e4 | 1250 | bond_change_active_slave(bond, best_slave); |
ff59c456 JV |
1251 | rv = bond_set_carrier(bond); |
1252 | if (!rv) | |
1253 | return; | |
1254 | ||
d66bd905 | 1255 | if (netif_carrier_ok(bond->dev)) |
b8bd72d3 | 1256 | netdev_info(bond->dev, "active interface up!\n"); |
d66bd905 | 1257 | else |
76444f50 | 1258 | netdev_info(bond->dev, "now running without any active interface!\n"); |
1da177e4 LT |
1259 | } |
1260 | } | |
1261 | ||
f6dc31a8 | 1262 | #ifdef CONFIG_NET_POLL_CONTROLLER |
8a8efa22 | 1263 | static inline int slave_enable_netpoll(struct slave *slave) |
f6dc31a8 | 1264 | { |
8a8efa22 AW |
1265 | struct netpoll *np; |
1266 | int err = 0; | |
f6dc31a8 | 1267 | |
a8779ec1 | 1268 | np = kzalloc(sizeof(*np), GFP_KERNEL); |
8a8efa22 AW |
1269 | err = -ENOMEM; |
1270 | if (!np) | |
1271 | goto out; | |
1272 | ||
a8779ec1 | 1273 | err = __netpoll_setup(np, slave->dev); |
8a8efa22 AW |
1274 | if (err) { |
1275 | kfree(np); | |
1276 | goto out; | |
f6dc31a8 | 1277 | } |
8a8efa22 AW |
1278 | slave->np = np; |
1279 | out: | |
1280 | return err; | |
1281 | } | |
1282 | static inline void slave_disable_netpoll(struct slave *slave) | |
1283 | { | |
1284 | struct netpoll *np = slave->np; | |
1285 | ||
1286 | if (!np) | |
1287 | return; | |
1288 | ||
1289 | slave->np = NULL; | |
c9fbd71f DB |
1290 | |
1291 | __netpoll_free(np); | |
8a8efa22 | 1292 | } |
f6dc31a8 WC |
1293 | |
1294 | static void bond_poll_controller(struct net_device *bond_dev) | |
1295 | { | |
616f4541 MB |
1296 | struct bonding *bond = netdev_priv(bond_dev); |
1297 | struct slave *slave = NULL; | |
1298 | struct list_head *iter; | |
1299 | struct ad_info ad_info; | |
616f4541 MB |
1300 | |
1301 | if (BOND_MODE(bond) == BOND_MODE_8023AD) | |
1302 | if (bond_3ad_get_active_agg_info(bond, &ad_info)) | |
1303 | return; | |
1304 | ||
616f4541 | 1305 | bond_for_each_slave_rcu(bond, slave, iter) { |
93f62ad5 | 1306 | if (!bond_slave_is_up(slave)) |
616f4541 MB |
1307 | continue; |
1308 | ||
1309 | if (BOND_MODE(bond) == BOND_MODE_8023AD) { | |
1310 | struct aggregator *agg = | |
1311 | SLAVE_AD_INFO(slave)->port.aggregator; | |
1312 | ||
1313 | if (agg && | |
1314 | agg->aggregator_identifier != ad_info.aggregator_id) | |
1315 | continue; | |
1316 | } | |
1317 | ||
93f62ad5 | 1318 | netpoll_poll_dev(slave->dev); |
616f4541 | 1319 | } |
8a8efa22 AW |
1320 | } |
1321 | ||
c4cdef9b | 1322 | static void bond_netpoll_cleanup(struct net_device *bond_dev) |
8a8efa22 | 1323 | { |
c4cdef9b | 1324 | struct bonding *bond = netdev_priv(bond_dev); |
9caff1e7 | 1325 | struct list_head *iter; |
c2355e1a | 1326 | struct slave *slave; |
c2355e1a | 1327 | |
9caff1e7 | 1328 | bond_for_each_slave(bond, slave, iter) |
b6adc610 | 1329 | if (bond_slave_is_up(slave)) |
8a8efa22 | 1330 | slave_disable_netpoll(slave); |
f6dc31a8 | 1331 | } |
8a8efa22 | 1332 | |
a8779ec1 | 1333 | static int bond_netpoll_setup(struct net_device *dev, struct netpoll_info *ni) |
8a8efa22 AW |
1334 | { |
1335 | struct bonding *bond = netdev_priv(dev); | |
9caff1e7 | 1336 | struct list_head *iter; |
f6dc31a8 | 1337 | struct slave *slave; |
dec1e90e | 1338 | int err = 0; |
f6dc31a8 | 1339 | |
9caff1e7 | 1340 | bond_for_each_slave(bond, slave, iter) { |
8a8efa22 AW |
1341 | err = slave_enable_netpoll(slave); |
1342 | if (err) { | |
c4cdef9b | 1343 | bond_netpoll_cleanup(dev); |
8a8efa22 | 1344 | break; |
f6dc31a8 WC |
1345 | } |
1346 | } | |
8a8efa22 | 1347 | return err; |
f6dc31a8 | 1348 | } |
8a8efa22 AW |
1349 | #else |
1350 | static inline int slave_enable_netpoll(struct slave *slave) | |
1351 | { | |
1352 | return 0; | |
1353 | } | |
1354 | static inline void slave_disable_netpoll(struct slave *slave) | |
1355 | { | |
1356 | } | |
f6dc31a8 WC |
1357 | static void bond_netpoll_cleanup(struct net_device *bond_dev) |
1358 | { | |
1359 | } | |
f6dc31a8 WC |
1360 | #endif |
1361 | ||
1da177e4 LT |
1362 | /*---------------------------------- IOCTL ----------------------------------*/ |
1363 | ||
c8f44aff | 1364 | static netdev_features_t bond_fix_features(struct net_device *dev, |
9b7b165a | 1365 | netdev_features_t features) |
8531c5ff | 1366 | { |
b2a103e6 | 1367 | struct bonding *bond = netdev_priv(dev); |
9caff1e7 | 1368 | struct list_head *iter; |
c8f44aff | 1369 | netdev_features_t mask; |
9b7b165a | 1370 | struct slave *slave; |
b63365a2 | 1371 | |
89df6a81 TT |
1372 | #if IS_ENABLED(CONFIG_TLS_DEVICE) |
1373 | if (bond_sk_check(bond)) | |
1374 | features |= BOND_TLS_FEATURES; | |
1375 | else | |
1376 | features &= ~BOND_TLS_FEATURES; | |
1377 | #endif | |
1378 | ||
7889cbee | 1379 | mask = features; |
c158cba3 | 1380 | |
b63365a2 | 1381 | features &= ~NETIF_F_ONE_FOR_ALL; |
b2a103e6 | 1382 | features |= NETIF_F_ALL_FOR_ALL; |
7f353bf2 | 1383 | |
9caff1e7 | 1384 | bond_for_each_slave(bond, slave, iter) { |
b63365a2 HX |
1385 | features = netdev_increment_features(features, |
1386 | slave->dev->features, | |
b2a103e6 MM |
1387 | mask); |
1388 | } | |
b0ce3508 | 1389 | features = netdev_add_tso_features(features, mask); |
b2a103e6 | 1390 | |
b2a103e6 MM |
1391 | return features; |
1392 | } | |
1393 | ||
a188222b | 1394 | #define BOND_VLAN_FEATURES (NETIF_F_HW_CSUM | NETIF_F_SG | \ |
ecb8fed4 | 1395 | NETIF_F_FRAGLIST | NETIF_F_GSO_SOFTWARE | \ |
62f2a3a4 | 1396 | NETIF_F_HIGHDMA | NETIF_F_LRO) |
b2a103e6 | 1397 | |
a188222b | 1398 | #define BOND_ENC_FEATURES (NETIF_F_HW_CSUM | NETIF_F_SG | \ |
ecb8fed4 | 1399 | NETIF_F_RXCSUM | NETIF_F_GSO_SOFTWARE) |
5a7baa78 | 1400 | |
2e770b50 | 1401 | #define BOND_MPLS_FEATURES (NETIF_F_HW_CSUM | NETIF_F_SG | \ |
ecb8fed4 | 1402 | NETIF_F_GSO_SOFTWARE) |
2e770b50 | 1403 | |
18cb261a | 1404 | |
b2a103e6 MM |
1405 | static void bond_compute_features(struct bonding *bond) |
1406 | { | |
02875878 ED |
1407 | unsigned int dst_release_flag = IFF_XMIT_DST_RELEASE | |
1408 | IFF_XMIT_DST_RELEASE_PERM; | |
c8f44aff | 1409 | netdev_features_t vlan_features = BOND_VLAN_FEATURES; |
5a7baa78 | 1410 | netdev_features_t enc_features = BOND_ENC_FEATURES; |
18cb261a JW |
1411 | #ifdef CONFIG_XFRM_OFFLOAD |
1412 | netdev_features_t xfrm_features = BOND_XFRM_FEATURES; | |
1413 | #endif /* CONFIG_XFRM_OFFLOAD */ | |
2e770b50 | 1414 | netdev_features_t mpls_features = BOND_MPLS_FEATURES; |
9caff1e7 VF |
1415 | struct net_device *bond_dev = bond->dev; |
1416 | struct list_head *iter; | |
1417 | struct slave *slave; | |
b2a103e6 | 1418 | unsigned short max_hard_header_len = ETH_HLEN; |
0e376bd0 SB |
1419 | unsigned int gso_max_size = GSO_MAX_SIZE; |
1420 | u16 gso_max_segs = GSO_MAX_SEGS; | |
b2a103e6 | 1421 | |
0965a1f3 | 1422 | if (!bond_has_slaves(bond)) |
b2a103e6 | 1423 | goto done; |
a9b3ace4 | 1424 | vlan_features &= NETIF_F_ALL_FOR_ALL; |
2e770b50 | 1425 | mpls_features &= NETIF_F_ALL_FOR_ALL; |
b2a103e6 | 1426 | |
9caff1e7 | 1427 | bond_for_each_slave(bond, slave, iter) { |
278339a4 | 1428 | vlan_features = netdev_increment_features(vlan_features, |
b2a103e6 MM |
1429 | slave->dev->vlan_features, BOND_VLAN_FEATURES); |
1430 | ||
5a7baa78 OG |
1431 | enc_features = netdev_increment_features(enc_features, |
1432 | slave->dev->hw_enc_features, | |
1433 | BOND_ENC_FEATURES); | |
2e770b50 | 1434 | |
18cb261a JW |
1435 | #ifdef CONFIG_XFRM_OFFLOAD |
1436 | xfrm_features = netdev_increment_features(xfrm_features, | |
1437 | slave->dev->hw_enc_features, | |
1438 | BOND_XFRM_FEATURES); | |
1439 | #endif /* CONFIG_XFRM_OFFLOAD */ | |
1440 | ||
2e770b50 AL |
1441 | mpls_features = netdev_increment_features(mpls_features, |
1442 | slave->dev->mpls_features, | |
1443 | BOND_MPLS_FEATURES); | |
1444 | ||
b6fe83e9 | 1445 | dst_release_flag &= slave->dev->priv_flags; |
54ef3137 JV |
1446 | if (slave->dev->hard_header_len > max_hard_header_len) |
1447 | max_hard_header_len = slave->dev->hard_header_len; | |
0e376bd0 SB |
1448 | |
1449 | gso_max_size = min(gso_max_size, slave->dev->gso_max_size); | |
1450 | gso_max_segs = min(gso_max_segs, slave->dev->gso_max_segs); | |
54ef3137 | 1451 | } |
19cdead3 | 1452 | bond_dev->hard_header_len = max_hard_header_len; |
8531c5ff | 1453 | |
b63365a2 | 1454 | done: |
b2a103e6 | 1455 | bond_dev->vlan_features = vlan_features; |
8eea1ca8 | 1456 | bond_dev->hw_enc_features = enc_features | NETIF_F_GSO_ENCAP_ALL | |
d595b03d | 1457 | NETIF_F_HW_VLAN_CTAG_TX | |
ecb8fed4 | 1458 | NETIF_F_HW_VLAN_STAG_TX; |
18cb261a JW |
1459 | #ifdef CONFIG_XFRM_OFFLOAD |
1460 | bond_dev->hw_enc_features |= xfrm_features; | |
1461 | #endif /* CONFIG_XFRM_OFFLOAD */ | |
2e770b50 | 1462 | bond_dev->mpls_features = mpls_features; |
0e376bd0 SB |
1463 | bond_dev->gso_max_segs = gso_max_segs; |
1464 | netif_set_gso_max_size(bond_dev, gso_max_size); | |
8531c5ff | 1465 | |
02875878 ED |
1466 | bond_dev->priv_flags &= ~IFF_XMIT_DST_RELEASE; |
1467 | if ((bond_dev->priv_flags & IFF_XMIT_DST_RELEASE_PERM) && | |
1468 | dst_release_flag == (IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM)) | |
1469 | bond_dev->priv_flags |= IFF_XMIT_DST_RELEASE; | |
b6fe83e9 | 1470 | |
b2a103e6 | 1471 | netdev_change_features(bond_dev); |
8531c5ff AK |
1472 | } |
1473 | ||
872254dd MS |
1474 | static void bond_setup_by_slave(struct net_device *bond_dev, |
1475 | struct net_device *slave_dev) | |
1476 | { | |
00829823 | 1477 | bond_dev->header_ops = slave_dev->header_ops; |
872254dd MS |
1478 | |
1479 | bond_dev->type = slave_dev->type; | |
1480 | bond_dev->hard_header_len = slave_dev->hard_header_len; | |
f32f1933 | 1481 | bond_dev->needed_headroom = slave_dev->needed_headroom; |
872254dd MS |
1482 | bond_dev->addr_len = slave_dev->addr_len; |
1483 | ||
1484 | memcpy(bond_dev->broadcast, slave_dev->broadcast, | |
1485 | slave_dev->addr_len); | |
1486 | } | |
1487 | ||
5b2c4dd2 | 1488 | /* On bonding slaves other than the currently active slave, suppress |
3aba891d | 1489 | * duplicates except for alb non-mcast/bcast. |
5b2c4dd2 JP |
1490 | */ |
1491 | static bool bond_should_deliver_exact_match(struct sk_buff *skb, | |
0bd80dad JP |
1492 | struct slave *slave, |
1493 | struct bonding *bond) | |
5b2c4dd2 | 1494 | { |
2d7011ca | 1495 | if (bond_is_slave_inactive(slave)) { |
01844098 | 1496 | if (BOND_MODE(bond) == BOND_MODE_ALB && |
5b2c4dd2 JP |
1497 | skb->pkt_type != PACKET_BROADCAST && |
1498 | skb->pkt_type != PACKET_MULTICAST) | |
5b2c4dd2 | 1499 | return false; |
5b2c4dd2 JP |
1500 | return true; |
1501 | } | |
1502 | return false; | |
1503 | } | |
1504 | ||
8a4eb573 | 1505 | static rx_handler_result_t bond_handle_frame(struct sk_buff **pskb) |
5b2c4dd2 | 1506 | { |
8a4eb573 | 1507 | struct sk_buff *skb = *pskb; |
f1c1775a | 1508 | struct slave *slave; |
0bd80dad | 1509 | struct bonding *bond; |
de063b70 ED |
1510 | int (*recv_probe)(const struct sk_buff *, struct bonding *, |
1511 | struct slave *); | |
13a8e0c8 | 1512 | int ret = RX_HANDLER_ANOTHER; |
5b2c4dd2 | 1513 | |
8a4eb573 JP |
1514 | skb = skb_share_check(skb, GFP_ATOMIC); |
1515 | if (unlikely(!skb)) | |
1516 | return RX_HANDLER_CONSUMED; | |
1517 | ||
1518 | *pskb = skb; | |
5b2c4dd2 | 1519 | |
35d48903 JP |
1520 | slave = bond_slave_get_rcu(skb->dev); |
1521 | bond = slave->bond; | |
0bd80dad | 1522 | |
6aa7de05 | 1523 | recv_probe = READ_ONCE(bond->recv_probe); |
4d97480b | 1524 | if (recv_probe) { |
de063b70 ED |
1525 | ret = recv_probe(skb, bond, slave); |
1526 | if (ret == RX_HANDLER_CONSUMED) { | |
1527 | consume_skb(skb); | |
1528 | return ret; | |
3aba891d JP |
1529 | } |
1530 | } | |
1531 | ||
3c963a33 MS |
1532 | /* |
1533 | * For packets determined by bond_should_deliver_exact_match() call to | |
1534 | * be suppressed we want to make an exception for link-local packets. | |
1535 | * This is necessary for e.g. LLDP daemons to be able to monitor | |
1536 | * inactive slave links without being forced to bind to them | |
1537 | * explicitly. | |
1538 | * | |
1539 | * At the same time, packets that are passed to the bonding master | |
1540 | * (including link-local ones) can have their originating interface | |
1541 | * determined via PACKET_ORIGDEV socket option. | |
6a9e461f | 1542 | */ |
3c963a33 MS |
1543 | if (bond_should_deliver_exact_match(skb, slave, bond)) { |
1544 | if (is_link_local_ether_addr(eth_hdr(skb)->h_dest)) | |
1545 | return RX_HANDLER_PASS; | |
8a4eb573 | 1546 | return RX_HANDLER_EXACT; |
3c963a33 | 1547 | } |
5b2c4dd2 | 1548 | |
35d48903 | 1549 | skb->dev = bond->dev; |
5b2c4dd2 | 1550 | |
01844098 | 1551 | if (BOND_MODE(bond) == BOND_MODE_ALB && |
2e92a2d0 | 1552 | netif_is_bridge_port(bond->dev) && |
5b2c4dd2 | 1553 | skb->pkt_type == PACKET_HOST) { |
5b2c4dd2 | 1554 | |
541ac7c9 CG |
1555 | if (unlikely(skb_cow_head(skb, |
1556 | skb->data - skb_mac_header(skb)))) { | |
1557 | kfree_skb(skb); | |
8a4eb573 | 1558 | return RX_HANDLER_CONSUMED; |
541ac7c9 | 1559 | } |
faeeb317 JW |
1560 | bond_hw_addr_copy(eth_hdr(skb)->h_dest, bond->dev->dev_addr, |
1561 | bond->dev->addr_len); | |
5b2c4dd2 JP |
1562 | } |
1563 | ||
13a8e0c8 | 1564 | return ret; |
5b2c4dd2 JP |
1565 | } |
1566 | ||
41f0b049 | 1567 | static enum netdev_lag_tx_type bond_lag_tx_type(struct bonding *bond) |
471cb5a3 | 1568 | { |
41f0b049 JP |
1569 | switch (BOND_MODE(bond)) { |
1570 | case BOND_MODE_ROUNDROBIN: | |
1571 | return NETDEV_LAG_TX_TYPE_ROUNDROBIN; | |
1572 | case BOND_MODE_ACTIVEBACKUP: | |
1573 | return NETDEV_LAG_TX_TYPE_ACTIVEBACKUP; | |
1574 | case BOND_MODE_BROADCAST: | |
1575 | return NETDEV_LAG_TX_TYPE_BROADCAST; | |
1576 | case BOND_MODE_XOR: | |
1577 | case BOND_MODE_8023AD: | |
1578 | return NETDEV_LAG_TX_TYPE_HASH; | |
1579 | default: | |
1580 | return NETDEV_LAG_TX_TYPE_UNKNOWN; | |
1581 | } | |
1582 | } | |
1583 | ||
f44aa9ef JH |
1584 | static enum netdev_lag_hash bond_lag_hash_type(struct bonding *bond, |
1585 | enum netdev_lag_tx_type type) | |
1586 | { | |
1587 | if (type != NETDEV_LAG_TX_TYPE_HASH) | |
1588 | return NETDEV_LAG_HASH_NONE; | |
1589 | ||
1590 | switch (bond->params.xmit_policy) { | |
1591 | case BOND_XMIT_POLICY_LAYER2: | |
1592 | return NETDEV_LAG_HASH_L2; | |
1593 | case BOND_XMIT_POLICY_LAYER34: | |
1594 | return NETDEV_LAG_HASH_L34; | |
1595 | case BOND_XMIT_POLICY_LAYER23: | |
1596 | return NETDEV_LAG_HASH_L23; | |
1597 | case BOND_XMIT_POLICY_ENCAP23: | |
1598 | return NETDEV_LAG_HASH_E23; | |
1599 | case BOND_XMIT_POLICY_ENCAP34: | |
1600 | return NETDEV_LAG_HASH_E34; | |
7b8fc010 JW |
1601 | case BOND_XMIT_POLICY_VLAN_SRCMAC: |
1602 | return NETDEV_LAG_HASH_VLAN_SRCMAC; | |
f44aa9ef JH |
1603 | default: |
1604 | return NETDEV_LAG_HASH_UNKNOWN; | |
1605 | } | |
1606 | } | |
1607 | ||
42ab19ee DA |
1608 | static int bond_master_upper_dev_link(struct bonding *bond, struct slave *slave, |
1609 | struct netlink_ext_ack *extack) | |
41f0b049 JP |
1610 | { |
1611 | struct netdev_lag_upper_info lag_upper_info; | |
f44aa9ef | 1612 | enum netdev_lag_tx_type type; |
471cb5a3 | 1613 | |
f44aa9ef JH |
1614 | type = bond_lag_tx_type(bond); |
1615 | lag_upper_info.tx_type = type; | |
1616 | lag_upper_info.hash_type = bond_lag_hash_type(bond, type); | |
4597efe3 XL |
1617 | |
1618 | return netdev_master_upper_dev_link(slave->dev, bond->dev, slave, | |
1619 | &lag_upper_info, extack); | |
471cb5a3 JP |
1620 | } |
1621 | ||
41f0b049 | 1622 | static void bond_upper_dev_unlink(struct bonding *bond, struct slave *slave) |
471cb5a3 | 1623 | { |
41f0b049 JP |
1624 | netdev_upper_dev_unlink(slave->dev, bond->dev); |
1625 | slave->dev->flags &= ~IFF_SLAVE; | |
471cb5a3 JP |
1626 | } |
1627 | ||
b9ad3e9f JI |
1628 | static void slave_kobj_release(struct kobject *kobj) |
1629 | { | |
1630 | struct slave *slave = to_slave(kobj); | |
1631 | struct bonding *bond = bond_get_bond_by_slave(slave); | |
1632 | ||
1633 | cancel_delayed_work_sync(&slave->notify_work); | |
1634 | if (BOND_MODE(bond) == BOND_MODE_8023AD) | |
1635 | kfree(SLAVE_AD_INFO(slave)); | |
1636 | ||
1637 | kfree(slave); | |
1638 | } | |
1639 | ||
1640 | static struct kobj_type slave_ktype = { | |
1641 | .release = slave_kobj_release, | |
1642 | #ifdef CONFIG_SYSFS | |
1643 | .sysfs_ops = &slave_sysfs_ops, | |
1644 | #endif | |
1645 | }; | |
1646 | ||
1647 | static int bond_kobj_init(struct slave *slave) | |
1648 | { | |
1649 | int err; | |
1650 | ||
1651 | err = kobject_init_and_add(&slave->kobj, &slave_ktype, | |
1652 | &(slave->dev->dev.kobj), "bonding_slave"); | |
1653 | if (err) | |
1654 | kobject_put(&slave->kobj); | |
1655 | ||
1656 | return err; | |
1657 | } | |
1658 | ||
1659 | static struct slave *bond_alloc_slave(struct bonding *bond, | |
1660 | struct net_device *slave_dev) | |
3fdddd85 | 1661 | { |
1662 | struct slave *slave = NULL; | |
1663 | ||
d66bd905 | 1664 | slave = kzalloc(sizeof(*slave), GFP_KERNEL); |
3fdddd85 | 1665 | if (!slave) |
1666 | return NULL; | |
1667 | ||
b9ad3e9f JI |
1668 | slave->bond = bond; |
1669 | slave->dev = slave_dev; | |
35d96e63 | 1670 | INIT_DELAYED_WORK(&slave->notify_work, bond_netdev_notify_work); |
b9ad3e9f JI |
1671 | |
1672 | if (bond_kobj_init(slave)) | |
1673 | return NULL; | |
1674 | ||
01844098 | 1675 | if (BOND_MODE(bond) == BOND_MODE_8023AD) { |
3fdddd85 | 1676 | SLAVE_AD_INFO(slave) = kzalloc(sizeof(struct ad_slave_info), |
1677 | GFP_KERNEL); | |
1678 | if (!SLAVE_AD_INFO(slave)) { | |
b9ad3e9f | 1679 | kobject_put(&slave->kobj); |
3fdddd85 | 1680 | return NULL; |
1681 | } | |
1682 | } | |
d4859d74 | 1683 | |
3fdddd85 | 1684 | return slave; |
1685 | } | |
1686 | ||
69a2338e MS |
1687 | static void bond_fill_ifbond(struct bonding *bond, struct ifbond *info) |
1688 | { | |
1689 | info->bond_mode = BOND_MODE(bond); | |
1690 | info->miimon = bond->params.miimon; | |
1691 | info->num_slaves = bond->slave_cnt; | |
1692 | } | |
1693 | ||
1694 | static void bond_fill_ifslave(struct slave *slave, struct ifslave *info) | |
1695 | { | |
1696 | strcpy(info->slave_name, slave->dev->name); | |
1697 | info->link = slave->link; | |
1698 | info->state = bond_slave_state(slave); | |
1699 | info->link_failure_count = slave->link_failure_count; | |
1700 | } | |
1701 | ||
69e61133 MS |
1702 | static void bond_netdev_notify_work(struct work_struct *_work) |
1703 | { | |
d4859d74 MB |
1704 | struct slave *slave = container_of(_work, struct slave, |
1705 | notify_work.work); | |
1706 | ||
1707 | if (rtnl_trylock()) { | |
1708 | struct netdev_bonding_info binfo; | |
69e61133 | 1709 | |
d4859d74 MB |
1710 | bond_fill_ifslave(slave, &binfo.slave); |
1711 | bond_fill_ifbond(slave->bond, &binfo.master); | |
1712 | netdev_bonding_info_change(slave->dev, &binfo); | |
1713 | rtnl_unlock(); | |
1714 | } else { | |
1715 | queue_delayed_work(slave->bond->wq, &slave->notify_work, 1); | |
1716 | } | |
69e61133 MS |
1717 | } |
1718 | ||
1719 | void bond_queue_slave_event(struct slave *slave) | |
1720 | { | |
d4859d74 | 1721 | queue_delayed_work(slave->bond->wq, &slave->notify_work, 0); |
69e61133 MS |
1722 | } |
1723 | ||
f7c7eb7f JP |
1724 | void bond_lower_state_changed(struct slave *slave) |
1725 | { | |
1726 | struct netdev_lag_lower_state_info info; | |
1727 | ||
1728 | info.link_up = slave->link == BOND_LINK_UP || | |
1729 | slave->link == BOND_LINK_FAIL; | |
1730 | info.tx_enabled = bond_is_active_slave(slave); | |
1731 | netdev_lower_state_changed(slave->dev, &info); | |
1732 | } | |
1733 | ||
6569fa2d JT |
1734 | #define BOND_NL_ERR(bond_dev, extack, errmsg) do { \ |
1735 | if (extack) \ | |
1736 | NL_SET_ERR_MSG(extack, errmsg); \ | |
1737 | else \ | |
1738 | netdev_err(bond_dev, "Error: %s\n", errmsg); \ | |
1739 | } while (0) | |
1740 | ||
1741 | #define SLAVE_NL_ERR(bond_dev, slave_dev, extack, errmsg) do { \ | |
1742 | if (extack) \ | |
1743 | NL_SET_ERR_MSG(extack, errmsg); \ | |
1744 | else \ | |
1745 | slave_err(bond_dev, slave_dev, "Error: %s\n", errmsg); \ | |
1746 | } while (0) | |
1747 | ||
1da177e4 | 1748 | /* enslave device <slave> to bond device <master> */ |
33eaf2a6 DA |
1749 | int bond_enslave(struct net_device *bond_dev, struct net_device *slave_dev, |
1750 | struct netlink_ext_ack *extack) | |
1da177e4 | 1751 | { |
454d7c9b | 1752 | struct bonding *bond = netdev_priv(bond_dev); |
eb7cc59a | 1753 | const struct net_device_ops *slave_ops = slave_dev->netdev_ops; |
c8c23903 | 1754 | struct slave *new_slave = NULL, *prev_slave; |
faeeb317 | 1755 | struct sockaddr_storage ss; |
1da177e4 | 1756 | int link_reporting; |
8599b52e | 1757 | int res = 0, i; |
1da177e4 | 1758 | |
4d293fe1 DZ |
1759 | if (slave_dev->flags & IFF_MASTER && |
1760 | !netif_is_bond_master(slave_dev)) { | |
6569fa2d | 1761 | BOND_NL_ERR(bond_dev, extack, |
1b3f78df | 1762 | "Device type (master device) cannot be enslaved"); |
3c9ef511 DZ |
1763 | return -EPERM; |
1764 | } | |
1765 | ||
c772dde3 BH |
1766 | if (!bond->params.use_carrier && |
1767 | slave_dev->ethtool_ops->get_link == NULL && | |
a7605370 | 1768 | slave_ops->ndo_eth_ioctl == NULL) { |
e2a7420d | 1769 | slave_warn(bond_dev, slave_dev, "no link monitoring support\n"); |
1da177e4 LT |
1770 | } |
1771 | ||
24b27fc4 MB |
1772 | /* already in-use? */ |
1773 | if (netdev_is_rx_handler_busy(slave_dev)) { | |
6569fa2d JT |
1774 | SLAVE_NL_ERR(bond_dev, slave_dev, extack, |
1775 | "Device is in use and cannot be enslaved"); | |
1da177e4 LT |
1776 | return -EBUSY; |
1777 | } | |
1778 | ||
09a89c21 | 1779 | if (bond_dev == slave_dev) { |
6569fa2d | 1780 | BOND_NL_ERR(bond_dev, extack, "Cannot enslave bond to itself."); |
09a89c21 JB |
1781 | return -EPERM; |
1782 | } | |
1783 | ||
1da177e4 LT |
1784 | /* vlan challenged mutual exclusion */ |
1785 | /* no need to lock since we're protected by rtnl_lock */ | |
1786 | if (slave_dev->features & NETIF_F_VLAN_CHALLENGED) { | |
e2a7420d | 1787 | slave_dbg(bond_dev, slave_dev, "is NETIF_F_VLAN_CHALLENGED\n"); |
55462cf3 | 1788 | if (vlan_uses_dev(bond_dev)) { |
6569fa2d JT |
1789 | SLAVE_NL_ERR(bond_dev, slave_dev, extack, |
1790 | "Can not enslave VLAN challenged device to VLAN enabled bond"); | |
1da177e4 LT |
1791 | return -EPERM; |
1792 | } else { | |
e2a7420d | 1793 | slave_warn(bond_dev, slave_dev, "enslaved VLAN challenged slave. Adding VLANs will be blocked as long as it is part of bond.\n"); |
1da177e4 LT |
1794 | } |
1795 | } else { | |
e2a7420d | 1796 | slave_dbg(bond_dev, slave_dev, "is !NETIF_F_VLAN_CHALLENGED\n"); |
1da177e4 LT |
1797 | } |
1798 | ||
18cb261a JW |
1799 | if (slave_dev->features & NETIF_F_HW_ESP) |
1800 | slave_dbg(bond_dev, slave_dev, "is esp-hw-offload capable\n"); | |
1801 | ||
547942ca | 1802 | /* Old ifenslave binaries are no longer supported. These can |
3d632c3f | 1803 | * be identified with moderate accuracy by the state of the slave: |
217df670 JV |
1804 | * the current ifenslave will set the interface down prior to |
1805 | * enslaving it; the old ifenslave will not. | |
1806 | */ | |
ce3ea1c7 | 1807 | if (slave_dev->flags & IFF_UP) { |
6569fa2d JT |
1808 | SLAVE_NL_ERR(bond_dev, slave_dev, extack, |
1809 | "Device can not be enslaved while up"); | |
1e2a8868 | 1810 | return -EPERM; |
217df670 | 1811 | } |
1da177e4 | 1812 | |
872254dd MS |
1813 | /* set bonding device ether type by slave - bonding netdevices are |
1814 | * created with ether_setup, so when the slave type is not ARPHRD_ETHER | |
1815 | * there is a need to override some of the type dependent attribs/funcs. | |
1816 | * | |
1817 | * bond ether type mutual exclusion - don't allow slaves of dissimilar | |
1818 | * ether type (eg ARPHRD_ETHER and ARPHRD_INFINIBAND) share the same bond | |
1819 | */ | |
0965a1f3 | 1820 | if (!bond_has_slaves(bond)) { |
e36b9d16 | 1821 | if (bond_dev->type != slave_dev->type) { |
e2a7420d JW |
1822 | slave_dbg(bond_dev, slave_dev, "change device type from %d to %d\n", |
1823 | bond_dev->type, slave_dev->type); | |
75c78500 | 1824 | |
b7bc2a5b AW |
1825 | res = call_netdevice_notifiers(NETDEV_PRE_TYPE_CHANGE, |
1826 | bond_dev); | |
3ca5b404 JP |
1827 | res = notifier_to_errno(res); |
1828 | if (res) { | |
e2a7420d | 1829 | slave_err(bond_dev, slave_dev, "refused to change device type\n"); |
1e2a8868 | 1830 | return -EBUSY; |
3ca5b404 | 1831 | } |
75c78500 | 1832 | |
32a806c1 | 1833 | /* Flush unicast and multicast addresses */ |
a748ee24 | 1834 | dev_uc_flush(bond_dev); |
22bedad3 | 1835 | dev_mc_flush(bond_dev); |
32a806c1 | 1836 | |
e36b9d16 MS |
1837 | if (slave_dev->type != ARPHRD_ETHER) |
1838 | bond_setup_by_slave(bond_dev, slave_dev); | |
550fd08c | 1839 | else { |
e36b9d16 | 1840 | ether_setup(bond_dev); |
550fd08c NH |
1841 | bond_dev->priv_flags &= ~IFF_TX_SKB_SHARING; |
1842 | } | |
75c78500 | 1843 | |
b7bc2a5b AW |
1844 | call_netdevice_notifiers(NETDEV_POST_TYPE_CHANGE, |
1845 | bond_dev); | |
e36b9d16 | 1846 | } |
872254dd | 1847 | } else if (bond_dev->type != slave_dev->type) { |
6569fa2d JT |
1848 | SLAVE_NL_ERR(bond_dev, slave_dev, extack, |
1849 | "Device type is different from other slaves"); | |
1e2a8868 | 1850 | return -EINVAL; |
872254dd MS |
1851 | } |
1852 | ||
1533e773 MB |
1853 | if (slave_dev->type == ARPHRD_INFINIBAND && |
1854 | BOND_MODE(bond) != BOND_MODE_ACTIVEBACKUP) { | |
6569fa2d JT |
1855 | SLAVE_NL_ERR(bond_dev, slave_dev, extack, |
1856 | "Only active-backup mode is supported for infiniband slaves"); | |
1533e773 MB |
1857 | res = -EOPNOTSUPP; |
1858 | goto err_undo_flags; | |
1859 | } | |
1860 | ||
1861 | if (!slave_ops->ndo_set_mac_address || | |
1862 | slave_dev->type == ARPHRD_INFINIBAND) { | |
e2a7420d | 1863 | slave_warn(bond_dev, slave_dev, "The slave device specified does not support setting the MAC address\n"); |
f5442441 VF |
1864 | if (BOND_MODE(bond) == BOND_MODE_ACTIVEBACKUP && |
1865 | bond->params.fail_over_mac != BOND_FOM_ACTIVE) { | |
1866 | if (!bond_has_slaves(bond)) { | |
00503b6f | 1867 | bond->params.fail_over_mac = BOND_FOM_ACTIVE; |
e2a7420d | 1868 | slave_warn(bond_dev, slave_dev, "Setting fail_over_mac to active for active-backup mode\n"); |
f5442441 | 1869 | } else { |
6569fa2d JT |
1870 | SLAVE_NL_ERR(bond_dev, slave_dev, extack, |
1871 | "Slave device does not support setting the MAC address, but fail_over_mac is not set to active"); | |
f5442441 VF |
1872 | res = -EOPNOTSUPP; |
1873 | goto err_undo_flags; | |
00503b6f | 1874 | } |
2ab82852 | 1875 | } |
1da177e4 LT |
1876 | } |
1877 | ||
8d8fc29d AW |
1878 | call_netdevice_notifiers(NETDEV_JOIN, slave_dev); |
1879 | ||
c20811a7 | 1880 | /* If this is the first slave, then we need to set the master's hardware |
547942ca NA |
1881 | * address to be the same as the slave's. |
1882 | */ | |
0965a1f3 | 1883 | if (!bond_has_slaves(bond) && |
b9245914 PM |
1884 | bond->dev->addr_assign_type == NET_ADDR_RANDOM) { |
1885 | res = bond_set_dev_addr(bond->dev, slave_dev); | |
1886 | if (res) | |
1887 | goto err_undo_flags; | |
1888 | } | |
c20811a7 | 1889 | |
b9ad3e9f | 1890 | new_slave = bond_alloc_slave(bond, slave_dev); |
1da177e4 LT |
1891 | if (!new_slave) { |
1892 | res = -ENOMEM; | |
1893 | goto err_undo_flags; | |
1894 | } | |
3fdddd85 | 1895 | |
547942ca | 1896 | /* Set the new_slave's queue_id to be zero. Queue ID mapping |
bb1d9123 AG |
1897 | * is set via sysfs or module option if desired. |
1898 | */ | |
1899 | new_slave->queue_id = 0; | |
1900 | ||
b15ba0fb JP |
1901 | /* Save slave's original mtu and then set it to match the bond */ |
1902 | new_slave->original_mtu = slave_dev->mtu; | |
1903 | res = dev_set_mtu(slave_dev, bond->dev->mtu); | |
1904 | if (res) { | |
e2a7420d | 1905 | slave_err(bond_dev, slave_dev, "Error %d calling dev_set_mtu\n", res); |
b15ba0fb JP |
1906 | goto err_free; |
1907 | } | |
1908 | ||
547942ca | 1909 | /* Save slave's original ("permanent") mac address for modes |
217df670 JV |
1910 | * that need it, and for restoring it upon release, and then |
1911 | * set it to the master's address | |
1912 | */ | |
faeeb317 JW |
1913 | bond_hw_addr_copy(new_slave->perm_hwaddr, slave_dev->dev_addr, |
1914 | slave_dev->addr_len); | |
1da177e4 | 1915 | |
00503b6f | 1916 | if (!bond->params.fail_over_mac || |
01844098 | 1917 | BOND_MODE(bond) != BOND_MODE_ACTIVEBACKUP) { |
547942ca | 1918 | /* Set slave to master's mac address. The application already |
2ab82852 MS |
1919 | * set the master's mac address to that of the first slave |
1920 | */ | |
faeeb317 JW |
1921 | memcpy(ss.__data, bond_dev->dev_addr, bond_dev->addr_len); |
1922 | ss.ss_family = slave_dev->type; | |
3a37a963 PM |
1923 | res = dev_set_mac_address(slave_dev, (struct sockaddr *)&ss, |
1924 | extack); | |
2ab82852 | 1925 | if (res) { |
e2a7420d | 1926 | slave_err(bond_dev, slave_dev, "Error %d calling set_mac_address\n", res); |
b15ba0fb | 1927 | goto err_restore_mtu; |
2ab82852 | 1928 | } |
217df670 | 1929 | } |
1da177e4 | 1930 | |
03d84a5f KH |
1931 | /* set slave flag before open to prevent IPv6 addrconf */ |
1932 | slave_dev->flags |= IFF_SLAVE; | |
1933 | ||
217df670 | 1934 | /* open the slave since the application closed it */ |
00f54e68 | 1935 | res = dev_open(slave_dev, extack); |
217df670 | 1936 | if (res) { |
e2a7420d | 1937 | slave_err(bond_dev, slave_dev, "Opening slave failed\n"); |
1f718f0f | 1938 | goto err_restore_mac; |
1da177e4 LT |
1939 | } |
1940 | ||
0b680e75 | 1941 | slave_dev->priv_flags |= IFF_BONDING; |
5f0c5f73 AG |
1942 | /* initialize slave stats */ |
1943 | dev_get_stats(new_slave->dev, &new_slave->slave_stats); | |
1da177e4 | 1944 | |
58402054 | 1945 | if (bond_is_lb(bond)) { |
1da177e4 LT |
1946 | /* bond_alb_init_slave() must be called before all other stages since |
1947 | * it might fail and we do not want to have to undo everything | |
1948 | */ | |
1949 | res = bond_alb_init_slave(bond, new_slave); | |
3d632c3f | 1950 | if (res) |
569f0c4d | 1951 | goto err_close; |
1da177e4 LT |
1952 | } |
1953 | ||
b8e2fde4 WY |
1954 | res = vlan_vids_add_by_dev(slave_dev, bond_dev); |
1955 | if (res) { | |
e2a7420d | 1956 | slave_err(bond_dev, slave_dev, "Couldn't add bond vlan ids\n"); |
ae42cc62 | 1957 | goto err_close; |
1ff412ad | 1958 | } |
1da177e4 | 1959 | |
c8c23903 | 1960 | prev_slave = bond_last_slave(bond); |
1da177e4 LT |
1961 | |
1962 | new_slave->delay = 0; | |
1963 | new_slave->link_failure_count = 0; | |
1964 | ||
ad729bc9 AB |
1965 | if (bond_update_speed_duplex(new_slave) && |
1966 | bond_needs_speed_duplex(bond)) | |
3f3c278c | 1967 | new_slave->link = BOND_LINK_DOWN; |
876254ae | 1968 | |
49f17de7 | 1969 | new_slave->last_rx = jiffies - |
f31c7937 | 1970 | (msecs_to_jiffies(bond->params.arp_interval) + 1); |
8599b52e | 1971 | for (i = 0; i < BOND_MAX_ARP_TARGETS; i++) |
49f17de7 | 1972 | new_slave->target_last_arp_rx[i] = new_slave->last_rx; |
f5b2b966 | 1973 | |
1da177e4 LT |
1974 | if (bond->params.miimon && !bond->params.use_carrier) { |
1975 | link_reporting = bond_check_dev_link(bond, slave_dev, 1); | |
1976 | ||
1977 | if ((link_reporting == -1) && !bond->params.arp_interval) { | |
547942ca | 1978 | /* miimon is set but a bonded network driver |
1da177e4 LT |
1979 | * does not support ETHTOOL/MII and |
1980 | * arp_interval is not set. Note: if | |
1981 | * use_carrier is enabled, we will never go | |
1982 | * here (because netif_carrier is always | |
1983 | * supported); thus, we don't need to change | |
1984 | * the messages for netif_carrier. | |
1985 | */ | |
e2a7420d | 1986 | slave_warn(bond_dev, slave_dev, "MII and ETHTOOL support not available for slave, and arp_interval/arp_ip_target module parameters not specified, thus bonding will not detect link failures! see bonding.txt for details\n"); |
1da177e4 LT |
1987 | } else if (link_reporting == -1) { |
1988 | /* unable get link status using mii/ethtool */ | |
e2a7420d | 1989 | slave_warn(bond_dev, slave_dev, "can't get link status from slave; 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"); |
1da177e4 LT |
1990 | } |
1991 | } | |
1992 | ||
1993 | /* check for initial state */ | |
a30b0168 | 1994 | new_slave->link = BOND_LINK_NOCHANGE; |
f31c7937 MK |
1995 | if (bond->params.miimon) { |
1996 | if (bond_check_dev_link(bond, slave_dev, 0) == BMSR_LSTATUS) { | |
1997 | if (bond->params.updelay) { | |
69a2338e | 1998 | bond_set_slave_link_state(new_slave, |
5d397061 JP |
1999 | BOND_LINK_BACK, |
2000 | BOND_SLAVE_NOTIFY_NOW); | |
f31c7937 MK |
2001 | new_slave->delay = bond->params.updelay; |
2002 | } else { | |
69a2338e | 2003 | bond_set_slave_link_state(new_slave, |
5d397061 JP |
2004 | BOND_LINK_UP, |
2005 | BOND_SLAVE_NOTIFY_NOW); | |
f31c7937 | 2006 | } |
1da177e4 | 2007 | } else { |
5d397061 JP |
2008 | bond_set_slave_link_state(new_slave, BOND_LINK_DOWN, |
2009 | BOND_SLAVE_NOTIFY_NOW); | |
1da177e4 | 2010 | } |
f31c7937 | 2011 | } else if (bond->params.arp_interval) { |
69a2338e MS |
2012 | bond_set_slave_link_state(new_slave, |
2013 | (netif_carrier_ok(slave_dev) ? | |
5d397061 JP |
2014 | BOND_LINK_UP : BOND_LINK_DOWN), |
2015 | BOND_SLAVE_NOTIFY_NOW); | |
1da177e4 | 2016 | } else { |
5d397061 JP |
2017 | bond_set_slave_link_state(new_slave, BOND_LINK_UP, |
2018 | BOND_SLAVE_NOTIFY_NOW); | |
1da177e4 LT |
2019 | } |
2020 | ||
f31c7937 | 2021 | if (new_slave->link != BOND_LINK_DOWN) |
8e603460 | 2022 | new_slave->last_link_up = jiffies; |
e2a7420d JW |
2023 | slave_dbg(bond_dev, slave_dev, "Initial state of slave is BOND_LINK_%s\n", |
2024 | new_slave->link == BOND_LINK_DOWN ? "DOWN" : | |
2025 | (new_slave->link == BOND_LINK_UP ? "UP" : "BACK")); | |
f31c7937 | 2026 | |
ec0865a9 | 2027 | if (bond_uses_primary(bond) && bond->params.primary[0]) { |
1da177e4 | 2028 | /* if there is a primary slave, remember it */ |
a549952a | 2029 | if (strcmp(bond->params.primary, new_slave->dev->name) == 0) { |
059b47e8 | 2030 | rcu_assign_pointer(bond->primary_slave, new_slave); |
a549952a JP |
2031 | bond->force_primary = true; |
2032 | } | |
1da177e4 LT |
2033 | } |
2034 | ||
01844098 | 2035 | switch (BOND_MODE(bond)) { |
1da177e4 | 2036 | case BOND_MODE_ACTIVEBACKUP: |
5e5b0665 | 2037 | bond_set_slave_inactive_flags(new_slave, |
2038 | BOND_SLAVE_NOTIFY_NOW); | |
1da177e4 LT |
2039 | break; |
2040 | case BOND_MODE_8023AD: | |
2041 | /* in 802.3ad mode, the internal mechanism | |
2042 | * will activate the slaves in the selected | |
2043 | * aggregator | |
2044 | */ | |
5e5b0665 | 2045 | bond_set_slave_inactive_flags(new_slave, BOND_SLAVE_NOTIFY_NOW); |
1da177e4 | 2046 | /* if this is the first slave */ |
23c147e0 | 2047 | if (!prev_slave) { |
3fdddd85 | 2048 | SLAVE_AD_INFO(new_slave)->id = 1; |
1da177e4 LT |
2049 | /* Initialize AD with the number of times that the AD timer is called in 1 second |
2050 | * can be called only after the mac address of the bond is set | |
2051 | */ | |
56d00c67 | 2052 | bond_3ad_initialize(bond, 1000/AD_TIMER_INTERVAL); |
1da177e4 | 2053 | } else { |
3fdddd85 | 2054 | SLAVE_AD_INFO(new_slave)->id = |
2055 | SLAVE_AD_INFO(prev_slave)->id + 1; | |
1da177e4 LT |
2056 | } |
2057 | ||
2058 | bond_3ad_bind_slave(new_slave); | |
2059 | break; | |
2060 | case BOND_MODE_TLB: | |
2061 | case BOND_MODE_ALB: | |
e30bc066 | 2062 | bond_set_active_slave(new_slave); |
5e5b0665 | 2063 | bond_set_slave_inactive_flags(new_slave, BOND_SLAVE_NOTIFY_NOW); |
1da177e4 LT |
2064 | break; |
2065 | default: | |
e2a7420d | 2066 | slave_dbg(bond_dev, slave_dev, "This slave is always active in trunk mode\n"); |
1da177e4 LT |
2067 | |
2068 | /* always active in trunk mode */ | |
e30bc066 | 2069 | bond_set_active_slave(new_slave); |
1da177e4 LT |
2070 | |
2071 | /* In trunking mode there is little meaning to curr_active_slave | |
2072 | * anyway (it holds no special properties of the bond device), | |
2073 | * so we can change it without calling change_active_interface() | |
2074 | */ | |
4740d638 ED |
2075 | if (!rcu_access_pointer(bond->curr_active_slave) && |
2076 | new_slave->link == BOND_LINK_UP) | |
278b2083 | 2077 | rcu_assign_pointer(bond->curr_active_slave, new_slave); |
3d632c3f | 2078 | |
1da177e4 LT |
2079 | break; |
2080 | } /* switch(bond_mode) */ | |
2081 | ||
f6dc31a8 | 2082 | #ifdef CONFIG_NET_POLL_CONTROLLER |
ddea788c | 2083 | if (bond->dev->npinfo) { |
8a8efa22 | 2084 | if (slave_enable_netpoll(new_slave)) { |
e2a7420d | 2085 | slave_info(bond_dev, slave_dev, "master_dev is using netpoll, but new slave device does not support netpoll\n"); |
8a8efa22 | 2086 | res = -EBUSY; |
f7d9821a | 2087 | goto err_detach; |
8a8efa22 | 2088 | } |
f6dc31a8 WC |
2089 | } |
2090 | #endif | |
8a8efa22 | 2091 | |
fbe168ba MK |
2092 | if (!(bond_dev->features & NETIF_F_LRO)) |
2093 | dev_disable_lro(slave_dev); | |
2094 | ||
35d48903 JP |
2095 | res = netdev_rx_handler_register(slave_dev, bond_handle_frame, |
2096 | new_slave); | |
2097 | if (res) { | |
e2a7420d | 2098 | slave_dbg(bond_dev, slave_dev, "Error %d calling netdev_rx_handler_register\n", res); |
5831d66e | 2099 | goto err_detach; |
35d48903 JP |
2100 | } |
2101 | ||
42ab19ee | 2102 | res = bond_master_upper_dev_link(bond, new_slave, extack); |
1f718f0f | 2103 | if (res) { |
e2a7420d | 2104 | slave_dbg(bond_dev, slave_dev, "Error %d calling bond_master_upper_dev_link\n", res); |
1f718f0f VF |
2105 | goto err_unregister; |
2106 | } | |
2107 | ||
32d4c564 TW |
2108 | bond_lower_state_changed(new_slave); |
2109 | ||
07699f9a | 2110 | res = bond_sysfs_slave_add(new_slave); |
2111 | if (res) { | |
e2a7420d | 2112 | slave_dbg(bond_dev, slave_dev, "Error %d calling bond_sysfs_slave_add\n", res); |
07699f9a | 2113 | goto err_upper_unlink; |
2114 | } | |
2115 | ||
ae42cc62 XL |
2116 | /* If the mode uses primary, then the following is handled by |
2117 | * bond_change_active_slave(). | |
2118 | */ | |
2119 | if (!bond_uses_primary(bond)) { | |
2120 | /* set promiscuity level to new slave */ | |
2121 | if (bond_dev->flags & IFF_PROMISC) { | |
2122 | res = dev_set_promiscuity(slave_dev, 1); | |
2123 | if (res) | |
2124 | goto err_sysfs_del; | |
2125 | } | |
2126 | ||
2127 | /* set allmulti level to new slave */ | |
2128 | if (bond_dev->flags & IFF_ALLMULTI) { | |
2129 | res = dev_set_allmulti(slave_dev, 1); | |
9f5a90c1 XL |
2130 | if (res) { |
2131 | if (bond_dev->flags & IFF_PROMISC) | |
2132 | dev_set_promiscuity(slave_dev, -1); | |
ae42cc62 | 2133 | goto err_sysfs_del; |
9f5a90c1 | 2134 | } |
ae42cc62 XL |
2135 | } |
2136 | ||
2137 | netif_addr_lock_bh(bond_dev); | |
2138 | dev_mc_sync_multiple(slave_dev, bond_dev); | |
2139 | dev_uc_sync_multiple(slave_dev, bond_dev); | |
2140 | netif_addr_unlock_bh(bond_dev); | |
2141 | ||
2142 | if (BOND_MODE(bond) == BOND_MODE_8023AD) { | |
2143 | /* add lacpdu mc addr to mc list */ | |
2144 | u8 lacpdu_multicast[ETH_ALEN] = MULTICAST_LACPDU_ADDR; | |
2145 | ||
2146 | dev_mc_add(slave_dev, lacpdu_multicast); | |
2147 | } | |
2148 | } | |
2149 | ||
5378c2e6 VF |
2150 | bond->slave_cnt++; |
2151 | bond_compute_features(bond); | |
2152 | bond_set_carrier(bond); | |
2153 | ||
ec0865a9 | 2154 | if (bond_uses_primary(bond)) { |
f80889a5 | 2155 | block_netpoll_tx(); |
5378c2e6 | 2156 | bond_select_active_slave(bond); |
f80889a5 | 2157 | unblock_netpoll_tx(); |
5378c2e6 | 2158 | } |
1f718f0f | 2159 | |
e79c1055 | 2160 | if (bond_mode_can_use_xmit_hash(bond)) |
ee637714 MB |
2161 | bond_update_slave_arr(bond, NULL); |
2162 | ||
21706ee8 | 2163 | |
9e2ee5c7 JM |
2164 | if (!slave_dev->netdev_ops->ndo_bpf || |
2165 | !slave_dev->netdev_ops->ndo_xdp_xmit) { | |
2166 | if (bond->xdp_prog) { | |
6569fa2d JT |
2167 | SLAVE_NL_ERR(bond_dev, slave_dev, extack, |
2168 | "Slave does not support XDP"); | |
9e2ee5c7 JM |
2169 | res = -EOPNOTSUPP; |
2170 | goto err_sysfs_del; | |
2171 | } | |
6d5f1ef8 | 2172 | } else if (bond->xdp_prog) { |
9e2ee5c7 JM |
2173 | struct netdev_bpf xdp = { |
2174 | .command = XDP_SETUP_PROG, | |
2175 | .flags = 0, | |
2176 | .prog = bond->xdp_prog, | |
2177 | .extack = extack, | |
2178 | }; | |
2179 | ||
2180 | if (dev_xdp_prog_count(slave_dev) > 0) { | |
6569fa2d JT |
2181 | SLAVE_NL_ERR(bond_dev, slave_dev, extack, |
2182 | "Slave has XDP program loaded, please unload before enslaving"); | |
9e2ee5c7 JM |
2183 | res = -EOPNOTSUPP; |
2184 | goto err_sysfs_del; | |
2185 | } | |
2186 | ||
2187 | res = slave_dev->netdev_ops->ndo_bpf(slave_dev, &xdp); | |
2188 | if (res < 0) { | |
2189 | /* ndo_bpf() sets extack error message */ | |
2190 | slave_dbg(bond_dev, slave_dev, "Error %d calling ndo_bpf\n", res); | |
2191 | goto err_sysfs_del; | |
2192 | } | |
2193 | if (bond->xdp_prog) | |
2194 | bpf_prog_inc(bond->xdp_prog); | |
2195 | } | |
2196 | ||
e2a7420d JW |
2197 | slave_info(bond_dev, slave_dev, "Enslaving as %s interface with %s link\n", |
2198 | bond_is_active_slave(new_slave) ? "an active" : "a backup", | |
2199 | new_slave->link != BOND_LINK_DOWN ? "an up" : "a down"); | |
1da177e4 LT |
2200 | |
2201 | /* enslave is successful */ | |
69e61133 | 2202 | bond_queue_slave_event(new_slave); |
1da177e4 LT |
2203 | return 0; |
2204 | ||
2205 | /* Undo stages on error */ | |
ae42cc62 XL |
2206 | err_sysfs_del: |
2207 | bond_sysfs_slave_del(new_slave); | |
2208 | ||
07699f9a | 2209 | err_upper_unlink: |
41f0b049 | 2210 | bond_upper_dev_unlink(bond, new_slave); |
07699f9a | 2211 | |
1f718f0f VF |
2212 | err_unregister: |
2213 | netdev_rx_handler_unregister(slave_dev); | |
2214 | ||
f7d9821a | 2215 | err_detach: |
1ff412ad | 2216 | vlan_vids_del_by_dev(slave_dev, bond_dev); |
059b47e8 NA |
2217 | if (rcu_access_pointer(bond->primary_slave) == new_slave) |
2218 | RCU_INIT_POINTER(bond->primary_slave, NULL); | |
4740d638 | 2219 | if (rcu_access_pointer(bond->curr_active_slave) == new_slave) { |
f80889a5 | 2220 | block_netpoll_tx(); |
c8517035 | 2221 | bond_change_active_slave(bond, NULL); |
3c5913b5 | 2222 | bond_select_active_slave(bond); |
f80889a5 | 2223 | unblock_netpoll_tx(); |
3c5913b5 | 2224 | } |
059b47e8 NA |
2225 | /* either primary_slave or curr_active_slave might've changed */ |
2226 | synchronize_rcu(); | |
fc7a72ac | 2227 | slave_disable_netpoll(new_slave); |
f7d9821a | 2228 | |
1da177e4 | 2229 | err_close: |
65de65d9 TY |
2230 | if (!netif_is_bond_master(slave_dev)) |
2231 | slave_dev->priv_flags &= ~IFF_BONDING; | |
1da177e4 LT |
2232 | dev_close(slave_dev); |
2233 | ||
2234 | err_restore_mac: | |
03d84a5f | 2235 | slave_dev->flags &= ~IFF_SLAVE; |
00503b6f | 2236 | if (!bond->params.fail_over_mac || |
01844098 | 2237 | BOND_MODE(bond) != BOND_MODE_ACTIVEBACKUP) { |
3915c1e8 JV |
2238 | /* XXX TODO - fom follow mode needs to change master's |
2239 | * MAC if this slave's MAC is in use by the bond, or at | |
2240 | * least print a warning. | |
2241 | */ | |
faeeb317 JW |
2242 | bond_hw_addr_copy(ss.__data, new_slave->perm_hwaddr, |
2243 | new_slave->dev->addr_len); | |
2244 | ss.ss_family = slave_dev->type; | |
3a37a963 | 2245 | dev_set_mac_address(slave_dev, (struct sockaddr *)&ss, NULL); |
2ab82852 | 2246 | } |
1da177e4 | 2247 | |
b15ba0fb JP |
2248 | err_restore_mtu: |
2249 | dev_set_mtu(slave_dev, new_slave->original_mtu); | |
2250 | ||
1da177e4 | 2251 | err_free: |
b9ad3e9f | 2252 | kobject_put(&new_slave->kobj); |
1da177e4 LT |
2253 | |
2254 | err_undo_flags: | |
b8fad459 | 2255 | /* Enslave of first slave has failed and we need to fix master's mac */ |
7d5cd2ce NA |
2256 | if (!bond_has_slaves(bond)) { |
2257 | if (ether_addr_equal_64bits(bond_dev->dev_addr, | |
2258 | slave_dev->dev_addr)) | |
2259 | eth_hw_addr_random(bond_dev); | |
2260 | if (bond_dev->type != ARPHRD_ETHER) { | |
40baec22 | 2261 | dev_close(bond_dev); |
7d5cd2ce NA |
2262 | ether_setup(bond_dev); |
2263 | bond_dev->flags |= IFF_MASTER; | |
2264 | bond_dev->priv_flags &= ~IFF_TX_SKB_SHARING; | |
2265 | } | |
2266 | } | |
3d632c3f | 2267 | |
1da177e4 LT |
2268 | return res; |
2269 | } | |
2270 | ||
547942ca | 2271 | /* Try to release the slave device <slave> from the bond device <master> |
1da177e4 | 2272 | * It is legal to access curr_active_slave without a lock because all the function |
8c0bc550 | 2273 | * is RTNL-locked. If "all" is true it means that the function is being called |
0896341a | 2274 | * while destroying a bond interface and all slaves are being released. |
1da177e4 LT |
2275 | * |
2276 | * The rules for slave state should be: | |
2277 | * for Active/Backup: | |
2278 | * Active stays on all backups go down | |
2279 | * for Bonded connections: | |
2280 | * The first up interface should be left on and all others downed. | |
2281 | */ | |
0896341a | 2282 | static int __bond_release_one(struct net_device *bond_dev, |
2283 | struct net_device *slave_dev, | |
f51048c3 | 2284 | bool all, bool unregister) |
1da177e4 | 2285 | { |
454d7c9b | 2286 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 | 2287 | struct slave *slave, *oldcurrent; |
faeeb317 | 2288 | struct sockaddr_storage ss; |
5a0068de | 2289 | int old_flags = bond_dev->flags; |
c8f44aff | 2290 | netdev_features_t old_features = bond_dev->features; |
1da177e4 LT |
2291 | |
2292 | /* slave is not a slave or master is not master of this slave */ | |
2293 | if (!(slave_dev->flags & IFF_SLAVE) || | |
471cb5a3 | 2294 | !netdev_has_upper_dev(slave_dev, bond_dev)) { |
e2a7420d | 2295 | slave_dbg(bond_dev, slave_dev, "cannot release slave\n"); |
1da177e4 LT |
2296 | return -EINVAL; |
2297 | } | |
2298 | ||
e843fa50 | 2299 | block_netpoll_tx(); |
1da177e4 LT |
2300 | |
2301 | slave = bond_get_slave_by_dev(bond, slave_dev); | |
2302 | if (!slave) { | |
2303 | /* not a slave of this bond */ | |
e2a7420d | 2304 | slave_info(bond_dev, slave_dev, "interface not enslaved\n"); |
e843fa50 | 2305 | unblock_netpoll_tx(); |
1da177e4 LT |
2306 | return -EINVAL; |
2307 | } | |
2308 | ||
57beaca8 JP |
2309 | bond_set_slave_inactive_flags(slave, BOND_SLAVE_NOTIFY_NOW); |
2310 | ||
07699f9a | 2311 | bond_sysfs_slave_del(slave); |
2312 | ||
5f0c5f73 AG |
2313 | /* recompute stats just before removing the slave */ |
2314 | bond_get_stats(bond->dev, &bond->bond_stats); | |
2315 | ||
9e2ee5c7 JM |
2316 | if (bond->xdp_prog) { |
2317 | struct netdev_bpf xdp = { | |
2318 | .command = XDP_SETUP_PROG, | |
2319 | .flags = 0, | |
2320 | .prog = NULL, | |
2321 | .extack = NULL, | |
2322 | }; | |
2323 | if (slave_dev->netdev_ops->ndo_bpf(slave_dev, &xdp)) | |
2324 | slave_warn(bond_dev, slave_dev, "failed to unload XDP program\n"); | |
2325 | } | |
2326 | ||
35d48903 JP |
2327 | /* unregister rx_handler early so bond_handle_frame wouldn't be called |
2328 | * for this slave anymore. | |
2329 | */ | |
2330 | netdev_rx_handler_unregister(slave_dev); | |
35d48903 | 2331 | |
e470259f | 2332 | if (BOND_MODE(bond) == BOND_MODE_8023AD) |
1da177e4 | 2333 | bond_3ad_unbind_slave(slave); |
1da177e4 | 2334 | |
220ade77 YM |
2335 | bond_upper_dev_unlink(bond, slave); |
2336 | ||
e79c1055 | 2337 | if (bond_mode_can_use_xmit_hash(bond)) |
ee637714 MB |
2338 | bond_update_slave_arr(bond, slave); |
2339 | ||
e2a7420d JW |
2340 | slave_info(bond_dev, slave_dev, "Releasing %s interface\n", |
2341 | bond_is_active_slave(slave) ? "active" : "backup"); | |
1da177e4 | 2342 | |
4740d638 | 2343 | oldcurrent = rcu_access_pointer(bond->curr_active_slave); |
1da177e4 | 2344 | |
85741718 | 2345 | RCU_INIT_POINTER(bond->current_arp_slave, NULL); |
1da177e4 | 2346 | |
00503b6f | 2347 | if (!all && (!bond->params.fail_over_mac || |
01844098 | 2348 | BOND_MODE(bond) != BOND_MODE_ACTIVEBACKUP)) { |
844223ab | 2349 | if (ether_addr_equal_64bits(bond_dev->dev_addr, slave->perm_hwaddr) && |
0965a1f3 | 2350 | bond_has_slaves(bond)) |
e2a7420d JW |
2351 | slave_warn(bond_dev, slave_dev, "the permanent HWaddr of slave - %pM - is still in use by bond - set the HWaddr of slave to a different address to avoid conflicts\n", |
2352 | slave->perm_hwaddr); | |
dec1e90e | 2353 | } |
2354 | ||
059b47e8 NA |
2355 | if (rtnl_dereference(bond->primary_slave) == slave) |
2356 | RCU_INIT_POINTER(bond->primary_slave, NULL); | |
1da177e4 | 2357 | |
1c72cfdc | 2358 | if (oldcurrent == slave) |
1da177e4 | 2359 | bond_change_active_slave(bond, NULL); |
1da177e4 | 2360 | |
58402054 | 2361 | if (bond_is_lb(bond)) { |
1da177e4 LT |
2362 | /* Must be called only after the slave has been |
2363 | * detached from the list and the curr_active_slave | |
2364 | * has been cleared (if our_slave == old_current), | |
2365 | * but before a new active slave is selected. | |
2366 | */ | |
2367 | bond_alb_deinit_slave(bond, slave); | |
2368 | } | |
2369 | ||
0896341a | 2370 | if (all) { |
36708b89 | 2371 | RCU_INIT_POINTER(bond->curr_active_slave, NULL); |
0896341a | 2372 | } else if (oldcurrent == slave) { |
547942ca | 2373 | /* Note that we hold RTNL over this sequence, so there |
059fe7a5 JV |
2374 | * is no concern that another slave add/remove event |
2375 | * will interfere. | |
2376 | */ | |
1da177e4 | 2377 | bond_select_active_slave(bond); |
059fe7a5 JV |
2378 | } |
2379 | ||
67b458a0 ZC |
2380 | bond_set_carrier(bond); |
2381 | if (!bond_has_slaves(bond)) | |
409cc1f8 | 2382 | eth_hw_addr_random(bond_dev); |
1da177e4 | 2383 | |
e843fa50 | 2384 | unblock_netpoll_tx(); |
278b2083 | 2385 | synchronize_rcu(); |
ee6154e1 | 2386 | bond->slave_cnt--; |
1da177e4 | 2387 | |
0965a1f3 | 2388 | if (!bond_has_slaves(bond)) { |
2af73d4b | 2389 | call_netdevice_notifiers(NETDEV_CHANGEADDR, bond->dev); |
80028ea1 VF |
2390 | call_netdevice_notifiers(NETDEV_RELEASE, bond->dev); |
2391 | } | |
2af73d4b | 2392 | |
b2a103e6 MM |
2393 | bond_compute_features(bond); |
2394 | if (!(bond_dev->features & NETIF_F_VLAN_CHALLENGED) && | |
2395 | (old_features & NETIF_F_VLAN_CHALLENGED)) | |
e2a7420d | 2396 | slave_info(bond_dev, slave_dev, "last VLAN challenged slave left bond - VLAN blocking is removed\n"); |
b2a103e6 | 2397 | |
1ff412ad | 2398 | vlan_vids_del_by_dev(slave_dev, bond_dev); |
1da177e4 | 2399 | |
547942ca | 2400 | /* If the mode uses primary, then this case was handled above by |
303d1cbf | 2401 | * bond_change_active_slave(..., NULL) |
1da177e4 | 2402 | */ |
ec0865a9 | 2403 | if (!bond_uses_primary(bond)) { |
5a0068de NH |
2404 | /* unset promiscuity level from slave |
2405 | * NOTE: The NETDEV_CHANGEADDR call above may change the value | |
2406 | * of the IFF_PROMISC flag in the bond_dev, but we need the | |
2407 | * value of that flag before that change, as that was the value | |
2408 | * when this slave was attached, so we cache at the start of the | |
2409 | * function and use it here. Same goes for ALLMULTI below | |
2410 | */ | |
2411 | if (old_flags & IFF_PROMISC) | |
1da177e4 | 2412 | dev_set_promiscuity(slave_dev, -1); |
1da177e4 LT |
2413 | |
2414 | /* unset allmulti level from slave */ | |
5a0068de | 2415 | if (old_flags & IFF_ALLMULTI) |
1da177e4 | 2416 | dev_set_allmulti(slave_dev, -1); |
1da177e4 | 2417 | |
303d1cbf | 2418 | bond_hw_addr_flush(bond_dev, slave_dev); |
1da177e4 LT |
2419 | } |
2420 | ||
8a8efa22 | 2421 | slave_disable_netpoll(slave); |
f6dc31a8 | 2422 | |
1da177e4 LT |
2423 | /* close slave before restoring its mac address */ |
2424 | dev_close(slave_dev); | |
2425 | ||
00503b6f | 2426 | if (bond->params.fail_over_mac != BOND_FOM_ACTIVE || |
01844098 | 2427 | BOND_MODE(bond) != BOND_MODE_ACTIVEBACKUP) { |
2ab82852 | 2428 | /* restore original ("permanent") mac address */ |
faeeb317 JW |
2429 | bond_hw_addr_copy(ss.__data, slave->perm_hwaddr, |
2430 | slave->dev->addr_len); | |
2431 | ss.ss_family = slave_dev->type; | |
3a37a963 | 2432 | dev_set_mac_address(slave_dev, (struct sockaddr *)&ss, NULL); |
2ab82852 | 2433 | } |
1da177e4 | 2434 | |
f51048c3 WC |
2435 | if (unregister) |
2436 | __dev_set_mtu(slave_dev, slave->original_mtu); | |
2437 | else | |
2438 | dev_set_mtu(slave_dev, slave->original_mtu); | |
b15ba0fb | 2439 | |
65de65d9 TY |
2440 | if (!netif_is_bond_master(slave_dev)) |
2441 | slave_dev->priv_flags &= ~IFF_BONDING; | |
1da177e4 | 2442 | |
b9ad3e9f | 2443 | kobject_put(&slave->kobj); |
1da177e4 | 2444 | |
547942ca | 2445 | return 0; |
1da177e4 LT |
2446 | } |
2447 | ||
0896341a | 2448 | /* A wrapper used because of ndo_del_link */ |
2449 | int bond_release(struct net_device *bond_dev, struct net_device *slave_dev) | |
2450 | { | |
f51048c3 | 2451 | return __bond_release_one(bond_dev, slave_dev, false, false); |
0896341a | 2452 | } |
2453 | ||
547942ca NA |
2454 | /* First release a slave and then destroy the bond if no more slaves are left. |
2455 | * Must be under rtnl_lock when this function is called. | |
2456 | */ | |
e2a7420d JW |
2457 | static int bond_release_and_destroy(struct net_device *bond_dev, |
2458 | struct net_device *slave_dev) | |
d90a162a | 2459 | { |
454d7c9b | 2460 | struct bonding *bond = netdev_priv(bond_dev); |
d90a162a MS |
2461 | int ret; |
2462 | ||
f51048c3 | 2463 | ret = __bond_release_one(bond_dev, slave_dev, false, true); |
83270702 CW |
2464 | if (ret == 0 && !bond_has_slaves(bond) && |
2465 | bond_dev->reg_state != NETREG_UNREGISTERING) { | |
8a8efa22 | 2466 | bond_dev->priv_flags |= IFF_DISABLE_NETPOLL; |
e2a7420d | 2467 | netdev_info(bond_dev, "Destroying bond\n"); |
06f6d109 | 2468 | bond_remove_proc_entry(bond); |
9e71626c | 2469 | unregister_netdevice(bond_dev); |
d90a162a MS |
2470 | } |
2471 | return ret; | |
2472 | } | |
2473 | ||
3d67576d | 2474 | static void bond_info_query(struct net_device *bond_dev, struct ifbond *info) |
1da177e4 | 2475 | { |
454d7c9b | 2476 | struct bonding *bond = netdev_priv(bond_dev); |
86a5ad0a | 2477 | |
69a2338e | 2478 | bond_fill_ifbond(bond, info); |
1da177e4 LT |
2479 | } |
2480 | ||
2481 | static int bond_slave_info_query(struct net_device *bond_dev, struct ifslave *info) | |
2482 | { | |
454d7c9b | 2483 | struct bonding *bond = netdev_priv(bond_dev); |
9caff1e7 | 2484 | struct list_head *iter; |
dec1e90e | 2485 | int i = 0, res = -ENODEV; |
1da177e4 | 2486 | struct slave *slave; |
1da177e4 | 2487 | |
9caff1e7 | 2488 | bond_for_each_slave(bond, slave, iter) { |
dec1e90e | 2489 | if (i++ == (int)info->slave_id) { |
689c96cc | 2490 | res = 0; |
69a2338e | 2491 | bond_fill_ifslave(slave, info); |
1da177e4 LT |
2492 | break; |
2493 | } | |
2494 | } | |
1da177e4 | 2495 | |
689c96cc | 2496 | return res; |
1da177e4 LT |
2497 | } |
2498 | ||
2499 | /*-------------------------------- Monitoring -------------------------------*/ | |
2500 | ||
4740d638 | 2501 | /* called with rcu_read_lock() */ |
f0c76d61 JV |
2502 | static int bond_miimon_inspect(struct bonding *bond) |
2503 | { | |
dec1e90e | 2504 | int link_state, commit = 0; |
9caff1e7 | 2505 | struct list_head *iter; |
f0c76d61 | 2506 | struct slave *slave; |
41f89100 JP |
2507 | bool ignore_updelay; |
2508 | ||
4740d638 | 2509 | ignore_updelay = !rcu_dereference(bond->curr_active_slave); |
1da177e4 | 2510 | |
4cb4f97b | 2511 | bond_for_each_slave_rcu(bond, slave, iter) { |
1899bb32 | 2512 | bond_propose_link_state(slave, BOND_LINK_NOCHANGE); |
1da177e4 | 2513 | |
f0c76d61 | 2514 | link_state = bond_check_dev_link(bond, slave->dev, 0); |
1da177e4 LT |
2515 | |
2516 | switch (slave->link) { | |
f0c76d61 JV |
2517 | case BOND_LINK_UP: |
2518 | if (link_state) | |
2519 | continue; | |
1da177e4 | 2520 | |
de77ecd4 | 2521 | bond_propose_link_state(slave, BOND_LINK_FAIL); |
d94708a5 | 2522 | commit++; |
f0c76d61 JV |
2523 | slave->delay = bond->params.downdelay; |
2524 | if (slave->delay) { | |
e2a7420d JW |
2525 | slave_info(bond->dev, slave->dev, "link status down for %sinterface, disabling it in %d ms\n", |
2526 | (BOND_MODE(bond) == | |
2527 | BOND_MODE_ACTIVEBACKUP) ? | |
2528 | (bond_is_active_slave(slave) ? | |
2529 | "active " : "backup ") : "", | |
2530 | bond->params.downdelay * bond->params.miimon); | |
1da177e4 | 2531 | } |
df561f66 | 2532 | fallthrough; |
f0c76d61 JV |
2533 | case BOND_LINK_FAIL: |
2534 | if (link_state) { | |
547942ca | 2535 | /* recovered before downdelay expired */ |
de77ecd4 | 2536 | bond_propose_link_state(slave, BOND_LINK_UP); |
8e603460 | 2537 | slave->last_link_up = jiffies; |
e2a7420d JW |
2538 | slave_info(bond->dev, slave->dev, "link status up again after %d ms\n", |
2539 | (bond->params.downdelay - slave->delay) * | |
2540 | bond->params.miimon); | |
fb9eb899 | 2541 | commit++; |
f0c76d61 | 2542 | continue; |
1da177e4 | 2543 | } |
f0c76d61 JV |
2544 | |
2545 | if (slave->delay <= 0) { | |
1899bb32 | 2546 | bond_propose_link_state(slave, BOND_LINK_DOWN); |
f0c76d61 JV |
2547 | commit++; |
2548 | continue; | |
1da177e4 | 2549 | } |
1da177e4 | 2550 | |
f0c76d61 JV |
2551 | slave->delay--; |
2552 | break; | |
2553 | ||
2554 | case BOND_LINK_DOWN: | |
2555 | if (!link_state) | |
2556 | continue; | |
2557 | ||
de77ecd4 | 2558 | bond_propose_link_state(slave, BOND_LINK_BACK); |
d94708a5 | 2559 | commit++; |
f0c76d61 JV |
2560 | slave->delay = bond->params.updelay; |
2561 | ||
2562 | if (slave->delay) { | |
e2a7420d JW |
2563 | slave_info(bond->dev, slave->dev, "link status up, enabling it in %d ms\n", |
2564 | ignore_updelay ? 0 : | |
2565 | bond->params.updelay * | |
2566 | bond->params.miimon); | |
f0c76d61 | 2567 | } |
df561f66 | 2568 | fallthrough; |
f0c76d61 JV |
2569 | case BOND_LINK_BACK: |
2570 | if (!link_state) { | |
de77ecd4 | 2571 | bond_propose_link_state(slave, BOND_LINK_DOWN); |
e2a7420d JW |
2572 | slave_info(bond->dev, slave->dev, "link status down again after %d ms\n", |
2573 | (bond->params.updelay - slave->delay) * | |
2574 | bond->params.miimon); | |
fb9eb899 | 2575 | commit++; |
f0c76d61 JV |
2576 | continue; |
2577 | } | |
2578 | ||
41f89100 JP |
2579 | if (ignore_updelay) |
2580 | slave->delay = 0; | |
2581 | ||
f0c76d61 | 2582 | if (slave->delay <= 0) { |
1899bb32 | 2583 | bond_propose_link_state(slave, BOND_LINK_UP); |
f0c76d61 | 2584 | commit++; |
41f89100 | 2585 | ignore_updelay = false; |
f0c76d61 | 2586 | continue; |
1da177e4 | 2587 | } |
f0c76d61 JV |
2588 | |
2589 | slave->delay--; | |
1da177e4 | 2590 | break; |
f0c76d61 JV |
2591 | } |
2592 | } | |
1da177e4 | 2593 | |
f0c76d61 JV |
2594 | return commit; |
2595 | } | |
1da177e4 | 2596 | |
7e878b60 TZ |
2597 | static void bond_miimon_link_change(struct bonding *bond, |
2598 | struct slave *slave, | |
2599 | char link) | |
2600 | { | |
2601 | switch (BOND_MODE(bond)) { | |
2602 | case BOND_MODE_8023AD: | |
2603 | bond_3ad_handle_link_change(slave, link); | |
2604 | break; | |
2605 | case BOND_MODE_TLB: | |
2606 | case BOND_MODE_ALB: | |
2607 | bond_alb_handle_link_change(bond, slave, link); | |
2608 | break; | |
2609 | case BOND_MODE_XOR: | |
2610 | bond_update_slave_arr(bond, NULL); | |
2611 | break; | |
2612 | } | |
2613 | } | |
2614 | ||
f0c76d61 JV |
2615 | static void bond_miimon_commit(struct bonding *bond) |
2616 | { | |
9caff1e7 | 2617 | struct list_head *iter; |
059b47e8 | 2618 | struct slave *slave, *primary; |
f0c76d61 | 2619 | |
9caff1e7 | 2620 | bond_for_each_slave(bond, slave, iter) { |
1899bb32 | 2621 | switch (slave->link_new_state) { |
f0c76d61 | 2622 | case BOND_LINK_NOCHANGE: |
12185dfe TF |
2623 | /* For 802.3ad mode, check current slave speed and |
2624 | * duplex again in case its port was disabled after | |
2625 | * invalid speed/duplex reporting but recovered before | |
2626 | * link monitoring could make a decision on the actual | |
2627 | * link status | |
2628 | */ | |
2629 | if (BOND_MODE(bond) == BOND_MODE_8023AD && | |
2630 | slave->link == BOND_LINK_UP) | |
2631 | bond_3ad_adapter_speed_duplex_changed(slave); | |
f0c76d61 | 2632 | continue; |
1da177e4 | 2633 | |
f0c76d61 | 2634 | case BOND_LINK_UP: |
ad729bc9 AB |
2635 | if (bond_update_speed_duplex(slave) && |
2636 | bond_needs_speed_duplex(bond)) { | |
3f3c278c | 2637 | slave->link = BOND_LINK_DOWN; |
11e9d782 | 2638 | if (net_ratelimit()) |
e2a7420d JW |
2639 | slave_warn(bond->dev, slave->dev, |
2640 | "failed to get link speed/duplex\n"); | |
b5bf0f5b MB |
2641 | continue; |
2642 | } | |
5d397061 JP |
2643 | bond_set_slave_link_state(slave, BOND_LINK_UP, |
2644 | BOND_SLAVE_NOTIFY_NOW); | |
8e603460 | 2645 | slave->last_link_up = jiffies; |
f0c76d61 | 2646 | |
059b47e8 | 2647 | primary = rtnl_dereference(bond->primary_slave); |
01844098 | 2648 | if (BOND_MODE(bond) == BOND_MODE_8023AD) { |
f0c76d61 | 2649 | /* prevent it from being the active one */ |
e30bc066 | 2650 | bond_set_backup_slave(slave); |
01844098 | 2651 | } else if (BOND_MODE(bond) != BOND_MODE_ACTIVEBACKUP) { |
f0c76d61 | 2652 | /* make it immediately active */ |
e30bc066 | 2653 | bond_set_active_slave(slave); |
1da177e4 | 2654 | } |
1da177e4 | 2655 | |
e2a7420d JW |
2656 | slave_info(bond->dev, slave->dev, "link status definitely up, %u Mbps %s duplex\n", |
2657 | slave->speed == SPEED_UNKNOWN ? 0 : slave->speed, | |
2658 | slave->duplex ? "full" : "half"); | |
1da177e4 | 2659 | |
7e878b60 | 2660 | bond_miimon_link_change(bond, slave, BOND_LINK_UP); |
ee637714 | 2661 | |
059b47e8 | 2662 | if (!bond->curr_active_slave || slave == primary) |
f0c76d61 | 2663 | goto do_failover; |
1da177e4 | 2664 | |
f0c76d61 | 2665 | continue; |
059fe7a5 | 2666 | |
f0c76d61 | 2667 | case BOND_LINK_DOWN: |
fba4acda JV |
2668 | if (slave->link_failure_count < UINT_MAX) |
2669 | slave->link_failure_count++; | |
2670 | ||
5d397061 JP |
2671 | bond_set_slave_link_state(slave, BOND_LINK_DOWN, |
2672 | BOND_SLAVE_NOTIFY_NOW); | |
1da177e4 | 2673 | |
01844098 VF |
2674 | if (BOND_MODE(bond) == BOND_MODE_ACTIVEBACKUP || |
2675 | BOND_MODE(bond) == BOND_MODE_8023AD) | |
5e5b0665 | 2676 | bond_set_slave_inactive_flags(slave, |
2677 | BOND_SLAVE_NOTIFY_NOW); | |
f0c76d61 | 2678 | |
e2a7420d | 2679 | slave_info(bond->dev, slave->dev, "link status definitely down, disabling slave\n"); |
f0c76d61 | 2680 | |
7e878b60 | 2681 | bond_miimon_link_change(bond, slave, BOND_LINK_DOWN); |
ee637714 | 2682 | |
4740d638 | 2683 | if (slave == rcu_access_pointer(bond->curr_active_slave)) |
f0c76d61 JV |
2684 | goto do_failover; |
2685 | ||
2686 | continue; | |
2687 | ||
2688 | default: | |
e2a7420d | 2689 | slave_err(bond->dev, slave->dev, "invalid new link %d on slave\n", |
1899bb32 JV |
2690 | slave->link_new_state); |
2691 | bond_propose_link_state(slave, BOND_LINK_NOCHANGE); | |
f0c76d61 JV |
2692 | |
2693 | continue; | |
2694 | } | |
2695 | ||
2696 | do_failover: | |
e843fa50 | 2697 | block_netpoll_tx(); |
f0c76d61 | 2698 | bond_select_active_slave(bond); |
e843fa50 | 2699 | unblock_netpoll_tx(); |
f0c76d61 JV |
2700 | } |
2701 | ||
2702 | bond_set_carrier(bond); | |
1da177e4 LT |
2703 | } |
2704 | ||
547942ca | 2705 | /* bond_mii_monitor |
0b0eef66 JV |
2706 | * |
2707 | * Really a wrapper that splits the mii monitor into two phases: an | |
f0c76d61 JV |
2708 | * inspection, then (if inspection indicates something needs to be done) |
2709 | * an acquisition of appropriate locks followed by a commit phase to | |
2710 | * implement whatever link state changes are indicated. | |
0b0eef66 | 2711 | */ |
6da67d26 | 2712 | static void bond_mii_monitor(struct work_struct *work) |
0b0eef66 JV |
2713 | { |
2714 | struct bonding *bond = container_of(work, struct bonding, | |
2715 | mii_work.work); | |
ad246c99 | 2716 | bool should_notify_peers = false; |
07a4ddec | 2717 | bool commit; |
1f2cd845 | 2718 | unsigned long delay; |
de77ecd4 MB |
2719 | struct slave *slave; |
2720 | struct list_head *iter; | |
0b0eef66 | 2721 | |
1f2cd845 DM |
2722 | delay = msecs_to_jiffies(bond->params.miimon); |
2723 | ||
2724 | if (!bond_has_slaves(bond)) | |
f0c76d61 | 2725 | goto re_arm; |
b59f9f74 | 2726 | |
4cb4f97b | 2727 | rcu_read_lock(); |
ad246c99 | 2728 | should_notify_peers = bond_should_notify_peers(bond); |
07a4ddec VB |
2729 | commit = !!bond_miimon_inspect(bond); |
2730 | if (bond->send_peer_notif) { | |
2731 | rcu_read_unlock(); | |
2732 | if (rtnl_trylock()) { | |
2733 | bond->send_peer_notif--; | |
2734 | rtnl_unlock(); | |
2735 | } | |
2736 | } else { | |
4cb4f97b | 2737 | rcu_read_unlock(); |
07a4ddec | 2738 | } |
f0c76d61 | 2739 | |
07a4ddec | 2740 | if (commit) { |
1f2cd845 DM |
2741 | /* Race avoidance with bond_close cancel of workqueue */ |
2742 | if (!rtnl_trylock()) { | |
1f2cd845 DM |
2743 | delay = 1; |
2744 | should_notify_peers = false; | |
2745 | goto re_arm; | |
2746 | } | |
6b6c5261 | 2747 | |
de77ecd4 MB |
2748 | bond_for_each_slave(bond, slave, iter) { |
2749 | bond_commit_link_state(slave, BOND_SLAVE_NOTIFY_LATER); | |
2750 | } | |
1f2cd845 DM |
2751 | bond_miimon_commit(bond); |
2752 | ||
1f2cd845 | 2753 | rtnl_unlock(); /* might sleep, hold no other locks */ |
07a4ddec | 2754 | } |
0b0eef66 | 2755 | |
f0c76d61 | 2756 | re_arm: |
e6d265e8 | 2757 | if (bond->params.miimon) |
1f2cd845 DM |
2758 | queue_delayed_work(bond->wq, &bond->mii_work, delay); |
2759 | ||
1f2cd845 DM |
2760 | if (should_notify_peers) { |
2761 | if (!rtnl_trylock()) | |
2762 | return; | |
2763 | call_netdevice_notifiers(NETDEV_NOTIFY_PEERS, bond->dev); | |
2764 | rtnl_unlock(); | |
2765 | } | |
0b0eef66 | 2766 | } |
c3ade5ca | 2767 | |
eff74233 TY |
2768 | static int bond_upper_dev_walk(struct net_device *upper, |
2769 | struct netdev_nested_priv *priv) | |
b3208b20 | 2770 | { |
eff74233 | 2771 | __be32 ip = *(__be32 *)priv->data; |
b3208b20 DA |
2772 | |
2773 | return ip == bond_confirm_addr(upper, 0, ip); | |
2774 | } | |
2775 | ||
50223ce4 | 2776 | static bool bond_has_this_ip(struct bonding *bond, __be32 ip) |
f5b2b966 | 2777 | { |
eff74233 TY |
2778 | struct netdev_nested_priv priv = { |
2779 | .data = (void *)&ip, | |
2780 | }; | |
50223ce4 | 2781 | bool ret = false; |
f5b2b966 | 2782 | |
eaddcd76 | 2783 | if (ip == bond_confirm_addr(bond->dev, 0, ip)) |
50223ce4 | 2784 | return true; |
f5b2b966 | 2785 | |
50223ce4 | 2786 | rcu_read_lock(); |
eff74233 | 2787 | if (netdev_walk_all_upper_dev_rcu(bond->dev, bond_upper_dev_walk, &priv)) |
b3208b20 | 2788 | ret = true; |
50223ce4 | 2789 | rcu_read_unlock(); |
f5b2b966 | 2790 | |
50223ce4 | 2791 | return ret; |
f5b2b966 JV |
2792 | } |
2793 | ||
547942ca | 2794 | /* We go to the (large) trouble of VLAN tagging ARP frames because |
c3ade5ca JV |
2795 | * switches in VLAN mode (especially if ports are configured as |
2796 | * "native" to a VLAN) might not pass non-tagged frames. | |
2797 | */ | |
e2a7420d JW |
2798 | static void bond_arp_send(struct slave *slave, int arp_op, __be32 dest_ip, |
2799 | __be32 src_ip, struct bond_vlan_tag *tags) | |
c3ade5ca JV |
2800 | { |
2801 | struct sk_buff *skb; | |
3e403a77 | 2802 | struct bond_vlan_tag *outer_tag = tags; |
e2a7420d JW |
2803 | struct net_device *slave_dev = slave->dev; |
2804 | struct net_device *bond_dev = slave->bond->dev; | |
c3ade5ca | 2805 | |
e2a7420d JW |
2806 | slave_dbg(bond_dev, slave_dev, "arp %d on slave: dst %pI4 src %pI4\n", |
2807 | arp_op, &dest_ip, &src_ip); | |
3d632c3f | 2808 | |
c3ade5ca JV |
2809 | skb = arp_create(arp_op, ETH_P_ARP, dest_ip, slave_dev, src_ip, |
2810 | NULL, slave_dev->dev_addr, NULL); | |
2811 | ||
2812 | if (!skb) { | |
4873ac3c | 2813 | net_err_ratelimited("ARP packet allocation failed\n"); |
c3ade5ca JV |
2814 | return; |
2815 | } | |
fbd929f2 | 2816 | |
3e403a77 VF |
2817 | if (!tags || tags->vlan_proto == VLAN_N_VID) |
2818 | goto xmit; | |
2819 | ||
2820 | tags++; | |
2821 | ||
44a40855 | 2822 | /* Go through all the tags backwards and add them to the packet */ |
3e403a77 VF |
2823 | while (tags->vlan_proto != VLAN_N_VID) { |
2824 | if (!tags->vlan_id) { | |
2825 | tags++; | |
44a40855 | 2826 | continue; |
3e403a77 | 2827 | } |
44a40855 | 2828 | |
e2a7420d JW |
2829 | slave_dbg(bond_dev, slave_dev, "inner tag: proto %X vid %X\n", |
2830 | ntohs(outer_tag->vlan_proto), tags->vlan_id); | |
62749e2c JP |
2831 | skb = vlan_insert_tag_set_proto(skb, tags->vlan_proto, |
2832 | tags->vlan_id); | |
44a40855 VY |
2833 | if (!skb) { |
2834 | net_err_ratelimited("failed to insert inner VLAN tag\n"); | |
2835 | return; | |
2836 | } | |
3e403a77 VF |
2837 | |
2838 | tags++; | |
44a40855 VY |
2839 | } |
2840 | /* Set the outer tag */ | |
3e403a77 | 2841 | if (outer_tag->vlan_id) { |
e2a7420d JW |
2842 | slave_dbg(bond_dev, slave_dev, "outer tag: proto %X vid %X\n", |
2843 | ntohs(outer_tag->vlan_proto), outer_tag->vlan_id); | |
b4bef1b5 JP |
2844 | __vlan_hwaccel_put_tag(skb, outer_tag->vlan_proto, |
2845 | outer_tag->vlan_id); | |
c3ade5ca | 2846 | } |
3e403a77 VF |
2847 | |
2848 | xmit: | |
c3ade5ca JV |
2849 | arp_xmit(skb); |
2850 | } | |
2851 | ||
44a40855 VY |
2852 | /* Validate the device path between the @start_dev and the @end_dev. |
2853 | * The path is valid if the @end_dev is reachable through device | |
2854 | * stacking. | |
2855 | * When the path is validated, collect any vlan information in the | |
2856 | * path. | |
2857 | */ | |
3e403a77 VF |
2858 | struct bond_vlan_tag *bond_verify_device_path(struct net_device *start_dev, |
2859 | struct net_device *end_dev, | |
2860 | int level) | |
44a40855 | 2861 | { |
3e403a77 | 2862 | struct bond_vlan_tag *tags; |
44a40855 VY |
2863 | struct net_device *upper; |
2864 | struct list_head *iter; | |
44a40855 | 2865 | |
3e403a77 | 2866 | if (start_dev == end_dev) { |
6396bb22 | 2867 | tags = kcalloc(level + 1, sizeof(*tags), GFP_ATOMIC); |
3e403a77 VF |
2868 | if (!tags) |
2869 | return ERR_PTR(-ENOMEM); | |
2870 | tags[level].vlan_proto = VLAN_N_VID; | |
2871 | return tags; | |
2872 | } | |
44a40855 VY |
2873 | |
2874 | netdev_for_each_upper_dev_rcu(start_dev, upper, iter) { | |
3e403a77 VF |
2875 | tags = bond_verify_device_path(upper, end_dev, level + 1); |
2876 | if (IS_ERR_OR_NULL(tags)) { | |
2877 | if (IS_ERR(tags)) | |
2878 | return tags; | |
2879 | continue; | |
44a40855 | 2880 | } |
3e403a77 VF |
2881 | if (is_vlan_dev(upper)) { |
2882 | tags[level].vlan_proto = vlan_dev_vlan_proto(upper); | |
2883 | tags[level].vlan_id = vlan_dev_vlan_id(upper); | |
2884 | } | |
2885 | ||
2886 | return tags; | |
44a40855 VY |
2887 | } |
2888 | ||
3e403a77 | 2889 | return NULL; |
44a40855 | 2890 | } |
c3ade5ca | 2891 | |
1da177e4 LT |
2892 | static void bond_arp_send_all(struct bonding *bond, struct slave *slave) |
2893 | { | |
c3ade5ca | 2894 | struct rtable *rt; |
3e403a77 | 2895 | struct bond_vlan_tag *tags; |
27bc11e6 | 2896 | __be32 *targets = bond->params.arp_targets, addr; |
fbd929f2 | 2897 | int i; |
1da177e4 | 2898 | |
27bc11e6 | 2899 | for (i = 0; i < BOND_MAX_ARP_TARGETS && targets[i]; i++) { |
e2a7420d JW |
2900 | slave_dbg(bond->dev, slave->dev, "%s: target %pI4\n", |
2901 | __func__, &targets[i]); | |
3e403a77 | 2902 | tags = NULL; |
c3ade5ca | 2903 | |
27bc11e6 | 2904 | /* Find out through which dev should the packet go */ |
78fbfd8a DM |
2905 | rt = ip_route_output(dev_net(bond->dev), targets[i], 0, |
2906 | RTO_ONLINK, 0); | |
b23dd4fe | 2907 | if (IS_ERR(rt)) { |
28572760 VF |
2908 | /* there's no route to target - try to send arp |
2909 | * probe to generate any traffic (arp_validate=0) | |
2910 | */ | |
4873ac3c | 2911 | if (bond->params.arp_validate) |
0a4fd8df DD |
2912 | pr_warn_once("%s: no route to arp_ip_target %pI4 and arp_validate is set\n", |
2913 | bond->dev->name, | |
2914 | &targets[i]); | |
e2a7420d | 2915 | bond_arp_send(slave, ARPOP_REQUEST, targets[i], |
44a40855 | 2916 | 0, tags); |
c3ade5ca JV |
2917 | continue; |
2918 | } | |
2919 | ||
27bc11e6 VF |
2920 | /* bond device itself */ |
2921 | if (rt->dst.dev == bond->dev) | |
2922 | goto found; | |
2923 | ||
2924 | rcu_read_lock(); | |
3e403a77 | 2925 | tags = bond_verify_device_path(bond->dev, rt->dst.dev, 0); |
27bc11e6 | 2926 | rcu_read_unlock(); |
c3ade5ca | 2927 | |
3e403a77 | 2928 | if (!IS_ERR_OR_NULL(tags)) |
44a40855 VY |
2929 | goto found; |
2930 | ||
27bc11e6 | 2931 | /* Not our device - skip */ |
e2a7420d | 2932 | slave_dbg(bond->dev, slave->dev, "no path to arp_ip_target %pI4 via rt.dev %s\n", |
76444f50 | 2933 | &targets[i], rt->dst.dev ? rt->dst.dev->name : "NULL"); |
3e32582f | 2934 | |
ed4b9f80 | 2935 | ip_rt_put(rt); |
27bc11e6 VF |
2936 | continue; |
2937 | ||
2938 | found: | |
2939 | addr = bond_confirm_addr(rt->dst.dev, targets[i], 0); | |
2940 | ip_rt_put(rt); | |
e2a7420d | 2941 | bond_arp_send(slave, ARPOP_REQUEST, targets[i], addr, tags); |
a67eed57 | 2942 | kfree(tags); |
c3ade5ca JV |
2943 | } |
2944 | } | |
2945 | ||
d3bb52b0 | 2946 | static void bond_validate_arp(struct bonding *bond, struct slave *slave, __be32 sip, __be32 tip) |
f5b2b966 | 2947 | { |
8599b52e VF |
2948 | int i; |
2949 | ||
87a7b84b | 2950 | if (!sip || !bond_has_this_ip(bond, tip)) { |
e2a7420d JW |
2951 | slave_dbg(bond->dev, slave->dev, "%s: sip %pI4 tip %pI4 not found\n", |
2952 | __func__, &sip, &tip); | |
87a7b84b VF |
2953 | return; |
2954 | } | |
f5b2b966 | 2955 | |
8599b52e VF |
2956 | i = bond_get_targets_ip(bond->params.arp_targets, sip); |
2957 | if (i == -1) { | |
e2a7420d JW |
2958 | slave_dbg(bond->dev, slave->dev, "%s: sip %pI4 not found in targets\n", |
2959 | __func__, &sip); | |
87a7b84b | 2960 | return; |
f5b2b966 | 2961 | } |
49f17de7 | 2962 | slave->last_rx = jiffies; |
8599b52e | 2963 | slave->target_last_arp_rx[i] = jiffies; |
f5b2b966 JV |
2964 | } |
2965 | ||
5bb9e0b5 | 2966 | int bond_arp_rcv(const struct sk_buff *skb, struct bonding *bond, |
2967 | struct slave *slave) | |
f5b2b966 | 2968 | { |
de063b70 | 2969 | struct arphdr *arp = (struct arphdr *)skb->data; |
21a75f09 | 2970 | struct slave *curr_active_slave, *curr_arp_slave; |
f5b2b966 | 2971 | unsigned char *arp_ptr; |
d3bb52b0 | 2972 | __be32 sip, tip; |
6ade97da AD |
2973 | int is_arp = skb->protocol == __cpu_to_be16(ETH_P_ARP); |
2974 | unsigned int alen; | |
f5b2b966 | 2975 | |
f2cb691a | 2976 | if (!slave_do_arp_validate(bond, slave)) { |
bedabf90 | 2977 | if ((slave_do_arp_validate_only(bond) && is_arp) || |
2978 | !slave_do_arp_validate_only(bond)) | |
49f17de7 | 2979 | slave->last_rx = jiffies; |
13a8e0c8 | 2980 | return RX_HANDLER_ANOTHER; |
f2cb691a VF |
2981 | } else if (!is_arp) { |
2982 | return RX_HANDLER_ANOTHER; | |
2983 | } | |
2c146102 | 2984 | |
de063b70 | 2985 | alen = arp_hdr_len(bond->dev); |
f5b2b966 | 2986 | |
e2a7420d JW |
2987 | slave_dbg(bond->dev, slave->dev, "%s: skb->dev %s\n", |
2988 | __func__, skb->dev->name); | |
f5b2b966 | 2989 | |
de063b70 ED |
2990 | if (alen > skb_headlen(skb)) { |
2991 | arp = kmalloc(alen, GFP_ATOMIC); | |
2992 | if (!arp) | |
2993 | goto out_unlock; | |
2994 | if (skb_copy_bits(skb, 0, arp, alen) < 0) | |
2995 | goto out_unlock; | |
2996 | } | |
f5b2b966 | 2997 | |
3aba891d | 2998 | if (arp->ar_hln != bond->dev->addr_len || |
f5b2b966 JV |
2999 | skb->pkt_type == PACKET_OTHERHOST || |
3000 | skb->pkt_type == PACKET_LOOPBACK || | |
3001 | arp->ar_hrd != htons(ARPHRD_ETHER) || | |
3002 | arp->ar_pro != htons(ETH_P_IP) || | |
3003 | arp->ar_pln != 4) | |
3004 | goto out_unlock; | |
3005 | ||
3006 | arp_ptr = (unsigned char *)(arp + 1); | |
3aba891d | 3007 | arp_ptr += bond->dev->addr_len; |
f5b2b966 | 3008 | memcpy(&sip, arp_ptr, 4); |
3aba891d | 3009 | arp_ptr += 4 + bond->dev->addr_len; |
f5b2b966 JV |
3010 | memcpy(&tip, arp_ptr, 4); |
3011 | ||
e2a7420d JW |
3012 | slave_dbg(bond->dev, slave->dev, "%s: %s/%d av %d sv %d sip %pI4 tip %pI4\n", |
3013 | __func__, slave->dev->name, bond_slave_state(slave), | |
3014 | bond->params.arp_validate, slave_do_arp_validate(bond, slave), | |
3015 | &sip, &tip); | |
f5b2b966 | 3016 | |
010d3c39 | 3017 | curr_active_slave = rcu_dereference(bond->curr_active_slave); |
21a75f09 | 3018 | curr_arp_slave = rcu_dereference(bond->current_arp_slave); |
010d3c39 | 3019 | |
21a75f09 | 3020 | /* We 'trust' the received ARP enough to validate it if: |
aeea64ac | 3021 | * |
21a75f09 JV |
3022 | * (a) the slave receiving the ARP is active (which includes the |
3023 | * current ARP slave, if any), or | |
3024 | * | |
3025 | * (b) the receiving slave isn't active, but there is a currently | |
3026 | * active slave and it received valid arp reply(s) after it became | |
3027 | * the currently active slave, or | |
3028 | * | |
3029 | * (c) there is an ARP slave that sent an ARP during the prior ARP | |
3030 | * interval, and we receive an ARP reply on any slave. We accept | |
3031 | * these because switch FDB update delays may deliver the ARP | |
3032 | * reply to a slave other than the sender of the ARP request. | |
3033 | * | |
3034 | * Note: for (b), backup slaves are receiving the broadcast ARP | |
3035 | * request, not a reply. This request passes from the sending | |
3036 | * slave through the L2 switch(es) to the receiving slave. Since | |
3037 | * this is checking the request, sip/tip are swapped for | |
3038 | * validation. | |
3039 | * | |
3040 | * This is done to avoid endless looping when we can't reach the | |
aeea64ac | 3041 | * arp_ip_target and fool ourselves with our own arp requests. |
f5b2b966 | 3042 | */ |
e30bc066 | 3043 | if (bond_is_active_slave(slave)) |
f5b2b966 | 3044 | bond_validate_arp(bond, slave, sip, tip); |
010d3c39 VF |
3045 | else if (curr_active_slave && |
3046 | time_after(slave_last_rx(bond, curr_active_slave), | |
3047 | curr_active_slave->last_link_up)) | |
f5b2b966 | 3048 | bond_validate_arp(bond, slave, tip, sip); |
21a75f09 JV |
3049 | else if (curr_arp_slave && (arp->ar_op == htons(ARPOP_REPLY)) && |
3050 | bond_time_in_interval(bond, | |
3051 | dev_trans_start(curr_arp_slave->dev), 1)) | |
3052 | bond_validate_arp(bond, slave, sip, tip); | |
f5b2b966 JV |
3053 | |
3054 | out_unlock: | |
de063b70 ED |
3055 | if (arp != (struct arphdr *)skb->data) |
3056 | kfree(arp); | |
13a8e0c8 | 3057 | return RX_HANDLER_ANOTHER; |
f5b2b966 JV |
3058 | } |
3059 | ||
e7f63f1d VF |
3060 | /* function to verify if we're in the arp_interval timeslice, returns true if |
3061 | * (last_act - arp_interval) <= jiffies <= (last_act + mod * arp_interval + | |
3062 | * arp_interval/2) . the arp_interval/2 is needed for really fast networks. | |
3063 | */ | |
3064 | static bool bond_time_in_interval(struct bonding *bond, unsigned long last_act, | |
3065 | int mod) | |
3066 | { | |
3067 | int delta_in_ticks = msecs_to_jiffies(bond->params.arp_interval); | |
3068 | ||
3069 | return time_in_range(jiffies, | |
3070 | last_act - delta_in_ticks, | |
3071 | last_act + mod * delta_in_ticks + delta_in_ticks/2); | |
3072 | } | |
3073 | ||
547942ca | 3074 | /* This function is called regularly to monitor each slave's link |
1da177e4 LT |
3075 | * ensuring that traffic is being sent and received when arp monitoring |
3076 | * is used in load-balancing mode. if the adapter has been dormant, then an | |
3077 | * arp is transmitted to generate traffic. see activebackup_arp_monitor for | |
3078 | * arp monitoring in active backup mode. | |
3079 | */ | |
d5e73f7b | 3080 | static void bond_loadbalance_arp_mon(struct bonding *bond) |
1da177e4 | 3081 | { |
1da177e4 | 3082 | struct slave *slave, *oldcurrent; |
9caff1e7 | 3083 | struct list_head *iter; |
6fde8f03 | 3084 | int do_failover = 0, slave_state_changed = 0; |
1da177e4 | 3085 | |
1f2cd845 | 3086 | if (!bond_has_slaves(bond)) |
1da177e4 | 3087 | goto re_arm; |
1da177e4 | 3088 | |
2e52f4fe | 3089 | rcu_read_lock(); |
3090 | ||
4740d638 | 3091 | oldcurrent = rcu_dereference(bond->curr_active_slave); |
1da177e4 LT |
3092 | /* see if any of the previous devices are up now (i.e. they have |
3093 | * xmt and rcv traffic). the curr_active_slave does not come into | |
8e603460 VF |
3094 | * the picture unless it is null. also, slave->last_link_up is not |
3095 | * needed here because we send an arp on each slave and give a slave | |
3096 | * as long as it needs to get the tx/rx within the delta. | |
1da177e4 LT |
3097 | * TODO: what about up/down delay in arp mode? it wasn't here before |
3098 | * so it can wait | |
3099 | */ | |
2e52f4fe | 3100 | bond_for_each_slave_rcu(bond, slave, iter) { |
cb32f2a0 JB |
3101 | unsigned long trans_start = dev_trans_start(slave->dev); |
3102 | ||
1899bb32 | 3103 | bond_propose_link_state(slave, BOND_LINK_NOCHANGE); |
797a9364 | 3104 | |
1da177e4 | 3105 | if (slave->link != BOND_LINK_UP) { |
e7f63f1d | 3106 | if (bond_time_in_interval(bond, trans_start, 1) && |
49f17de7 | 3107 | bond_time_in_interval(bond, slave->last_rx, 1)) { |
1da177e4 | 3108 | |
1899bb32 | 3109 | bond_propose_link_state(slave, BOND_LINK_UP); |
6fde8f03 | 3110 | slave_state_changed = 1; |
1da177e4 LT |
3111 | |
3112 | /* primary_slave has no meaning in round-robin | |
3113 | * mode. the window of a slave being up and | |
3114 | * curr_active_slave being null after enslaving | |
3115 | * is closed. | |
3116 | */ | |
3117 | if (!oldcurrent) { | |
e2a7420d | 3118 | slave_info(bond->dev, slave->dev, "link status definitely up\n"); |
1da177e4 LT |
3119 | do_failover = 1; |
3120 | } else { | |
e2a7420d | 3121 | slave_info(bond->dev, slave->dev, "interface is now up\n"); |
1da177e4 LT |
3122 | } |
3123 | } | |
3124 | } else { | |
3125 | /* slave->link == BOND_LINK_UP */ | |
3126 | ||
3127 | /* not all switches will respond to an arp request | |
3128 | * when the source ip is 0, so don't take the link down | |
3129 | * if we don't know our ip yet | |
3130 | */ | |
e7f63f1d | 3131 | if (!bond_time_in_interval(bond, trans_start, 2) || |
49f17de7 | 3132 | !bond_time_in_interval(bond, slave->last_rx, 2)) { |
1da177e4 | 3133 | |
1899bb32 | 3134 | bond_propose_link_state(slave, BOND_LINK_DOWN); |
6fde8f03 | 3135 | slave_state_changed = 1; |
1da177e4 | 3136 | |
3d632c3f | 3137 | if (slave->link_failure_count < UINT_MAX) |
1da177e4 | 3138 | slave->link_failure_count++; |
1da177e4 | 3139 | |
e2a7420d | 3140 | slave_info(bond->dev, slave->dev, "interface is now down\n"); |
1da177e4 | 3141 | |
3d632c3f | 3142 | if (slave == oldcurrent) |
1da177e4 | 3143 | do_failover = 1; |
1da177e4 LT |
3144 | } |
3145 | } | |
3146 | ||
3147 | /* note: if switch is in round-robin mode, all links | |
3148 | * must tx arp to ensure all links rx an arp - otherwise | |
3149 | * links may oscillate or not come up at all; if switch is | |
3150 | * in something like xor mode, there is nothing we can | |
3151 | * do - all replies will be rx'ed on same link causing slaves | |
3152 | * to be unstable during low/no traffic periods | |
3153 | */ | |
b6adc610 | 3154 | if (bond_slave_is_up(slave)) |
1da177e4 | 3155 | bond_arp_send_all(bond, slave); |
1da177e4 LT |
3156 | } |
3157 | ||
2e52f4fe | 3158 | rcu_read_unlock(); |
3159 | ||
6fde8f03 | 3160 | if (do_failover || slave_state_changed) { |
2e52f4fe | 3161 | if (!rtnl_trylock()) |
3162 | goto re_arm; | |
1da177e4 | 3163 | |
797a9364 | 3164 | bond_for_each_slave(bond, slave, iter) { |
1899bb32 JV |
3165 | if (slave->link_new_state != BOND_LINK_NOCHANGE) |
3166 | slave->link = slave->link_new_state; | |
797a9364 NS |
3167 | } |
3168 | ||
6fde8f03 DT |
3169 | if (slave_state_changed) { |
3170 | bond_slave_state_change(bond); | |
ee637714 MB |
3171 | if (BOND_MODE(bond) == BOND_MODE_XOR) |
3172 | bond_update_slave_arr(bond, NULL); | |
b8e4500f NA |
3173 | } |
3174 | if (do_failover) { | |
6fde8f03 | 3175 | block_netpoll_tx(); |
6fde8f03 | 3176 | bond_select_active_slave(bond); |
6fde8f03 DT |
3177 | unblock_netpoll_tx(); |
3178 | } | |
2e52f4fe | 3179 | rtnl_unlock(); |
1da177e4 LT |
3180 | } |
3181 | ||
3182 | re_arm: | |
e6d265e8 | 3183 | if (bond->params.arp_interval) |
e7f63f1d VF |
3184 | queue_delayed_work(bond->wq, &bond->arp_work, |
3185 | msecs_to_jiffies(bond->params.arp_interval)); | |
1da177e4 LT |
3186 | } |
3187 | ||
547942ca | 3188 | /* Called to inspect slaves for active-backup mode ARP monitor link state |
1899bb32 JV |
3189 | * changes. Sets proposed link state in slaves to specify what action |
3190 | * should take place for the slave. Returns 0 if no changes are found, >0 | |
3191 | * if changes to link states must be committed. | |
b2220cad | 3192 | * |
8c0bc550 | 3193 | * Called with rcu_read_lock held. |
1da177e4 | 3194 | */ |
e7f63f1d | 3195 | static int bond_ab_arp_inspect(struct bonding *bond) |
1da177e4 | 3196 | { |
def4460c | 3197 | unsigned long trans_start, last_rx; |
9caff1e7 | 3198 | struct list_head *iter; |
dec1e90e | 3199 | struct slave *slave; |
dec1e90e | 3200 | int commit = 0; |
da210f55 | 3201 | |
eb9fa4b0 | 3202 | bond_for_each_slave_rcu(bond, slave, iter) { |
1899bb32 | 3203 | bond_propose_link_state(slave, BOND_LINK_NOCHANGE); |
def4460c | 3204 | last_rx = slave_last_rx(bond, slave); |
1da177e4 | 3205 | |
b2220cad | 3206 | if (slave->link != BOND_LINK_UP) { |
e7f63f1d | 3207 | if (bond_time_in_interval(bond, last_rx, 1)) { |
1899bb32 | 3208 | bond_propose_link_state(slave, BOND_LINK_UP); |
b2220cad | 3209 | commit++; |
0410d071 JW |
3210 | } else if (slave->link == BOND_LINK_BACK) { |
3211 | bond_propose_link_state(slave, BOND_LINK_FAIL); | |
3212 | commit++; | |
b2220cad | 3213 | } |
b2220cad JV |
3214 | continue; |
3215 | } | |
1da177e4 | 3216 | |
547942ca | 3217 | /* Give slaves 2*delta after being enslaved or made |
b2220cad JV |
3218 | * active. This avoids bouncing, as the last receive |
3219 | * times need a full ARP monitor cycle to be updated. | |
3220 | */ | |
8e603460 | 3221 | if (bond_time_in_interval(bond, slave->last_link_up, 2)) |
b2220cad JV |
3222 | continue; |
3223 | ||
547942ca | 3224 | /* Backup slave is down if: |
b2220cad JV |
3225 | * - No current_arp_slave AND |
3226 | * - more than 3*delta since last receive AND | |
3227 | * - the bond has an IP address | |
3228 | * | |
3229 | * Note: a non-null current_arp_slave indicates | |
3230 | * the curr_active_slave went down and we are | |
3231 | * searching for a new one; under this condition | |
3232 | * we only take the curr_active_slave down - this | |
3233 | * gives each slave a chance to tx/rx traffic | |
3234 | * before being taken out | |
3235 | */ | |
e30bc066 | 3236 | if (!bond_is_active_slave(slave) && |
85741718 | 3237 | !rcu_access_pointer(bond->current_arp_slave) && |
e7f63f1d | 3238 | !bond_time_in_interval(bond, last_rx, 3)) { |
1899bb32 | 3239 | bond_propose_link_state(slave, BOND_LINK_DOWN); |
b2220cad JV |
3240 | commit++; |
3241 | } | |
3242 | ||
547942ca | 3243 | /* Active slave is down if: |
b2220cad JV |
3244 | * - more than 2*delta since transmitting OR |
3245 | * - (more than 2*delta since receive AND | |
3246 | * the bond has an IP address) | |
3247 | */ | |
cb32f2a0 | 3248 | trans_start = dev_trans_start(slave->dev); |
e30bc066 | 3249 | if (bond_is_active_slave(slave) && |
e7f63f1d VF |
3250 | (!bond_time_in_interval(bond, trans_start, 2) || |
3251 | !bond_time_in_interval(bond, last_rx, 2))) { | |
1899bb32 | 3252 | bond_propose_link_state(slave, BOND_LINK_DOWN); |
b2220cad JV |
3253 | commit++; |
3254 | } | |
1da177e4 LT |
3255 | } |
3256 | ||
b2220cad JV |
3257 | return commit; |
3258 | } | |
1da177e4 | 3259 | |
547942ca | 3260 | /* Called to commit link state changes noted by inspection step of |
b2220cad JV |
3261 | * active-backup mode ARP monitor. |
3262 | * | |
eb9fa4b0 | 3263 | * Called with RTNL hold. |
b2220cad | 3264 | */ |
e7f63f1d | 3265 | static void bond_ab_arp_commit(struct bonding *bond) |
b2220cad | 3266 | { |
cb32f2a0 | 3267 | unsigned long trans_start; |
9caff1e7 | 3268 | struct list_head *iter; |
dec1e90e | 3269 | struct slave *slave; |
1da177e4 | 3270 | |
9caff1e7 | 3271 | bond_for_each_slave(bond, slave, iter) { |
1899bb32 | 3272 | switch (slave->link_new_state) { |
b2220cad JV |
3273 | case BOND_LINK_NOCHANGE: |
3274 | continue; | |
ff59c456 | 3275 | |
b2220cad | 3276 | case BOND_LINK_UP: |
cb32f2a0 | 3277 | trans_start = dev_trans_start(slave->dev); |
4740d638 ED |
3278 | if (rtnl_dereference(bond->curr_active_slave) != slave || |
3279 | (!rtnl_dereference(bond->curr_active_slave) && | |
e7f63f1d | 3280 | bond_time_in_interval(bond, trans_start, 1))) { |
85741718 ED |
3281 | struct slave *current_arp_slave; |
3282 | ||
3283 | current_arp_slave = rtnl_dereference(bond->current_arp_slave); | |
5d397061 JP |
3284 | bond_set_slave_link_state(slave, BOND_LINK_UP, |
3285 | BOND_SLAVE_NOTIFY_NOW); | |
85741718 | 3286 | if (current_arp_slave) { |
5a430974 | 3287 | bond_set_slave_inactive_flags( |
85741718 | 3288 | current_arp_slave, |
5e5b0665 | 3289 | BOND_SLAVE_NOTIFY_NOW); |
85741718 | 3290 | RCU_INIT_POINTER(bond->current_arp_slave, NULL); |
5a430974 | 3291 | } |
b2220cad | 3292 | |
e2a7420d | 3293 | slave_info(bond->dev, slave->dev, "link status definitely up\n"); |
b2220cad | 3294 | |
4740d638 | 3295 | if (!rtnl_dereference(bond->curr_active_slave) || |
059b47e8 | 3296 | slave == rtnl_dereference(bond->primary_slave)) |
b9f60253 | 3297 | goto do_failover; |
1da177e4 | 3298 | |
b9f60253 | 3299 | } |
1da177e4 | 3300 | |
b9f60253 | 3301 | continue; |
1da177e4 | 3302 | |
b2220cad JV |
3303 | case BOND_LINK_DOWN: |
3304 | if (slave->link_failure_count < UINT_MAX) | |
3305 | slave->link_failure_count++; | |
3306 | ||
5d397061 JP |
3307 | bond_set_slave_link_state(slave, BOND_LINK_DOWN, |
3308 | BOND_SLAVE_NOTIFY_NOW); | |
5e5b0665 | 3309 | bond_set_slave_inactive_flags(slave, |
3310 | BOND_SLAVE_NOTIFY_NOW); | |
b2220cad | 3311 | |
e2a7420d | 3312 | slave_info(bond->dev, slave->dev, "link status definitely down, disabling slave\n"); |
b2220cad | 3313 | |
4740d638 | 3314 | if (slave == rtnl_dereference(bond->curr_active_slave)) { |
85741718 | 3315 | RCU_INIT_POINTER(bond->current_arp_slave, NULL); |
b9f60253 | 3316 | goto do_failover; |
1da177e4 | 3317 | } |
b9f60253 JP |
3318 | |
3319 | continue; | |
b2220cad | 3320 | |
0410d071 JW |
3321 | case BOND_LINK_FAIL: |
3322 | bond_set_slave_link_state(slave, BOND_LINK_FAIL, | |
3323 | BOND_SLAVE_NOTIFY_NOW); | |
3324 | bond_set_slave_inactive_flags(slave, | |
3325 | BOND_SLAVE_NOTIFY_NOW); | |
3326 | ||
3327 | /* A slave has just been enslaved and has become | |
3328 | * the current active slave. | |
3329 | */ | |
3330 | if (rtnl_dereference(bond->curr_active_slave)) | |
3331 | RCU_INIT_POINTER(bond->current_arp_slave, NULL); | |
3332 | continue; | |
3333 | ||
b2220cad | 3334 | default: |
1899bb32 JV |
3335 | slave_err(bond->dev, slave->dev, |
3336 | "impossible: link_new_state %d on slave\n", | |
3337 | slave->link_new_state); | |
b9f60253 | 3338 | continue; |
1da177e4 | 3339 | } |
1da177e4 | 3340 | |
b9f60253 | 3341 | do_failover: |
e843fa50 | 3342 | block_netpoll_tx(); |
b9f60253 | 3343 | bond_select_active_slave(bond); |
e843fa50 | 3344 | unblock_netpoll_tx(); |
b2220cad | 3345 | } |
1da177e4 | 3346 | |
b2220cad JV |
3347 | bond_set_carrier(bond); |
3348 | } | |
1da177e4 | 3349 | |
547942ca | 3350 | /* Send ARP probes for active-backup mode ARP monitor. |
b0929915 | 3351 | * |
8c0bc550 | 3352 | * Called with rcu_read_lock held. |
b2220cad | 3353 | */ |
f2ebd477 | 3354 | static bool bond_ab_arp_probe(struct bonding *bond) |
b2220cad | 3355 | { |
eb9fa4b0 | 3356 | struct slave *slave, *before = NULL, *new_slave = NULL, |
b0929915 | 3357 | *curr_arp_slave = rcu_dereference(bond->current_arp_slave), |
3358 | *curr_active_slave = rcu_dereference(bond->curr_active_slave); | |
4087df87 VF |
3359 | struct list_head *iter; |
3360 | bool found = false; | |
b0929915 | 3361 | bool should_notify_rtnl = BOND_SLAVE_NOTIFY_LATER; |
f2ebd477 | 3362 | |
98b90f26 | 3363 | if (curr_arp_slave && curr_active_slave) |
76444f50 VF |
3364 | netdev_info(bond->dev, "PROBE: c_arp %s && cas %s BAD\n", |
3365 | curr_arp_slave->dev->name, | |
3366 | curr_active_slave->dev->name); | |
1da177e4 | 3367 | |
98b90f26 VF |
3368 | if (curr_active_slave) { |
3369 | bond_arp_send_all(bond, curr_active_slave); | |
b0929915 | 3370 | return should_notify_rtnl; |
b2220cad | 3371 | } |
1da177e4 | 3372 | |
b2220cad JV |
3373 | /* if we don't have a curr_active_slave, search for the next available |
3374 | * backup slave from the current_arp_slave and make it the candidate | |
3375 | * for becoming the curr_active_slave | |
3376 | */ | |
1da177e4 | 3377 | |
eb9fa4b0 | 3378 | if (!curr_arp_slave) { |
b0929915 | 3379 | curr_arp_slave = bond_first_slave_rcu(bond); |
3380 | if (!curr_arp_slave) | |
3381 | return should_notify_rtnl; | |
b2220cad | 3382 | } |
1da177e4 | 3383 | |
b0929915 | 3384 | bond_for_each_slave_rcu(bond, slave, iter) { |
b6adc610 | 3385 | if (!found && !before && bond_slave_is_up(slave)) |
4087df87 | 3386 | before = slave; |
1da177e4 | 3387 | |
b6adc610 | 3388 | if (found && !new_slave && bond_slave_is_up(slave)) |
4087df87 | 3389 | new_slave = slave; |
b2220cad JV |
3390 | /* if the link state is up at this point, we |
3391 | * mark it down - this can happen if we have | |
3392 | * simultaneous link failures and | |
3393 | * reselect_active_interface doesn't make this | |
3394 | * one the current slave so it is still marked | |
3395 | * up when it is actually down | |
1da177e4 | 3396 | */ |
b6adc610 | 3397 | if (!bond_slave_is_up(slave) && slave->link == BOND_LINK_UP) { |
5d397061 JP |
3398 | bond_set_slave_link_state(slave, BOND_LINK_DOWN, |
3399 | BOND_SLAVE_NOTIFY_LATER); | |
b2220cad JV |
3400 | if (slave->link_failure_count < UINT_MAX) |
3401 | slave->link_failure_count++; | |
1da177e4 | 3402 | |
5e5b0665 | 3403 | bond_set_slave_inactive_flags(slave, |
b0929915 | 3404 | BOND_SLAVE_NOTIFY_LATER); |
b2220cad | 3405 | |
e2a7420d | 3406 | slave_info(bond->dev, slave->dev, "backup interface is now down\n"); |
1da177e4 | 3407 | } |
eb9fa4b0 | 3408 | if (slave == curr_arp_slave) |
4087df87 | 3409 | found = true; |
b2220cad | 3410 | } |
4087df87 VF |
3411 | |
3412 | if (!new_slave && before) | |
3413 | new_slave = before; | |
3414 | ||
b0929915 | 3415 | if (!new_slave) |
3416 | goto check_state; | |
4087df87 | 3417 | |
5d397061 JP |
3418 | bond_set_slave_link_state(new_slave, BOND_LINK_BACK, |
3419 | BOND_SLAVE_NOTIFY_LATER); | |
b0929915 | 3420 | bond_set_slave_active_flags(new_slave, BOND_SLAVE_NOTIFY_LATER); |
4087df87 | 3421 | bond_arp_send_all(bond, new_slave); |
8e603460 | 3422 | new_slave->last_link_up = jiffies; |
eb9fa4b0 | 3423 | rcu_assign_pointer(bond->current_arp_slave, new_slave); |
f2ebd477 | 3424 | |
b0929915 | 3425 | check_state: |
3426 | bond_for_each_slave_rcu(bond, slave, iter) { | |
5d397061 | 3427 | if (slave->should_notify || slave->should_notify_link) { |
b0929915 | 3428 | should_notify_rtnl = BOND_SLAVE_NOTIFY_NOW; |
3429 | break; | |
3430 | } | |
3431 | } | |
3432 | return should_notify_rtnl; | |
b2220cad | 3433 | } |
1da177e4 | 3434 | |
d5e73f7b | 3435 | static void bond_activebackup_arp_mon(struct bonding *bond) |
b2220cad | 3436 | { |
b0929915 | 3437 | bool should_notify_peers = false; |
3438 | bool should_notify_rtnl = false; | |
1f2cd845 | 3439 | int delta_in_ticks; |
1da177e4 | 3440 | |
1f2cd845 DM |
3441 | delta_in_ticks = msecs_to_jiffies(bond->params.arp_interval); |
3442 | ||
3443 | if (!bond_has_slaves(bond)) | |
b2220cad JV |
3444 | goto re_arm; |
3445 | ||
eb9fa4b0 | 3446 | rcu_read_lock(); |
b0929915 | 3447 | |
ad246c99 BH |
3448 | should_notify_peers = bond_should_notify_peers(bond); |
3449 | ||
b0929915 | 3450 | if (bond_ab_arp_inspect(bond)) { |
3451 | rcu_read_unlock(); | |
3452 | ||
1f2cd845 DM |
3453 | /* Race avoidance with bond_close flush of workqueue */ |
3454 | if (!rtnl_trylock()) { | |
1f2cd845 DM |
3455 | delta_in_ticks = 1; |
3456 | should_notify_peers = false; | |
3457 | goto re_arm; | |
3458 | } | |
b2220cad | 3459 | |
1f2cd845 | 3460 | bond_ab_arp_commit(bond); |
b0929915 | 3461 | |
1f2cd845 | 3462 | rtnl_unlock(); |
b0929915 | 3463 | rcu_read_lock(); |
1f2cd845 DM |
3464 | } |
3465 | ||
b0929915 | 3466 | should_notify_rtnl = bond_ab_arp_probe(bond); |
3467 | rcu_read_unlock(); | |
ad246c99 | 3468 | |
80b9d236 | 3469 | re_arm: |
3470 | if (bond->params.arp_interval) | |
1f2cd845 DM |
3471 | queue_delayed_work(bond->wq, &bond->arp_work, delta_in_ticks); |
3472 | ||
b0929915 | 3473 | if (should_notify_peers || should_notify_rtnl) { |
1f2cd845 DM |
3474 | if (!rtnl_trylock()) |
3475 | return; | |
b0929915 | 3476 | |
3477 | if (should_notify_peers) | |
3478 | call_netdevice_notifiers(NETDEV_NOTIFY_PEERS, | |
3479 | bond->dev); | |
5d397061 | 3480 | if (should_notify_rtnl) { |
b0929915 | 3481 | bond_slave_state_notify(bond); |
5d397061 JP |
3482 | bond_slave_link_notify(bond); |
3483 | } | |
b0929915 | 3484 | |
1f2cd845 DM |
3485 | rtnl_unlock(); |
3486 | } | |
1da177e4 LT |
3487 | } |
3488 | ||
d5e73f7b MB |
3489 | static void bond_arp_monitor(struct work_struct *work) |
3490 | { | |
3491 | struct bonding *bond = container_of(work, struct bonding, | |
3492 | arp_work.work); | |
3493 | ||
3494 | if (BOND_MODE(bond) == BOND_MODE_ACTIVEBACKUP) | |
3495 | bond_activebackup_arp_mon(bond); | |
3496 | else | |
3497 | bond_loadbalance_arp_mon(bond); | |
3498 | } | |
3499 | ||
1da177e4 LT |
3500 | /*-------------------------- netdev event handling --------------------------*/ |
3501 | ||
547942ca | 3502 | /* Change device name */ |
1da177e4 LT |
3503 | static int bond_event_changename(struct bonding *bond) |
3504 | { | |
1da177e4 LT |
3505 | bond_remove_proc_entry(bond); |
3506 | bond_create_proc_entry(bond); | |
7e083840 | 3507 | |
f073c7ca TI |
3508 | bond_debug_reregister(bond); |
3509 | ||
1da177e4 LT |
3510 | return NOTIFY_DONE; |
3511 | } | |
3512 | ||
3d632c3f SH |
3513 | static int bond_master_netdev_event(unsigned long event, |
3514 | struct net_device *bond_dev) | |
1da177e4 | 3515 | { |
454d7c9b | 3516 | struct bonding *event_bond = netdev_priv(bond_dev); |
1da177e4 | 3517 | |
e2a7420d JW |
3518 | netdev_dbg(bond_dev, "%s called\n", __func__); |
3519 | ||
1da177e4 LT |
3520 | switch (event) { |
3521 | case NETDEV_CHANGENAME: | |
3522 | return bond_event_changename(event_bond); | |
a64d49c3 EB |
3523 | case NETDEV_UNREGISTER: |
3524 | bond_remove_proc_entry(event_bond); | |
5b69874f | 3525 | #ifdef CONFIG_XFRM_OFFLOAD |
9a560550 | 3526 | xfrm_dev_state_flush(dev_net(bond_dev), bond_dev, true); |
5b69874f | 3527 | #endif /* CONFIG_XFRM_OFFLOAD */ |
a64d49c3 EB |
3528 | break; |
3529 | case NETDEV_REGISTER: | |
3530 | bond_create_proc_entry(event_bond); | |
3531 | break; | |
1da177e4 LT |
3532 | default: |
3533 | break; | |
3534 | } | |
3535 | ||
3536 | return NOTIFY_DONE; | |
3537 | } | |
3538 | ||
3d632c3f SH |
3539 | static int bond_slave_netdev_event(unsigned long event, |
3540 | struct net_device *slave_dev) | |
1da177e4 | 3541 | { |
059b47e8 | 3542 | struct slave *slave = bond_slave_get_rtnl(slave_dev), *primary; |
6101391d | 3543 | struct bonding *bond; |
3544 | struct net_device *bond_dev; | |
1da177e4 | 3545 | |
6101391d | 3546 | /* A netdev event can be generated while enslaving a device |
3547 | * before netdev_rx_handler_register is called in which case | |
3548 | * slave will be NULL | |
3549 | */ | |
e2a7420d JW |
3550 | if (!slave) { |
3551 | netdev_dbg(slave_dev, "%s called on NULL slave\n", __func__); | |
6101391d | 3552 | return NOTIFY_DONE; |
e2a7420d JW |
3553 | } |
3554 | ||
6101391d | 3555 | bond_dev = slave->bond->dev; |
3556 | bond = slave->bond; | |
059b47e8 | 3557 | primary = rtnl_dereference(bond->primary_slave); |
6101391d | 3558 | |
e2a7420d JW |
3559 | slave_dbg(bond_dev, slave_dev, "%s called\n", __func__); |
3560 | ||
1da177e4 LT |
3561 | switch (event) { |
3562 | case NETDEV_UNREGISTER: | |
8d2ada77 | 3563 | if (bond_dev->type != ARPHRD_ETHER) |
471cb5a3 JP |
3564 | bond_release_and_destroy(bond_dev, slave_dev); |
3565 | else | |
f51048c3 | 3566 | __bond_release_one(bond_dev, slave_dev, false, true); |
1da177e4 | 3567 | break; |
98f41f69 | 3568 | case NETDEV_UP: |
1da177e4 | 3569 | case NETDEV_CHANGE: |
4d2c0cda MB |
3570 | /* For 802.3ad mode only: |
3571 | * Getting invalid Speed/Duplex values here will put slave | |
33403121 JW |
3572 | * in weird state. Mark it as link-fail if the link was |
3573 | * previously up or link-down if it hasn't yet come up, and | |
3574 | * let link-monitoring (miimon) set it right when correct | |
3575 | * speeds/duplex are available. | |
4d2c0cda MB |
3576 | */ |
3577 | if (bond_update_speed_duplex(slave) && | |
33403121 JW |
3578 | BOND_MODE(bond) == BOND_MODE_8023AD) { |
3579 | if (slave->last_link_up) | |
3580 | slave->link = BOND_LINK_FAIL; | |
3581 | else | |
3582 | slave->link = BOND_LINK_DOWN; | |
3583 | } | |
4d2c0cda | 3584 | |
52bc6716 MB |
3585 | if (BOND_MODE(bond) == BOND_MODE_8023AD) |
3586 | bond_3ad_adapter_speed_duplex_changed(slave); | |
df561f66 | 3587 | fallthrough; |
950ddcb1 | 3588 | case NETDEV_DOWN: |
ee637714 MB |
3589 | /* Refresh slave-array if applicable! |
3590 | * If the setup does not use miimon or arpmon (mode-specific!), | |
3591 | * then these events will not cause the slave-array to be | |
3592 | * refreshed. This will cause xmit to use a slave that is not | |
3593 | * usable. Avoid such situation by refeshing the array at these | |
3594 | * events. If these (miimon/arpmon) parameters are configured | |
3595 | * then array gets refreshed twice and that should be fine! | |
3596 | */ | |
e79c1055 | 3597 | if (bond_mode_can_use_xmit_hash(bond)) |
ee637714 | 3598 | bond_update_slave_arr(bond, NULL); |
1da177e4 | 3599 | break; |
1da177e4 | 3600 | case NETDEV_CHANGEMTU: |
547942ca | 3601 | /* TODO: Should slaves be allowed to |
1da177e4 LT |
3602 | * independently alter their MTU? For |
3603 | * an active-backup bond, slaves need | |
3604 | * not be the same type of device, so | |
3605 | * MTUs may vary. For other modes, | |
3606 | * slaves arguably should have the | |
3607 | * same MTUs. To do this, we'd need to | |
3608 | * take over the slave's change_mtu | |
3609 | * function for the duration of their | |
3610 | * servitude. | |
3611 | */ | |
3612 | break; | |
3613 | case NETDEV_CHANGENAME: | |
3ec775b9 | 3614 | /* we don't care if we don't have primary set */ |
ec0865a9 | 3615 | if (!bond_uses_primary(bond) || |
3ec775b9 VF |
3616 | !bond->params.primary[0]) |
3617 | break; | |
3618 | ||
059b47e8 | 3619 | if (slave == primary) { |
3ec775b9 | 3620 | /* slave's name changed - he's no longer primary */ |
059b47e8 | 3621 | RCU_INIT_POINTER(bond->primary_slave, NULL); |
3ec775b9 VF |
3622 | } else if (!strcmp(slave_dev->name, bond->params.primary)) { |
3623 | /* we have a new primary slave */ | |
059b47e8 | 3624 | rcu_assign_pointer(bond->primary_slave, slave); |
3ec775b9 VF |
3625 | } else { /* we didn't change primary - exit */ |
3626 | break; | |
3627 | } | |
3628 | ||
76444f50 | 3629 | netdev_info(bond->dev, "Primary slave changed to %s, reselecting active slave\n", |
059b47e8 | 3630 | primary ? slave_dev->name : "none"); |
f80889a5 | 3631 | |
3632 | block_netpoll_tx(); | |
3ec775b9 | 3633 | bond_select_active_slave(bond); |
f80889a5 | 3634 | unblock_netpoll_tx(); |
1da177e4 | 3635 | break; |
8531c5ff AK |
3636 | case NETDEV_FEAT_CHANGE: |
3637 | bond_compute_features(bond); | |
3638 | break; | |
4aa5dee4 JP |
3639 | case NETDEV_RESEND_IGMP: |
3640 | /* Propagate to master device */ | |
3641 | call_netdevice_notifiers(event, slave->bond->dev); | |
3642 | break; | |
1da177e4 LT |
3643 | default: |
3644 | break; | |
3645 | } | |
3646 | ||
3647 | return NOTIFY_DONE; | |
3648 | } | |
3649 | ||
547942ca | 3650 | /* bond_netdev_event: handle netdev notifier chain events. |
1da177e4 LT |
3651 | * |
3652 | * This function receives events for the netdev chain. The caller (an | |
e041c683 | 3653 | * ioctl handler calling blocking_notifier_call_chain) holds the necessary |
1da177e4 LT |
3654 | * locks for us to safely manipulate the slave devices (RTNL lock, |
3655 | * dev_probe_lock). | |
3656 | */ | |
3d632c3f SH |
3657 | static int bond_netdev_event(struct notifier_block *this, |
3658 | unsigned long event, void *ptr) | |
1da177e4 | 3659 | { |
351638e7 | 3660 | struct net_device *event_dev = netdev_notifier_info_to_dev(ptr); |
1da177e4 | 3661 | |
75466dce JW |
3662 | netdev_dbg(event_dev, "%s received %s\n", |
3663 | __func__, netdev_cmd_to_name(event)); | |
1da177e4 | 3664 | |
0b680e75 JV |
3665 | if (!(event_dev->priv_flags & IFF_BONDING)) |
3666 | return NOTIFY_DONE; | |
3667 | ||
1da177e4 | 3668 | if (event_dev->flags & IFF_MASTER) { |
92480b39 SD |
3669 | int ret; |
3670 | ||
92480b39 SD |
3671 | ret = bond_master_netdev_event(event, event_dev); |
3672 | if (ret != NOTIFY_DONE) | |
3673 | return ret; | |
1da177e4 LT |
3674 | } |
3675 | ||
e2a7420d | 3676 | if (event_dev->flags & IFF_SLAVE) |
1da177e4 | 3677 | return bond_slave_netdev_event(event, event_dev); |
1da177e4 LT |
3678 | |
3679 | return NOTIFY_DONE; | |
3680 | } | |
3681 | ||
3682 | static struct notifier_block bond_netdev_notifier = { | |
3683 | .notifier_call = bond_netdev_event, | |
3684 | }; | |
3685 | ||
169a3e66 JV |
3686 | /*---------------------------- Hashing Policies -----------------------------*/ |
3687 | ||
a815bde5 JM |
3688 | /* Helper to access data in a packet, with or without a backing skb. |
3689 | * If skb is given the data is linearized if necessary via pskb_may_pull. | |
3690 | */ | |
3691 | static inline const void *bond_pull_data(struct sk_buff *skb, | |
3692 | const void *data, int hlen, int n) | |
3693 | { | |
3694 | if (likely(n <= hlen)) | |
3695 | return data; | |
3696 | else if (skb && likely(pskb_may_pull(skb, n))) | |
3697 | return skb->head; | |
3698 | ||
3699 | return NULL; | |
3700 | } | |
3701 | ||
32819dc1 | 3702 | /* L2 hash helper */ |
a815bde5 | 3703 | static inline u32 bond_eth_hash(struct sk_buff *skb, const void *data, int mhoff, int hlen) |
6b923cb7 | 3704 | { |
a815bde5 | 3705 | struct ethhdr *ep; |
6b923cb7 | 3706 | |
a815bde5 JM |
3707 | data = bond_pull_data(skb, data, hlen, mhoff + sizeof(struct ethhdr)); |
3708 | if (!data) | |
3709 | return 0; | |
3710 | ||
3711 | ep = (struct ethhdr *)(data + mhoff); | |
9e2ee5c7 | 3712 | return ep->h_dest[5] ^ ep->h_source[5] ^ be16_to_cpu(ep->h_proto); |
6b923cb7 JE |
3713 | } |
3714 | ||
a815bde5 JM |
3715 | static bool bond_flow_ip(struct sk_buff *skb, struct flow_keys *fk, const void *data, |
3716 | int hlen, __be16 l2_proto, int *nhoff, int *ip_proto, bool l34) | |
df98be06 MC |
3717 | { |
3718 | const struct ipv6hdr *iph6; | |
3719 | const struct iphdr *iph; | |
3720 | ||
a815bde5 JM |
3721 | if (l2_proto == htons(ETH_P_IP)) { |
3722 | data = bond_pull_data(skb, data, hlen, *nhoff + sizeof(*iph)); | |
3723 | if (!data) | |
df98be06 | 3724 | return false; |
a815bde5 JM |
3725 | |
3726 | iph = (const struct iphdr *)(data + *nhoff); | |
df98be06 | 3727 | iph_to_flow_copy_v4addrs(fk, iph); |
a815bde5 | 3728 | *nhoff += iph->ihl << 2; |
df98be06 | 3729 | if (!ip_is_fragment(iph)) |
a815bde5 JM |
3730 | *ip_proto = iph->protocol; |
3731 | } else if (l2_proto == htons(ETH_P_IPV6)) { | |
3732 | data = bond_pull_data(skb, data, hlen, *nhoff + sizeof(*iph6)); | |
3733 | if (!data) | |
df98be06 | 3734 | return false; |
a815bde5 JM |
3735 | |
3736 | iph6 = (const struct ipv6hdr *)(data + *nhoff); | |
df98be06 | 3737 | iph_to_flow_copy_v6addrs(fk, iph6); |
a815bde5 JM |
3738 | *nhoff += sizeof(*iph6); |
3739 | *ip_proto = iph6->nexthdr; | |
df98be06 MC |
3740 | } else { |
3741 | return false; | |
3742 | } | |
3743 | ||
a815bde5 JM |
3744 | if (l34 && *ip_proto >= 0) |
3745 | fk->ports.ports = __skb_flow_get_ports(skb, *nhoff, *ip_proto, data, hlen); | |
df98be06 MC |
3746 | |
3747 | return true; | |
3748 | } | |
3749 | ||
a815bde5 | 3750 | static u32 bond_vlan_srcmac_hash(struct sk_buff *skb, const void *data, int mhoff, int hlen) |
7b8fc010 | 3751 | { |
7b8fc010 | 3752 | u32 srcmac_vendor = 0, srcmac_dev = 0; |
39a0876d JM |
3753 | struct ethhdr *mac_hdr; |
3754 | u16 vlan = 0; | |
7b8fc010 JW |
3755 | int i; |
3756 | ||
a815bde5 JM |
3757 | data = bond_pull_data(skb, data, hlen, mhoff + sizeof(struct ethhdr)); |
3758 | if (!data) | |
3759 | return 0; | |
3760 | mac_hdr = (struct ethhdr *)(data + mhoff); | |
3761 | ||
7b8fc010 JW |
3762 | for (i = 0; i < 3; i++) |
3763 | srcmac_vendor = (srcmac_vendor << 8) | mac_hdr->h_source[i]; | |
3764 | ||
3765 | for (i = 3; i < ETH_ALEN; i++) | |
3766 | srcmac_dev = (srcmac_dev << 8) | mac_hdr->h_source[i]; | |
3767 | ||
39a0876d JM |
3768 | if (skb && skb_vlan_tag_present(skb)) |
3769 | vlan = skb_vlan_tag_get(skb); | |
7b8fc010 JW |
3770 | |
3771 | return vlan ^ srcmac_vendor ^ srcmac_dev; | |
3772 | } | |
3773 | ||
32819dc1 | 3774 | /* Extract the appropriate headers based on bond's xmit policy */ |
a815bde5 JM |
3775 | static bool bond_flow_dissect(struct bonding *bond, struct sk_buff *skb, const void *data, |
3776 | __be16 l2_proto, int nhoff, int hlen, struct flow_keys *fk) | |
6f6652be | 3777 | { |
df98be06 | 3778 | bool l34 = bond->params.xmit_policy == BOND_XMIT_POLICY_LAYER34; |
a815bde5 | 3779 | int ip_proto = -1; |
6b923cb7 | 3780 | |
7b8fc010 JW |
3781 | switch (bond->params.xmit_policy) { |
3782 | case BOND_XMIT_POLICY_ENCAP23: | |
3783 | case BOND_XMIT_POLICY_ENCAP34: | |
58deb77c MC |
3784 | memset(fk, 0, sizeof(*fk)); |
3785 | return __skb_flow_dissect(NULL, skb, &flow_keys_bonding, | |
a815bde5 | 3786 | fk, data, l2_proto, nhoff, hlen, 0); |
7b8fc010 JW |
3787 | default: |
3788 | break; | |
58deb77c | 3789 | } |
32819dc1 | 3790 | |
06635a35 | 3791 | fk->ports.ports = 0; |
58deb77c | 3792 | memset(&fk->icmp, 0, sizeof(fk->icmp)); |
a815bde5 | 3793 | if (!bond_flow_ip(skb, fk, data, hlen, l2_proto, &nhoff, &ip_proto, l34)) |
32819dc1 | 3794 | return false; |
df98be06 MC |
3795 | |
3796 | /* ICMP error packets contains at least 8 bytes of the header | |
3797 | * of the packet which generated the error. Use this information | |
3798 | * to correlate ICMP error packets within the same flow which | |
3799 | * generated the error. | |
3800 | */ | |
a815bde5 JM |
3801 | if (ip_proto == IPPROTO_ICMP || ip_proto == IPPROTO_ICMPV6) { |
3802 | skb_flow_get_icmp_tci(skb, &fk->icmp, data, nhoff, hlen); | |
3803 | if (ip_proto == IPPROTO_ICMP) { | |
df98be06 MC |
3804 | if (!icmp_is_err(fk->icmp.type)) |
3805 | return true; | |
3806 | ||
a815bde5 JM |
3807 | nhoff += sizeof(struct icmphdr); |
3808 | } else if (ip_proto == IPPROTO_ICMPV6) { | |
df98be06 MC |
3809 | if (!icmpv6_is_err(fk->icmp.type)) |
3810 | return true; | |
3811 | ||
a815bde5 | 3812 | nhoff += sizeof(struct icmp6hdr); |
df98be06 | 3813 | } |
a815bde5 | 3814 | return bond_flow_ip(skb, fk, data, hlen, l2_proto, &nhoff, &ip_proto, l34); |
58deb77c | 3815 | } |
6f6652be | 3816 | |
32819dc1 | 3817 | return true; |
6f6652be JV |
3818 | } |
3819 | ||
5b998545 TT |
3820 | static u32 bond_ip_hash(u32 hash, struct flow_keys *flow) |
3821 | { | |
3822 | hash ^= (__force u32)flow_get_u32_dst(flow) ^ | |
3823 | (__force u32)flow_get_u32_src(flow); | |
3824 | hash ^= (hash >> 16); | |
3825 | hash ^= (hash >> 8); | |
3826 | /* discard lowest hash bit to deal with the common even ports pattern */ | |
3827 | return hash >> 1; | |
3828 | } | |
3829 | ||
a815bde5 JM |
3830 | /* Generate hash based on xmit policy. If @skb is given it is used to linearize |
3831 | * the data as required, but this function can be used without it if the data is | |
3832 | * known to be linear (e.g. with xdp_buff). | |
169a3e66 | 3833 | */ |
a815bde5 JM |
3834 | static u32 __bond_xmit_hash(struct bonding *bond, struct sk_buff *skb, const void *data, |
3835 | __be16 l2_proto, int mhoff, int nhoff, int hlen) | |
169a3e66 | 3836 | { |
32819dc1 NA |
3837 | struct flow_keys flow; |
3838 | u32 hash; | |
4394542c | 3839 | |
7b8fc010 | 3840 | if (bond->params.xmit_policy == BOND_XMIT_POLICY_VLAN_SRCMAC) |
a815bde5 | 3841 | return bond_vlan_srcmac_hash(skb, data, mhoff, hlen); |
7b8fc010 | 3842 | |
32819dc1 | 3843 | if (bond->params.xmit_policy == BOND_XMIT_POLICY_LAYER2 || |
a815bde5 JM |
3844 | !bond_flow_dissect(bond, skb, data, l2_proto, nhoff, hlen, &flow)) |
3845 | return bond_eth_hash(skb, data, mhoff, hlen); | |
169a3e66 | 3846 | |
32819dc1 | 3847 | if (bond->params.xmit_policy == BOND_XMIT_POLICY_LAYER23 || |
58deb77c | 3848 | bond->params.xmit_policy == BOND_XMIT_POLICY_ENCAP23) { |
a815bde5 | 3849 | hash = bond_eth_hash(skb, data, mhoff, hlen); |
58deb77c MC |
3850 | } else { |
3851 | if (flow.icmp.id) | |
3852 | memcpy(&hash, &flow.icmp, sizeof(hash)); | |
3853 | else | |
3854 | memcpy(&hash, &flow.ports.ports, sizeof(hash)); | |
3855 | } | |
32819dc1 | 3856 | |
5b998545 | 3857 | return bond_ip_hash(hash, &flow); |
169a3e66 JV |
3858 | } |
3859 | ||
a815bde5 JM |
3860 | /** |
3861 | * bond_xmit_hash - generate a hash value based on the xmit policy | |
3862 | * @bond: bonding device | |
3863 | * @skb: buffer to use for headers | |
3864 | * | |
3865 | * This function will extract the necessary headers from the skb buffer and use | |
3866 | * them to generate a hash based on the xmit_policy set in the bonding device | |
3867 | */ | |
3868 | u32 bond_xmit_hash(struct bonding *bond, struct sk_buff *skb) | |
3869 | { | |
3870 | if (bond->params.xmit_policy == BOND_XMIT_POLICY_ENCAP34 && | |
3871 | skb->l4_hash) | |
3872 | return skb->hash; | |
3873 | ||
63b97a37 MT |
3874 | return __bond_xmit_hash(bond, skb, skb->data, skb->protocol, |
3875 | skb_mac_offset(skb), skb_network_offset(skb), | |
a815bde5 JM |
3876 | skb_headlen(skb)); |
3877 | } | |
3878 | ||
9e2ee5c7 JM |
3879 | /** |
3880 | * bond_xmit_hash_xdp - generate a hash value based on the xmit policy | |
3881 | * @bond: bonding device | |
3882 | * @xdp: buffer to use for headers | |
3883 | * | |
3884 | * The XDP variant of bond_xmit_hash. | |
3885 | */ | |
3886 | static u32 bond_xmit_hash_xdp(struct bonding *bond, struct xdp_buff *xdp) | |
3887 | { | |
3888 | struct ethhdr *eth; | |
3889 | ||
3890 | if (xdp->data + sizeof(struct ethhdr) > xdp->data_end) | |
3891 | return 0; | |
3892 | ||
3893 | eth = (struct ethhdr *)xdp->data; | |
3894 | ||
3895 | return __bond_xmit_hash(bond, NULL, xdp->data, eth->h_proto, 0, | |
3896 | sizeof(struct ethhdr), xdp->data_end - xdp->data); | |
3897 | } | |
3898 | ||
1da177e4 LT |
3899 | /*-------------------------- Device entry points ----------------------------*/ |
3900 | ||
ea8ffc08 | 3901 | void bond_work_init_all(struct bonding *bond) |
fbb0c41b | 3902 | { |
3903 | INIT_DELAYED_WORK(&bond->mcast_work, | |
3904 | bond_resend_igmp_join_requests_delayed); | |
3905 | INIT_DELAYED_WORK(&bond->alb_work, bond_alb_monitor); | |
3906 | INIT_DELAYED_WORK(&bond->mii_work, bond_mii_monitor); | |
d5e73f7b | 3907 | INIT_DELAYED_WORK(&bond->arp_work, bond_arp_monitor); |
fbb0c41b | 3908 | INIT_DELAYED_WORK(&bond->ad_work, bond_3ad_state_machine_handler); |
ee637714 | 3909 | INIT_DELAYED_WORK(&bond->slave_arr_work, bond_slave_arr_handler); |
fbb0c41b | 3910 | } |
3911 | ||
3912 | static void bond_work_cancel_all(struct bonding *bond) | |
3913 | { | |
3914 | cancel_delayed_work_sync(&bond->mii_work); | |
3915 | cancel_delayed_work_sync(&bond->arp_work); | |
3916 | cancel_delayed_work_sync(&bond->alb_work); | |
3917 | cancel_delayed_work_sync(&bond->ad_work); | |
3918 | cancel_delayed_work_sync(&bond->mcast_work); | |
ee637714 | 3919 | cancel_delayed_work_sync(&bond->slave_arr_work); |
fbb0c41b | 3920 | } |
3921 | ||
1da177e4 LT |
3922 | static int bond_open(struct net_device *bond_dev) |
3923 | { | |
454d7c9b | 3924 | struct bonding *bond = netdev_priv(bond_dev); |
9caff1e7 | 3925 | struct list_head *iter; |
ba3211cc | 3926 | struct slave *slave; |
1da177e4 | 3927 | |
ba3211cc | 3928 | /* reset slave->backup and slave->inactive */ |
0965a1f3 | 3929 | if (bond_has_slaves(bond)) { |
9caff1e7 | 3930 | bond_for_each_slave(bond, slave, iter) { |
4740d638 ED |
3931 | if (bond_uses_primary(bond) && |
3932 | slave != rcu_access_pointer(bond->curr_active_slave)) { | |
5e5b0665 | 3933 | bond_set_slave_inactive_flags(slave, |
3934 | BOND_SLAVE_NOTIFY_NOW); | |
8bbe71a5 | 3935 | } else if (BOND_MODE(bond) != BOND_MODE_8023AD) { |
5e5b0665 | 3936 | bond_set_slave_active_flags(slave, |
3937 | BOND_SLAVE_NOTIFY_NOW); | |
ba3211cc PP |
3938 | } |
3939 | } | |
ba3211cc | 3940 | } |
ba3211cc | 3941 | |
58402054 | 3942 | if (bond_is_lb(bond)) { |
1da177e4 LT |
3943 | /* bond_alb_initialize must be called before the timer |
3944 | * is started. | |
3945 | */ | |
01844098 | 3946 | if (bond_alb_initialize(bond, (BOND_MODE(bond) == BOND_MODE_ALB))) |
b473946a | 3947 | return -ENOMEM; |
e79c1055 | 3948 | if (bond->params.tlb_dynamic_lb || BOND_MODE(bond) == BOND_MODE_ALB) |
e9f0fb88 | 3949 | queue_delayed_work(bond->wq, &bond->alb_work, 0); |
1da177e4 LT |
3950 | } |
3951 | ||
fbb0c41b | 3952 | if (bond->params.miimon) /* link check interval, in milliseconds. */ |
1b76b316 | 3953 | queue_delayed_work(bond->wq, &bond->mii_work, 0); |
1da177e4 LT |
3954 | |
3955 | if (bond->params.arp_interval) { /* arp interval, in milliseconds. */ | |
1b76b316 | 3956 | queue_delayed_work(bond->wq, &bond->arp_work, 0); |
3fe68df9 | 3957 | bond->recv_probe = bond_arp_rcv; |
1da177e4 LT |
3958 | } |
3959 | ||
01844098 | 3960 | if (BOND_MODE(bond) == BOND_MODE_8023AD) { |
1b76b316 | 3961 | queue_delayed_work(bond->wq, &bond->ad_work, 0); |
1da177e4 | 3962 | /* register to receive LACPDUs */ |
3aba891d | 3963 | bond->recv_probe = bond_3ad_lacpdu_recv; |
fd989c83 | 3964 | bond_3ad_initiate_agg_selection(bond, 1); |
1da177e4 LT |
3965 | } |
3966 | ||
e79c1055 | 3967 | if (bond_mode_can_use_xmit_hash(bond)) |
ee637714 MB |
3968 | bond_update_slave_arr(bond, NULL); |
3969 | ||
1da177e4 LT |
3970 | return 0; |
3971 | } | |
3972 | ||
3973 | static int bond_close(struct net_device *bond_dev) | |
3974 | { | |
454d7c9b | 3975 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 | 3976 | |
fbb0c41b | 3977 | bond_work_cancel_all(bond); |
6c8d23f7 | 3978 | bond->send_peer_notif = 0; |
ee8487c0 | 3979 | if (bond_is_lb(bond)) |
1da177e4 | 3980 | bond_alb_deinitialize(bond); |
3aba891d | 3981 | bond->recv_probe = NULL; |
1da177e4 LT |
3982 | |
3983 | return 0; | |
3984 | } | |
3985 | ||
fe30937b ED |
3986 | /* fold stats, assuming all rtnl_link_stats64 fields are u64, but |
3987 | * that some drivers can provide 32bit values only. | |
3988 | */ | |
3989 | static void bond_fold_stats(struct rtnl_link_stats64 *_res, | |
3990 | const struct rtnl_link_stats64 *_new, | |
3991 | const struct rtnl_link_stats64 *_old) | |
3992 | { | |
3993 | const u64 *new = (const u64 *)_new; | |
3994 | const u64 *old = (const u64 *)_old; | |
3995 | u64 *res = (u64 *)_res; | |
3996 | int i; | |
3997 | ||
3998 | for (i = 0; i < sizeof(*_res) / sizeof(u64); i++) { | |
3999 | u64 nv = new[i]; | |
4000 | u64 ov = old[i]; | |
142c6594 | 4001 | s64 delta = nv - ov; |
fe30937b ED |
4002 | |
4003 | /* detects if this particular field is 32bit only */ | |
4004 | if (((nv | ov) >> 32) == 0) | |
142c6594 ED |
4005 | delta = (s64)(s32)((u32)nv - (u32)ov); |
4006 | ||
4007 | /* filter anomalies, some drivers reset their stats | |
4008 | * at down/up events. | |
4009 | */ | |
4010 | if (delta > 0) | |
4011 | res[i] += delta; | |
fe30937b ED |
4012 | } |
4013 | } | |
4014 | ||
b3e80d44 TY |
4015 | #ifdef CONFIG_LOCKDEP |
4016 | static int bond_get_lowest_level_rcu(struct net_device *dev) | |
4017 | { | |
4018 | struct net_device *ldev, *next, *now, *dev_stack[MAX_NEST_DEV + 1]; | |
4019 | struct list_head *niter, *iter, *iter_stack[MAX_NEST_DEV + 1]; | |
4020 | int cur = 0, max = 0; | |
4021 | ||
4022 | now = dev; | |
4023 | iter = &dev->adj_list.lower; | |
4024 | ||
4025 | while (1) { | |
4026 | next = NULL; | |
4027 | while (1) { | |
4028 | ldev = netdev_next_lower_dev_rcu(now, &iter); | |
4029 | if (!ldev) | |
4030 | break; | |
4031 | ||
4032 | next = ldev; | |
4033 | niter = &ldev->adj_list.lower; | |
4034 | dev_stack[cur] = now; | |
4035 | iter_stack[cur++] = iter; | |
4036 | if (max <= cur) | |
4037 | max = cur; | |
4038 | break; | |
4039 | } | |
4040 | ||
4041 | if (!next) { | |
4042 | if (!cur) | |
4043 | return max; | |
4044 | next = dev_stack[--cur]; | |
4045 | niter = iter_stack[cur]; | |
4046 | } | |
4047 | ||
4048 | now = next; | |
4049 | iter = niter; | |
4050 | } | |
4051 | ||
4052 | return max; | |
4053 | } | |
4054 | #endif | |
4055 | ||
bc1f4470 | 4056 | static void bond_get_stats(struct net_device *bond_dev, |
4057 | struct rtnl_link_stats64 *stats) | |
1da177e4 | 4058 | { |
454d7c9b | 4059 | struct bonding *bond = netdev_priv(bond_dev); |
28172739 | 4060 | struct rtnl_link_stats64 temp; |
9caff1e7 | 4061 | struct list_head *iter; |
1da177e4 | 4062 | struct slave *slave; |
b3e80d44 | 4063 | int nest_level = 0; |
1da177e4 | 4064 | |
1da177e4 | 4065 | |
fe30937b | 4066 | rcu_read_lock(); |
b3e80d44 TY |
4067 | #ifdef CONFIG_LOCKDEP |
4068 | nest_level = bond_get_lowest_level_rcu(bond_dev); | |
4069 | #endif | |
4070 | ||
4071 | spin_lock_nested(&bond->stats_lock, nest_level); | |
4072 | memcpy(stats, &bond->bond_stats, sizeof(*stats)); | |
4073 | ||
fe30937b ED |
4074 | bond_for_each_slave_rcu(bond, slave, iter) { |
4075 | const struct rtnl_link_stats64 *new = | |
28172739 | 4076 | dev_get_stats(slave->dev, &temp); |
fe30937b ED |
4077 | |
4078 | bond_fold_stats(stats, new, &slave->slave_stats); | |
5f0c5f73 AG |
4079 | |
4080 | /* save off the slave stats for the next run */ | |
fe30937b | 4081 | memcpy(&slave->slave_stats, new, sizeof(*new)); |
5f0c5f73 | 4082 | } |
fe30937b | 4083 | |
5f0c5f73 | 4084 | memcpy(&bond->bond_stats, stats, sizeof(*stats)); |
fe30937b | 4085 | spin_unlock(&bond->stats_lock); |
b3e80d44 | 4086 | rcu_read_unlock(); |
1da177e4 LT |
4087 | } |
4088 | ||
a7605370 | 4089 | static int bond_eth_ioctl(struct net_device *bond_dev, struct ifreq *ifr, int cmd) |
1da177e4 | 4090 | { |
080a06e1 | 4091 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 | 4092 | struct mii_ioctl_data *mii = NULL; |
a7605370 | 4093 | int res; |
1da177e4 | 4094 | |
a7605370 | 4095 | netdev_dbg(bond_dev, "bond_eth_ioctl: cmd=%d\n", cmd); |
1da177e4 LT |
4096 | |
4097 | switch (cmd) { | |
1da177e4 LT |
4098 | case SIOCGMIIPHY: |
4099 | mii = if_mii(ifr); | |
3d632c3f | 4100 | if (!mii) |
1da177e4 | 4101 | return -EINVAL; |
3d632c3f | 4102 | |
1da177e4 | 4103 | mii->phy_id = 0; |
df561f66 | 4104 | fallthrough; |
1da177e4 | 4105 | case SIOCGMIIREG: |
547942ca | 4106 | /* We do this again just in case we were called by SIOCGMIIREG |
1da177e4 LT |
4107 | * instead of SIOCGMIIPHY. |
4108 | */ | |
4109 | mii = if_mii(ifr); | |
3d632c3f | 4110 | if (!mii) |
1da177e4 | 4111 | return -EINVAL; |
3d632c3f | 4112 | |
1da177e4 | 4113 | if (mii->reg_num == 1) { |
1da177e4 | 4114 | mii->val_out = 0; |
3d632c3f | 4115 | if (netif_carrier_ok(bond->dev)) |
1da177e4 | 4116 | mii->val_out = BMSR_LSTATUS; |
1da177e4 LT |
4117 | } |
4118 | ||
4119 | return 0; | |
a7605370 AB |
4120 | default: |
4121 | res = -EOPNOTSUPP; | |
4122 | } | |
4123 | ||
4124 | return res; | |
4125 | } | |
4126 | ||
4127 | static int bond_do_ioctl(struct net_device *bond_dev, struct ifreq *ifr, int cmd) | |
4128 | { | |
4129 | struct bonding *bond = netdev_priv(bond_dev); | |
4130 | struct net_device *slave_dev = NULL; | |
4131 | struct ifbond k_binfo; | |
4132 | struct ifbond __user *u_binfo = NULL; | |
4133 | struct ifslave k_sinfo; | |
4134 | struct ifslave __user *u_sinfo = NULL; | |
4135 | struct bond_opt_value newval; | |
4136 | struct net *net; | |
4137 | int res = 0; | |
4138 | ||
4139 | netdev_dbg(bond_dev, "bond_ioctl: cmd=%d\n", cmd); | |
4140 | ||
4141 | switch (cmd) { | |
1da177e4 LT |
4142 | case SIOCBONDINFOQUERY: |
4143 | u_binfo = (struct ifbond __user *)ifr->ifr_data; | |
4144 | ||
3d632c3f | 4145 | if (copy_from_user(&k_binfo, u_binfo, sizeof(ifbond))) |
1da177e4 | 4146 | return -EFAULT; |
1da177e4 | 4147 | |
3d67576d ZY |
4148 | bond_info_query(bond_dev, &k_binfo); |
4149 | if (copy_to_user(u_binfo, &k_binfo, sizeof(ifbond))) | |
3d632c3f | 4150 | return -EFAULT; |
1da177e4 | 4151 | |
3d67576d | 4152 | return 0; |
1da177e4 LT |
4153 | case SIOCBONDSLAVEINFOQUERY: |
4154 | u_sinfo = (struct ifslave __user *)ifr->ifr_data; | |
4155 | ||
3d632c3f | 4156 | if (copy_from_user(&k_sinfo, u_sinfo, sizeof(ifslave))) |
1da177e4 | 4157 | return -EFAULT; |
1da177e4 LT |
4158 | |
4159 | res = bond_slave_info_query(bond_dev, &k_sinfo); | |
3d632c3f SH |
4160 | if (res == 0 && |
4161 | copy_to_user(u_sinfo, &k_sinfo, sizeof(ifslave))) | |
4162 | return -EFAULT; | |
1da177e4 LT |
4163 | |
4164 | return res; | |
4165 | default: | |
1da177e4 LT |
4166 | break; |
4167 | } | |
4168 | ||
387ff911 G |
4169 | net = dev_net(bond_dev); |
4170 | ||
4171 | if (!ns_capable(net->user_ns, CAP_NET_ADMIN)) | |
1da177e4 | 4172 | return -EPERM; |
1da177e4 | 4173 | |
0917b933 | 4174 | slave_dev = __dev_get_by_name(net, ifr->ifr_slave); |
1da177e4 | 4175 | |
e2a7420d | 4176 | slave_dbg(bond_dev, slave_dev, "slave_dev=%p:\n", slave_dev); |
1da177e4 | 4177 | |
3d632c3f | 4178 | if (!slave_dev) |
0917b933 | 4179 | return -ENODEV; |
1da177e4 | 4180 | |
0917b933 | 4181 | switch (cmd) { |
0917b933 | 4182 | case SIOCBONDENSLAVE: |
33eaf2a6 | 4183 | res = bond_enslave(bond_dev, slave_dev, NULL); |
0917b933 | 4184 | break; |
0917b933 YX |
4185 | case SIOCBONDRELEASE: |
4186 | res = bond_release(bond_dev, slave_dev); | |
4187 | break; | |
0917b933 | 4188 | case SIOCBONDSETHWADDR: |
b9245914 | 4189 | res = bond_set_dev_addr(bond_dev, slave_dev); |
0917b933 | 4190 | break; |
0917b933 | 4191 | case SIOCBONDCHANGEACTIVE: |
d1fbd3ed | 4192 | bond_opt_initstr(&newval, slave_dev->name); |
7a7e96e0 VY |
4193 | res = __bond_opt_set_notify(bond, BOND_OPT_ACTIVE_SLAVE, |
4194 | &newval); | |
0917b933 YX |
4195 | break; |
4196 | default: | |
4197 | res = -EOPNOTSUPP; | |
1da177e4 LT |
4198 | } |
4199 | ||
1da177e4 LT |
4200 | return res; |
4201 | } | |
4202 | ||
232ec98e AB |
4203 | static int bond_siocdevprivate(struct net_device *bond_dev, struct ifreq *ifr, |
4204 | void __user *data, int cmd) | |
4205 | { | |
4206 | struct ifreq ifrdata = { .ifr_data = data }; | |
4207 | ||
4208 | switch (cmd) { | |
4209 | case BOND_INFO_QUERY_OLD: | |
4210 | return bond_do_ioctl(bond_dev, &ifrdata, SIOCBONDINFOQUERY); | |
4211 | case BOND_SLAVE_INFO_QUERY_OLD: | |
4212 | return bond_do_ioctl(bond_dev, &ifrdata, SIOCBONDSLAVEINFOQUERY); | |
4213 | case BOND_ENSLAVE_OLD: | |
4214 | return bond_do_ioctl(bond_dev, ifr, SIOCBONDENSLAVE); | |
4215 | case BOND_RELEASE_OLD: | |
4216 | return bond_do_ioctl(bond_dev, ifr, SIOCBONDRELEASE); | |
4217 | case BOND_SETHWADDR_OLD: | |
4218 | return bond_do_ioctl(bond_dev, ifr, SIOCBONDSETHWADDR); | |
4219 | case BOND_CHANGE_ACTIVE_OLD: | |
4220 | return bond_do_ioctl(bond_dev, ifr, SIOCBONDCHANGEACTIVE); | |
4221 | } | |
4222 | ||
4223 | return -EOPNOTSUPP; | |
4224 | } | |
4225 | ||
d03462b9 | 4226 | static void bond_change_rx_flags(struct net_device *bond_dev, int change) |
1da177e4 | 4227 | { |
454d7c9b | 4228 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 | 4229 | |
d03462b9 JP |
4230 | if (change & IFF_PROMISC) |
4231 | bond_set_promiscuity(bond, | |
4232 | bond_dev->flags & IFF_PROMISC ? 1 : -1); | |
3d632c3f | 4233 | |
d03462b9 JP |
4234 | if (change & IFF_ALLMULTI) |
4235 | bond_set_allmulti(bond, | |
4236 | bond_dev->flags & IFF_ALLMULTI ? 1 : -1); | |
4237 | } | |
1da177e4 | 4238 | |
303d1cbf | 4239 | static void bond_set_rx_mode(struct net_device *bond_dev) |
d03462b9 JP |
4240 | { |
4241 | struct bonding *bond = netdev_priv(bond_dev); | |
9caff1e7 | 4242 | struct list_head *iter; |
303d1cbf | 4243 | struct slave *slave; |
1da177e4 | 4244 | |
b3241870 | 4245 | rcu_read_lock(); |
ec0865a9 | 4246 | if (bond_uses_primary(bond)) { |
b3241870 | 4247 | slave = rcu_dereference(bond->curr_active_slave); |
303d1cbf JV |
4248 | if (slave) { |
4249 | dev_uc_sync(slave->dev, bond_dev); | |
4250 | dev_mc_sync(slave->dev, bond_dev); | |
4251 | } | |
303d1cbf | 4252 | } else { |
b3241870 | 4253 | bond_for_each_slave_rcu(bond, slave, iter) { |
303d1cbf JV |
4254 | dev_uc_sync_multiple(slave->dev, bond_dev); |
4255 | dev_mc_sync_multiple(slave->dev, bond_dev); | |
4256 | } | |
1da177e4 | 4257 | } |
b3241870 | 4258 | rcu_read_unlock(); |
1da177e4 LT |
4259 | } |
4260 | ||
234bcf8a | 4261 | static int bond_neigh_init(struct neighbour *n) |
00829823 | 4262 | { |
234bcf8a | 4263 | struct bonding *bond = netdev_priv(n->dev); |
234bcf8a SP |
4264 | const struct net_device_ops *slave_ops; |
4265 | struct neigh_parms parms; | |
dec1e90e | 4266 | struct slave *slave; |
9e99bfef | 4267 | int ret = 0; |
234bcf8a | 4268 | |
9e99bfef ED |
4269 | rcu_read_lock(); |
4270 | slave = bond_first_slave_rcu(bond); | |
080bfa1e | 4271 | if (!slave) |
9e99bfef | 4272 | goto out; |
234bcf8a | 4273 | slave_ops = slave->dev->netdev_ops; |
080bfa1e | 4274 | if (!slave_ops->ndo_neigh_setup) |
9e99bfef | 4275 | goto out; |
234bcf8a | 4276 | |
9e99bfef ED |
4277 | /* TODO: find another way [1] to implement this. |
4278 | * Passing a zeroed structure is fragile, | |
4279 | * but at least we do not pass garbage. | |
4280 | * | |
4281 | * [1] One way would be that ndo_neigh_setup() never touch | |
4282 | * struct neigh_parms, but propagate the new neigh_setup() | |
4283 | * back to ___neigh_create() / neigh_parms_alloc() | |
4284 | */ | |
4285 | memset(&parms, 0, sizeof(parms)); | |
234bcf8a | 4286 | ret = slave_ops->ndo_neigh_setup(slave->dev, &parms); |
234bcf8a | 4287 | |
9e99bfef ED |
4288 | if (ret) |
4289 | goto out; | |
234bcf8a | 4290 | |
9e99bfef ED |
4291 | if (parms.neigh_setup) |
4292 | ret = parms.neigh_setup(n); | |
4293 | out: | |
4294 | rcu_read_unlock(); | |
4295 | return ret; | |
234bcf8a SP |
4296 | } |
4297 | ||
547942ca | 4298 | /* The bonding ndo_neigh_setup is called at init time beofre any |
234bcf8a SP |
4299 | * slave exists. So we must declare proxy setup function which will |
4300 | * be used at run time to resolve the actual slave neigh param setup. | |
9918d5bf VF |
4301 | * |
4302 | * It's also called by master devices (such as vlans) to setup their | |
4303 | * underlying devices. In that case - do nothing, we're already set up from | |
4304 | * our init. | |
234bcf8a SP |
4305 | */ |
4306 | static int bond_neigh_setup(struct net_device *dev, | |
4307 | struct neigh_parms *parms) | |
4308 | { | |
9918d5bf VF |
4309 | /* modify only our neigh_parms */ |
4310 | if (parms->dev == dev) | |
4311 | parms->neigh_setup = bond_neigh_init; | |
00829823 | 4312 | |
00829823 SH |
4313 | return 0; |
4314 | } | |
4315 | ||
547942ca | 4316 | /* Change the MTU of all of a master's slaves to match the master */ |
1da177e4 LT |
4317 | static int bond_change_mtu(struct net_device *bond_dev, int new_mtu) |
4318 | { | |
454d7c9b | 4319 | struct bonding *bond = netdev_priv(bond_dev); |
81f23b13 | 4320 | struct slave *slave, *rollback_slave; |
9caff1e7 | 4321 | struct list_head *iter; |
1da177e4 | 4322 | int res = 0; |
1da177e4 | 4323 | |
76444f50 | 4324 | netdev_dbg(bond_dev, "bond=%p, new_mtu=%d\n", bond, new_mtu); |
1da177e4 | 4325 | |
9caff1e7 | 4326 | bond_for_each_slave(bond, slave, iter) { |
e2a7420d | 4327 | slave_dbg(bond_dev, slave->dev, "s %p c_m %p\n", |
76444f50 | 4328 | slave, slave->dev->netdev_ops->ndo_change_mtu); |
e944ef79 | 4329 | |
1da177e4 LT |
4330 | res = dev_set_mtu(slave->dev, new_mtu); |
4331 | ||
4332 | if (res) { | |
4333 | /* If we failed to set the slave's mtu to the new value | |
4334 | * we must abort the operation even in ACTIVE_BACKUP | |
4335 | * mode, because if we allow the backup slaves to have | |
4336 | * different mtu values than the active slave we'll | |
4337 | * need to change their mtu when doing a failover. That | |
4338 | * means changing their mtu from timer context, which | |
4339 | * is probably not a good idea. | |
4340 | */ | |
e2a7420d JW |
4341 | slave_dbg(bond_dev, slave->dev, "err %d setting mtu to %d\n", |
4342 | res, new_mtu); | |
1da177e4 LT |
4343 | goto unwind; |
4344 | } | |
4345 | } | |
4346 | ||
4347 | bond_dev->mtu = new_mtu; | |
4348 | ||
4349 | return 0; | |
4350 | ||
4351 | unwind: | |
4352 | /* unwind from head to the slave that failed */ | |
9caff1e7 | 4353 | bond_for_each_slave(bond, rollback_slave, iter) { |
1da177e4 LT |
4354 | int tmp_res; |
4355 | ||
81f23b13 VF |
4356 | if (rollback_slave == slave) |
4357 | break; | |
4358 | ||
4359 | tmp_res = dev_set_mtu(rollback_slave->dev, bond_dev->mtu); | |
e2a7420d JW |
4360 | if (tmp_res) |
4361 | slave_dbg(bond_dev, rollback_slave->dev, "unwind err %d\n", | |
4362 | tmp_res); | |
1da177e4 LT |
4363 | } |
4364 | ||
4365 | return res; | |
4366 | } | |
4367 | ||
547942ca | 4368 | /* Change HW address |
1da177e4 LT |
4369 | * |
4370 | * Note that many devices must be down to change the HW address, and | |
4371 | * downing the master releases all slaves. We can make bonds full of | |
4372 | * bonding devices to test this, however. | |
4373 | */ | |
4374 | static int bond_set_mac_address(struct net_device *bond_dev, void *addr) | |
4375 | { | |
454d7c9b | 4376 | struct bonding *bond = netdev_priv(bond_dev); |
81f23b13 | 4377 | struct slave *slave, *rollback_slave; |
faeeb317 | 4378 | struct sockaddr_storage *ss = addr, tmp_ss; |
9caff1e7 | 4379 | struct list_head *iter; |
1da177e4 | 4380 | int res = 0; |
1da177e4 | 4381 | |
01844098 | 4382 | if (BOND_MODE(bond) == BOND_MODE_ALB) |
eb7cc59a SH |
4383 | return bond_alb_set_mac_address(bond_dev, addr); |
4384 | ||
4385 | ||
e2a7420d | 4386 | netdev_dbg(bond_dev, "%s: bond=%p\n", __func__, bond); |
1da177e4 | 4387 | |
1b5acd29 JV |
4388 | /* If fail_over_mac is enabled, do nothing and return success. |
4389 | * Returning an error causes ifenslave to fail. | |
dd957c57 | 4390 | */ |
cc689aaa | 4391 | if (bond->params.fail_over_mac && |
01844098 | 4392 | BOND_MODE(bond) == BOND_MODE_ACTIVEBACKUP) |
dd957c57 | 4393 | return 0; |
2ab82852 | 4394 | |
faeeb317 | 4395 | if (!is_valid_ether_addr(ss->__data)) |
1da177e4 | 4396 | return -EADDRNOTAVAIL; |
1da177e4 | 4397 | |
9caff1e7 | 4398 | bond_for_each_slave(bond, slave, iter) { |
e2a7420d JW |
4399 | slave_dbg(bond_dev, slave->dev, "%s: slave=%p\n", |
4400 | __func__, slave); | |
3a37a963 | 4401 | res = dev_set_mac_address(slave->dev, addr, NULL); |
1da177e4 LT |
4402 | if (res) { |
4403 | /* TODO: consider downing the slave | |
4404 | * and retry ? | |
4405 | * User should expect communications | |
4406 | * breakage anyway until ARP finish | |
4407 | * updating, so... | |
4408 | */ | |
e2a7420d JW |
4409 | slave_dbg(bond_dev, slave->dev, "%s: err %d\n", |
4410 | __func__, res); | |
1da177e4 LT |
4411 | goto unwind; |
4412 | } | |
4413 | } | |
4414 | ||
4415 | /* success */ | |
faeeb317 | 4416 | memcpy(bond_dev->dev_addr, ss->__data, bond_dev->addr_len); |
1da177e4 LT |
4417 | return 0; |
4418 | ||
4419 | unwind: | |
faeeb317 JW |
4420 | memcpy(tmp_ss.__data, bond_dev->dev_addr, bond_dev->addr_len); |
4421 | tmp_ss.ss_family = bond_dev->type; | |
1da177e4 LT |
4422 | |
4423 | /* unwind from head to the slave that failed */ | |
9caff1e7 | 4424 | bond_for_each_slave(bond, rollback_slave, iter) { |
1da177e4 LT |
4425 | int tmp_res; |
4426 | ||
81f23b13 VF |
4427 | if (rollback_slave == slave) |
4428 | break; | |
4429 | ||
faeeb317 | 4430 | tmp_res = dev_set_mac_address(rollback_slave->dev, |
3a37a963 | 4431 | (struct sockaddr *)&tmp_ss, NULL); |
1da177e4 | 4432 | if (tmp_res) { |
e2a7420d JW |
4433 | slave_dbg(bond_dev, rollback_slave->dev, "%s: unwind err %d\n", |
4434 | __func__, tmp_res); | |
1da177e4 LT |
4435 | } |
4436 | } | |
4437 | ||
4438 | return res; | |
4439 | } | |
4440 | ||
15077228 | 4441 | /** |
29d5bbcc | 4442 | * bond_get_slave_by_id - get xmit slave with slave_id |
15077228 | 4443 | * @bond: bonding device that is transmitting |
15077228 NA |
4444 | * @slave_id: slave id up to slave_cnt-1 through which to transmit |
4445 | * | |
29d5bbcc | 4446 | * This function tries to get slave with slave_id but in case |
15077228 | 4447 | * it fails, it tries to find the first available slave for transmission. |
15077228 | 4448 | */ |
29d5bbcc MG |
4449 | static struct slave *bond_get_slave_by_id(struct bonding *bond, |
4450 | int slave_id) | |
15077228 | 4451 | { |
9caff1e7 | 4452 | struct list_head *iter; |
15077228 NA |
4453 | struct slave *slave; |
4454 | int i = slave_id; | |
4455 | ||
4456 | /* Here we start from the slave with slave_id */ | |
9caff1e7 | 4457 | bond_for_each_slave_rcu(bond, slave, iter) { |
15077228 | 4458 | if (--i < 0) { |
ae46f184 | 4459 | if (bond_slave_can_tx(slave)) |
29d5bbcc | 4460 | return slave; |
15077228 NA |
4461 | } |
4462 | } | |
4463 | ||
4464 | /* Here we start from the first slave up to slave_id */ | |
4465 | i = slave_id; | |
9caff1e7 | 4466 | bond_for_each_slave_rcu(bond, slave, iter) { |
15077228 NA |
4467 | if (--i < 0) |
4468 | break; | |
ae46f184 | 4469 | if (bond_slave_can_tx(slave)) |
29d5bbcc | 4470 | return slave; |
15077228 NA |
4471 | } |
4472 | /* no slave that can tx has been found */ | |
29d5bbcc | 4473 | return NULL; |
15077228 NA |
4474 | } |
4475 | ||
73958329 NA |
4476 | /** |
4477 | * bond_rr_gen_slave_id - generate slave id based on packets_per_slave | |
4478 | * @bond: bonding device to use | |
4479 | * | |
4480 | * Based on the value of the bonding device's packets_per_slave parameter | |
4481 | * this function generates a slave id, which is usually used as the next | |
4482 | * slave to transmit through. | |
4483 | */ | |
4484 | static u32 bond_rr_gen_slave_id(struct bonding *bond) | |
4485 | { | |
73958329 | 4486 | u32 slave_id; |
809fa972 HFS |
4487 | struct reciprocal_value reciprocal_packets_per_slave; |
4488 | int packets_per_slave = bond->params.packets_per_slave; | |
73958329 NA |
4489 | |
4490 | switch (packets_per_slave) { | |
4491 | case 0: | |
4492 | slave_id = prandom_u32(); | |
4493 | break; | |
4494 | case 1: | |
848ca918 | 4495 | slave_id = this_cpu_inc_return(*bond->rr_tx_counter); |
73958329 NA |
4496 | break; |
4497 | default: | |
809fa972 HFS |
4498 | reciprocal_packets_per_slave = |
4499 | bond->params.reciprocal_packets_per_slave; | |
848ca918 JM |
4500 | slave_id = this_cpu_inc_return(*bond->rr_tx_counter); |
4501 | slave_id = reciprocal_divide(slave_id, | |
809fa972 | 4502 | reciprocal_packets_per_slave); |
73958329 NA |
4503 | break; |
4504 | } | |
73958329 NA |
4505 | |
4506 | return slave_id; | |
4507 | } | |
4508 | ||
29d5bbcc MG |
4509 | static struct slave *bond_xmit_roundrobin_slave_get(struct bonding *bond, |
4510 | struct sk_buff *skb) | |
1da177e4 | 4511 | { |
15077228 | 4512 | struct slave *slave; |
9d1bc24b | 4513 | int slave_cnt; |
73958329 | 4514 | u32 slave_id; |
1da177e4 | 4515 | |
73958329 | 4516 | /* Start with the curr_active_slave that joined the bond as the |
a2fd940f AG |
4517 | * default for sending IGMP traffic. For failover purposes one |
4518 | * needs to maintain some consistency for the interface that will | |
4519 | * send the join/membership reports. The curr_active_slave found | |
4520 | * will send all of this type of traffic. | |
cf5f9044 | 4521 | */ |
9d1bc24b CW |
4522 | if (skb->protocol == htons(ETH_P_IP)) { |
4523 | int noff = skb_network_offset(skb); | |
4524 | struct iphdr *iph; | |
9a72c2da | 4525 | |
9d1bc24b CW |
4526 | if (unlikely(!pskb_may_pull(skb, noff + sizeof(*iph)))) |
4527 | goto non_igmp; | |
4528 | ||
4529 | iph = ip_hdr(skb); | |
4530 | if (iph->protocol == IPPROTO_IGMP) { | |
4531 | slave = rcu_dereference(bond->curr_active_slave); | |
4532 | if (slave) | |
29d5bbcc MG |
4533 | return slave; |
4534 | return bond_get_slave_by_id(bond, 0); | |
9a72c2da | 4535 | } |
1da177e4 | 4536 | } |
0693e88e | 4537 | |
9d1bc24b CW |
4538 | non_igmp: |
4539 | slave_cnt = READ_ONCE(bond->slave_cnt); | |
4540 | if (likely(slave_cnt)) { | |
29d5bbcc MG |
4541 | slave_id = bond_rr_gen_slave_id(bond) % slave_cnt; |
4542 | return bond_get_slave_by_id(bond, slave_id); | |
9d1bc24b | 4543 | } |
29d5bbcc MG |
4544 | return NULL; |
4545 | } | |
4546 | ||
9e2ee5c7 JM |
4547 | static struct slave *bond_xdp_xmit_roundrobin_slave_get(struct bonding *bond, |
4548 | struct xdp_buff *xdp) | |
4549 | { | |
4550 | struct slave *slave; | |
4551 | int slave_cnt; | |
4552 | u32 slave_id; | |
4553 | const struct ethhdr *eth; | |
4554 | void *data = xdp->data; | |
4555 | ||
4556 | if (data + sizeof(struct ethhdr) > xdp->data_end) | |
4557 | goto non_igmp; | |
4558 | ||
4559 | eth = (struct ethhdr *)data; | |
4560 | data += sizeof(struct ethhdr); | |
4561 | ||
4562 | /* See comment on IGMP in bond_xmit_roundrobin_slave_get() */ | |
4563 | if (eth->h_proto == htons(ETH_P_IP)) { | |
4564 | const struct iphdr *iph; | |
4565 | ||
4566 | if (data + sizeof(struct iphdr) > xdp->data_end) | |
4567 | goto non_igmp; | |
4568 | ||
4569 | iph = (struct iphdr *)data; | |
4570 | ||
4571 | if (iph->protocol == IPPROTO_IGMP) { | |
4572 | slave = rcu_dereference(bond->curr_active_slave); | |
4573 | if (slave) | |
4574 | return slave; | |
4575 | return bond_get_slave_by_id(bond, 0); | |
4576 | } | |
4577 | } | |
4578 | ||
4579 | non_igmp: | |
4580 | slave_cnt = READ_ONCE(bond->slave_cnt); | |
4581 | if (likely(slave_cnt)) { | |
4582 | slave_id = bond_rr_gen_slave_id(bond) % slave_cnt; | |
4583 | return bond_get_slave_by_id(bond, slave_id); | |
4584 | } | |
4585 | return NULL; | |
4586 | } | |
4587 | ||
29d5bbcc MG |
4588 | static netdev_tx_t bond_xmit_roundrobin(struct sk_buff *skb, |
4589 | struct net_device *bond_dev) | |
4590 | { | |
4591 | struct bonding *bond = netdev_priv(bond_dev); | |
4592 | struct slave *slave; | |
4593 | ||
4594 | slave = bond_xmit_roundrobin_slave_get(bond, skb); | |
76cd622f SM |
4595 | if (likely(slave)) |
4596 | return bond_dev_queue_xmit(bond, skb, slave->dev); | |
4597 | ||
ae46f184 | 4598 | return bond_tx_drop(bond_dev, skb); |
1da177e4 LT |
4599 | } |
4600 | ||
a815bde5 | 4601 | static struct slave *bond_xmit_activebackup_slave_get(struct bonding *bond) |
5a19f1c1 MG |
4602 | { |
4603 | return rcu_dereference(bond->curr_active_slave); | |
4604 | } | |
4605 | ||
547942ca | 4606 | /* In active-backup mode, we know that bond->curr_active_slave is always valid if |
1da177e4 LT |
4607 | * the bond has a usable interface. |
4608 | */ | |
dbdc8a21 TZ |
4609 | static netdev_tx_t bond_xmit_activebackup(struct sk_buff *skb, |
4610 | struct net_device *bond_dev) | |
1da177e4 | 4611 | { |
454d7c9b | 4612 | struct bonding *bond = netdev_priv(bond_dev); |
71bc3b2d | 4613 | struct slave *slave; |
1da177e4 | 4614 | |
a815bde5 | 4615 | slave = bond_xmit_activebackup_slave_get(bond); |
71bc3b2d | 4616 | if (slave) |
ae46f184 | 4617 | return bond_dev_queue_xmit(bond, skb, slave->dev); |
0693e88e | 4618 | |
ae46f184 | 4619 | return bond_tx_drop(bond_dev, skb); |
1da177e4 LT |
4620 | } |
4621 | ||
ee637714 MB |
4622 | /* Use this to update slave_array when (a) it's not appropriate to update |
4623 | * slave_array right away (note that update_slave_array() may sleep) | |
4624 | * and / or (b) RTNL is not held. | |
1da177e4 | 4625 | */ |
ee637714 | 4626 | void bond_slave_arr_work_rearm(struct bonding *bond, unsigned long delay) |
1da177e4 | 4627 | { |
ee637714 MB |
4628 | queue_delayed_work(bond->wq, &bond->slave_arr_work, delay); |
4629 | } | |
1da177e4 | 4630 | |
ee637714 MB |
4631 | /* Slave array work handler. Holds only RTNL */ |
4632 | static void bond_slave_arr_handler(struct work_struct *work) | |
4633 | { | |
4634 | struct bonding *bond = container_of(work, struct bonding, | |
4635 | slave_arr_work.work); | |
4636 | int ret; | |
4637 | ||
4638 | if (!rtnl_trylock()) | |
4639 | goto err; | |
4640 | ||
4641 | ret = bond_update_slave_arr(bond, NULL); | |
4642 | rtnl_unlock(); | |
4643 | if (ret) { | |
4644 | pr_warn_ratelimited("Failed to update slave array from WT\n"); | |
4645 | goto err; | |
4646 | } | |
4647 | return; | |
4648 | ||
4649 | err: | |
4650 | bond_slave_arr_work_rearm(bond, 1); | |
4651 | } | |
4652 | ||
119d48fd MG |
4653 | static void bond_skip_slave(struct bond_up_slave *slaves, |
4654 | struct slave *skipslave) | |
4655 | { | |
4656 | int idx; | |
4657 | ||
4658 | /* Rare situation where caller has asked to skip a specific | |
4659 | * slave but allocation failed (most likely!). BTW this is | |
4660 | * only possible when the call is initiated from | |
4661 | * __bond_release_one(). In this situation; overwrite the | |
4662 | * skipslave entry in the array with the last entry from the | |
4663 | * array to avoid a situation where the xmit path may choose | |
4664 | * this to-be-skipped slave to send a packet out. | |
4665 | */ | |
4666 | for (idx = 0; slaves && idx < slaves->count; idx++) { | |
4667 | if (skipslave == slaves->arr[idx]) { | |
4668 | slaves->arr[idx] = | |
4669 | slaves->arr[slaves->count - 1]; | |
4670 | slaves->count--; | |
4671 | break; | |
4672 | } | |
4673 | } | |
4674 | } | |
4675 | ||
6b447e76 MG |
4676 | static void bond_set_slave_arr(struct bonding *bond, |
4677 | struct bond_up_slave *usable_slaves, | |
4678 | struct bond_up_slave *all_slaves) | |
4679 | { | |
4680 | struct bond_up_slave *usable, *all; | |
4681 | ||
4682 | usable = rtnl_dereference(bond->usable_slaves); | |
4683 | rcu_assign_pointer(bond->usable_slaves, usable_slaves); | |
4684 | kfree_rcu(usable, rcu); | |
4685 | ||
4686 | all = rtnl_dereference(bond->all_slaves); | |
4687 | rcu_assign_pointer(bond->all_slaves, all_slaves); | |
4688 | kfree_rcu(all, rcu); | |
4689 | } | |
4690 | ||
4691 | static void bond_reset_slave_arr(struct bonding *bond) | |
4692 | { | |
4693 | struct bond_up_slave *usable, *all; | |
4694 | ||
4695 | usable = rtnl_dereference(bond->usable_slaves); | |
4696 | if (usable) { | |
4697 | RCU_INIT_POINTER(bond->usable_slaves, NULL); | |
4698 | kfree_rcu(usable, rcu); | |
4699 | } | |
4700 | ||
4701 | all = rtnl_dereference(bond->all_slaves); | |
4702 | if (all) { | |
4703 | RCU_INIT_POINTER(bond->all_slaves, NULL); | |
4704 | kfree_rcu(all, rcu); | |
4705 | } | |
4706 | } | |
4707 | ||
ee637714 MB |
4708 | /* Build the usable slaves array in control path for modes that use xmit-hash |
4709 | * to determine the slave interface - | |
4710 | * (a) BOND_MODE_8023AD | |
4711 | * (b) BOND_MODE_XOR | |
e79c1055 | 4712 | * (c) (BOND_MODE_TLB || BOND_MODE_ALB) && tlb_dynamic_lb == 0 |
ee637714 MB |
4713 | * |
4714 | * The caller is expected to hold RTNL only and NO other lock! | |
4715 | */ | |
4716 | int bond_update_slave_arr(struct bonding *bond, struct slave *skipslave) | |
4717 | { | |
6b447e76 | 4718 | struct bond_up_slave *usable_slaves = NULL, *all_slaves = NULL; |
ee637714 MB |
4719 | struct slave *slave; |
4720 | struct list_head *iter; | |
ee637714 MB |
4721 | int agg_id = 0; |
4722 | int ret = 0; | |
4723 | ||
83d686a6 | 4724 | might_sleep(); |
ee637714 | 4725 | |
ed7d4f02 MG |
4726 | usable_slaves = kzalloc(struct_size(usable_slaves, arr, |
4727 | bond->slave_cnt), GFP_KERNEL); | |
6b447e76 MG |
4728 | all_slaves = kzalloc(struct_size(all_slaves, arr, |
4729 | bond->slave_cnt), GFP_KERNEL); | |
4730 | if (!usable_slaves || !all_slaves) { | |
ee637714 | 4731 | ret = -ENOMEM; |
ee637714 MB |
4732 | goto out; |
4733 | } | |
4734 | if (BOND_MODE(bond) == BOND_MODE_8023AD) { | |
4735 | struct ad_info ad_info; | |
4736 | ||
83d686a6 | 4737 | spin_lock_bh(&bond->mode_lock); |
ee637714 | 4738 | if (bond_3ad_get_active_agg_info(bond, &ad_info)) { |
83d686a6 | 4739 | spin_unlock_bh(&bond->mode_lock); |
ee637714 | 4740 | pr_debug("bond_3ad_get_active_agg_info failed\n"); |
ee637714 MB |
4741 | /* No active aggragator means it's not safe to use |
4742 | * the previous array. | |
4743 | */ | |
6b447e76 | 4744 | bond_reset_slave_arr(bond); |
ee637714 MB |
4745 | goto out; |
4746 | } | |
83d686a6 | 4747 | spin_unlock_bh(&bond->mode_lock); |
ee637714 MB |
4748 | agg_id = ad_info.aggregator_id; |
4749 | } | |
4750 | bond_for_each_slave(bond, slave, iter) { | |
6b447e76 MG |
4751 | if (skipslave == slave) |
4752 | continue; | |
4753 | ||
4754 | all_slaves->arr[all_slaves->count++] = slave; | |
ee637714 MB |
4755 | if (BOND_MODE(bond) == BOND_MODE_8023AD) { |
4756 | struct aggregator *agg; | |
4757 | ||
4758 | agg = SLAVE_AD_INFO(slave)->port.aggregator; | |
4759 | if (!agg || agg->aggregator_identifier != agg_id) | |
4760 | continue; | |
4761 | } | |
4762 | if (!bond_slave_can_tx(slave)) | |
4763 | continue; | |
e79c1055 | 4764 | |
e2a7420d | 4765 | slave_dbg(bond->dev, slave->dev, "Adding slave to tx hash array[%d]\n", |
ed7d4f02 | 4766 | usable_slaves->count); |
e79c1055 | 4767 | |
ed7d4f02 | 4768 | usable_slaves->arr[usable_slaves->count++] = slave; |
ee637714 MB |
4769 | } |
4770 | ||
6b447e76 MG |
4771 | bond_set_slave_arr(bond, usable_slaves, all_slaves); |
4772 | return ret; | |
ee637714 MB |
4773 | out: |
4774 | if (ret != 0 && skipslave) { | |
6b447e76 MG |
4775 | bond_skip_slave(rtnl_dereference(bond->all_slaves), |
4776 | skipslave); | |
ed7d4f02 MG |
4777 | bond_skip_slave(rtnl_dereference(bond->usable_slaves), |
4778 | skipslave); | |
ee637714 | 4779 | } |
6b447e76 MG |
4780 | kfree_rcu(all_slaves, rcu); |
4781 | kfree_rcu(usable_slaves, rcu); | |
119d48fd | 4782 | |
ee637714 MB |
4783 | return ret; |
4784 | } | |
4785 | ||
c071d91d MG |
4786 | static struct slave *bond_xmit_3ad_xor_slave_get(struct bonding *bond, |
4787 | struct sk_buff *skb, | |
4788 | struct bond_up_slave *slaves) | |
4789 | { | |
4790 | struct slave *slave; | |
4791 | unsigned int count; | |
4792 | u32 hash; | |
4793 | ||
4794 | hash = bond_xmit_hash(bond, skb); | |
4795 | count = slaves ? READ_ONCE(slaves->count) : 0; | |
4796 | if (unlikely(!count)) | |
4797 | return NULL; | |
4798 | ||
4799 | slave = slaves->arr[hash % count]; | |
4800 | return slave; | |
4801 | } | |
4802 | ||
9e2ee5c7 JM |
4803 | static struct slave *bond_xdp_xmit_3ad_xor_slave_get(struct bonding *bond, |
4804 | struct xdp_buff *xdp) | |
4805 | { | |
4806 | struct bond_up_slave *slaves; | |
4807 | unsigned int count; | |
4808 | u32 hash; | |
4809 | ||
4810 | hash = bond_xmit_hash_xdp(bond, xdp); | |
4811 | slaves = rcu_dereference(bond->usable_slaves); | |
4812 | count = slaves ? READ_ONCE(slaves->count) : 0; | |
4813 | if (unlikely(!count)) | |
4814 | return NULL; | |
4815 | ||
4816 | return slaves->arr[hash % count]; | |
4817 | } | |
4818 | ||
ee637714 MB |
4819 | /* Use this Xmit function for 3AD as well as XOR modes. The current |
4820 | * usable slave array is formed in the control path. The xmit function | |
4821 | * just calculates hash and sends the packet out. | |
4822 | */ | |
dbdc8a21 TZ |
4823 | static netdev_tx_t bond_3ad_xor_xmit(struct sk_buff *skb, |
4824 | struct net_device *dev) | |
ee637714 MB |
4825 | { |
4826 | struct bonding *bond = netdev_priv(dev); | |
ee637714 | 4827 | struct bond_up_slave *slaves; |
c071d91d | 4828 | struct slave *slave; |
ee637714 | 4829 | |
ed7d4f02 | 4830 | slaves = rcu_dereference(bond->usable_slaves); |
c071d91d MG |
4831 | slave = bond_xmit_3ad_xor_slave_get(bond, skb, slaves); |
4832 | if (likely(slave)) | |
ae46f184 | 4833 | return bond_dev_queue_xmit(bond, skb, slave->dev); |
0693e88e | 4834 | |
ae46f184 | 4835 | return bond_tx_drop(dev, skb); |
1da177e4 LT |
4836 | } |
4837 | ||
78a646ce | 4838 | /* in broadcast mode, we send everything to all usable interfaces. */ |
dbdc8a21 TZ |
4839 | static netdev_tx_t bond_xmit_broadcast(struct sk_buff *skb, |
4840 | struct net_device *bond_dev) | |
1da177e4 | 4841 | { |
454d7c9b | 4842 | struct bonding *bond = netdev_priv(bond_dev); |
78a646ce | 4843 | struct slave *slave = NULL; |
9caff1e7 | 4844 | struct list_head *iter; |
763672b7 JW |
4845 | bool xmit_suc = false; |
4846 | bool skb_used = false; | |
1da177e4 | 4847 | |
9caff1e7 | 4848 | bond_for_each_slave_rcu(bond, slave, iter) { |
763672b7 JW |
4849 | struct sk_buff *skb2; |
4850 | ||
4851 | if (!(bond_slave_is_up(slave) && slave->link == BOND_LINK_UP)) | |
4852 | continue; | |
1da177e4 | 4853 | |
763672b7 JW |
4854 | if (bond_is_last_slave(bond, slave)) { |
4855 | skb2 = skb; | |
4856 | skb_used = true; | |
4857 | } else { | |
4858 | skb2 = skb_clone(skb, GFP_ATOMIC); | |
78a646ce | 4859 | if (!skb2) { |
9152e26d | 4860 | net_err_ratelimited("%s: Error: %s: skb_clone() failed\n", |
4861 | bond_dev->name, __func__); | |
78a646ce | 4862 | continue; |
1da177e4 | 4863 | } |
1da177e4 | 4864 | } |
763672b7 JW |
4865 | |
4866 | if (bond_dev_queue_xmit(bond, skb2, slave->dev) == NETDEV_TX_OK) | |
4867 | xmit_suc = true; | |
1da177e4 | 4868 | } |
3d632c3f | 4869 | |
763672b7 JW |
4870 | if (!skb_used) |
4871 | dev_kfree_skb_any(skb); | |
4872 | ||
4873 | if (xmit_suc) | |
4874 | return NETDEV_TX_OK; | |
4875 | ||
4876 | atomic_long_inc(&bond_dev->tx_dropped); | |
4877 | return NET_XMIT_DROP; | |
1da177e4 LT |
4878 | } |
4879 | ||
4880 | /*------------------------- Device initialization ---------------------------*/ | |
4881 | ||
547942ca | 4882 | /* Lookup the slave that corresponds to a qid */ |
bb1d9123 AG |
4883 | static inline int bond_slave_override(struct bonding *bond, |
4884 | struct sk_buff *skb) | |
4885 | { | |
bb1d9123 | 4886 | struct slave *slave = NULL; |
9caff1e7 | 4887 | struct list_head *iter; |
bb1d9123 | 4888 | |
ae35c6f7 | 4889 | if (!skb_rx_queue_recorded(skb)) |
0693e88e | 4890 | return 1; |
bb1d9123 AG |
4891 | |
4892 | /* Find out if any slaves have the same mapping as this skb. */ | |
3900f290 | 4893 | bond_for_each_slave_rcu(bond, slave, iter) { |
ae35c6f7 | 4894 | if (slave->queue_id == skb_get_queue_mapping(skb)) { |
f5e2dc5d AN |
4895 | if (bond_slave_is_up(slave) && |
4896 | slave->link == BOND_LINK_UP) { | |
3900f290 | 4897 | bond_dev_queue_xmit(bond, skb, slave->dev); |
4898 | return 0; | |
4899 | } | |
4900 | /* If the slave isn't UP, use default transmit policy. */ | |
bb1d9123 AG |
4901 | break; |
4902 | } | |
4903 | } | |
4904 | ||
3900f290 | 4905 | return 1; |
bb1d9123 AG |
4906 | } |
4907 | ||
374eeb5a | 4908 | |
f663dd9a | 4909 | static u16 bond_select_queue(struct net_device *dev, struct sk_buff *skb, |
a350ecce | 4910 | struct net_device *sb_dev) |
bb1d9123 | 4911 | { |
547942ca | 4912 | /* This helper function exists to help dev_pick_tx get the correct |
fd0e435b | 4913 | * destination queue. Using a helper function skips a call to |
bb1d9123 AG |
4914 | * skb_tx_hash and will put the skbs in the queue we expect on their |
4915 | * way down to the bonding driver. | |
4916 | */ | |
fd0e435b PO |
4917 | u16 txq = skb_rx_queue_recorded(skb) ? skb_get_rx_queue(skb) : 0; |
4918 | ||
547942ca | 4919 | /* Save the original txq to restore before passing to the driver */ |
ae35c6f7 | 4920 | qdisc_skb_cb(skb)->slave_dev_queue_mapping = skb_get_queue_mapping(skb); |
374eeb5a | 4921 | |
fd0e435b | 4922 | if (unlikely(txq >= dev->real_num_tx_queues)) { |
d30ee670 | 4923 | do { |
fd0e435b | 4924 | txq -= dev->real_num_tx_queues; |
d30ee670 | 4925 | } while (txq >= dev->real_num_tx_queues); |
fd0e435b PO |
4926 | } |
4927 | return txq; | |
bb1d9123 AG |
4928 | } |
4929 | ||
33720aaf MG |
4930 | static struct net_device *bond_xmit_get_slave(struct net_device *master_dev, |
4931 | struct sk_buff *skb, | |
4932 | bool all_slaves) | |
4933 | { | |
4934 | struct bonding *bond = netdev_priv(master_dev); | |
4935 | struct bond_up_slave *slaves; | |
4936 | struct slave *slave = NULL; | |
4937 | ||
4938 | switch (BOND_MODE(bond)) { | |
4939 | case BOND_MODE_ROUNDROBIN: | |
4940 | slave = bond_xmit_roundrobin_slave_get(bond, skb); | |
4941 | break; | |
4942 | case BOND_MODE_ACTIVEBACKUP: | |
a815bde5 | 4943 | slave = bond_xmit_activebackup_slave_get(bond); |
33720aaf MG |
4944 | break; |
4945 | case BOND_MODE_8023AD: | |
4946 | case BOND_MODE_XOR: | |
4947 | if (all_slaves) | |
4948 | slaves = rcu_dereference(bond->all_slaves); | |
4949 | else | |
4950 | slaves = rcu_dereference(bond->usable_slaves); | |
4951 | slave = bond_xmit_3ad_xor_slave_get(bond, skb, slaves); | |
4952 | break; | |
4953 | case BOND_MODE_BROADCAST: | |
4954 | break; | |
4955 | case BOND_MODE_ALB: | |
4956 | slave = bond_xmit_alb_slave_get(bond, skb); | |
4957 | break; | |
4958 | case BOND_MODE_TLB: | |
4959 | slave = bond_xmit_tlb_slave_get(bond, skb); | |
4960 | break; | |
4961 | default: | |
4962 | /* Should never happen, mode already checked */ | |
4963 | WARN_ONCE(true, "Unknown bonding mode"); | |
4964 | break; | |
4965 | } | |
4966 | ||
4967 | if (slave) | |
4968 | return slave->dev; | |
4969 | return NULL; | |
4970 | } | |
4971 | ||
007feb87 TT |
4972 | static void bond_sk_to_flow(struct sock *sk, struct flow_keys *flow) |
4973 | { | |
4974 | switch (sk->sk_family) { | |
4975 | #if IS_ENABLED(CONFIG_IPV6) | |
4976 | case AF_INET6: | |
4977 | if (sk->sk_ipv6only || | |
4978 | ipv6_addr_type(&sk->sk_v6_daddr) != IPV6_ADDR_MAPPED) { | |
4979 | flow->control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS; | |
4980 | flow->addrs.v6addrs.src = inet6_sk(sk)->saddr; | |
4981 | flow->addrs.v6addrs.dst = sk->sk_v6_daddr; | |
4982 | break; | |
4983 | } | |
4984 | fallthrough; | |
4985 | #endif | |
4986 | default: /* AF_INET */ | |
4987 | flow->control.addr_type = FLOW_DISSECTOR_KEY_IPV4_ADDRS; | |
4988 | flow->addrs.v4addrs.src = inet_sk(sk)->inet_rcv_saddr; | |
4989 | flow->addrs.v4addrs.dst = inet_sk(sk)->inet_daddr; | |
4990 | break; | |
4991 | } | |
4992 | ||
4993 | flow->ports.src = inet_sk(sk)->inet_sport; | |
4994 | flow->ports.dst = inet_sk(sk)->inet_dport; | |
4995 | } | |
4996 | ||
4997 | /** | |
4998 | * bond_sk_hash_l34 - generate a hash value based on the socket's L3 and L4 fields | |
4999 | * @sk: socket to use for headers | |
5000 | * | |
5001 | * This function will extract the necessary field from the socket and use | |
5002 | * them to generate a hash based on the LAYER34 xmit_policy. | |
5003 | * Assumes that sk is a TCP or UDP socket. | |
5004 | */ | |
5005 | static u32 bond_sk_hash_l34(struct sock *sk) | |
5006 | { | |
5007 | struct flow_keys flow; | |
5008 | u32 hash; | |
5009 | ||
5010 | bond_sk_to_flow(sk, &flow); | |
5011 | ||
5012 | /* L4 */ | |
5013 | memcpy(&hash, &flow.ports.ports, sizeof(hash)); | |
5014 | /* L3 */ | |
5015 | return bond_ip_hash(hash, &flow); | |
5016 | } | |
5017 | ||
5018 | static struct net_device *__bond_sk_get_lower_dev(struct bonding *bond, | |
5019 | struct sock *sk) | |
5020 | { | |
5021 | struct bond_up_slave *slaves; | |
5022 | struct slave *slave; | |
5023 | unsigned int count; | |
5024 | u32 hash; | |
5025 | ||
5026 | slaves = rcu_dereference(bond->usable_slaves); | |
5027 | count = slaves ? READ_ONCE(slaves->count) : 0; | |
5028 | if (unlikely(!count)) | |
5029 | return NULL; | |
5030 | ||
5031 | hash = bond_sk_hash_l34(sk); | |
5032 | slave = slaves->arr[hash % count]; | |
5033 | ||
5034 | return slave->dev; | |
5035 | } | |
5036 | ||
5037 | static struct net_device *bond_sk_get_lower_dev(struct net_device *dev, | |
5038 | struct sock *sk) | |
5039 | { | |
5040 | struct bonding *bond = netdev_priv(dev); | |
5041 | struct net_device *lower = NULL; | |
5042 | ||
5043 | rcu_read_lock(); | |
5044 | if (bond_sk_check(bond)) | |
5045 | lower = __bond_sk_get_lower_dev(bond, sk); | |
5046 | rcu_read_unlock(); | |
5047 | ||
5048 | return lower; | |
5049 | } | |
5050 | ||
89df6a81 TT |
5051 | #if IS_ENABLED(CONFIG_TLS_DEVICE) |
5052 | static netdev_tx_t bond_tls_device_xmit(struct bonding *bond, struct sk_buff *skb, | |
5053 | struct net_device *dev) | |
5054 | { | |
5055 | if (likely(bond_get_slave_by_dev(bond, tls_get_ctx(skb->sk)->netdev))) | |
5056 | return bond_dev_queue_xmit(bond, skb, tls_get_ctx(skb->sk)->netdev); | |
5057 | return bond_tx_drop(dev, skb); | |
5058 | } | |
5059 | #endif | |
5060 | ||
0693e88e | 5061 | static netdev_tx_t __bond_start_xmit(struct sk_buff *skb, struct net_device *dev) |
00829823 | 5062 | { |
bb1d9123 AG |
5063 | struct bonding *bond = netdev_priv(dev); |
5064 | ||
d1e2e5cd VF |
5065 | if (bond_should_override_tx_queue(bond) && |
5066 | !bond_slave_override(bond, skb)) | |
5067 | return NETDEV_TX_OK; | |
00829823 | 5068 | |
89df6a81 TT |
5069 | #if IS_ENABLED(CONFIG_TLS_DEVICE) |
5070 | if (skb->sk && tls_is_sk_tx_device_offloaded(skb->sk)) | |
5071 | return bond_tls_device_xmit(bond, skb, dev); | |
5072 | #endif | |
5073 | ||
01844098 | 5074 | switch (BOND_MODE(bond)) { |
00829823 SH |
5075 | case BOND_MODE_ROUNDROBIN: |
5076 | return bond_xmit_roundrobin(skb, dev); | |
5077 | case BOND_MODE_ACTIVEBACKUP: | |
5078 | return bond_xmit_activebackup(skb, dev); | |
ee637714 | 5079 | case BOND_MODE_8023AD: |
00829823 | 5080 | case BOND_MODE_XOR: |
ee637714 | 5081 | return bond_3ad_xor_xmit(skb, dev); |
00829823 SH |
5082 | case BOND_MODE_BROADCAST: |
5083 | return bond_xmit_broadcast(skb, dev); | |
00829823 | 5084 | case BOND_MODE_ALB: |
00829823 | 5085 | return bond_alb_xmit(skb, dev); |
f05b42ea MB |
5086 | case BOND_MODE_TLB: |
5087 | return bond_tlb_xmit(skb, dev); | |
00829823 SH |
5088 | default: |
5089 | /* Should never happen, mode already checked */ | |
76444f50 | 5090 | netdev_err(dev, "Unknown bonding mode %d\n", BOND_MODE(bond)); |
00829823 | 5091 | WARN_ON_ONCE(1); |
ae46f184 | 5092 | return bond_tx_drop(dev, skb); |
00829823 SH |
5093 | } |
5094 | } | |
5095 | ||
0693e88e MM |
5096 | static netdev_tx_t bond_start_xmit(struct sk_buff *skb, struct net_device *dev) |
5097 | { | |
5098 | struct bonding *bond = netdev_priv(dev); | |
5099 | netdev_tx_t ret = NETDEV_TX_OK; | |
5100 | ||
547942ca | 5101 | /* If we risk deadlock from transmitting this in the |
0693e88e MM |
5102 | * netpoll path, tell netpoll to queue the frame for later tx |
5103 | */ | |
054bb880 | 5104 | if (unlikely(is_netpoll_tx_blocked(dev))) |
0693e88e MM |
5105 | return NETDEV_TX_BUSY; |
5106 | ||
278b2083 | 5107 | rcu_read_lock(); |
0965a1f3 | 5108 | if (bond_has_slaves(bond)) |
0693e88e MM |
5109 | ret = __bond_start_xmit(skb, dev); |
5110 | else | |
ae46f184 | 5111 | ret = bond_tx_drop(dev, skb); |
278b2083 | 5112 | rcu_read_unlock(); |
0693e88e MM |
5113 | |
5114 | return ret; | |
5115 | } | |
00829823 | 5116 | |
9e2ee5c7 JM |
5117 | static struct net_device * |
5118 | bond_xdp_get_xmit_slave(struct net_device *bond_dev, struct xdp_buff *xdp) | |
5119 | { | |
5120 | struct bonding *bond = netdev_priv(bond_dev); | |
5121 | struct slave *slave; | |
5122 | ||
5123 | /* Caller needs to hold rcu_read_lock() */ | |
5124 | ||
5125 | switch (BOND_MODE(bond)) { | |
5126 | case BOND_MODE_ROUNDROBIN: | |
5127 | slave = bond_xdp_xmit_roundrobin_slave_get(bond, xdp); | |
5128 | break; | |
5129 | ||
5130 | case BOND_MODE_ACTIVEBACKUP: | |
5131 | slave = bond_xmit_activebackup_slave_get(bond); | |
5132 | break; | |
5133 | ||
5134 | case BOND_MODE_8023AD: | |
5135 | case BOND_MODE_XOR: | |
5136 | slave = bond_xdp_xmit_3ad_xor_slave_get(bond, xdp); | |
5137 | break; | |
5138 | ||
5139 | default: | |
5140 | /* Should never happen. Mode guarded by bond_xdp_check() */ | |
5141 | netdev_err(bond_dev, "Unknown bonding mode %d for xdp xmit\n", BOND_MODE(bond)); | |
5142 | WARN_ON_ONCE(1); | |
5143 | return NULL; | |
5144 | } | |
5145 | ||
5146 | if (slave) | |
5147 | return slave->dev; | |
5148 | ||
5149 | return NULL; | |
5150 | } | |
5151 | ||
5152 | static int bond_xdp_xmit(struct net_device *bond_dev, | |
5153 | int n, struct xdp_frame **frames, u32 flags) | |
5154 | { | |
5155 | int nxmit, err = -ENXIO; | |
5156 | ||
5157 | rcu_read_lock(); | |
5158 | ||
5159 | for (nxmit = 0; nxmit < n; nxmit++) { | |
5160 | struct xdp_frame *frame = frames[nxmit]; | |
5161 | struct xdp_frame *frames1[] = {frame}; | |
5162 | struct net_device *slave_dev; | |
5163 | struct xdp_buff xdp; | |
5164 | ||
5165 | xdp_convert_frame_to_buff(frame, &xdp); | |
5166 | ||
5167 | slave_dev = bond_xdp_get_xmit_slave(bond_dev, &xdp); | |
5168 | if (!slave_dev) { | |
5169 | err = -ENXIO; | |
5170 | break; | |
5171 | } | |
5172 | ||
5173 | err = slave_dev->netdev_ops->ndo_xdp_xmit(slave_dev, 1, frames1, flags); | |
5174 | if (err < 1) | |
5175 | break; | |
5176 | } | |
5177 | ||
5178 | rcu_read_unlock(); | |
5179 | ||
5180 | /* If error happened on the first frame then we can pass the error up, otherwise | |
5181 | * report the number of frames that were xmitted. | |
5182 | */ | |
5183 | if (err < 0) | |
5184 | return (nxmit == 0 ? err : nxmit); | |
5185 | ||
5186 | return nxmit; | |
5187 | } | |
5188 | ||
5189 | static int bond_xdp_set(struct net_device *dev, struct bpf_prog *prog, | |
5190 | struct netlink_ext_ack *extack) | |
5191 | { | |
5192 | struct bonding *bond = netdev_priv(dev); | |
5193 | struct list_head *iter; | |
5194 | struct slave *slave, *rollback_slave; | |
5195 | struct bpf_prog *old_prog; | |
5196 | struct netdev_bpf xdp = { | |
5197 | .command = XDP_SETUP_PROG, | |
5198 | .flags = 0, | |
5199 | .prog = prog, | |
5200 | .extack = extack, | |
5201 | }; | |
5202 | int err; | |
5203 | ||
5204 | ASSERT_RTNL(); | |
5205 | ||
5206 | if (!bond_xdp_check(bond)) | |
5207 | return -EOPNOTSUPP; | |
5208 | ||
5209 | old_prog = bond->xdp_prog; | |
5210 | bond->xdp_prog = prog; | |
5211 | ||
5212 | bond_for_each_slave(bond, slave, iter) { | |
5213 | struct net_device *slave_dev = slave->dev; | |
5214 | ||
5215 | if (!slave_dev->netdev_ops->ndo_bpf || | |
5216 | !slave_dev->netdev_ops->ndo_xdp_xmit) { | |
6569fa2d JT |
5217 | SLAVE_NL_ERR(dev, slave_dev, extack, |
5218 | "Slave device does not support XDP"); | |
9e2ee5c7 JM |
5219 | err = -EOPNOTSUPP; |
5220 | goto err; | |
5221 | } | |
5222 | ||
5223 | if (dev_xdp_prog_count(slave_dev) > 0) { | |
6569fa2d JT |
5224 | SLAVE_NL_ERR(dev, slave_dev, extack, |
5225 | "Slave has XDP program loaded, please unload before enslaving"); | |
9e2ee5c7 JM |
5226 | err = -EOPNOTSUPP; |
5227 | goto err; | |
5228 | } | |
5229 | ||
5230 | err = slave_dev->netdev_ops->ndo_bpf(slave_dev, &xdp); | |
5231 | if (err < 0) { | |
5232 | /* ndo_bpf() sets extack error message */ | |
5233 | slave_err(dev, slave_dev, "Error %d calling ndo_bpf\n", err); | |
5234 | goto err; | |
5235 | } | |
5236 | if (prog) | |
5237 | bpf_prog_inc(prog); | |
5238 | } | |
5239 | ||
6d5f1ef8 | 5240 | if (prog) { |
9e2ee5c7 | 5241 | static_branch_inc(&bpf_master_redirect_enabled_key); |
6d5f1ef8 JM |
5242 | } else if (old_prog) { |
5243 | bpf_prog_put(old_prog); | |
9e2ee5c7 | 5244 | static_branch_dec(&bpf_master_redirect_enabled_key); |
6d5f1ef8 | 5245 | } |
9e2ee5c7 JM |
5246 | |
5247 | return 0; | |
5248 | ||
5249 | err: | |
5250 | /* unwind the program changes */ | |
5251 | bond->xdp_prog = old_prog; | |
5252 | xdp.prog = old_prog; | |
5253 | xdp.extack = NULL; /* do not overwrite original error */ | |
5254 | ||
5255 | bond_for_each_slave(bond, rollback_slave, iter) { | |
5256 | struct net_device *slave_dev = rollback_slave->dev; | |
5257 | int err_unwind; | |
5258 | ||
5259 | if (slave == rollback_slave) | |
5260 | break; | |
5261 | ||
5262 | err_unwind = slave_dev->netdev_ops->ndo_bpf(slave_dev, &xdp); | |
5263 | if (err_unwind < 0) | |
5264 | slave_err(dev, slave_dev, | |
5265 | "Error %d when unwinding XDP program change\n", err_unwind); | |
5266 | else if (xdp.prog) | |
5267 | bpf_prog_inc(xdp.prog); | |
5268 | } | |
5269 | return err; | |
5270 | } | |
5271 | ||
5272 | static int bond_xdp(struct net_device *dev, struct netdev_bpf *xdp) | |
5273 | { | |
5274 | switch (xdp->command) { | |
5275 | case XDP_SETUP_PROG: | |
5276 | return bond_xdp_set(dev, xdp->prog, xdp->extack); | |
5277 | default: | |
5278 | return -EINVAL; | |
5279 | } | |
5280 | } | |
5281 | ||
4ca0d9ac JW |
5282 | static u32 bond_mode_bcast_speed(struct slave *slave, u32 speed) |
5283 | { | |
5284 | if (speed == 0 || speed == SPEED_UNKNOWN) | |
5285 | speed = slave->speed; | |
5286 | else | |
5287 | speed = min(speed, slave->speed); | |
5288 | ||
5289 | return speed; | |
5290 | } | |
5291 | ||
d46b6349 PR |
5292 | static int bond_ethtool_get_link_ksettings(struct net_device *bond_dev, |
5293 | struct ethtool_link_ksettings *cmd) | |
bb5b052f AG |
5294 | { |
5295 | struct bonding *bond = netdev_priv(bond_dev); | |
9caff1e7 | 5296 | struct list_head *iter; |
dec1e90e | 5297 | struct slave *slave; |
4ca0d9ac | 5298 | u32 speed = 0; |
bb5b052f | 5299 | |
d46b6349 PR |
5300 | cmd->base.duplex = DUPLEX_UNKNOWN; |
5301 | cmd->base.port = PORT_OTHER; | |
bb5b052f | 5302 | |
8557cd74 | 5303 | /* Since bond_slave_can_tx returns false for all inactive or down slaves, we |
bb5b052f AG |
5304 | * do not need to check mode. Though link speed might not represent |
5305 | * the true receive or transmit bandwidth (not all modes are symmetric) | |
5306 | * this is an accurate maximum. | |
5307 | */ | |
9caff1e7 | 5308 | bond_for_each_slave(bond, slave, iter) { |
8557cd74 | 5309 | if (bond_slave_can_tx(slave)) { |
4ca0d9ac JW |
5310 | if (slave->speed != SPEED_UNKNOWN) { |
5311 | if (BOND_MODE(bond) == BOND_MODE_BROADCAST) | |
5312 | speed = bond_mode_bcast_speed(slave, | |
5313 | speed); | |
5314 | else | |
5315 | speed += slave->speed; | |
5316 | } | |
d46b6349 | 5317 | if (cmd->base.duplex == DUPLEX_UNKNOWN && |
bb5b052f | 5318 | slave->duplex != DUPLEX_UNKNOWN) |
d46b6349 | 5319 | cmd->base.duplex = slave->duplex; |
bb5b052f AG |
5320 | } |
5321 | } | |
d46b6349 | 5322 | cmd->base.speed = speed ? : SPEED_UNKNOWN; |
dec1e90e | 5323 | |
bb5b052f AG |
5324 | return 0; |
5325 | } | |
5326 | ||
217df670 | 5327 | static void bond_ethtool_get_drvinfo(struct net_device *bond_dev, |
7826d43f | 5328 | struct ethtool_drvinfo *drvinfo) |
217df670 | 5329 | { |
7826d43f | 5330 | strlcpy(drvinfo->driver, DRV_NAME, sizeof(drvinfo->driver)); |
7826d43f JP |
5331 | snprintf(drvinfo->fw_version, sizeof(drvinfo->fw_version), "%d", |
5332 | BOND_ABI_VERSION); | |
217df670 JV |
5333 | } |
5334 | ||
7282d491 | 5335 | static const struct ethtool_ops bond_ethtool_ops = { |
217df670 | 5336 | .get_drvinfo = bond_ethtool_get_drvinfo, |
fa53ebac | 5337 | .get_link = ethtool_op_get_link, |
d46b6349 | 5338 | .get_link_ksettings = bond_ethtool_get_link_ksettings, |
8531c5ff AK |
5339 | }; |
5340 | ||
eb7cc59a | 5341 | static const struct net_device_ops bond_netdev_ops = { |
181470fc | 5342 | .ndo_init = bond_init, |
9e71626c | 5343 | .ndo_uninit = bond_uninit, |
eb7cc59a SH |
5344 | .ndo_open = bond_open, |
5345 | .ndo_stop = bond_close, | |
00829823 | 5346 | .ndo_start_xmit = bond_start_xmit, |
bb1d9123 | 5347 | .ndo_select_queue = bond_select_queue, |
be1f3c2c | 5348 | .ndo_get_stats64 = bond_get_stats, |
a7605370 | 5349 | .ndo_eth_ioctl = bond_eth_ioctl, |
3d9d00bd | 5350 | .ndo_siocbond = bond_do_ioctl, |
232ec98e | 5351 | .ndo_siocdevprivate = bond_siocdevprivate, |
d03462b9 | 5352 | .ndo_change_rx_flags = bond_change_rx_flags, |
303d1cbf | 5353 | .ndo_set_rx_mode = bond_set_rx_mode, |
eb7cc59a | 5354 | .ndo_change_mtu = bond_change_mtu, |
cc0e4070 | 5355 | .ndo_set_mac_address = bond_set_mac_address, |
00829823 | 5356 | .ndo_neigh_setup = bond_neigh_setup, |
cc0e4070 | 5357 | .ndo_vlan_rx_add_vid = bond_vlan_rx_add_vid, |
eb7cc59a | 5358 | .ndo_vlan_rx_kill_vid = bond_vlan_rx_kill_vid, |
f6dc31a8 | 5359 | #ifdef CONFIG_NET_POLL_CONTROLLER |
8a8efa22 | 5360 | .ndo_netpoll_setup = bond_netpoll_setup, |
f6dc31a8 WC |
5361 | .ndo_netpoll_cleanup = bond_netpoll_cleanup, |
5362 | .ndo_poll_controller = bond_poll_controller, | |
5363 | #endif | |
9232ecca JP |
5364 | .ndo_add_slave = bond_enslave, |
5365 | .ndo_del_slave = bond_release, | |
b2a103e6 | 5366 | .ndo_fix_features = bond_fix_features, |
4847f049 | 5367 | .ndo_features_check = passthru_features_check, |
33720aaf | 5368 | .ndo_get_xmit_slave = bond_xmit_get_slave, |
007feb87 | 5369 | .ndo_sk_get_lower_dev = bond_sk_get_lower_dev, |
9e2ee5c7 JM |
5370 | .ndo_bpf = bond_xdp, |
5371 | .ndo_xdp_xmit = bond_xdp_xmit, | |
5372 | .ndo_xdp_get_xmit_slave = bond_xdp_get_xmit_slave, | |
eb7cc59a SH |
5373 | }; |
5374 | ||
b3f92b63 DG |
5375 | static const struct device_type bond_type = { |
5376 | .name = "bond", | |
5377 | }; | |
5378 | ||
9e2e61fb AW |
5379 | static void bond_destructor(struct net_device *bond_dev) |
5380 | { | |
5381 | struct bonding *bond = netdev_priv(bond_dev); | |
86a5ad0a | 5382 | |
9e2e61fb AW |
5383 | if (bond->wq) |
5384 | destroy_workqueue(bond->wq); | |
848ca918 JM |
5385 | |
5386 | if (bond->rr_tx_counter) | |
5387 | free_percpu(bond->rr_tx_counter); | |
9e2e61fb AW |
5388 | } |
5389 | ||
0a2a78c4 | 5390 | void bond_setup(struct net_device *bond_dev) |
1da177e4 | 5391 | { |
454d7c9b | 5392 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 | 5393 | |
b7435628 | 5394 | spin_lock_init(&bond->mode_lock); |
d2991f75 | 5395 | bond->params = bonding_defaults; |
1da177e4 LT |
5396 | |
5397 | /* Initialize pointers */ | |
1da177e4 | 5398 | bond->dev = bond_dev; |
1da177e4 LT |
5399 | |
5400 | /* Initialize the device entry points */ | |
181470fc | 5401 | ether_setup(bond_dev); |
31c05415 | 5402 | bond_dev->max_mtu = ETH_MAX_MTU; |
eb7cc59a | 5403 | bond_dev->netdev_ops = &bond_netdev_ops; |
8531c5ff | 5404 | bond_dev->ethtool_ops = &bond_ethtool_ops; |
1da177e4 | 5405 | |
cf124db5 DM |
5406 | bond_dev->needs_free_netdev = true; |
5407 | bond_dev->priv_destructor = bond_destructor; | |
1da177e4 | 5408 | |
b3f92b63 DG |
5409 | SET_NETDEV_DEVTYPE(bond_dev, &bond_type); |
5410 | ||
1da177e4 | 5411 | /* Initialize the device options */ |
1098cee6 | 5412 | bond_dev->flags |= IFF_MASTER; |
1e6f20ca | 5413 | bond_dev->priv_flags |= IFF_BONDING | IFF_UNICAST_FLT | IFF_NO_QUEUE; |
550fd08c | 5414 | bond_dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_TX_SKB_SHARING); |
181470fc | 5415 | |
18cb261a JW |
5416 | #ifdef CONFIG_XFRM_OFFLOAD |
5417 | /* set up xfrm device ops (only supported in active-backup right now) */ | |
a3b658cf | 5418 | bond_dev->xfrmdev_ops = &bond_xfrmdev_ops; |
9a560550 TY |
5419 | INIT_LIST_HEAD(&bond->ipsec_list); |
5420 | spin_lock_init(&bond->ipsec_lock); | |
18cb261a JW |
5421 | #endif /* CONFIG_XFRM_OFFLOAD */ |
5422 | ||
547942ca | 5423 | /* don't acquire bond device's netif_tx_lock when transmitting */ |
1da177e4 LT |
5424 | bond_dev->features |= NETIF_F_LLTX; |
5425 | ||
5426 | /* By default, we declare the bond to be fully | |
5427 | * VLAN hardware accelerated capable. Special | |
5428 | * care is taken in the various xmit functions | |
5429 | * when there are slaves that are not hw accel | |
5430 | * capable | |
5431 | */ | |
1da177e4 | 5432 | |
f9399814 WC |
5433 | /* Don't allow bond devices to change network namespaces. */ |
5434 | bond_dev->features |= NETIF_F_NETNS_LOCAL; | |
5435 | ||
b2a103e6 | 5436 | bond_dev->hw_features = BOND_VLAN_FEATURES | |
f646968f PM |
5437 | NETIF_F_HW_VLAN_CTAG_RX | |
5438 | NETIF_F_HW_VLAN_CTAG_FILTER; | |
b2a103e6 | 5439 | |
ecb8fed4 | 5440 | bond_dev->hw_features |= NETIF_F_GSO_ENCAP_ALL; |
b2a103e6 | 5441 | bond_dev->features |= bond_dev->hw_features; |
30d8177e | 5442 | bond_dev->features |= NETIF_F_HW_VLAN_CTAG_TX | NETIF_F_HW_VLAN_STAG_TX; |
a3b658cf | 5443 | #ifdef CONFIG_XFRM_OFFLOAD |
d241b382 JW |
5444 | bond_dev->hw_features |= BOND_XFRM_FEATURES; |
5445 | /* Only enable XFRM features if this is an active-backup config */ | |
5446 | if (BOND_MODE(bond) == BOND_MODE_ACTIVEBACKUP) | |
5447 | bond_dev->features |= BOND_XFRM_FEATURES; | |
a3b658cf | 5448 | #endif /* CONFIG_XFRM_OFFLOAD */ |
89df6a81 TT |
5449 | #if IS_ENABLED(CONFIG_TLS_DEVICE) |
5450 | if (bond_sk_check(bond)) | |
5451 | bond_dev->features |= BOND_TLS_FEATURES; | |
5452 | #endif | |
1da177e4 LT |
5453 | } |
5454 | ||
547942ca NA |
5455 | /* Destroy a bonding device. |
5456 | * Must be under rtnl_lock when this function is called. | |
5457 | */ | |
c67dfb29 | 5458 | static void bond_uninit(struct net_device *bond_dev) |
a434e43f | 5459 | { |
454d7c9b | 5460 | struct bonding *bond = netdev_priv(bond_dev); |
6b447e76 | 5461 | struct bond_up_slave *usable, *all; |
544a028e VF |
5462 | struct list_head *iter; |
5463 | struct slave *slave; | |
a434e43f | 5464 | |
f6dc31a8 WC |
5465 | bond_netpoll_cleanup(bond_dev); |
5466 | ||
c67dfb29 | 5467 | /* Release the bonded slaves */ |
544a028e | 5468 | bond_for_each_slave(bond, slave, iter) |
f51048c3 | 5469 | __bond_release_one(bond_dev, slave->dev, true, true); |
76444f50 | 5470 | netdev_info(bond_dev, "Released all slaves\n"); |
c67dfb29 | 5471 | |
6b447e76 MG |
5472 | usable = rtnl_dereference(bond->usable_slaves); |
5473 | if (usable) { | |
ed7d4f02 | 5474 | RCU_INIT_POINTER(bond->usable_slaves, NULL); |
6b447e76 MG |
5475 | kfree_rcu(usable, rcu); |
5476 | } | |
5477 | ||
5478 | all = rtnl_dereference(bond->all_slaves); | |
5479 | if (all) { | |
5480 | RCU_INIT_POINTER(bond->all_slaves, NULL); | |
5481 | kfree_rcu(all, rcu); | |
ee637714 MB |
5482 | } |
5483 | ||
a434e43f JV |
5484 | list_del(&bond->bond_list); |
5485 | ||
f073c7ca | 5486 | bond_debug_unregister(bond); |
a434e43f JV |
5487 | } |
5488 | ||
1da177e4 LT |
5489 | /*------------------------- Module initialization ---------------------------*/ |
5490 | ||
1da177e4 LT |
5491 | static int bond_check_params(struct bond_params *params) |
5492 | { | |
5a5c5fd4 | 5493 | int arp_validate_value, fail_over_mac_value, primary_reselect_value, i; |
f3253339 | 5494 | struct bond_opt_value newval; |
5495 | const struct bond_opt_value *valptr; | |
72ccc471 | 5496 | int arp_all_targets_value = 0; |
6791e466 | 5497 | u16 ad_actor_sys_prio = 0; |
d22a5fc0 | 5498 | u16 ad_user_port_key = 0; |
72ccc471 | 5499 | __be32 arp_target[BOND_MAX_ARP_TARGETS] = { 0 }; |
dc9c4d0f MB |
5500 | int arp_ip_count; |
5501 | int bond_mode = BOND_MODE_ROUNDROBIN; | |
5502 | int xmit_hashtype = BOND_XMIT_POLICY_LAYER2; | |
5503 | int lacp_fast = 0; | |
f13ad104 | 5504 | int tlb_dynamic_lb; |
f5b2b966 | 5505 | |
547942ca | 5506 | /* Convert string parameters. */ |
1da177e4 | 5507 | if (mode) { |
2b3798d5 NA |
5508 | bond_opt_initstr(&newval, mode); |
5509 | valptr = bond_opt_parse(bond_opt_get(BOND_OPT_MODE), &newval); | |
5510 | if (!valptr) { | |
5511 | pr_err("Error: Invalid bonding mode \"%s\"\n", mode); | |
1da177e4 LT |
5512 | return -EINVAL; |
5513 | } | |
2b3798d5 | 5514 | bond_mode = valptr->value; |
1da177e4 LT |
5515 | } |
5516 | ||
169a3e66 | 5517 | if (xmit_hash_policy) { |
e79c1055 DB |
5518 | if (bond_mode == BOND_MODE_ROUNDROBIN || |
5519 | bond_mode == BOND_MODE_ACTIVEBACKUP || | |
5520 | bond_mode == BOND_MODE_BROADCAST) { | |
a4aee5c8 | 5521 | pr_info("xmit_hash_policy param is irrelevant in mode %s\n", |
90194264 | 5522 | bond_mode_name(bond_mode)); |
169a3e66 | 5523 | } else { |
a4b32ce7 NA |
5524 | bond_opt_initstr(&newval, xmit_hash_policy); |
5525 | valptr = bond_opt_parse(bond_opt_get(BOND_OPT_XMIT_HASH), | |
5526 | &newval); | |
5527 | if (!valptr) { | |
a4aee5c8 | 5528 | pr_err("Error: Invalid xmit_hash_policy \"%s\"\n", |
169a3e66 JV |
5529 | xmit_hash_policy); |
5530 | return -EINVAL; | |
5531 | } | |
a4b32ce7 | 5532 | xmit_hashtype = valptr->value; |
169a3e66 JV |
5533 | } |
5534 | } | |
5535 | ||
1da177e4 LT |
5536 | if (lacp_rate) { |
5537 | if (bond_mode != BOND_MODE_8023AD) { | |
a4aee5c8 JP |
5538 | pr_info("lacp_rate param is irrelevant in mode %s\n", |
5539 | bond_mode_name(bond_mode)); | |
1da177e4 | 5540 | } else { |
d3131de7 NA |
5541 | bond_opt_initstr(&newval, lacp_rate); |
5542 | valptr = bond_opt_parse(bond_opt_get(BOND_OPT_LACP_RATE), | |
5543 | &newval); | |
5544 | if (!valptr) { | |
a4aee5c8 | 5545 | pr_err("Error: Invalid lacp rate \"%s\"\n", |
d3131de7 | 5546 | lacp_rate); |
1da177e4 LT |
5547 | return -EINVAL; |
5548 | } | |
d3131de7 | 5549 | lacp_fast = valptr->value; |
1da177e4 LT |
5550 | } |
5551 | } | |
5552 | ||
fd989c83 | 5553 | if (ad_select) { |
548d28bd | 5554 | bond_opt_initstr(&newval, ad_select); |
9e5f5eeb NA |
5555 | valptr = bond_opt_parse(bond_opt_get(BOND_OPT_AD_SELECT), |
5556 | &newval); | |
5557 | if (!valptr) { | |
5558 | pr_err("Error: Invalid ad_select \"%s\"\n", ad_select); | |
fd989c83 JV |
5559 | return -EINVAL; |
5560 | } | |
9e5f5eeb NA |
5561 | params->ad_select = valptr->value; |
5562 | if (bond_mode != BOND_MODE_8023AD) | |
91565ebb | 5563 | pr_warn("ad_select param only affects 802.3ad mode\n"); |
fd989c83 JV |
5564 | } else { |
5565 | params->ad_select = BOND_AD_STABLE; | |
5566 | } | |
5567 | ||
f5841306 | 5568 | if (max_bonds < 0) { |
91565ebb JP |
5569 | pr_warn("Warning: max_bonds (%d) not in range %d-%d, so it was reset to BOND_DEFAULT_MAX_BONDS (%d)\n", |
5570 | max_bonds, 0, INT_MAX, BOND_DEFAULT_MAX_BONDS); | |
1da177e4 LT |
5571 | max_bonds = BOND_DEFAULT_MAX_BONDS; |
5572 | } | |
5573 | ||
5574 | if (miimon < 0) { | |
91565ebb JP |
5575 | pr_warn("Warning: miimon module parameter (%d), not in range 0-%d, so it was reset to 0\n", |
5576 | miimon, INT_MAX); | |
b98d9c66 | 5577 | miimon = 0; |
1da177e4 LT |
5578 | } |
5579 | ||
5580 | if (updelay < 0) { | |
91565ebb JP |
5581 | pr_warn("Warning: updelay module parameter (%d), not in range 0-%d, so it was reset to 0\n", |
5582 | updelay, INT_MAX); | |
1da177e4 LT |
5583 | updelay = 0; |
5584 | } | |
5585 | ||
5586 | if (downdelay < 0) { | |
91565ebb JP |
5587 | pr_warn("Warning: downdelay module parameter (%d), not in range 0-%d, so it was reset to 0\n", |
5588 | downdelay, INT_MAX); | |
1da177e4 LT |
5589 | downdelay = 0; |
5590 | } | |
5591 | ||
b3c898e2 DB |
5592 | if ((use_carrier != 0) && (use_carrier != 1)) { |
5593 | pr_warn("Warning: use_carrier module parameter (%d), not of valid value (0/1), so it was set to 1\n", | |
91565ebb | 5594 | use_carrier); |
1da177e4 LT |
5595 | use_carrier = 1; |
5596 | } | |
5597 | ||
ad246c99 | 5598 | if (num_peer_notif < 0 || num_peer_notif > 255) { |
91565ebb JP |
5599 | pr_warn("Warning: num_grat_arp/num_unsol_na (%d) not in range 0-255 so it was reset to 1\n", |
5600 | num_peer_notif); | |
ad246c99 BH |
5601 | num_peer_notif = 1; |
5602 | } | |
5603 | ||
834db4bc | 5604 | /* reset values for 802.3ad/TLB/ALB */ |
267bed77 | 5605 | if (!bond_mode_uses_arp(bond_mode)) { |
1da177e4 | 5606 | if (!miimon) { |
91565ebb JP |
5607 | pr_warn("Warning: miimon must be specified, otherwise bonding will not detect link failure, speed and duplex which are essential for 802.3ad operation\n"); |
5608 | pr_warn("Forcing miimon to 100msec\n"); | |
fe9d04af | 5609 | miimon = BOND_DEFAULT_MIIMON; |
1da177e4 LT |
5610 | } |
5611 | } | |
5612 | ||
bb1d9123 | 5613 | if (tx_queues < 1 || tx_queues > 255) { |
91565ebb JP |
5614 | pr_warn("Warning: tx_queues (%d) should be between 1 and 255, resetting to %d\n", |
5615 | tx_queues, BOND_DEFAULT_TX_QUEUES); | |
bb1d9123 AG |
5616 | tx_queues = BOND_DEFAULT_TX_QUEUES; |
5617 | } | |
5618 | ||
ebd8e497 | 5619 | if ((all_slaves_active != 0) && (all_slaves_active != 1)) { |
91565ebb JP |
5620 | pr_warn("Warning: all_slaves_active module parameter (%d), not of valid value (0/1), so it was set to 0\n", |
5621 | all_slaves_active); | |
ebd8e497 AG |
5622 | all_slaves_active = 0; |
5623 | } | |
5624 | ||
c2952c31 | 5625 | if (resend_igmp < 0 || resend_igmp > 255) { |
91565ebb JP |
5626 | pr_warn("Warning: resend_igmp (%d) should be between 0 and 255, resetting to %d\n", |
5627 | resend_igmp, BOND_DEFAULT_RESEND_IGMP); | |
c2952c31 FL |
5628 | resend_igmp = BOND_DEFAULT_RESEND_IGMP; |
5629 | } | |
5630 | ||
aa59d851 NA |
5631 | bond_opt_initval(&newval, packets_per_slave); |
5632 | if (!bond_opt_parse(bond_opt_get(BOND_OPT_PACKETS_PER_SLAVE), &newval)) { | |
73958329 NA |
5633 | pr_warn("Warning: packets_per_slave (%d) should be between 0 and %u resetting to 1\n", |
5634 | packets_per_slave, USHRT_MAX); | |
5635 | packets_per_slave = 1; | |
5636 | } | |
5637 | ||
1da177e4 | 5638 | if (bond_mode == BOND_MODE_ALB) { |
a4aee5c8 JP |
5639 | 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", |
5640 | updelay); | |
1da177e4 LT |
5641 | } |
5642 | ||
5643 | if (!miimon) { | |
5644 | if (updelay || downdelay) { | |
5645 | /* just warn the user the up/down delay will have | |
5646 | * no effect since miimon is zero... | |
5647 | */ | |
91565ebb JP |
5648 | pr_warn("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", |
5649 | updelay, downdelay); | |
1da177e4 LT |
5650 | } |
5651 | } else { | |
5652 | /* don't allow arp monitoring */ | |
5653 | if (arp_interval) { | |
91565ebb JP |
5654 | pr_warn("Warning: miimon (%d) and arp_interval (%d) can't be used simultaneously, disabling ARP monitoring\n", |
5655 | miimon, arp_interval); | |
1da177e4 LT |
5656 | arp_interval = 0; |
5657 | } | |
5658 | ||
5659 | if ((updelay % miimon) != 0) { | |
91565ebb JP |
5660 | pr_warn("Warning: updelay (%d) is not a multiple of miimon (%d), updelay rounded to %d ms\n", |
5661 | updelay, miimon, (updelay / miimon) * miimon); | |
1da177e4 LT |
5662 | } |
5663 | ||
5664 | updelay /= miimon; | |
5665 | ||
5666 | if ((downdelay % miimon) != 0) { | |
91565ebb JP |
5667 | pr_warn("Warning: downdelay (%d) is not a multiple of miimon (%d), downdelay rounded to %d ms\n", |
5668 | downdelay, miimon, | |
5669 | (downdelay / miimon) * miimon); | |
1da177e4 LT |
5670 | } |
5671 | ||
5672 | downdelay /= miimon; | |
5673 | } | |
5674 | ||
5675 | if (arp_interval < 0) { | |
91565ebb JP |
5676 | pr_warn("Warning: arp_interval module parameter (%d), not in range 0-%d, so it was reset to 0\n", |
5677 | arp_interval, INT_MAX); | |
7bdb04ed | 5678 | arp_interval = 0; |
1da177e4 LT |
5679 | } |
5680 | ||
5a5c5fd4 | 5681 | for (arp_ip_count = 0, i = 0; |
5682 | (arp_ip_count < BOND_MAX_ARP_TARGETS) && arp_ip_target[i]; i++) { | |
89015c18 | 5683 | __be32 ip; |
547942ca NA |
5684 | |
5685 | /* not a complete check, but good enough to catch mistakes */ | |
89015c18 | 5686 | if (!in4_pton(arp_ip_target[i], -1, (u8 *)&ip, -1, NULL) || |
2807a9fe | 5687 | !bond_is_ip_target_ok(ip)) { |
91565ebb JP |
5688 | pr_warn("Warning: bad arp_ip_target module parameter (%s), ARP monitoring will not be performed\n", |
5689 | arp_ip_target[i]); | |
1da177e4 LT |
5690 | arp_interval = 0; |
5691 | } else { | |
0afee4e8 VF |
5692 | if (bond_get_targets_ip(arp_target, ip) == -1) |
5693 | arp_target[arp_ip_count++] = ip; | |
5694 | else | |
91565ebb JP |
5695 | pr_warn("Warning: duplicate address %pI4 in arp_ip_target, skipping\n", |
5696 | &ip); | |
1da177e4 LT |
5697 | } |
5698 | } | |
5699 | ||
5700 | if (arp_interval && !arp_ip_count) { | |
5701 | /* don't allow arping if no arp_ip_target given... */ | |
91565ebb JP |
5702 | pr_warn("Warning: arp_interval module parameter (%d) specified without providing an arp_ip_target parameter, arp_interval was reset to 0\n", |
5703 | arp_interval); | |
1da177e4 LT |
5704 | arp_interval = 0; |
5705 | } | |
5706 | ||
f5b2b966 | 5707 | if (arp_validate) { |
f5b2b966 | 5708 | if (!arp_interval) { |
a4aee5c8 | 5709 | pr_err("arp_validate requires arp_interval\n"); |
f5b2b966 JV |
5710 | return -EINVAL; |
5711 | } | |
5712 | ||
16228881 NA |
5713 | bond_opt_initstr(&newval, arp_validate); |
5714 | valptr = bond_opt_parse(bond_opt_get(BOND_OPT_ARP_VALIDATE), | |
5715 | &newval); | |
5716 | if (!valptr) { | |
a4aee5c8 | 5717 | pr_err("Error: invalid arp_validate \"%s\"\n", |
16228881 | 5718 | arp_validate); |
f5b2b966 JV |
5719 | return -EINVAL; |
5720 | } | |
16228881 NA |
5721 | arp_validate_value = valptr->value; |
5722 | } else { | |
f5b2b966 | 5723 | arp_validate_value = 0; |
16228881 | 5724 | } |
f5b2b966 | 5725 | |
8599b52e | 5726 | if (arp_all_targets) { |
edf36b24 NA |
5727 | bond_opt_initstr(&newval, arp_all_targets); |
5728 | valptr = bond_opt_parse(bond_opt_get(BOND_OPT_ARP_ALL_TARGETS), | |
5729 | &newval); | |
5730 | if (!valptr) { | |
8599b52e VF |
5731 | pr_err("Error: invalid arp_all_targets_value \"%s\"\n", |
5732 | arp_all_targets); | |
5733 | arp_all_targets_value = 0; | |
edf36b24 NA |
5734 | } else { |
5735 | arp_all_targets_value = valptr->value; | |
8599b52e VF |
5736 | } |
5737 | } | |
5738 | ||
1da177e4 | 5739 | if (miimon) { |
a4aee5c8 | 5740 | pr_info("MII link monitoring set to %d ms\n", miimon); |
1da177e4 | 5741 | } else if (arp_interval) { |
16228881 NA |
5742 | valptr = bond_opt_get_val(BOND_OPT_ARP_VALIDATE, |
5743 | arp_validate_value); | |
a4aee5c8 | 5744 | pr_info("ARP monitoring set to %d ms, validate %s, with %d target(s):", |
16228881 | 5745 | arp_interval, valptr->string, arp_ip_count); |
1da177e4 LT |
5746 | |
5747 | for (i = 0; i < arp_ip_count; i++) | |
90194264 | 5748 | pr_cont(" %s", arp_ip_target[i]); |
1da177e4 | 5749 | |
90194264 | 5750 | pr_cont("\n"); |
1da177e4 | 5751 | |
b8a9787e | 5752 | } else if (max_bonds) { |
1da177e4 LT |
5753 | /* miimon and arp_interval not set, we need one so things |
5754 | * work as expected, see bonding.txt for details | |
5755 | */ | |
90194264 | 5756 | 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 |
5757 | } |
5758 | ||
ec0865a9 | 5759 | if (primary && !bond_mode_uses_primary(bond_mode)) { |
1da177e4 LT |
5760 | /* currently, using a primary only makes sense |
5761 | * in active backup, TLB or ALB modes | |
5762 | */ | |
91565ebb JP |
5763 | pr_warn("Warning: %s primary device specified but has no effect in %s mode\n", |
5764 | primary, bond_mode_name(bond_mode)); | |
1da177e4 LT |
5765 | primary = NULL; |
5766 | } | |
5767 | ||
a549952a | 5768 | if (primary && primary_reselect) { |
388d3a6d NA |
5769 | bond_opt_initstr(&newval, primary_reselect); |
5770 | valptr = bond_opt_parse(bond_opt_get(BOND_OPT_PRIMARY_RESELECT), | |
5771 | &newval); | |
5772 | if (!valptr) { | |
a4aee5c8 | 5773 | pr_err("Error: Invalid primary_reselect \"%s\"\n", |
388d3a6d | 5774 | primary_reselect); |
a549952a JP |
5775 | return -EINVAL; |
5776 | } | |
388d3a6d | 5777 | primary_reselect_value = valptr->value; |
a549952a JP |
5778 | } else { |
5779 | primary_reselect_value = BOND_PRI_RESELECT_ALWAYS; | |
5780 | } | |
5781 | ||
3915c1e8 | 5782 | if (fail_over_mac) { |
1df6b6aa NA |
5783 | bond_opt_initstr(&newval, fail_over_mac); |
5784 | valptr = bond_opt_parse(bond_opt_get(BOND_OPT_FAIL_OVER_MAC), | |
5785 | &newval); | |
5786 | if (!valptr) { | |
a4aee5c8 | 5787 | pr_err("Error: invalid fail_over_mac \"%s\"\n", |
1df6b6aa | 5788 | fail_over_mac); |
3915c1e8 JV |
5789 | return -EINVAL; |
5790 | } | |
1df6b6aa | 5791 | fail_over_mac_value = valptr->value; |
3915c1e8 | 5792 | if (bond_mode != BOND_MODE_ACTIVEBACKUP) |
91565ebb | 5793 | pr_warn("Warning: fail_over_mac only affects active-backup mode\n"); |
3915c1e8 JV |
5794 | } else { |
5795 | fail_over_mac_value = BOND_FOM_NONE; | |
5796 | } | |
dd957c57 | 5797 | |
6791e466 MB |
5798 | bond_opt_initstr(&newval, "default"); |
5799 | valptr = bond_opt_parse( | |
5800 | bond_opt_get(BOND_OPT_AD_ACTOR_SYS_PRIO), | |
5801 | &newval); | |
5802 | if (!valptr) { | |
5803 | pr_err("Error: No ad_actor_sys_prio default value"); | |
5804 | return -EINVAL; | |
5805 | } | |
5806 | ad_actor_sys_prio = valptr->value; | |
5807 | ||
d22a5fc0 MB |
5808 | valptr = bond_opt_parse(bond_opt_get(BOND_OPT_AD_USER_PORT_KEY), |
5809 | &newval); | |
5810 | if (!valptr) { | |
5811 | pr_err("Error: No ad_user_port_key default value"); | |
5812 | return -EINVAL; | |
5813 | } | |
5814 | ad_user_port_key = valptr->value; | |
5815 | ||
f13ad104 NA |
5816 | bond_opt_initstr(&newval, "default"); |
5817 | valptr = bond_opt_parse(bond_opt_get(BOND_OPT_TLB_DYNAMIC_LB), &newval); | |
5818 | if (!valptr) { | |
5819 | pr_err("Error: No tlb_dynamic_lb default value"); | |
5820 | return -EINVAL; | |
8b426dc5 | 5821 | } |
f13ad104 | 5822 | tlb_dynamic_lb = valptr->value; |
8b426dc5 | 5823 | |
3a7129e5 | 5824 | if (lp_interval == 0) { |
91565ebb JP |
5825 | pr_warn("Warning: ip_interval must be between 1 and %d, so it was reset to %d\n", |
5826 | INT_MAX, BOND_ALB_DEFAULT_LP_INTERVAL); | |
3a7129e5 | 5827 | lp_interval = BOND_ALB_DEFAULT_LP_INTERVAL; |
5828 | } | |
5829 | ||
1da177e4 LT |
5830 | /* fill params struct with the proper values */ |
5831 | params->mode = bond_mode; | |
169a3e66 | 5832 | params->xmit_policy = xmit_hashtype; |
1da177e4 | 5833 | params->miimon = miimon; |
ad246c99 | 5834 | params->num_peer_notif = num_peer_notif; |
1da177e4 | 5835 | params->arp_interval = arp_interval; |
f5b2b966 | 5836 | params->arp_validate = arp_validate_value; |
8599b52e | 5837 | params->arp_all_targets = arp_all_targets_value; |
1da177e4 LT |
5838 | params->updelay = updelay; |
5839 | params->downdelay = downdelay; | |
07a4ddec | 5840 | params->peer_notif_delay = 0; |
1da177e4 | 5841 | params->use_carrier = use_carrier; |
3a755cd8 | 5842 | params->lacp_active = 1; |
1da177e4 LT |
5843 | params->lacp_fast = lacp_fast; |
5844 | params->primary[0] = 0; | |
a549952a | 5845 | params->primary_reselect = primary_reselect_value; |
3915c1e8 | 5846 | params->fail_over_mac = fail_over_mac_value; |
bb1d9123 | 5847 | params->tx_queues = tx_queues; |
ebd8e497 | 5848 | params->all_slaves_active = all_slaves_active; |
c2952c31 | 5849 | params->resend_igmp = resend_igmp; |
655f8919 | 5850 | params->min_links = min_links; |
3a7129e5 | 5851 | params->lp_interval = lp_interval; |
809fa972 | 5852 | params->packets_per_slave = packets_per_slave; |
8b426dc5 | 5853 | params->tlb_dynamic_lb = tlb_dynamic_lb; |
6791e466 | 5854 | params->ad_actor_sys_prio = ad_actor_sys_prio; |
74514957 | 5855 | eth_zero_addr(params->ad_actor_system); |
d22a5fc0 | 5856 | params->ad_user_port_key = ad_user_port_key; |
809fa972 HFS |
5857 | if (packets_per_slave > 0) { |
5858 | params->reciprocal_packets_per_slave = | |
5859 | reciprocal_value(packets_per_slave); | |
5860 | } else { | |
5861 | /* reciprocal_packets_per_slave is unused if | |
5862 | * packets_per_slave is 0 or 1, just initialize it | |
5863 | */ | |
5864 | params->reciprocal_packets_per_slave = | |
5865 | (struct reciprocal_value) { 0 }; | |
5866 | } | |
5867 | ||
43902070 KC |
5868 | if (primary) |
5869 | strscpy_pad(params->primary, primary, sizeof(params->primary)); | |
1da177e4 LT |
5870 | |
5871 | memcpy(params->arp_targets, arp_target, sizeof(arp_target)); | |
5872 | ||
5873 | return 0; | |
5874 | } | |
5875 | ||
547942ca | 5876 | /* Called from registration process */ |
181470fc SH |
5877 | static int bond_init(struct net_device *bond_dev) |
5878 | { | |
5879 | struct bonding *bond = netdev_priv(bond_dev); | |
ec87fd3b | 5880 | struct bond_net *bn = net_generic(dev_net(bond_dev), bond_net_id); |
181470fc | 5881 | |
76444f50 | 5882 | netdev_dbg(bond_dev, "Begin bond_init\n"); |
181470fc | 5883 | |
f9f225eb | 5884 | bond->wq = alloc_ordered_workqueue(bond_dev->name, WQ_MEM_RECLAIM); |
181470fc SH |
5885 | if (!bond->wq) |
5886 | return -ENOMEM; | |
5887 | ||
848ca918 JM |
5888 | if (BOND_MODE(bond) == BOND_MODE_ROUNDROBIN) { |
5889 | bond->rr_tx_counter = alloc_percpu(u32); | |
5890 | if (!bond->rr_tx_counter) { | |
5891 | destroy_workqueue(bond->wq); | |
5892 | bond->wq = NULL; | |
5893 | return -ENOMEM; | |
5894 | } | |
5895 | } | |
5896 | ||
089bca2c | 5897 | spin_lock_init(&bond->stats_lock); |
1a33e10e | 5898 | netdev_lockdep_set_classes(bond_dev); |
181470fc | 5899 | |
ec87fd3b | 5900 | list_add_tail(&bond->bond_list, &bn->dev_list); |
181470fc | 5901 | |
6151b3d4 | 5902 | bond_prepare_sysfs_group(bond); |
22bedad3 | 5903 | |
f073c7ca TI |
5904 | bond_debug_register(bond); |
5905 | ||
409cc1f8 JP |
5906 | /* Ensure valid dev_addr */ |
5907 | if (is_zero_ether_addr(bond_dev->dev_addr) && | |
97a1e639 | 5908 | bond_dev->addr_assign_type == NET_ADDR_PERM) |
409cc1f8 | 5909 | eth_hw_addr_random(bond_dev); |
409cc1f8 | 5910 | |
181470fc SH |
5911 | return 0; |
5912 | } | |
5913 | ||
0a2a78c4 | 5914 | unsigned int bond_get_num_tx_queues(void) |
d5da4510 | 5915 | { |
efacb309 | 5916 | return tx_queues; |
d5da4510 JP |
5917 | } |
5918 | ||
dfe60397 | 5919 | /* Create a new bond based on the specified name and bonding parameters. |
e4b91c48 | 5920 | * If name is NULL, obtain a suitable "bond%d" name for us. |
dfe60397 MW |
5921 | * Caller must NOT hold rtnl_lock; we need to release it here before we |
5922 | * set up our sysfs entries. | |
5923 | */ | |
ec87fd3b | 5924 | int bond_create(struct net *net, const char *name) |
dfe60397 MW |
5925 | { |
5926 | struct net_device *bond_dev; | |
e913fb27 P |
5927 | struct bonding *bond; |
5928 | struct alb_bond_info *bond_info; | |
dfe60397 MW |
5929 | int res; |
5930 | ||
5931 | rtnl_lock(); | |
027ea041 | 5932 | |
1c5cae81 | 5933 | bond_dev = alloc_netdev_mq(sizeof(struct bonding), |
c835a677 | 5934 | name ? name : "bond%d", NET_NAME_UNKNOWN, |
1c5cae81 | 5935 | bond_setup, tx_queues); |
dfe60397 | 5936 | if (!bond_dev) { |
a4aee5c8 | 5937 | pr_err("%s: eek! can't alloc netdev!\n", name); |
9e2e61fb AW |
5938 | rtnl_unlock(); |
5939 | return -ENOMEM; | |
dfe60397 MW |
5940 | } |
5941 | ||
e913fb27 P |
5942 | /* |
5943 | * Initialize rx_hashtbl_used_head to RLB_NULL_INDEX. | |
5944 | * It is set to 0 by default which is wrong. | |
5945 | */ | |
5946 | bond = netdev_priv(bond_dev); | |
5947 | bond_info = &(BOND_ALB_INFO(bond)); | |
5948 | bond_info->rx_hashtbl_used_head = RLB_NULL_INDEX; | |
5949 | ||
ec87fd3b | 5950 | dev_net_set(bond_dev, net); |
88ead977 EB |
5951 | bond_dev->rtnl_link_ops = &bond_link_ops; |
5952 | ||
dfe60397 | 5953 | res = register_netdevice(bond_dev); |
544f287b TY |
5954 | if (res < 0) { |
5955 | free_netdev(bond_dev); | |
5956 | rtnl_unlock(); | |
5957 | ||
5958 | return res; | |
5959 | } | |
0daa2303 | 5960 | |
e826eafa PO |
5961 | netif_carrier_off(bond_dev); |
5962 | ||
4493b81b MB |
5963 | bond_work_init_all(bond); |
5964 | ||
7e083840 | 5965 | rtnl_unlock(); |
544f287b | 5966 | return 0; |
dfe60397 MW |
5967 | } |
5968 | ||
2c8c1e72 | 5969 | static int __net_init bond_net_init(struct net *net) |
ec87fd3b | 5970 | { |
15449745 | 5971 | struct bond_net *bn = net_generic(net, bond_net_id); |
ec87fd3b EB |
5972 | |
5973 | bn->net = net; | |
5974 | INIT_LIST_HEAD(&bn->dev_list); | |
5975 | ||
ec87fd3b | 5976 | bond_create_proc_dir(bn); |
4c22400a | 5977 | bond_create_sysfs(bn); |
87a7b84b | 5978 | |
15449745 | 5979 | return 0; |
ec87fd3b EB |
5980 | } |
5981 | ||
2c8c1e72 | 5982 | static void __net_exit bond_net_exit(struct net *net) |
ec87fd3b | 5983 | { |
15449745 | 5984 | struct bond_net *bn = net_generic(net, bond_net_id); |
69b0216a | 5985 | struct bonding *bond, *tmp_bond; |
5986 | LIST_HEAD(list); | |
ec87fd3b | 5987 | |
4c22400a | 5988 | bond_destroy_sysfs(bn); |
69b0216a | 5989 | |
5990 | /* Kill off any bonds created after unregistering bond rtnl ops */ | |
5991 | rtnl_lock(); | |
5992 | list_for_each_entry_safe(bond, tmp_bond, &bn->dev_list, bond_list) | |
5993 | unregister_netdevice_queue(bond->dev, &list); | |
5994 | unregister_netdevice_many(&list); | |
5995 | rtnl_unlock(); | |
23fa5c2c VF |
5996 | |
5997 | bond_destroy_proc_dir(bn); | |
ec87fd3b EB |
5998 | } |
5999 | ||
6000 | static struct pernet_operations bond_net_ops = { | |
6001 | .init = bond_net_init, | |
6002 | .exit = bond_net_exit, | |
15449745 EB |
6003 | .id = &bond_net_id, |
6004 | .size = sizeof(struct bond_net), | |
ec87fd3b EB |
6005 | }; |
6006 | ||
1da177e4 LT |
6007 | static int __init bonding_init(void) |
6008 | { | |
1da177e4 LT |
6009 | int i; |
6010 | int res; | |
6011 | ||
dfe60397 | 6012 | res = bond_check_params(&bonding_defaults); |
3d632c3f | 6013 | if (res) |
dfe60397 | 6014 | goto out; |
1da177e4 | 6015 | |
15449745 | 6016 | res = register_pernet_subsys(&bond_net_ops); |
ec87fd3b EB |
6017 | if (res) |
6018 | goto out; | |
027ea041 | 6019 | |
0a2a78c4 | 6020 | res = bond_netlink_init(); |
88ead977 | 6021 | if (res) |
6639104b | 6022 | goto err_link; |
88ead977 | 6023 | |
f073c7ca TI |
6024 | bond_create_debugfs(); |
6025 | ||
1da177e4 | 6026 | for (i = 0; i < max_bonds; i++) { |
ec87fd3b | 6027 | res = bond_create(&init_net, NULL); |
dfe60397 MW |
6028 | if (res) |
6029 | goto err; | |
1da177e4 LT |
6030 | } |
6031 | ||
58deb77c MC |
6032 | skb_flow_dissector_init(&flow_keys_bonding, |
6033 | flow_keys_bonding_keys, | |
6034 | ARRAY_SIZE(flow_keys_bonding_keys)); | |
6035 | ||
1da177e4 | 6036 | register_netdevice_notifier(&bond_netdev_notifier); |
dfe60397 | 6037 | out: |
1da177e4 | 6038 | return res; |
88ead977 | 6039 | err: |
db298686 | 6040 | bond_destroy_debugfs(); |
0a2a78c4 | 6041 | bond_netlink_fini(); |
6639104b | 6042 | err_link: |
15449745 | 6043 | unregister_pernet_subsys(&bond_net_ops); |
88ead977 | 6044 | goto out; |
dfe60397 | 6045 | |
1da177e4 LT |
6046 | } |
6047 | ||
6048 | static void __exit bonding_exit(void) | |
6049 | { | |
6050 | unregister_netdevice_notifier(&bond_netdev_notifier); | |
6051 | ||
f073c7ca | 6052 | bond_destroy_debugfs(); |
ae68c398 | 6053 | |
0a2a78c4 | 6054 | bond_netlink_fini(); |
ffcdedb6 | 6055 | unregister_pernet_subsys(&bond_net_ops); |
e843fa50 NH |
6056 | |
6057 | #ifdef CONFIG_NET_POLL_CONTROLLER | |
547942ca | 6058 | /* Make sure we don't have an imbalance on our netpoll blocking */ |
fb4fa76a | 6059 | WARN_ON(atomic_read(&netpoll_block_tx)); |
e843fa50 | 6060 | #endif |
1da177e4 LT |
6061 | } |
6062 | ||
6063 | module_init(bonding_init); | |
6064 | module_exit(bonding_exit); | |
6065 | MODULE_LICENSE("GPL"); | |
2b526b56 | 6066 | MODULE_DESCRIPTION(DRV_DESCRIPTION); |
1da177e4 | 6067 | MODULE_AUTHOR("Thomas Davis, tadavis@lbl.gov and many others"); |