blob: fd01138c411e1de4221c3a333e3dbbc457271c20 [file] [log] [blame]
Linus Torvalds1da177e2005-04-16 15:20:36 -07001/*
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 *
Linus Torvalds1da177e2005-04-16 15:20:36 -070032 */
33
Linus Torvalds1da177e2005-04-16 15:20:36 -070034#include <linux/kernel.h>
35#include <linux/module.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070036#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>
Jay Vosburgh169a3e62005-06-26 17:54:11 -040042#include <net/ip.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070043#include <linux/ip.h>
Jay Vosburgh169a3e62005-06-26 17:54:11 -040044#include <linux/tcp.h>
45#include <linux/udp.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070046#include <linux/slab.h>
47#include <linux/string.h>
48#include <linux/init.h>
49#include <linux/timer.h>
50#include <linux/socket.h>
51#include <linux/ctype.h>
52#include <linux/inet.h>
53#include <linux/bitops.h>
Stephen Hemminger3d632c32009-06-12 19:02:48 +000054#include <linux/io.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070055#include <asm/dma.h>
Stephen Hemminger3d632c32009-06-12 19:02:48 +000056#include <linux/uaccess.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070057#include <linux/errno.h>
58#include <linux/netdevice.h>
59#include <linux/inetdevice.h>
Jay Vosburgha816c7c2007-02-28 17:03:37 -080060#include <linux/igmp.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070061#include <linux/etherdevice.h>
62#include <linux/skbuff.h>
63#include <net/sock.h>
64#include <linux/rtnetlink.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070065#include <linux/smp.h>
66#include <linux/if_ether.h>
67#include <net/arp.h>
68#include <linux/mii.h>
69#include <linux/ethtool.h>
70#include <linux/if_vlan.h>
71#include <linux/if_bonding.h>
David Sterbab63bb732007-12-06 23:40:33 -080072#include <linux/jiffies.h>
Neil Hormane843fa52010-10-13 16:01:50 +000073#include <linux/preempt.h>
Jay Vosburghc3ade5c2005-06-26 17:52:20 -040074#include <net/route.h>
Eric W. Biederman457c4cb2007-09-12 12:01:34 +020075#include <net/net_namespace.h>
Eric W. Biedermanec87fd32009-10-29 14:18:26 +000076#include <net/netns/generic.h>
Eric Dumazet5ee31c682012-06-12 06:03:51 +000077#include <net/pkt_sched.h>
nikolay@redhat.com278b2082013-08-01 16:54:51 +020078#include <linux/rculist.h>
Jiri Pirko1bd758e2015-05-12 14:56:07 +020079#include <net/flow_dissector.h>
Roopa Prabhuc158cba2015-01-29 22:40:16 -080080#include <net/switchdev.h>
David S. Miller1ef80192014-11-10 13:27:49 -050081#include <net/bonding.h>
82#include <net/bond_3ad.h>
83#include <net/bond_alb.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070084
Matan Barak73b5a6f22015-04-26 15:55:57 +030085#include "bonding_priv.h"
86
Linus Torvalds1da177e2005-04-16 15:20:36 -070087/*---------------------------- Module parameters ----------------------------*/
88
89/* monitor all links that often (in milliseconds). <=0 disables monitoring */
Linus Torvalds1da177e2005-04-16 15:20:36 -070090
91static int max_bonds = BOND_DEFAULT_MAX_BONDS;
Andy Gospodarekbb1d9122010-06-02 08:40:18 +000092static int tx_queues = BOND_DEFAULT_TX_QUEUES;
Ben Hutchingsad246c92011-04-26 15:25:52 +000093static int num_peer_notif = 1;
Nikolay Aleksandrovb98d9c62014-01-22 14:53:31 +010094static int miimon;
Stephen Hemminger3d632c32009-06-12 19:02:48 +000095static int updelay;
96static int downdelay;
Linus Torvalds1da177e2005-04-16 15:20:36 -070097static int use_carrier = 1;
Stephen Hemminger3d632c32009-06-12 19:02:48 +000098static char *mode;
99static char *primary;
Jiri Pirkoa5499522009-09-25 03:28:09 +0000100static char *primary_reselect;
Stephen Hemminger3d632c32009-06-12 19:02:48 +0000101static char *lacp_rate;
stephen hemminger655f8912011-06-22 09:54:39 +0000102static int min_links;
Stephen Hemminger3d632c32009-06-12 19:02:48 +0000103static char *ad_select;
104static char *xmit_hash_policy;
Nikolay Aleksandrov7bdb04e2014-01-22 14:53:23 +0100105static int arp_interval;
Stephen Hemminger3d632c32009-06-12 19:02:48 +0000106static char *arp_ip_target[BOND_MAX_ARP_TARGETS];
107static char *arp_validate;
Veaceslav Falico8599b522013-06-24 11:49:34 +0200108static char *arp_all_targets;
Stephen Hemminger3d632c32009-06-12 19:02:48 +0000109static char *fail_over_mac;
dingtianhongb07ea072013-07-23 15:25:47 +0800110static int all_slaves_active;
Stephen Hemmingerd2991f72009-06-12 19:02:44 +0000111static struct bond_params bonding_defaults;
Flavio Leitnerc2952c32010-10-05 14:23:59 +0000112static int resend_igmp = BOND_DEFAULT_RESEND_IGMP;
Nikolay Aleksandrov73958322013-11-05 13:51:41 +0100113static int packets_per_slave = 1;
dingtianhong3a7129e2013-12-21 14:40:12 +0800114static int lp_interval = BOND_ALB_DEFAULT_LP_INTERVAL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700115
116module_param(max_bonds, int, 0);
117MODULE_PARM_DESC(max_bonds, "Max number of bonded devices");
Andy Gospodarekbb1d9122010-06-02 08:40:18 +0000118module_param(tx_queues, int, 0);
119MODULE_PARM_DESC(tx_queues, "Max number of transmit queues (default = 16)");
Ben Hutchingsad246c92011-04-26 15:25:52 +0000120module_param_named(num_grat_arp, num_peer_notif, int, 0644);
Andy Gospodarek90e62472011-05-25 04:41:59 +0000121MODULE_PARM_DESC(num_grat_arp, "Number of peer notifications to send on "
122 "failover event (alias of num_unsol_na)");
Ben Hutchingsad246c92011-04-26 15:25:52 +0000123module_param_named(num_unsol_na, num_peer_notif, int, 0644);
Andy Gospodarek90e62472011-05-25 04:41:59 +0000124MODULE_PARM_DESC(num_unsol_na, "Number of peer notifications to send on "
125 "failover event (alias of num_grat_arp)");
Linus Torvalds1da177e2005-04-16 15:20:36 -0700126module_param(miimon, int, 0);
127MODULE_PARM_DESC(miimon, "Link check interval in milliseconds");
128module_param(updelay, int, 0);
129MODULE_PARM_DESC(updelay, "Delay before considering link up, in milliseconds");
130module_param(downdelay, int, 0);
Mitch Williams2ac47662005-11-09 10:35:03 -0800131MODULE_PARM_DESC(downdelay, "Delay before considering link down, "
132 "in milliseconds");
Linus Torvalds1da177e2005-04-16 15:20:36 -0700133module_param(use_carrier, int, 0);
Mitch Williams2ac47662005-11-09 10:35:03 -0800134MODULE_PARM_DESC(use_carrier, "Use netif_carrier_ok (vs MII ioctls) in miimon; "
135 "0 for off, 1 for on (default)");
Linus Torvalds1da177e2005-04-16 15:20:36 -0700136module_param(mode, charp, 0);
Andy Gospodarek90e62472011-05-25 04:41:59 +0000137MODULE_PARM_DESC(mode, "Mode of operation; 0 for balance-rr, "
Mitch Williams2ac47662005-11-09 10:35:03 -0800138 "1 for active-backup, 2 for balance-xor, "
139 "3 for broadcast, 4 for 802.3ad, 5 for balance-tlb, "
140 "6 for balance-alb");
Linus Torvalds1da177e2005-04-16 15:20:36 -0700141module_param(primary, charp, 0);
142MODULE_PARM_DESC(primary, "Primary network device to use");
Jiri Pirkoa5499522009-09-25 03:28:09 +0000143module_param(primary_reselect, charp, 0);
144MODULE_PARM_DESC(primary_reselect, "Reselect primary slave "
145 "once it comes up; "
146 "0 for always (default), "
147 "1 for only if speed of primary is "
148 "better, "
149 "2 for only on active slave "
150 "failure");
Linus Torvalds1da177e2005-04-16 15:20:36 -0700151module_param(lacp_rate, charp, 0);
Andy Gospodarek90e62472011-05-25 04:41:59 +0000152MODULE_PARM_DESC(lacp_rate, "LACPDU tx rate to request from 802.3ad partner; "
153 "0 for slow, 1 for fast");
Jay Vosburghfd989c82008-11-04 17:51:16 -0800154module_param(ad_select, charp, 0);
Zhu Yanjun0d039f32016-08-09 21:36:04 +0800155MODULE_PARM_DESC(ad_select, "802.3ad aggregation selection logic; "
Andy Gospodarek90e62472011-05-25 04:41:59 +0000156 "0 for stable (default), 1 for bandwidth, "
157 "2 for count");
stephen hemminger655f8912011-06-22 09:54:39 +0000158module_param(min_links, int, 0);
159MODULE_PARM_DESC(min_links, "Minimum number of available links before turning on carrier");
160
Jay Vosburgh169a3e62005-06-26 17:54:11 -0400161module_param(xmit_hash_policy, charp, 0);
Andy Gospodarek90e62472011-05-25 04:41:59 +0000162MODULE_PARM_DESC(xmit_hash_policy, "balance-xor and 802.3ad hashing method; "
163 "0 for layer 2 (default), 1 for layer 3+4, "
Nikolay Aleksandrov32819dc2013-10-02 13:39:25 +0200164 "2 for layer 2+3, 3 for encap layer 2+3, "
165 "4 for encap layer 3+4");
Linus Torvalds1da177e2005-04-16 15:20:36 -0700166module_param(arp_interval, int, 0);
167MODULE_PARM_DESC(arp_interval, "arp interval in milliseconds");
168module_param_array(arp_ip_target, charp, NULL, 0);
169MODULE_PARM_DESC(arp_ip_target, "arp targets in n.n.n.n form");
Jay Vosburghf5b2b962006-09-22 21:54:53 -0700170module_param(arp_validate, charp, 0);
Andy Gospodarek90e62472011-05-25 04:41:59 +0000171MODULE_PARM_DESC(arp_validate, "validate src/dst of ARP probes; "
172 "0 for none (default), 1 for active, "
173 "2 for backup, 3 for all");
Veaceslav Falico8599b522013-06-24 11:49:34 +0200174module_param(arp_all_targets, charp, 0);
175MODULE_PARM_DESC(arp_all_targets, "fail on any/all arp targets timeout; 0 for any (default), 1 for all");
Jay Vosburgh3915c1e82008-05-17 21:10:14 -0700176module_param(fail_over_mac, charp, 0);
Andy Gospodarek90e62472011-05-25 04:41:59 +0000177MODULE_PARM_DESC(fail_over_mac, "For active-backup, do not set all slaves to "
178 "the same MAC; 0 for none (default), "
179 "1 for active, 2 for follow");
Andy Gospodarekebd8e492010-06-02 08:39:21 +0000180module_param(all_slaves_active, int, 0);
Masanari Iida37b70212014-09-09 18:07:55 +0900181MODULE_PARM_DESC(all_slaves_active, "Keep all frames received on an interface "
Andy Gospodarek90e62472011-05-25 04:41:59 +0000182 "by setting active flag for all slaves; "
Andy Gospodarekebd8e492010-06-02 08:39:21 +0000183 "0 for never (default), 1 for always.");
Flavio Leitnerc2952c32010-10-05 14:23:59 +0000184module_param(resend_igmp, int, 0);
Andy Gospodarek90e62472011-05-25 04:41:59 +0000185MODULE_PARM_DESC(resend_igmp, "Number of IGMP membership reports to send on "
186 "link failure");
Nikolay Aleksandrov73958322013-11-05 13:51:41 +0100187module_param(packets_per_slave, int, 0);
188MODULE_PARM_DESC(packets_per_slave, "Packets to send per slave in balance-rr "
189 "mode; 0 for a random slave, 1 packet per "
190 "slave (default), >1 packets per slave.");
dingtianhong3a7129e2013-12-21 14:40:12 +0800191module_param(lp_interval, uint, 0);
192MODULE_PARM_DESC(lp_interval, "The number of seconds between instances where "
193 "the bonding driver sends learning packets to "
194 "each slaves peer switch. The default is 1.");
Linus Torvalds1da177e2005-04-16 15:20:36 -0700195
196/*----------------------------- Global variables ----------------------------*/
197
Neil Hormane843fa52010-10-13 16:01:50 +0000198#ifdef CONFIG_NET_POLL_CONTROLLER
Neil Hormanfb4fa762010-12-06 09:05:50 +0000199atomic_t netpoll_block_tx = ATOMIC_INIT(0);
Neil Hormane843fa52010-10-13 16:01:50 +0000200#endif
201
Eric Dumazetf99189b2009-11-17 10:42:49 +0000202int bond_net_id __read_mostly;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700203
Stephen Hemminger3d632c32009-06-12 19:02:48 +0000204static __be32 arp_target[BOND_MAX_ARP_TARGETS];
205static int arp_ip_count;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700206static int bond_mode = BOND_MODE_ROUNDROBIN;
Stephen Hemminger3d632c32009-06-12 19:02:48 +0000207static int xmit_hashtype = BOND_XMIT_POLICY_LAYER2;
208static int lacp_fast;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700209
Linus Torvalds1da177e2005-04-16 15:20:36 -0700210/*-------------------------- Forward declarations ---------------------------*/
211
Stephen Hemminger181470f2009-06-12 19:02:52 +0000212static int bond_init(struct net_device *bond_dev);
Eric W. Biedermanc67dfb22009-10-29 14:18:24 +0000213static void bond_uninit(struct net_device *bond_dev);
Andy Gospodarek5f0c5f72014-09-28 22:34:37 -0400214static struct rtnl_link_stats64 *bond_get_stats(struct net_device *bond_dev,
215 struct rtnl_link_stats64 *stats);
Mahesh Bandewaree637712014-10-04 17:45:01 -0700216static void bond_slave_arr_handler(struct work_struct *work);
Jay Vosburgh21a75f02016-02-02 13:35:56 -0800217static bool bond_time_in_interval(struct bonding *bond, unsigned long last_act,
218 int mod);
Mahesh Bandewar22ad2932018-09-24 14:40:11 -0700219static void bond_netdev_notify_work(struct work_struct *work);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700220
221/*---------------------------- General routines -----------------------------*/
222
Amerigo Wangbd33acc2011-03-06 21:58:46 +0000223const char *bond_mode_name(int mode)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700224{
Holger Eitzenberger77afc922008-12-09 23:08:09 -0800225 static const char *names[] = {
226 [BOND_MODE_ROUNDROBIN] = "load balancing (round-robin)",
227 [BOND_MODE_ACTIVEBACKUP] = "fault-tolerance (active-backup)",
228 [BOND_MODE_XOR] = "load balancing (xor)",
229 [BOND_MODE_BROADCAST] = "fault-tolerance (broadcast)",
Stephen Hemminger3d632c32009-06-12 19:02:48 +0000230 [BOND_MODE_8023AD] = "IEEE 802.3ad Dynamic link aggregation",
Holger Eitzenberger77afc922008-12-09 23:08:09 -0800231 [BOND_MODE_TLB] = "transmit load balancing",
232 [BOND_MODE_ALB] = "adaptive load balancing",
233 };
234
Wang Sheng-Huif5280942013-07-24 14:53:26 +0800235 if (mode < BOND_MODE_ROUNDROBIN || mode > BOND_MODE_ALB)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700236 return "unknown";
Holger Eitzenberger77afc922008-12-09 23:08:09 -0800237
238 return names[mode];
Linus Torvalds1da177e2005-04-16 15:20:36 -0700239}
240
241/*---------------------------------- VLAN -----------------------------------*/
242
243/**
Linus Torvalds1da177e2005-04-16 15:20:36 -0700244 * bond_dev_queue_xmit - Prepare skb for xmit.
Stephen Hemminger3d632c32009-06-12 19:02:48 +0000245 *
Linus Torvalds1da177e2005-04-16 15:20:36 -0700246 * @bond: bond device that got this skb for tx.
247 * @skb: hw accel VLAN tagged skb to transmit
248 * @slave_dev: slave that is supposed to xmit this skbuff
Linus Torvalds1da177e2005-04-16 15:20:36 -0700249 */
dingtianhongd316ded2014-01-02 09:13:09 +0800250void bond_dev_queue_xmit(struct bonding *bond, struct sk_buff *skb,
Stephen Hemminger3d632c32009-06-12 19:02:48 +0000251 struct net_device *slave_dev)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700252{
Ben Hutchings83874512010-12-13 08:19:28 +0000253 skb->dev = slave_dev;
Neil Horman374eeb52011-06-03 10:35:52 +0000254
Eric Dumazet5ee31c682012-06-12 06:03:51 +0000255 BUILD_BUG_ON(sizeof(skb->queue_mapping) !=
Jiri Pirkodf4ab5b2012-07-20 02:28:49 +0000256 sizeof(qdisc_skb_cb(skb)->slave_dev_queue_mapping));
257 skb->queue_mapping = qdisc_skb_cb(skb)->slave_dev_queue_mapping;
Neil Horman374eeb52011-06-03 10:35:52 +0000258
Amerigo Wange15c3c222012-08-10 01:24:45 +0000259 if (unlikely(netpoll_tx_running(bond->dev)))
Amerigo Wang8a8efa22011-02-17 23:43:32 +0000260 bond_netpoll_send_skb(bond_get_slave_by_dev(bond, slave_dev), skb);
Amerigo Wang080e4132011-02-17 23:43:33 +0000261 else
WANG Congf6dc31a2010-05-06 00:48:51 -0700262 dev_queue_xmit(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700263}
264
Nikolay Aleksandrov547942c2014-09-15 17:19:34 +0200265/* In the following 2 functions, bond_vlan_rx_add_vid and bond_vlan_rx_kill_vid,
Jiri Pirkocc0e4072011-07-20 04:54:46 +0000266 * We don't protect the slave list iteration with a lock because:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700267 * a. This operation is performed in IOCTL context,
268 * b. The operation is protected by the RTNL semaphore in the 8021q code,
269 * c. Holding a lock with BH disabled while directly calling a base driver
270 * entry point is generally a BAD idea.
Stephen Hemminger3d632c32009-06-12 19:02:48 +0000271 *
Linus Torvalds1da177e2005-04-16 15:20:36 -0700272 * The design of synchronization/protection for this operation in the 8021q
273 * module is good for one or more VLAN devices over a single physical device
274 * and cannot be extended for a teaming solution like bonding, so there is a
275 * potential race condition here where a net device from the vlan group might
276 * be referenced (either by a base driver or the 8021q code) while it is being
277 * removed from the system. However, it turns out we're not making matters
278 * worse, and if it works for regular VLAN usage it will work here too.
279*/
280
281/**
Linus Torvalds1da177e2005-04-16 15:20:36 -0700282 * bond_vlan_rx_add_vid - Propagates adding an id to slaves
283 * @bond_dev: bonding net device that got called
284 * @vid: vlan id being added
285 */
Patrick McHardy80d5c362013-04-19 02:04:28 +0000286static int bond_vlan_rx_add_vid(struct net_device *bond_dev,
287 __be16 proto, u16 vid)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700288{
Wang Chen454d7c92008-11-12 23:37:49 -0800289 struct bonding *bond = netdev_priv(bond_dev);
Veaceslav Falico81f23b12013-09-25 09:20:13 +0200290 struct slave *slave, *rollback_slave;
Veaceslav Falico9caff1e2013-09-25 09:20:14 +0200291 struct list_head *iter;
nikolay@redhat.comdec1e902013-08-01 16:54:47 +0200292 int res;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700293
Veaceslav Falico9caff1e2013-09-25 09:20:14 +0200294 bond_for_each_slave(bond, slave, iter) {
Patrick McHardy80d5c362013-04-19 02:04:28 +0000295 res = vlan_vid_add(slave->dev, proto, vid);
Jiri Pirko87002b02011-12-08 04:11:17 +0000296 if (res)
297 goto unwind;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700298 }
299
Jiri Pirko8e586132011-12-08 19:52:37 -0500300 return 0;
Jiri Pirko87002b02011-12-08 04:11:17 +0000301
302unwind:
Veaceslav Falico81f23b12013-09-25 09:20:13 +0200303 /* unwind to the slave that failed */
Veaceslav Falico9caff1e2013-09-25 09:20:14 +0200304 bond_for_each_slave(bond, rollback_slave, iter) {
Veaceslav Falico81f23b12013-09-25 09:20:13 +0200305 if (rollback_slave == slave)
306 break;
307
308 vlan_vid_del(rollback_slave->dev, proto, vid);
309 }
Jiri Pirko87002b02011-12-08 04:11:17 +0000310
311 return res;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700312}
313
314/**
315 * bond_vlan_rx_kill_vid - Propagates deleting an id to slaves
316 * @bond_dev: bonding net device that got called
317 * @vid: vlan id being removed
318 */
Patrick McHardy80d5c362013-04-19 02:04:28 +0000319static int bond_vlan_rx_kill_vid(struct net_device *bond_dev,
320 __be16 proto, u16 vid)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700321{
Wang Chen454d7c92008-11-12 23:37:49 -0800322 struct bonding *bond = netdev_priv(bond_dev);
Veaceslav Falico9caff1e2013-09-25 09:20:14 +0200323 struct list_head *iter;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700324 struct slave *slave;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700325
Veaceslav Falico9caff1e2013-09-25 09:20:14 +0200326 bond_for_each_slave(bond, slave, iter)
Patrick McHardy80d5c362013-04-19 02:04:28 +0000327 vlan_vid_del(slave->dev, proto, vid);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700328
Veaceslav Falicoe868b0c2013-08-28 23:25:15 +0200329 if (bond_is_lb(bond))
330 bond_alb_clear_vlan(bond, vid);
Jiri Pirko8e586132011-12-08 19:52:37 -0500331
332 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700333}
334
Linus Torvalds1da177e2005-04-16 15:20:36 -0700335/*------------------------------- Link status -------------------------------*/
336
Nikolay Aleksandrov547942c2014-09-15 17:19:34 +0200337/* Set the carrier state for the master according to the state of its
Jay Vosburghff59c452006-03-27 13:27:43 -0800338 * slaves. If any slaves are up, the master is up. In 802.3ad mode,
339 * do special 802.3ad magic.
340 *
341 * Returns zero if carrier state does not change, nonzero if it does.
342 */
Jonathan Toppins2477bc92015-01-26 01:16:57 -0500343int bond_set_carrier(struct bonding *bond)
Jay Vosburghff59c452006-03-27 13:27:43 -0800344{
Veaceslav Falico9caff1e2013-09-25 09:20:14 +0200345 struct list_head *iter;
Jay Vosburghff59c452006-03-27 13:27:43 -0800346 struct slave *slave;
Jay Vosburghff59c452006-03-27 13:27:43 -0800347
Veaceslav Falico0965a1f2013-09-25 09:20:21 +0200348 if (!bond_has_slaves(bond))
Jay Vosburghff59c452006-03-27 13:27:43 -0800349 goto down;
350
Veaceslav Falico01844092014-05-15 21:39:55 +0200351 if (BOND_MODE(bond) == BOND_MODE_8023AD)
Jay Vosburghff59c452006-03-27 13:27:43 -0800352 return bond_3ad_set_carrier(bond);
353
Veaceslav Falico9caff1e2013-09-25 09:20:14 +0200354 bond_for_each_slave(bond, slave, iter) {
Jay Vosburghff59c452006-03-27 13:27:43 -0800355 if (slave->link == BOND_LINK_UP) {
356 if (!netif_carrier_ok(bond->dev)) {
357 netif_carrier_on(bond->dev);
358 return 1;
359 }
360 return 0;
361 }
362 }
363
364down:
365 if (netif_carrier_ok(bond->dev)) {
366 netif_carrier_off(bond->dev);
367 return 1;
368 }
369 return 0;
370}
371
Nikolay Aleksandrov547942c2014-09-15 17:19:34 +0200372/* Get link speed and duplex from the slave's base driver
Linus Torvalds1da177e2005-04-16 15:20:36 -0700373 * using ethtool. If for some reason the call fails or the
Weiping Pan98f41f62011-10-31 17:20:48 +0000374 * values are invalid, set speed and duplex to -1,
Mahesh Bandewar851eb312017-03-27 11:37:35 -0700375 * and return. Return 1 if speed or duplex settings are
376 * UNKNOWN; 0 otherwise.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700377 */
Mahesh Bandewar851eb312017-03-27 11:37:35 -0700378static int bond_update_speed_duplex(struct slave *slave)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700379{
380 struct net_device *slave_dev = slave->dev;
David Decotigny98569092016-02-24 10:58:02 -0800381 struct ethtool_link_ksettings ecmd;
Matthew Wilcox61a44b92007-07-31 14:00:02 -0700382 int res;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700383
Dan Carpenter589665f2011-11-04 08:21:38 +0000384 slave->speed = SPEED_UNKNOWN;
385 slave->duplex = DUPLEX_UNKNOWN;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700386
David Decotigny98569092016-02-24 10:58:02 -0800387 res = __ethtool_get_link_ksettings(slave_dev, &ecmd);
Mahesh Bandewarbc5ad402017-04-03 18:38:39 -0700388 if (res < 0)
Mahesh Bandewar851eb312017-03-27 11:37:35 -0700389 return 1;
Mahesh Bandewarbc5ad402017-04-03 18:38:39 -0700390 if (ecmd.base.speed == 0 || ecmd.base.speed == ((__u32)-1))
Mahesh Bandewar851eb312017-03-27 11:37:35 -0700391 return 1;
David Decotigny98569092016-02-24 10:58:02 -0800392 switch (ecmd.base.duplex) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700393 case DUPLEX_FULL:
394 case DUPLEX_HALF:
395 break;
396 default:
Mahesh Bandewar851eb312017-03-27 11:37:35 -0700397 return 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700398 }
399
David Decotigny98569092016-02-24 10:58:02 -0800400 slave->speed = ecmd.base.speed;
401 slave->duplex = ecmd.base.duplex;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700402
Mahesh Bandewar851eb312017-03-27 11:37:35 -0700403 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700404}
405
sfeldma@cumulusnetworks.com07699f92014-01-16 22:57:49 -0800406const char *bond_slave_link_status(s8 link)
407{
408 switch (link) {
409 case BOND_LINK_UP:
410 return "up";
411 case BOND_LINK_FAIL:
412 return "going down";
413 case BOND_LINK_DOWN:
414 return "down";
415 case BOND_LINK_BACK:
416 return "going back";
417 default:
418 return "unknown";
419 }
420}
421
Nikolay Aleksandrov547942c2014-09-15 17:19:34 +0200422/* if <dev> supports MII link status reporting, check its link status.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700423 *
424 * We either do MII/ETHTOOL ioctls, or check netif_carrier_ok(),
Stephen Hemminger3d632c32009-06-12 19:02:48 +0000425 * depending upon the setting of the use_carrier parameter.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700426 *
427 * Return either BMSR_LSTATUS, meaning that the link is up (or we
428 * can't tell and just pretend it is), or 0, meaning that the link is
429 * down.
430 *
431 * If reporting is non-zero, instead of faking link up, return -1 if
432 * both ETHTOOL and MII ioctls fail (meaning the device does not
433 * support them). If use_carrier is set, return whatever it says.
434 * It'd be nice if there was a good way to tell if a driver supports
435 * netif_carrier, but there really isn't.
436 */
Stephen Hemminger3d632c32009-06-12 19:02:48 +0000437static int bond_check_dev_link(struct bonding *bond,
438 struct net_device *slave_dev, int reporting)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700439{
Stephen Hemmingereb7cc592008-11-19 21:56:05 -0800440 const struct net_device_ops *slave_ops = slave_dev->netdev_ops;
Jiri Bohacd9d52832009-10-28 22:23:54 -0700441 int (*ioctl)(struct net_device *, struct ifreq *, int);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700442 struct ifreq ifr;
443 struct mii_ioctl_data *mii;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700444
Petri Gynther6c988852009-08-28 12:05:15 +0000445 if (!reporting && !netif_running(slave_dev))
446 return 0;
447
Stephen Hemmingereb7cc592008-11-19 21:56:05 -0800448 if (bond->params.use_carrier)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700449 return netif_carrier_ok(slave_dev) ? BMSR_LSTATUS : 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700450
Jiri Pirko29112f42009-04-24 01:58:23 +0000451 /* Try to get link status using Ethtool first. */
Ben Hutchingsc772dde2012-12-07 06:15:32 +0000452 if (slave_dev->ethtool_ops->get_link)
453 return slave_dev->ethtool_ops->get_link(slave_dev) ?
454 BMSR_LSTATUS : 0;
Jiri Pirko29112f42009-04-24 01:58:23 +0000455
Stephen Hemminger3d632c32009-06-12 19:02:48 +0000456 /* Ethtool can't be used, fallback to MII ioctls. */
Stephen Hemmingereb7cc592008-11-19 21:56:05 -0800457 ioctl = slave_ops->ndo_do_ioctl;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700458 if (ioctl) {
Nikolay Aleksandrov547942c2014-09-15 17:19:34 +0200459 /* TODO: set pointer to correct ioctl on a per team member
460 * bases to make this more efficient. that is, once
461 * we determine the correct ioctl, we will always
462 * call it and not the others for that team
463 * member.
464 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700465
Nikolay Aleksandrov547942c2014-09-15 17:19:34 +0200466 /* We cannot assume that SIOCGMIIPHY will also read a
Linus Torvalds1da177e2005-04-16 15:20:36 -0700467 * register; not all network drivers (e.g., e100)
468 * support that.
469 */
470
471 /* Yes, the mii is overlaid on the ifreq.ifr_ifru */
472 strncpy(ifr.ifr_name, slave_dev->name, IFNAMSIZ);
473 mii = if_mii(&ifr);
Al Viro4ad41c12016-09-03 19:37:25 -0400474 if (ioctl(slave_dev, &ifr, SIOCGMIIPHY) == 0) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700475 mii->reg_num = MII_BMSR;
Al Viro4ad41c12016-09-03 19:37:25 -0400476 if (ioctl(slave_dev, &ifr, SIOCGMIIREG) == 0)
Stephen Hemminger3d632c32009-06-12 19:02:48 +0000477 return mii->val_out & BMSR_LSTATUS;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700478 }
479 }
480
Nikolay Aleksandrov547942c2014-09-15 17:19:34 +0200481 /* If reporting, report that either there's no dev->do_ioctl,
Matthew Wilcox61a44b92007-07-31 14:00:02 -0700482 * or both SIOCGMIIREG and get_link failed (meaning that we
Linus Torvalds1da177e2005-04-16 15:20:36 -0700483 * cannot report link status). If not reporting, pretend
484 * we're ok.
485 */
Stephen Hemminger3d632c32009-06-12 19:02:48 +0000486 return reporting ? -1 : BMSR_LSTATUS;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700487}
488
489/*----------------------------- Multicast list ------------------------------*/
490
Nikolay Aleksandrov547942c2014-09-15 17:19:34 +0200491/* Push the promiscuity flag down to appropriate slaves */
Wang Chen7e1a1ac2008-07-14 20:51:36 -0700492static int bond_set_promiscuity(struct bonding *bond, int inc)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700493{
Veaceslav Falico9caff1e2013-09-25 09:20:14 +0200494 struct list_head *iter;
Wang Chen7e1a1ac2008-07-14 20:51:36 -0700495 int err = 0;
Veaceslav Falico9caff1e2013-09-25 09:20:14 +0200496
Veaceslav Falicoec0865a2014-05-15 21:39:54 +0200497 if (bond_uses_primary(bond)) {
Veaceslav Falico14056e72014-07-16 18:32:01 +0200498 struct slave *curr_active = rtnl_dereference(bond->curr_active_slave);
Eric Dumazet4740d632014-07-15 06:56:55 -0700499
Eric Dumazet4740d632014-07-15 06:56:55 -0700500 if (curr_active)
501 err = dev_set_promiscuity(curr_active->dev, inc);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700502 } else {
503 struct slave *slave;
nikolay@redhat.comdec1e902013-08-01 16:54:47 +0200504
Veaceslav Falico9caff1e2013-09-25 09:20:14 +0200505 bond_for_each_slave(bond, slave, iter) {
Wang Chen7e1a1ac2008-07-14 20:51:36 -0700506 err = dev_set_promiscuity(slave->dev, inc);
507 if (err)
508 return err;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700509 }
510 }
Wang Chen7e1a1ac2008-07-14 20:51:36 -0700511 return err;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700512}
513
Nikolay Aleksandrov547942c2014-09-15 17:19:34 +0200514/* Push the allmulti flag down to all slaves */
Wang Chen7e1a1ac2008-07-14 20:51:36 -0700515static int bond_set_allmulti(struct bonding *bond, int inc)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700516{
Veaceslav Falico9caff1e2013-09-25 09:20:14 +0200517 struct list_head *iter;
Wang Chen7e1a1ac2008-07-14 20:51:36 -0700518 int err = 0;
Veaceslav Falico9caff1e2013-09-25 09:20:14 +0200519
Veaceslav Falicoec0865a2014-05-15 21:39:54 +0200520 if (bond_uses_primary(bond)) {
Veaceslav Falico14056e72014-07-16 18:32:01 +0200521 struct slave *curr_active = rtnl_dereference(bond->curr_active_slave);
Eric Dumazet4740d632014-07-15 06:56:55 -0700522
Eric Dumazet4740d632014-07-15 06:56:55 -0700523 if (curr_active)
524 err = dev_set_allmulti(curr_active->dev, inc);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700525 } else {
526 struct slave *slave;
nikolay@redhat.comdec1e902013-08-01 16:54:47 +0200527
Veaceslav Falico9caff1e2013-09-25 09:20:14 +0200528 bond_for_each_slave(bond, slave, iter) {
Wang Chen7e1a1ac2008-07-14 20:51:36 -0700529 err = dev_set_allmulti(slave->dev, inc);
530 if (err)
531 return err;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700532 }
533 }
Wang Chen7e1a1ac2008-07-14 20:51:36 -0700534 return err;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700535}
536
Nikolay Aleksandrov547942c2014-09-15 17:19:34 +0200537/* Retrieve the list of registered multicast addresses for the bonding
Jay Vosburgha816c7c2007-02-28 17:03:37 -0800538 * device and retransmit an IGMP JOIN request to the current active
539 * slave.
540 */
stephen hemminger379b7382010-10-15 11:02:56 +0000541static void bond_resend_igmp_join_requests_delayed(struct work_struct *work)
Flavio Leitner5a37e8c2010-10-05 14:23:57 +0000542{
543 struct bonding *bond = container_of(work, struct bonding,
Flavio Leitner94265cf2011-05-25 08:38:58 +0000544 mcast_work.work);
Veaceslav Falicoad999ee2013-03-26 04:10:02 +0000545
dingtianhongf2369102013-12-13 10:20:26 +0800546 if (!rtnl_trylock()) {
547 queue_delayed_work(bond->wq, &bond->mcast_work, 1);
548 return;
549 }
550 call_netdevice_notifiers(NETDEV_RESEND_IGMP, bond->dev);
551
552 if (bond->igmp_retrans > 1) {
553 bond->igmp_retrans--;
554 queue_delayed_work(bond->wq, &bond->mcast_work, HZ/5);
555 }
556 rtnl_unlock();
Jay Vosburgha816c7c2007-02-28 17:03:37 -0800557}
558
Nikolay Aleksandrov547942c2014-09-15 17:19:34 +0200559/* Flush bond's hardware addresses from slave */
Jay Vosburgh303d1cb2013-05-31 11:57:30 +0000560static void bond_hw_addr_flush(struct net_device *bond_dev,
Stephen Hemminger3d632c32009-06-12 19:02:48 +0000561 struct net_device *slave_dev)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700562{
Wang Chen454d7c92008-11-12 23:37:49 -0800563 struct bonding *bond = netdev_priv(bond_dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700564
Jay Vosburgh303d1cb2013-05-31 11:57:30 +0000565 dev_uc_unsync(slave_dev, bond_dev);
566 dev_mc_unsync(slave_dev, bond_dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700567
Veaceslav Falico01844092014-05-15 21:39:55 +0200568 if (BOND_MODE(bond) == BOND_MODE_8023AD) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700569 /* del lacpdu mc addr from mc list */
570 u8 lacpdu_multicast[ETH_ALEN] = MULTICAST_LACPDU_ADDR;
571
Jiri Pirko22bedad32010-04-01 21:22:57 +0000572 dev_mc_del(slave_dev, lacpdu_multicast);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700573 }
574}
575
576/*--------------------------- Active slave change ---------------------------*/
577
Jay Vosburgh303d1cb2013-05-31 11:57:30 +0000578/* Update the hardware address list and promisc/allmulti for the new and
Veaceslav Falicoec0865a2014-05-15 21:39:54 +0200579 * old active slaves (if any). Modes that are not using primary keep all
580 * slaves up date at all times; only the modes that use primary need to call
Jay Vosburgh303d1cb2013-05-31 11:57:30 +0000581 * this function to swap these settings during a failover.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700582 */
Jay Vosburgh303d1cb2013-05-31 11:57:30 +0000583static void bond_hw_addr_swap(struct bonding *bond, struct slave *new_active,
584 struct slave *old_active)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700585{
Linus Torvalds1da177e2005-04-16 15:20:36 -0700586 if (old_active) {
Stephen Hemminger3d632c32009-06-12 19:02:48 +0000587 if (bond->dev->flags & IFF_PROMISC)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700588 dev_set_promiscuity(old_active->dev, -1);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700589
Stephen Hemminger3d632c32009-06-12 19:02:48 +0000590 if (bond->dev->flags & IFF_ALLMULTI)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700591 dev_set_allmulti(old_active->dev, -1);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700592
Jay Vosburgh303d1cb2013-05-31 11:57:30 +0000593 bond_hw_addr_flush(bond->dev, old_active->dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700594 }
595
596 if (new_active) {
Wang Chen7e1a1ac2008-07-14 20:51:36 -0700597 /* FIXME: Signal errors upstream. */
Stephen Hemminger3d632c32009-06-12 19:02:48 +0000598 if (bond->dev->flags & IFF_PROMISC)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700599 dev_set_promiscuity(new_active->dev, 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700600
Stephen Hemminger3d632c32009-06-12 19:02:48 +0000601 if (bond->dev->flags & IFF_ALLMULTI)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700602 dev_set_allmulti(new_active->dev, 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700603
nikolay@redhat.comd632ce92013-04-18 07:33:38 +0000604 netif_addr_lock_bh(bond->dev);
Jay Vosburgh303d1cb2013-05-31 11:57:30 +0000605 dev_uc_sync(new_active->dev, bond->dev);
606 dev_mc_sync(new_active->dev, bond->dev);
nikolay@redhat.comd632ce92013-04-18 07:33:38 +0000607 netif_addr_unlock_bh(bond->dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700608 }
609}
610
nikolay@redhat.comae0d6752013-06-26 17:13:39 +0200611/**
612 * bond_set_dev_addr - clone slave's address to bond
613 * @bond_dev: bond net device
614 * @slave_dev: slave net device
615 *
616 * Should be called with RTNL held.
617 */
618static void bond_set_dev_addr(struct net_device *bond_dev,
619 struct net_device *slave_dev)
620{
Zhang Shengjuc6140a22016-02-02 08:32:55 +0000621 netdev_dbg(bond_dev, "bond_dev=%p slave_dev=%p slave_dev->name=%s slave_dev->addr_len=%d\n",
622 bond_dev, slave_dev, slave_dev->name, slave_dev->addr_len);
nikolay@redhat.comae0d6752013-06-26 17:13:39 +0200623 memcpy(bond_dev->dev_addr, slave_dev->dev_addr, slave_dev->addr_len);
624 bond_dev->addr_assign_type = NET_ADDR_STOLEN;
625 call_netdevice_notifiers(NETDEV_CHANGEADDR, bond_dev);
626}
627
dingtianhonga951bc12015-07-16 16:30:02 +0800628static struct slave *bond_get_old_active(struct bonding *bond,
629 struct slave *new_active)
630{
631 struct slave *slave;
632 struct list_head *iter;
633
634 bond_for_each_slave(bond, slave, iter) {
635 if (slave == new_active)
636 continue;
637
638 if (ether_addr_equal(bond->dev->dev_addr, slave->dev->dev_addr))
639 return slave;
640 }
641
642 return NULL;
643}
644
Nikolay Aleksandrov547942c2014-09-15 17:19:34 +0200645/* bond_do_fail_over_mac
Jay Vosburgh3915c1e82008-05-17 21:10:14 -0700646 *
647 * Perform special MAC address swapping for fail_over_mac settings
648 *
Nikolay Aleksandrov1c72cfdc92014-09-11 22:49:24 +0200649 * Called with RTNL
Jay Vosburgh3915c1e82008-05-17 21:10:14 -0700650 */
651static void bond_do_fail_over_mac(struct bonding *bond,
652 struct slave *new_active,
653 struct slave *old_active)
654{
655 u8 tmp_mac[ETH_ALEN];
656 struct sockaddr saddr;
657 int rv;
658
659 switch (bond->params.fail_over_mac) {
660 case BOND_FOM_ACTIVE:
Nikolay Aleksandrov1c72cfdc92014-09-11 22:49:24 +0200661 if (new_active)
nikolay@redhat.comae0d6752013-06-26 17:13:39 +0200662 bond_set_dev_addr(bond->dev, new_active->dev);
Jay Vosburgh3915c1e82008-05-17 21:10:14 -0700663 break;
664 case BOND_FOM_FOLLOW:
Nikolay Aleksandrov547942c2014-09-15 17:19:34 +0200665 /* if new_active && old_active, swap them
Jay Vosburgh3915c1e82008-05-17 21:10:14 -0700666 * if just old_active, do nothing (going to no active slave)
667 * if just new_active, set new_active to bond's MAC
668 */
669 if (!new_active)
670 return;
671
dingtianhonga951bc12015-07-16 16:30:02 +0800672 if (!old_active)
673 old_active = bond_get_old_active(bond, new_active);
674
Jay Vosburgh3915c1e82008-05-17 21:10:14 -0700675 if (old_active) {
Joe Perchesada0f862014-02-15 16:02:17 -0800676 ether_addr_copy(tmp_mac, new_active->dev->dev_addr);
Joe Perches2a7c1832014-02-18 09:42:45 -0800677 ether_addr_copy(saddr.sa_data,
678 old_active->dev->dev_addr);
Jay Vosburgh3915c1e82008-05-17 21:10:14 -0700679 saddr.sa_family = new_active->dev->type;
680 } else {
Joe Perchesada0f862014-02-15 16:02:17 -0800681 ether_addr_copy(saddr.sa_data, bond->dev->dev_addr);
Jay Vosburgh3915c1e82008-05-17 21:10:14 -0700682 saddr.sa_family = bond->dev->type;
683 }
684
685 rv = dev_set_mac_address(new_active->dev, &saddr);
686 if (rv) {
Veaceslav Falico76444f52014-07-15 19:35:58 +0200687 netdev_err(bond->dev, "Error %d setting MAC of slave %s\n",
688 -rv, new_active->dev->name);
Jay Vosburgh3915c1e82008-05-17 21:10:14 -0700689 goto out;
690 }
691
692 if (!old_active)
693 goto out;
694
Joe Perchesada0f862014-02-15 16:02:17 -0800695 ether_addr_copy(saddr.sa_data, tmp_mac);
Jay Vosburgh3915c1e82008-05-17 21:10:14 -0700696 saddr.sa_family = old_active->dev->type;
697
698 rv = dev_set_mac_address(old_active->dev, &saddr);
699 if (rv)
Veaceslav Falico76444f52014-07-15 19:35:58 +0200700 netdev_err(bond->dev, "Error %d setting MAC of slave %s\n",
701 -rv, new_active->dev->name);
Jay Vosburgh3915c1e82008-05-17 21:10:14 -0700702out:
Jay Vosburgh3915c1e82008-05-17 21:10:14 -0700703 break;
704 default:
Veaceslav Falico76444f52014-07-15 19:35:58 +0200705 netdev_err(bond->dev, "bond_do_fail_over_mac impossible: bad policy %d\n",
706 bond->params.fail_over_mac);
Jay Vosburgh3915c1e82008-05-17 21:10:14 -0700707 break;
708 }
709
710}
711
Mazhar Ranab5a983f2015-07-07 15:04:50 +0530712static struct slave *bond_choose_primary_or_current(struct bonding *bond)
Jiri Pirkoa5499522009-09-25 03:28:09 +0000713{
Nikolay Aleksandrov059b47e2014-09-09 23:17:00 +0200714 struct slave *prim = rtnl_dereference(bond->primary_slave);
Nikolay Aleksandrov1c72cfdc92014-09-11 22:49:24 +0200715 struct slave *curr = rtnl_dereference(bond->curr_active_slave);
Jiri Pirkoa5499522009-09-25 03:28:09 +0000716
Mazhar Ranab5a983f2015-07-07 15:04:50 +0530717 if (!prim || prim->link != BOND_LINK_UP) {
718 if (!curr || curr->link != BOND_LINK_UP)
719 return NULL;
720 return curr;
721 }
722
Jiri Pirkoa5499522009-09-25 03:28:09 +0000723 if (bond->force_primary) {
724 bond->force_primary = false;
Mazhar Ranab5a983f2015-07-07 15:04:50 +0530725 return prim;
Jiri Pirkoa5499522009-09-25 03:28:09 +0000726 }
Mazhar Ranab5a983f2015-07-07 15:04:50 +0530727
728 if (!curr || curr->link != BOND_LINK_UP)
729 return prim;
730
731 /* At this point, prim and curr are both up */
732 switch (bond->params.primary_reselect) {
733 case BOND_PRI_RESELECT_ALWAYS:
734 return prim;
735 case BOND_PRI_RESELECT_BETTER:
736 if (prim->speed < curr->speed)
737 return curr;
738 if (prim->speed == curr->speed && prim->duplex <= curr->duplex)
739 return curr;
740 return prim;
741 case BOND_PRI_RESELECT_FAILURE:
742 return curr;
743 default:
744 netdev_err(bond->dev, "impossible primary_reselect %d\n",
745 bond->params.primary_reselect);
746 return curr;
747 }
Jiri Pirkoa5499522009-09-25 03:28:09 +0000748}
Jay Vosburgh3915c1e82008-05-17 21:10:14 -0700749
Linus Torvalds1da177e2005-04-16 15:20:36 -0700750/**
Mazhar Ranab5a983f2015-07-07 15:04:50 +0530751 * bond_find_best_slave - select the best available slave to be the active one
Linus Torvalds1da177e2005-04-16 15:20:36 -0700752 * @bond: our bonding struct
Linus Torvalds1da177e2005-04-16 15:20:36 -0700753 */
754static struct slave *bond_find_best_slave(struct bonding *bond)
755{
Mazhar Ranab5a983f2015-07-07 15:04:50 +0530756 struct slave *slave, *bestslave = NULL;
Veaceslav Falico77140d22013-09-25 09:20:18 +0200757 struct list_head *iter;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700758 int mintime = bond->params.updelay;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700759
Mazhar Ranab5a983f2015-07-07 15:04:50 +0530760 slave = bond_choose_primary_or_current(bond);
761 if (slave)
762 return slave;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700763
Veaceslav Falico77140d22013-09-25 09:20:18 +0200764 bond_for_each_slave(bond, slave, iter) {
765 if (slave->link == BOND_LINK_UP)
766 return slave;
Veaceslav Falicob6adc612014-05-15 21:39:57 +0200767 if (slave->link == BOND_LINK_BACK && bond_slave_is_up(slave) &&
Veaceslav Falico77140d22013-09-25 09:20:18 +0200768 slave->delay < mintime) {
769 mintime = slave->delay;
770 bestslave = slave;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700771 }
772 }
773
774 return bestslave;
775}
776
Ben Hutchingsad246c92011-04-26 15:25:52 +0000777static bool bond_should_notify_peers(struct bonding *bond)
778{
dingtianhong4cb4f972013-12-13 10:19:39 +0800779 struct slave *slave;
780
781 rcu_read_lock();
782 slave = rcu_dereference(bond->curr_active_slave);
783 rcu_read_unlock();
Ben Hutchingsad246c92011-04-26 15:25:52 +0000784
Veaceslav Falico76444f52014-07-15 19:35:58 +0200785 netdev_dbg(bond->dev, "bond_should_notify_peers: slave %s\n",
786 slave ? slave->dev->name : "NULL");
Ben Hutchingsad246c92011-04-26 15:25:52 +0000787
788 if (!slave || !bond->send_peer_notif ||
Venkat Venkatsubrab02e3e92015-08-11 07:57:23 -0700789 !netif_carrier_ok(bond->dev) ||
Ben Hutchingsad246c92011-04-26 15:25:52 +0000790 test_bit(__LINK_STATE_LINKWATCH_PENDING, &slave->dev->state))
791 return false;
792
Ben Hutchingsad246c92011-04-26 15:25:52 +0000793 return true;
794}
795
Linus Torvalds1da177e2005-04-16 15:20:36 -0700796/**
797 * change_active_interface - change the active slave into the specified one
798 * @bond: our bonding struct
799 * @new: the new slave to make the active one
800 *
801 * Set the new slave to the bond's settings and unset them on the old
802 * curr_active_slave.
803 * Setting include flags, mc-list, promiscuity, allmulti, etc.
804 *
805 * If @new's link state is %BOND_LINK_BACK we'll set it to %BOND_LINK_UP,
806 * because it is apparently the best available slave we have, even though its
807 * updelay hasn't timed out yet.
808 *
Nikolay Aleksandrov1c72cfdc92014-09-11 22:49:24 +0200809 * Caller must hold RTNL.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700810 */
Mitch Williamsa77b5322005-11-09 10:35:51 -0800811void bond_change_active_slave(struct bonding *bond, struct slave *new_active)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700812{
Eric Dumazet4740d632014-07-15 06:56:55 -0700813 struct slave *old_active;
814
Nikolay Aleksandrov1c72cfdc92014-09-11 22:49:24 +0200815 ASSERT_RTNL();
816
817 old_active = rtnl_dereference(bond->curr_active_slave);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700818
Stephen Hemminger3d632c32009-06-12 19:02:48 +0000819 if (old_active == new_active)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700820 return;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700821
822 if (new_active) {
Veaceslav Falico8e603462014-02-18 07:48:46 +0100823 new_active->last_link_up = jiffies;
Jay Vosburghb2220ca2008-05-17 21:10:13 -0700824
Linus Torvalds1da177e2005-04-16 15:20:36 -0700825 if (new_active->link == BOND_LINK_BACK) {
Veaceslav Falicoec0865a2014-05-15 21:39:54 +0200826 if (bond_uses_primary(bond)) {
Veaceslav Falico76444f52014-07-15 19:35:58 +0200827 netdev_info(bond->dev, "making interface %s the new active one %d ms earlier\n",
828 new_active->dev->name,
829 (bond->params.updelay - new_active->delay) * bond->params.miimon);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700830 }
831
832 new_active->delay = 0;
Jiri Pirko5d397062015-12-03 12:12:19 +0100833 bond_set_slave_link_state(new_active, BOND_LINK_UP,
834 BOND_SLAVE_NOTIFY_NOW);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700835
Veaceslav Falico01844092014-05-15 21:39:55 +0200836 if (BOND_MODE(bond) == BOND_MODE_8023AD)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700837 bond_3ad_handle_link_change(new_active, BOND_LINK_UP);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700838
Holger Eitzenberger58402052008-12-09 23:07:13 -0800839 if (bond_is_lb(bond))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700840 bond_alb_handle_link_change(bond, new_active, BOND_LINK_UP);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700841 } else {
Veaceslav Falicoec0865a2014-05-15 21:39:54 +0200842 if (bond_uses_primary(bond)) {
Veaceslav Falico76444f52014-07-15 19:35:58 +0200843 netdev_info(bond->dev, "making interface %s the new active one\n",
844 new_active->dev->name);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700845 }
846 }
847 }
848
Veaceslav Falicoec0865a2014-05-15 21:39:54 +0200849 if (bond_uses_primary(bond))
Jay Vosburgh303d1cb2013-05-31 11:57:30 +0000850 bond_hw_addr_swap(bond, new_active, old_active);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700851
Holger Eitzenberger58402052008-12-09 23:07:13 -0800852 if (bond_is_lb(bond)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700853 bond_alb_handle_active_change(bond, new_active);
Jay Vosburgh8f903c72006-02-21 16:36:44 -0800854 if (old_active)
dingtianhong5e5b0662014-02-26 11:05:22 +0800855 bond_set_slave_inactive_flags(old_active,
856 BOND_SLAVE_NOTIFY_NOW);
Jay Vosburgh8f903c72006-02-21 16:36:44 -0800857 if (new_active)
dingtianhong5e5b0662014-02-26 11:05:22 +0800858 bond_set_slave_active_flags(new_active,
859 BOND_SLAVE_NOTIFY_NOW);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700860 } else {
nikolay@redhat.com278b2082013-08-01 16:54:51 +0200861 rcu_assign_pointer(bond->curr_active_slave, new_active);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700862 }
Jay Vosburghc3ade5c2005-06-26 17:52:20 -0400863
Veaceslav Falico01844092014-05-15 21:39:55 +0200864 if (BOND_MODE(bond) == BOND_MODE_ACTIVEBACKUP) {
Stephen Hemminger3d632c32009-06-12 19:02:48 +0000865 if (old_active)
dingtianhong5e5b0662014-02-26 11:05:22 +0800866 bond_set_slave_inactive_flags(old_active,
867 BOND_SLAVE_NOTIFY_NOW);
Jay Vosburghc3ade5c2005-06-26 17:52:20 -0400868
869 if (new_active) {
Ben Hutchingsad246c92011-04-26 15:25:52 +0000870 bool should_notify_peers = false;
871
dingtianhong5e5b0662014-02-26 11:05:22 +0800872 bond_set_slave_active_flags(new_active,
873 BOND_SLAVE_NOTIFY_NOW);
Moni Shoua2ab82852007-10-09 19:43:39 -0700874
Or Gerlitz709f8a42008-06-13 18:12:01 -0700875 if (bond->params.fail_over_mac)
876 bond_do_fail_over_mac(bond, new_active,
877 old_active);
Jay Vosburgh3915c1e82008-05-17 21:10:14 -0700878
Ben Hutchingsad246c92011-04-26 15:25:52 +0000879 if (netif_running(bond->dev)) {
880 bond->send_peer_notif =
881 bond->params.num_peer_notif;
882 should_notify_peers =
883 bond_should_notify_peers(bond);
884 }
885
Amerigo Wangb7bc2a52012-08-09 22:14:57 +0000886 call_netdevice_notifiers(NETDEV_BONDING_FAILOVER, bond->dev);
Ben Hutchingsad246c92011-04-26 15:25:52 +0000887 if (should_notify_peers)
Amerigo Wangb7bc2a52012-08-09 22:14:57 +0000888 call_netdevice_notifiers(NETDEV_NOTIFY_PEERS,
889 bond->dev);
Moni Shoua7893b242008-05-17 21:10:12 -0700890 }
Jay Vosburghc3ade5c2005-06-26 17:52:20 -0400891 }
Andy Gospodareka2fd9402010-03-25 14:49:05 +0000892
Flavio Leitner5a37e8c2010-10-05 14:23:57 +0000893 /* resend IGMP joins since active slave has changed or
Flavio Leitner94265cf2011-05-25 08:38:58 +0000894 * all were sent on curr_active_slave.
895 * resend only if bond is brought up with the affected
Nikolay Aleksandrov547942c2014-09-15 17:19:34 +0200896 * bonding modes and the retransmission is enabled
897 */
Flavio Leitner94265cf2011-05-25 08:38:58 +0000898 if (netif_running(bond->dev) && (bond->params.resend_igmp > 0) &&
Veaceslav Falicoec0865a2014-05-15 21:39:54 +0200899 ((bond_uses_primary(bond) && new_active) ||
Veaceslav Falico01844092014-05-15 21:39:55 +0200900 BOND_MODE(bond) == BOND_MODE_ROUNDROBIN)) {
Flavio Leitnerc2952c32010-10-05 14:23:59 +0000901 bond->igmp_retrans = bond->params.resend_igmp;
Nikolay Aleksandrov4beac022013-08-01 11:51:42 +0200902 queue_delayed_work(bond->wq, &bond->mcast_work, 1);
Andy Gospodareka2fd9402010-03-25 14:49:05 +0000903 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700904}
905
906/**
907 * bond_select_active_slave - select a new active slave, if needed
908 * @bond: our bonding struct
909 *
Stephen Hemminger3d632c32009-06-12 19:02:48 +0000910 * This functions should be called when one of the following occurs:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700911 * - The old curr_active_slave has been released or lost its link.
912 * - The primary_slave has got its link back.
913 * - A slave has got its link back and there's no old curr_active_slave.
914 *
Nikolay Aleksandrov1c72cfdc92014-09-11 22:49:24 +0200915 * Caller must hold RTNL.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700916 */
Mitch Williamsa77b5322005-11-09 10:35:51 -0800917void bond_select_active_slave(struct bonding *bond)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700918{
919 struct slave *best_slave;
Jay Vosburghff59c452006-03-27 13:27:43 -0800920 int rv;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700921
Nikolay Aleksandrove0974582014-09-15 17:19:35 +0200922 ASSERT_RTNL();
923
Linus Torvalds1da177e2005-04-16 15:20:36 -0700924 best_slave = bond_find_best_slave(bond);
Nikolay Aleksandrov1c72cfdc92014-09-11 22:49:24 +0200925 if (best_slave != rtnl_dereference(bond->curr_active_slave)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700926 bond_change_active_slave(bond, best_slave);
Jay Vosburghff59c452006-03-27 13:27:43 -0800927 rv = bond_set_carrier(bond);
928 if (!rv)
929 return;
930
Zhang Shengjud66bd902016-02-03 02:02:32 +0000931 if (netif_carrier_ok(bond->dev))
Veaceslav Falico76444f52014-07-15 19:35:58 +0200932 netdev_info(bond->dev, "first active interface up!\n");
Zhang Shengjud66bd902016-02-03 02:02:32 +0000933 else
Veaceslav Falico76444f52014-07-15 19:35:58 +0200934 netdev_info(bond->dev, "now running without any active interface!\n");
Linus Torvalds1da177e2005-04-16 15:20:36 -0700935 }
936}
937
WANG Congf6dc31a2010-05-06 00:48:51 -0700938#ifdef CONFIG_NET_POLL_CONTROLLER
Amerigo Wang8a8efa22011-02-17 23:43:32 +0000939static inline int slave_enable_netpoll(struct slave *slave)
WANG Congf6dc31a2010-05-06 00:48:51 -0700940{
Amerigo Wang8a8efa22011-02-17 23:43:32 +0000941 struct netpoll *np;
942 int err = 0;
WANG Congf6dc31a2010-05-06 00:48:51 -0700943
Eric W. Biedermana8779ec2014-03-27 15:36:38 -0700944 np = kzalloc(sizeof(*np), GFP_KERNEL);
Amerigo Wang8a8efa22011-02-17 23:43:32 +0000945 err = -ENOMEM;
946 if (!np)
947 goto out;
948
Eric W. Biedermana8779ec2014-03-27 15:36:38 -0700949 err = __netpoll_setup(np, slave->dev);
Amerigo Wang8a8efa22011-02-17 23:43:32 +0000950 if (err) {
951 kfree(np);
952 goto out;
WANG Congf6dc31a2010-05-06 00:48:51 -0700953 }
Amerigo Wang8a8efa22011-02-17 23:43:32 +0000954 slave->np = np;
955out:
956 return err;
957}
958static inline void slave_disable_netpoll(struct slave *slave)
959{
960 struct netpoll *np = slave->np;
961
962 if (!np)
963 return;
964
965 slave->np = NULL;
Neil Horman2cde6ac2013-02-11 10:25:30 +0000966 __netpoll_free_async(np);
Amerigo Wang8a8efa22011-02-17 23:43:32 +0000967}
WANG Congf6dc31a2010-05-06 00:48:51 -0700968
969static void bond_poll_controller(struct net_device *bond_dev)
970{
Mahesh Bandewar616f4542015-03-04 21:57:52 -0800971 struct bonding *bond = netdev_priv(bond_dev);
972 struct slave *slave = NULL;
973 struct list_head *iter;
974 struct ad_info ad_info;
975 struct netpoll_info *ni;
976 const struct net_device_ops *ops;
977
978 if (BOND_MODE(bond) == BOND_MODE_8023AD)
979 if (bond_3ad_get_active_agg_info(bond, &ad_info))
980 return;
981
Mahesh Bandewar616f4542015-03-04 21:57:52 -0800982 bond_for_each_slave_rcu(bond, slave, iter) {
983 ops = slave->dev->netdev_ops;
984 if (!bond_slave_is_up(slave) || !ops->ndo_poll_controller)
985 continue;
986
987 if (BOND_MODE(bond) == BOND_MODE_8023AD) {
988 struct aggregator *agg =
989 SLAVE_AD_INFO(slave)->port.aggregator;
990
991 if (agg &&
992 agg->aggregator_identifier != ad_info.aggregator_id)
993 continue;
994 }
995
996 ni = rcu_dereference_bh(slave->dev->npinfo);
997 if (down_trylock(&ni->dev_lock))
998 continue;
999 ops->ndo_poll_controller(slave->dev);
1000 up(&ni->dev_lock);
1001 }
Amerigo Wang8a8efa22011-02-17 23:43:32 +00001002}
1003
dingtianhongc4cdef92013-07-23 15:25:27 +08001004static void bond_netpoll_cleanup(struct net_device *bond_dev)
Amerigo Wang8a8efa22011-02-17 23:43:32 +00001005{
dingtianhongc4cdef92013-07-23 15:25:27 +08001006 struct bonding *bond = netdev_priv(bond_dev);
Veaceslav Falico9caff1e2013-09-25 09:20:14 +02001007 struct list_head *iter;
Neil Hormanc2355e12010-10-13 16:01:49 +00001008 struct slave *slave;
Neil Hormanc2355e12010-10-13 16:01:49 +00001009
Veaceslav Falico9caff1e2013-09-25 09:20:14 +02001010 bond_for_each_slave(bond, slave, iter)
Veaceslav Falicob6adc612014-05-15 21:39:57 +02001011 if (bond_slave_is_up(slave))
Amerigo Wang8a8efa22011-02-17 23:43:32 +00001012 slave_disable_netpoll(slave);
WANG Congf6dc31a2010-05-06 00:48:51 -07001013}
WANG Congf6dc31a2010-05-06 00:48:51 -07001014
Eric W. Biedermana8779ec2014-03-27 15:36:38 -07001015static int bond_netpoll_setup(struct net_device *dev, struct netpoll_info *ni)
Amerigo Wang8a8efa22011-02-17 23:43:32 +00001016{
1017 struct bonding *bond = netdev_priv(dev);
Veaceslav Falico9caff1e2013-09-25 09:20:14 +02001018 struct list_head *iter;
Amerigo Wang8a8efa22011-02-17 23:43:32 +00001019 struct slave *slave;
nikolay@redhat.comdec1e902013-08-01 16:54:47 +02001020 int err = 0;
WANG Congf6dc31a2010-05-06 00:48:51 -07001021
Veaceslav Falico9caff1e2013-09-25 09:20:14 +02001022 bond_for_each_slave(bond, slave, iter) {
Amerigo Wang8a8efa22011-02-17 23:43:32 +00001023 err = slave_enable_netpoll(slave);
1024 if (err) {
dingtianhongc4cdef92013-07-23 15:25:27 +08001025 bond_netpoll_cleanup(dev);
Amerigo Wang8a8efa22011-02-17 23:43:32 +00001026 break;
1027 }
1028 }
Amerigo Wang8a8efa22011-02-17 23:43:32 +00001029 return err;
1030}
Amerigo Wang8a8efa22011-02-17 23:43:32 +00001031#else
1032static inline int slave_enable_netpoll(struct slave *slave)
1033{
1034 return 0;
1035}
1036static inline void slave_disable_netpoll(struct slave *slave)
1037{
1038}
WANG Congf6dc31a2010-05-06 00:48:51 -07001039static void bond_netpoll_cleanup(struct net_device *bond_dev)
1040{
1041}
WANG Congf6dc31a2010-05-06 00:48:51 -07001042#endif
1043
Linus Torvalds1da177e2005-04-16 15:20:36 -07001044/*---------------------------------- IOCTL ----------------------------------*/
1045
Michał Mirosławc8f44af2011-11-15 15:29:55 +00001046static netdev_features_t bond_fix_features(struct net_device *dev,
nikolay@redhat.com9b7b1652013-09-02 13:51:41 +02001047 netdev_features_t features)
Michał Mirosławb2a103e2011-05-07 03:22:17 +00001048{
Michał Mirosławb2a103e2011-05-07 03:22:17 +00001049 struct bonding *bond = netdev_priv(dev);
Veaceslav Falico9caff1e2013-09-25 09:20:14 +02001050 struct list_head *iter;
Michał Mirosławc8f44af2011-11-15 15:29:55 +00001051 netdev_features_t mask;
nikolay@redhat.com9b7b1652013-09-02 13:51:41 +02001052 struct slave *slave;
Michał Mirosławb2a103e2011-05-07 03:22:17 +00001053
Scott Feldman7889cbe2015-05-10 09:48:07 -07001054 mask = features;
Roopa Prabhuc158cba2015-01-29 22:40:16 -08001055
Michał Mirosławb2a103e2011-05-07 03:22:17 +00001056 features &= ~NETIF_F_ONE_FOR_ALL;
1057 features |= NETIF_F_ALL_FOR_ALL;
1058
Veaceslav Falico9caff1e2013-09-25 09:20:14 +02001059 bond_for_each_slave(bond, slave, iter) {
Michał Mirosławb2a103e2011-05-07 03:22:17 +00001060 features = netdev_increment_features(features,
1061 slave->dev->features,
1062 mask);
1063 }
Eric Dumazetb0ce3502013-05-16 07:34:53 +00001064 features = netdev_add_tso_features(features, mask);
Michał Mirosławb2a103e2011-05-07 03:22:17 +00001065
Michał Mirosławb2a103e2011-05-07 03:22:17 +00001066 return features;
1067}
1068
Tom Herberta1882222015-12-14 11:19:43 -08001069#define BOND_VLAN_FEATURES (NETIF_F_HW_CSUM | NETIF_F_SG | \
Michał Mirosław62f2a3a2011-07-13 14:10:29 +00001070 NETIF_F_FRAGLIST | NETIF_F_ALL_TSO | \
1071 NETIF_F_HIGHDMA | NETIF_F_LRO)
Michał Mirosławb2a103e2011-05-07 03:22:17 +00001072
Tom Herberta1882222015-12-14 11:19:43 -08001073#define BOND_ENC_FEATURES (NETIF_F_HW_CSUM | NETIF_F_SG | \
1074 NETIF_F_RXCSUM | NETIF_F_ALL_TSO)
Or Gerlitz5a7baa72014-06-17 16:11:09 +03001075
Michał Mirosławb2a103e2011-05-07 03:22:17 +00001076static void bond_compute_features(struct bonding *bond)
Arthur Kepner8531c5f2005-08-23 01:34:53 -04001077{
Eric Dumazet02875872014-10-05 18:38:35 -07001078 unsigned int dst_release_flag = IFF_XMIT_DST_RELEASE |
1079 IFF_XMIT_DST_RELEASE_PERM;
Michał Mirosławc8f44af2011-11-15 15:29:55 +00001080 netdev_features_t vlan_features = BOND_VLAN_FEATURES;
Or Gerlitz5a7baa72014-06-17 16:11:09 +03001081 netdev_features_t enc_features = BOND_ENC_FEATURES;
Veaceslav Falico9caff1e2013-09-25 09:20:14 +02001082 struct net_device *bond_dev = bond->dev;
1083 struct list_head *iter;
1084 struct slave *slave;
Michał Mirosławb2a103e2011-05-07 03:22:17 +00001085 unsigned short max_hard_header_len = ETH_HLEN;
Sarveshwar Bandi0e376bd2012-11-21 04:35:03 +00001086 unsigned int gso_max_size = GSO_MAX_SIZE;
1087 u16 gso_max_segs = GSO_MAX_SEGS;
Herbert Xub63365a2008-10-23 01:11:29 -07001088
Veaceslav Falico0965a1f2013-09-25 09:20:21 +02001089 if (!bond_has_slaves(bond))
Herbert Xub63365a2008-10-23 01:11:29 -07001090 goto done;
Michal Kubečeka9b3ace2014-05-20 08:29:35 +02001091 vlan_features &= NETIF_F_ALL_FOR_ALL;
Herbert Xub63365a2008-10-23 01:11:29 -07001092
Veaceslav Falico9caff1e2013-09-25 09:20:14 +02001093 bond_for_each_slave(bond, slave, iter) {
Jay Vosburgh278339a2009-08-28 12:05:12 +00001094 vlan_features = netdev_increment_features(vlan_features,
Michał Mirosławb2a103e2011-05-07 03:22:17 +00001095 slave->dev->vlan_features, BOND_VLAN_FEATURES);
1096
Or Gerlitz5a7baa72014-06-17 16:11:09 +03001097 enc_features = netdev_increment_features(enc_features,
1098 slave->dev->hw_enc_features,
1099 BOND_ENC_FEATURES);
Eric Dumazetb6fe83e2012-07-17 12:19:48 +00001100 dst_release_flag &= slave->dev->priv_flags;
Jay Vosburgh54ef3132006-09-22 21:53:39 -07001101 if (slave->dev->hard_header_len > max_hard_header_len)
1102 max_hard_header_len = slave->dev->hard_header_len;
Sarveshwar Bandi0e376bd2012-11-21 04:35:03 +00001103
1104 gso_max_size = min(gso_max_size, slave->dev->gso_max_size);
1105 gso_max_segs = min(gso_max_segs, slave->dev->gso_max_segs);
Jay Vosburgh54ef3132006-09-22 21:53:39 -07001106 }
Arthur Kepner8531c5f2005-08-23 01:34:53 -04001107
Herbert Xub63365a2008-10-23 01:11:29 -07001108done:
Michał Mirosławb2a103e2011-05-07 03:22:17 +00001109 bond_dev->vlan_features = vlan_features;
Eric Dumazet24f87d42015-01-25 23:27:14 -08001110 bond_dev->hw_enc_features = enc_features | NETIF_F_GSO_ENCAP_ALL;
Jay Vosburgh54ef3132006-09-22 21:53:39 -07001111 bond_dev->hard_header_len = max_hard_header_len;
Sarveshwar Bandi0e376bd2012-11-21 04:35:03 +00001112 bond_dev->gso_max_segs = gso_max_segs;
1113 netif_set_gso_max_size(bond_dev, gso_max_size);
Arthur Kepner8531c5f2005-08-23 01:34:53 -04001114
Eric Dumazet02875872014-10-05 18:38:35 -07001115 bond_dev->priv_flags &= ~IFF_XMIT_DST_RELEASE;
1116 if ((bond_dev->priv_flags & IFF_XMIT_DST_RELEASE_PERM) &&
1117 dst_release_flag == (IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM))
1118 bond_dev->priv_flags |= IFF_XMIT_DST_RELEASE;
Eric Dumazetb6fe83e2012-07-17 12:19:48 +00001119
Michał Mirosławb2a103e2011-05-07 03:22:17 +00001120 netdev_change_features(bond_dev);
Arthur Kepner8531c5f2005-08-23 01:34:53 -04001121}
1122
Moni Shoua872254d2007-10-09 19:43:38 -07001123static void bond_setup_by_slave(struct net_device *bond_dev,
1124 struct net_device *slave_dev)
1125{
Stephen Hemminger00829822008-11-20 20:14:53 -08001126 bond_dev->header_ops = slave_dev->header_ops;
Moni Shoua872254d2007-10-09 19:43:38 -07001127
1128 bond_dev->type = slave_dev->type;
1129 bond_dev->hard_header_len = slave_dev->hard_header_len;
1130 bond_dev->addr_len = slave_dev->addr_len;
1131
1132 memcpy(bond_dev->broadcast, slave_dev->broadcast,
1133 slave_dev->addr_len);
1134}
1135
Jiri Pirko5b2c4dd2011-02-23 09:05:42 +00001136/* On bonding slaves other than the currently active slave, suppress
Jiri Pirko3aba8912011-04-19 03:48:16 +00001137 * duplicates except for alb non-mcast/bcast.
Jiri Pirko5b2c4dd2011-02-23 09:05:42 +00001138 */
1139static bool bond_should_deliver_exact_match(struct sk_buff *skb,
Jiri Pirko0bd80da2011-03-16 08:45:23 +00001140 struct slave *slave,
1141 struct bonding *bond)
Jiri Pirko5b2c4dd2011-02-23 09:05:42 +00001142{
Jiri Pirko2d7011c2011-03-16 08:46:43 +00001143 if (bond_is_slave_inactive(slave)) {
Veaceslav Falico01844092014-05-15 21:39:55 +02001144 if (BOND_MODE(bond) == BOND_MODE_ALB &&
Jiri Pirko5b2c4dd2011-02-23 09:05:42 +00001145 skb->pkt_type != PACKET_BROADCAST &&
1146 skb->pkt_type != PACKET_MULTICAST)
Jiri Pirko5b2c4dd2011-02-23 09:05:42 +00001147 return false;
Jiri Pirko5b2c4dd2011-02-23 09:05:42 +00001148 return true;
1149 }
1150 return false;
1151}
1152
Jiri Pirko8a4eb572011-03-12 03:14:39 +00001153static rx_handler_result_t bond_handle_frame(struct sk_buff **pskb)
Jiri Pirko5b2c4dd2011-02-23 09:05:42 +00001154{
Jiri Pirko8a4eb572011-03-12 03:14:39 +00001155 struct sk_buff *skb = *pskb;
Jiri Pirkof1c17752011-03-12 03:14:35 +00001156 struct slave *slave;
Jiri Pirko0bd80da2011-03-16 08:45:23 +00001157 struct bonding *bond;
Eric Dumazetde063b72012-06-11 19:23:07 +00001158 int (*recv_probe)(const struct sk_buff *, struct bonding *,
1159 struct slave *);
Jiri Bohac13a8e0c2012-05-09 01:01:40 +00001160 int ret = RX_HANDLER_ANOTHER;
Jiri Pirko5b2c4dd2011-02-23 09:05:42 +00001161
Jiri Pirko8a4eb572011-03-12 03:14:39 +00001162 skb = skb_share_check(skb, GFP_ATOMIC);
1163 if (unlikely(!skb))
1164 return RX_HANDLER_CONSUMED;
1165
1166 *pskb = skb;
Jiri Pirko5b2c4dd2011-02-23 09:05:42 +00001167
Jiri Pirko35d48902011-03-22 02:38:12 +00001168 slave = bond_slave_get_rcu(skb->dev);
1169 bond = slave->bond;
Jiri Pirko0bd80da2011-03-16 08:45:23 +00001170
Mitsuo Hayasaka4d974802011-10-12 16:04:29 +00001171 recv_probe = ACCESS_ONCE(bond->recv_probe);
1172 if (recv_probe) {
Eric Dumazetde063b72012-06-11 19:23:07 +00001173 ret = recv_probe(skb, bond, slave);
1174 if (ret == RX_HANDLER_CONSUMED) {
1175 consume_skb(skb);
1176 return ret;
Jiri Pirko3aba8912011-04-19 03:48:16 +00001177 }
1178 }
1179
Zhang Shengjud66bd902016-02-03 02:02:32 +00001180 if (bond_should_deliver_exact_match(skb, slave, bond))
Jiri Pirko8a4eb572011-03-12 03:14:39 +00001181 return RX_HANDLER_EXACT;
Jiri Pirko5b2c4dd2011-02-23 09:05:42 +00001182
Jiri Pirko35d48902011-03-22 02:38:12 +00001183 skb->dev = bond->dev;
Jiri Pirko5b2c4dd2011-02-23 09:05:42 +00001184
Veaceslav Falico01844092014-05-15 21:39:55 +02001185 if (BOND_MODE(bond) == BOND_MODE_ALB &&
Jiri Pirko35d48902011-03-22 02:38:12 +00001186 bond->dev->priv_flags & IFF_BRIDGE_PORT &&
Jiri Pirko5b2c4dd2011-02-23 09:05:42 +00001187 skb->pkt_type == PACKET_HOST) {
Jiri Pirko5b2c4dd2011-02-23 09:05:42 +00001188
Changli Gao541ac7c2011-03-02 21:07:14 +00001189 if (unlikely(skb_cow_head(skb,
1190 skb->data - skb_mac_header(skb)))) {
1191 kfree_skb(skb);
Jiri Pirko8a4eb572011-03-12 03:14:39 +00001192 return RX_HANDLER_CONSUMED;
Changli Gao541ac7c2011-03-02 21:07:14 +00001193 }
Joe Perches2a7c1832014-02-18 09:42:45 -08001194 ether_addr_copy(eth_hdr(skb)->h_dest, bond->dev->dev_addr);
Jiri Pirko5b2c4dd2011-02-23 09:05:42 +00001195 }
1196
Jiri Bohac13a8e0c2012-05-09 01:01:40 +00001197 return ret;
Jiri Pirko5b2c4dd2011-02-23 09:05:42 +00001198}
1199
Jiri Pirko41f0b042015-12-03 12:12:14 +01001200static enum netdev_lag_tx_type bond_lag_tx_type(struct bonding *bond)
Jiri Pirko471cb5a2013-01-03 22:49:01 +00001201{
Jiri Pirko41f0b042015-12-03 12:12:14 +01001202 switch (BOND_MODE(bond)) {
1203 case BOND_MODE_ROUNDROBIN:
1204 return NETDEV_LAG_TX_TYPE_ROUNDROBIN;
1205 case BOND_MODE_ACTIVEBACKUP:
1206 return NETDEV_LAG_TX_TYPE_ACTIVEBACKUP;
1207 case BOND_MODE_BROADCAST:
1208 return NETDEV_LAG_TX_TYPE_BROADCAST;
1209 case BOND_MODE_XOR:
1210 case BOND_MODE_8023AD:
1211 return NETDEV_LAG_TX_TYPE_HASH;
1212 default:
1213 return NETDEV_LAG_TX_TYPE_UNKNOWN;
1214 }
1215}
1216
1217static int bond_master_upper_dev_link(struct bonding *bond, struct slave *slave)
1218{
1219 struct netdev_lag_upper_info lag_upper_info;
Jiri Pirko471cb5a2013-01-03 22:49:01 +00001220 int err;
1221
Jiri Pirko41f0b042015-12-03 12:12:14 +01001222 lag_upper_info.tx_type = bond_lag_tx_type(bond);
1223 err = netdev_master_upper_dev_link(slave->dev, bond->dev, slave,
1224 &lag_upper_info);
Jiri Pirko471cb5a2013-01-03 22:49:01 +00001225 if (err)
1226 return err;
Jiri Pirko41f0b042015-12-03 12:12:14 +01001227 rtmsg_ifinfo(RTM_NEWLINK, slave->dev, IFF_SLAVE, GFP_KERNEL);
Jiri Pirko471cb5a2013-01-03 22:49:01 +00001228 return 0;
1229}
1230
Jiri Pirko41f0b042015-12-03 12:12:14 +01001231static void bond_upper_dev_unlink(struct bonding *bond, struct slave *slave)
Jiri Pirko471cb5a2013-01-03 22:49:01 +00001232{
Jiri Pirko41f0b042015-12-03 12:12:14 +01001233 netdev_upper_dev_unlink(slave->dev, bond->dev);
1234 slave->dev->flags &= ~IFF_SLAVE;
1235 rtmsg_ifinfo(RTM_NEWLINK, slave->dev, IFF_SLAVE, GFP_KERNEL);
Jiri Pirko471cb5a2013-01-03 22:49:01 +00001236}
1237
dingtianhong3fdddd82014-05-12 15:08:43 +08001238static struct slave *bond_alloc_slave(struct bonding *bond)
1239{
1240 struct slave *slave = NULL;
1241
Zhang Shengjud66bd902016-02-03 02:02:32 +00001242 slave = kzalloc(sizeof(*slave), GFP_KERNEL);
dingtianhong3fdddd82014-05-12 15:08:43 +08001243 if (!slave)
1244 return NULL;
1245
Veaceslav Falico01844092014-05-15 21:39:55 +02001246 if (BOND_MODE(bond) == BOND_MODE_8023AD) {
dingtianhong3fdddd82014-05-12 15:08:43 +08001247 SLAVE_AD_INFO(slave) = kzalloc(sizeof(struct ad_slave_info),
1248 GFP_KERNEL);
1249 if (!SLAVE_AD_INFO(slave)) {
1250 kfree(slave);
1251 return NULL;
1252 }
1253 }
Mahesh Bandewar22ad2932018-09-24 14:40:11 -07001254 INIT_DELAYED_WORK(&slave->notify_work, bond_netdev_notify_work);
1255
dingtianhong3fdddd82014-05-12 15:08:43 +08001256 return slave;
1257}
1258
1259static void bond_free_slave(struct slave *slave)
1260{
1261 struct bonding *bond = bond_get_bond_by_slave(slave);
1262
Mahesh Bandewar22ad2932018-09-24 14:40:11 -07001263 cancel_delayed_work_sync(&slave->notify_work);
Veaceslav Falico01844092014-05-15 21:39:55 +02001264 if (BOND_MODE(bond) == BOND_MODE_8023AD)
dingtianhong3fdddd82014-05-12 15:08:43 +08001265 kfree(SLAVE_AD_INFO(slave));
1266
1267 kfree(slave);
1268}
1269
Moni Shoua69a23382015-02-03 16:48:30 +02001270static void bond_fill_ifbond(struct bonding *bond, struct ifbond *info)
1271{
1272 info->bond_mode = BOND_MODE(bond);
1273 info->miimon = bond->params.miimon;
1274 info->num_slaves = bond->slave_cnt;
1275}
1276
1277static void bond_fill_ifslave(struct slave *slave, struct ifslave *info)
1278{
1279 strcpy(info->slave_name, slave->dev->name);
1280 info->link = slave->link;
1281 info->state = bond_slave_state(slave);
1282 info->link_failure_count = slave->link_failure_count;
1283}
1284
Moni Shoua69e61132015-02-03 16:48:31 +02001285static void bond_netdev_notify_work(struct work_struct *_work)
1286{
Mahesh Bandewar22ad2932018-09-24 14:40:11 -07001287 struct slave *slave = container_of(_work, struct slave,
1288 notify_work.work);
Moni Shoua69e61132015-02-03 16:48:31 +02001289
Mahesh Bandewar22ad2932018-09-24 14:40:11 -07001290 if (rtnl_trylock()) {
1291 struct netdev_bonding_info binfo;
1292
1293 bond_fill_ifslave(slave, &binfo.slave);
1294 bond_fill_ifbond(slave->bond, &binfo.master);
1295 netdev_bonding_info_change(slave->dev, &binfo);
1296 rtnl_unlock();
1297 } else {
1298 queue_delayed_work(slave->bond->wq, &slave->notify_work, 1);
1299 }
Moni Shoua69e61132015-02-03 16:48:31 +02001300}
1301
1302void bond_queue_slave_event(struct slave *slave)
1303{
Mahesh Bandewar22ad2932018-09-24 14:40:11 -07001304 queue_delayed_work(slave->bond->wq, &slave->notify_work, 0);
Moni Shoua69e61132015-02-03 16:48:31 +02001305}
1306
Jiri Pirkof7c7eb72015-12-03 12:12:20 +01001307void bond_lower_state_changed(struct slave *slave)
1308{
1309 struct netdev_lag_lower_state_info info;
1310
1311 info.link_up = slave->link == BOND_LINK_UP ||
1312 slave->link == BOND_LINK_FAIL;
1313 info.tx_enabled = bond_is_active_slave(slave);
1314 netdev_lower_state_changed(slave->dev, &info);
1315}
1316
Linus Torvalds1da177e2005-04-16 15:20:36 -07001317/* enslave device <slave> to bond device <master> */
Mitch Williamsa77b5322005-11-09 10:35:51 -08001318int bond_enslave(struct net_device *bond_dev, struct net_device *slave_dev)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001319{
Wang Chen454d7c92008-11-12 23:37:49 -08001320 struct bonding *bond = netdev_priv(bond_dev);
Stephen Hemmingereb7cc592008-11-19 21:56:05 -08001321 const struct net_device_ops *slave_ops = slave_dev->netdev_ops;
Veaceslav Falicoc8c23902013-09-25 09:20:25 +02001322 struct slave *new_slave = NULL, *prev_slave;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001323 struct sockaddr addr;
1324 int link_reporting;
Veaceslav Falico8599b522013-06-24 11:49:34 +02001325 int res = 0, i;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001326
Ben Hutchingsc772dde2012-12-07 06:15:32 +00001327 if (!bond->params.use_carrier &&
1328 slave_dev->ethtool_ops->get_link == NULL &&
1329 slave_ops->ndo_do_ioctl == NULL) {
Veaceslav Falico76444f52014-07-15 19:35:58 +02001330 netdev_warn(bond_dev, "no link monitoring support for %s\n",
1331 slave_dev->name);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001332 }
1333
Mahesh Bandewar24b27fc2016-09-01 22:18:34 -07001334 /* already in-use? */
1335 if (netdev_is_rx_handler_busy(slave_dev)) {
1336 netdev_err(bond_dev,
1337 "Error: Device is in use and cannot be enslaved\n");
Linus Torvalds1da177e2005-04-16 15:20:36 -07001338 return -EBUSY;
1339 }
1340
Jiri Bohac09a89c22014-02-26 18:20:13 +01001341 if (bond_dev == slave_dev) {
Veaceslav Falico76444f52014-07-15 19:35:58 +02001342 netdev_err(bond_dev, "cannot enslave bond to itself.\n");
Jiri Bohac09a89c22014-02-26 18:20:13 +01001343 return -EPERM;
1344 }
1345
Linus Torvalds1da177e2005-04-16 15:20:36 -07001346 /* vlan challenged mutual exclusion */
1347 /* no need to lock since we're protected by rtnl_lock */
1348 if (slave_dev->features & NETIF_F_VLAN_CHALLENGED) {
Veaceslav Falico76444f52014-07-15 19:35:58 +02001349 netdev_dbg(bond_dev, "%s is NETIF_F_VLAN_CHALLENGED\n",
1350 slave_dev->name);
Jiri Pirko55462cf2012-10-14 04:30:56 +00001351 if (vlan_uses_dev(bond_dev)) {
Veaceslav Falico76444f52014-07-15 19:35:58 +02001352 netdev_err(bond_dev, "Error: cannot enslave VLAN challenged slave %s on VLAN enabled bond %s\n",
1353 slave_dev->name, bond_dev->name);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001354 return -EPERM;
1355 } else {
Veaceslav Falico76444f52014-07-15 19:35:58 +02001356 netdev_warn(bond_dev, "enslaved VLAN challenged slave %s. Adding VLANs will be blocked as long as %s is part of bond %s\n",
1357 slave_dev->name, slave_dev->name,
1358 bond_dev->name);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001359 }
1360 } else {
Veaceslav Falico76444f52014-07-15 19:35:58 +02001361 netdev_dbg(bond_dev, "%s is !NETIF_F_VLAN_CHALLENGED\n",
1362 slave_dev->name);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001363 }
1364
Nikolay Aleksandrov547942c2014-09-15 17:19:34 +02001365 /* Old ifenslave binaries are no longer supported. These can
Stephen Hemminger3d632c32009-06-12 19:02:48 +00001366 * be identified with moderate accuracy by the state of the slave:
Jay Vosburgh217df672005-09-26 16:11:50 -07001367 * the current ifenslave will set the interface down prior to
1368 * enslaving it; the old ifenslave will not.
1369 */
yzhu1ce3ea1c2015-12-03 18:00:55 +08001370 if (slave_dev->flags & IFF_UP) {
Veaceslav Falico76444f52014-07-15 19:35:58 +02001371 netdev_err(bond_dev, "%s is up - this may be due to an out of date ifenslave\n",
1372 slave_dev->name);
Zhang Shengju1e2a8862016-02-09 10:37:46 +00001373 return -EPERM;
Jay Vosburgh217df672005-09-26 16:11:50 -07001374 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001375
Moni Shoua872254d2007-10-09 19:43:38 -07001376 /* set bonding device ether type by slave - bonding netdevices are
1377 * created with ether_setup, so when the slave type is not ARPHRD_ETHER
1378 * there is a need to override some of the type dependent attribs/funcs.
1379 *
1380 * bond ether type mutual exclusion - don't allow slaves of dissimilar
1381 * ether type (eg ARPHRD_ETHER and ARPHRD_INFINIBAND) share the same bond
1382 */
Veaceslav Falico0965a1f2013-09-25 09:20:21 +02001383 if (!bond_has_slaves(bond)) {
Moni Shouae36b9d12009-07-15 04:56:31 +00001384 if (bond_dev->type != slave_dev->type) {
Veaceslav Falico76444f52014-07-15 19:35:58 +02001385 netdev_dbg(bond_dev, "change device type from %d to %d\n",
1386 bond_dev->type, slave_dev->type);
Moni Shoua75c78502009-09-15 02:37:40 -07001387
Amerigo Wangb7bc2a52012-08-09 22:14:57 +00001388 res = call_netdevice_notifiers(NETDEV_PRE_TYPE_CHANGE,
1389 bond_dev);
Jiri Pirko3ca5b402010-03-10 10:29:35 +00001390 res = notifier_to_errno(res);
1391 if (res) {
Veaceslav Falico76444f52014-07-15 19:35:58 +02001392 netdev_err(bond_dev, "refused to change device type\n");
Zhang Shengju1e2a8862016-02-09 10:37:46 +00001393 return -EBUSY;
Jiri Pirko3ca5b402010-03-10 10:29:35 +00001394 }
Moni Shoua75c78502009-09-15 02:37:40 -07001395
Jiri Pirko32a806c2010-03-19 04:00:23 +00001396 /* Flush unicast and multicast addresses */
Jiri Pirkoa748ee22010-04-01 21:22:09 +00001397 dev_uc_flush(bond_dev);
Jiri Pirko22bedad32010-04-01 21:22:57 +00001398 dev_mc_flush(bond_dev);
Jiri Pirko32a806c2010-03-19 04:00:23 +00001399
Moni Shouae36b9d12009-07-15 04:56:31 +00001400 if (slave_dev->type != ARPHRD_ETHER)
1401 bond_setup_by_slave(bond_dev, slave_dev);
Neil Horman550fd082011-07-26 06:05:38 +00001402 else {
Moni Shouae36b9d12009-07-15 04:56:31 +00001403 ether_setup(bond_dev);
Neil Horman550fd082011-07-26 06:05:38 +00001404 bond_dev->priv_flags &= ~IFF_TX_SKB_SHARING;
1405 }
Moni Shoua75c78502009-09-15 02:37:40 -07001406
Amerigo Wangb7bc2a52012-08-09 22:14:57 +00001407 call_netdevice_notifiers(NETDEV_POST_TYPE_CHANGE,
1408 bond_dev);
Moni Shouae36b9d12009-07-15 04:56:31 +00001409 }
Moni Shoua872254d2007-10-09 19:43:38 -07001410 } else if (bond_dev->type != slave_dev->type) {
Veaceslav Falico76444f52014-07-15 19:35:58 +02001411 netdev_err(bond_dev, "%s ether type (%d) is different from other slaves (%d), can not enslave it\n",
1412 slave_dev->name, slave_dev->type, bond_dev->type);
Zhang Shengju1e2a8862016-02-09 10:37:46 +00001413 return -EINVAL;
Moni Shoua872254d2007-10-09 19:43:38 -07001414 }
1415
Mark Bloch1533e772016-07-21 11:52:55 +03001416 if (slave_dev->type == ARPHRD_INFINIBAND &&
1417 BOND_MODE(bond) != BOND_MODE_ACTIVEBACKUP) {
1418 netdev_warn(bond_dev, "Type (%d) supports only active-backup mode\n",
1419 slave_dev->type);
1420 res = -EOPNOTSUPP;
1421 goto err_undo_flags;
1422 }
1423
1424 if (!slave_ops->ndo_set_mac_address ||
1425 slave_dev->type == ARPHRD_INFINIBAND) {
Veaceslav Falico76444f52014-07-15 19:35:58 +02001426 netdev_warn(bond_dev, "The slave device specified does not support setting the MAC address\n");
Veaceslav Falicof5442442014-07-15 13:26:01 +02001427 if (BOND_MODE(bond) == BOND_MODE_ACTIVEBACKUP &&
1428 bond->params.fail_over_mac != BOND_FOM_ACTIVE) {
1429 if (!bond_has_slaves(bond)) {
dingtianhong00503b62014-01-25 13:00:29 +08001430 bond->params.fail_over_mac = BOND_FOM_ACTIVE;
Veaceslav Falico76444f52014-07-15 19:35:58 +02001431 netdev_warn(bond_dev, "Setting fail_over_mac to active for active-backup mode\n");
Veaceslav Falicof5442442014-07-15 13:26:01 +02001432 } else {
Veaceslav Falico76444f52014-07-15 19:35:58 +02001433 netdev_err(bond_dev, "The slave device specified does not support setting the MAC address, but fail_over_mac is not set to active\n");
Veaceslav Falicof5442442014-07-15 13:26:01 +02001434 res = -EOPNOTSUPP;
1435 goto err_undo_flags;
dingtianhong00503b62014-01-25 13:00:29 +08001436 }
Moni Shoua2ab82852007-10-09 19:43:39 -07001437 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001438 }
1439
Amerigo Wang8d8fc292011-05-19 21:39:10 +00001440 call_netdevice_notifiers(NETDEV_JOIN, slave_dev);
1441
Jiri Pirkoc20811a2010-05-19 01:14:29 +00001442 /* If this is the first slave, then we need to set the master's hardware
Nikolay Aleksandrov547942c2014-09-15 17:19:34 +02001443 * address to be the same as the slave's.
1444 */
Veaceslav Falico0965a1f2013-09-25 09:20:21 +02001445 if (!bond_has_slaves(bond) &&
nikolay@redhat.comdec1e902013-08-01 16:54:47 +02001446 bond->dev->addr_assign_type == NET_ADDR_RANDOM)
Jiri Pirko409cc1f2013-01-30 11:08:11 +01001447 bond_set_dev_addr(bond->dev, slave_dev);
Jiri Pirkoc20811a2010-05-19 01:14:29 +00001448
dingtianhong3fdddd82014-05-12 15:08:43 +08001449 new_slave = bond_alloc_slave(bond);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001450 if (!new_slave) {
1451 res = -ENOMEM;
1452 goto err_undo_flags;
1453 }
dingtianhong3fdddd82014-05-12 15:08:43 +08001454
Veaceslav Falicodc73c412014-05-21 17:42:00 +02001455 new_slave->bond = bond;
1456 new_slave->dev = slave_dev;
Nikolay Aleksandrov547942c2014-09-15 17:19:34 +02001457 /* Set the new_slave's queue_id to be zero. Queue ID mapping
Andy Gospodarekbb1d9122010-06-02 08:40:18 +00001458 * is set via sysfs or module option if desired.
1459 */
1460 new_slave->queue_id = 0;
1461
Jiri Pirkob15ba0f2010-05-18 05:42:40 +00001462 /* Save slave's original mtu and then set it to match the bond */
1463 new_slave->original_mtu = slave_dev->mtu;
1464 res = dev_set_mtu(slave_dev, bond->dev->mtu);
1465 if (res) {
Veaceslav Falico76444f52014-07-15 19:35:58 +02001466 netdev_dbg(bond_dev, "Error %d calling dev_set_mtu\n", res);
Jiri Pirkob15ba0f2010-05-18 05:42:40 +00001467 goto err_free;
1468 }
1469
Nikolay Aleksandrov547942c2014-09-15 17:19:34 +02001470 /* Save slave's original ("permanent") mac address for modes
Jay Vosburgh217df672005-09-26 16:11:50 -07001471 * that need it, and for restoring it upon release, and then
1472 * set it to the master's address
1473 */
Joe Perchesada0f862014-02-15 16:02:17 -08001474 ether_addr_copy(new_slave->perm_hwaddr, slave_dev->dev_addr);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001475
dingtianhong00503b62014-01-25 13:00:29 +08001476 if (!bond->params.fail_over_mac ||
Veaceslav Falico01844092014-05-15 21:39:55 +02001477 BOND_MODE(bond) != BOND_MODE_ACTIVEBACKUP) {
Nikolay Aleksandrov547942c2014-09-15 17:19:34 +02001478 /* Set slave to master's mac address. The application already
Moni Shoua2ab82852007-10-09 19:43:39 -07001479 * set the master's mac address to that of the first slave
1480 */
1481 memcpy(addr.sa_data, bond_dev->dev_addr, bond_dev->addr_len);
1482 addr.sa_family = slave_dev->type;
1483 res = dev_set_mac_address(slave_dev, &addr);
1484 if (res) {
Veaceslav Falico76444f52014-07-15 19:35:58 +02001485 netdev_dbg(bond_dev, "Error %d calling set_mac_address\n", res);
Jiri Pirkob15ba0f2010-05-18 05:42:40 +00001486 goto err_restore_mtu;
Moni Shoua2ab82852007-10-09 19:43:39 -07001487 }
Jay Vosburgh217df672005-09-26 16:11:50 -07001488 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001489
Karl Heiss03d84a52016-01-11 08:28:43 -05001490 /* set slave flag before open to prevent IPv6 addrconf */
1491 slave_dev->flags |= IFF_SLAVE;
1492
Jay Vosburgh217df672005-09-26 16:11:50 -07001493 /* open the slave since the application closed it */
1494 res = dev_open(slave_dev);
1495 if (res) {
Veaceslav Falico76444f52014-07-15 19:35:58 +02001496 netdev_dbg(bond_dev, "Opening slave %s failed\n", slave_dev->name);
Veaceslav Falico1f718f02013-09-25 09:20:10 +02001497 goto err_restore_mac;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001498 }
1499
Jay Vosburgh0b680e72006-09-22 21:54:10 -07001500 slave_dev->priv_flags |= IFF_BONDING;
Andy Gospodarek5f0c5f72014-09-28 22:34:37 -04001501 /* initialize slave stats */
1502 dev_get_stats(new_slave->dev, &new_slave->slave_stats);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001503
Holger Eitzenberger58402052008-12-09 23:07:13 -08001504 if (bond_is_lb(bond)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001505 /* bond_alb_init_slave() must be called before all other stages since
1506 * it might fail and we do not want to have to undo everything
1507 */
1508 res = bond_alb_init_slave(bond, new_slave);
Stephen Hemminger3d632c32009-06-12 19:02:48 +00001509 if (res)
Jay Vosburgh569f0c42008-05-02 18:06:02 -07001510 goto err_close;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001511 }
1512
Wei Yongjunb8e2fde2013-08-23 10:45:07 +08001513 res = vlan_vids_add_by_dev(slave_dev, bond_dev);
1514 if (res) {
Veaceslav Falico76444f52014-07-15 19:35:58 +02001515 netdev_err(bond_dev, "Couldn't add bond vlan ids to %s\n",
1516 slave_dev->name);
Xin Longd6d65b42018-03-26 01:16:46 +08001517 goto err_close;
nikolay@redhat.com1ff412a2013-08-06 12:40:15 +02001518 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001519
Veaceslav Falicoc8c23902013-09-25 09:20:25 +02001520 prev_slave = bond_last_slave(bond);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001521
1522 new_slave->delay = 0;
1523 new_slave->link_failure_count = 0;
1524
Andreas Bornae0c8ee2017-08-10 06:41:44 +02001525 if (bond_update_speed_duplex(new_slave) &&
1526 bond_needs_speed_duplex(bond))
Mahesh Bandewarbc5ad402017-04-03 18:38:39 -07001527 new_slave->link = BOND_LINK_DOWN;
Veaceslav Falico876254a2013-03-12 06:31:32 +00001528
Veaceslav Falico49f17de2014-02-18 07:48:47 +01001529 new_slave->last_rx = jiffies -
Michal Kubečekf31c7932012-04-17 02:02:06 +00001530 (msecs_to_jiffies(bond->params.arp_interval) + 1);
Veaceslav Falico8599b522013-06-24 11:49:34 +02001531 for (i = 0; i < BOND_MAX_ARP_TARGETS; i++)
Veaceslav Falico49f17de2014-02-18 07:48:47 +01001532 new_slave->target_last_arp_rx[i] = new_slave->last_rx;
Jay Vosburghf5b2b962006-09-22 21:54:53 -07001533
Linus Torvalds1da177e2005-04-16 15:20:36 -07001534 if (bond->params.miimon && !bond->params.use_carrier) {
1535 link_reporting = bond_check_dev_link(bond, slave_dev, 1);
1536
1537 if ((link_reporting == -1) && !bond->params.arp_interval) {
Nikolay Aleksandrov547942c2014-09-15 17:19:34 +02001538 /* miimon is set but a bonded network driver
Linus Torvalds1da177e2005-04-16 15:20:36 -07001539 * does not support ETHTOOL/MII and
1540 * arp_interval is not set. Note: if
1541 * use_carrier is enabled, we will never go
1542 * here (because netif_carrier is always
1543 * supported); thus, we don't need to change
1544 * the messages for netif_carrier.
1545 */
Veaceslav Falico76444f52014-07-15 19:35:58 +02001546 netdev_warn(bond_dev, "MII and ETHTOOL support not available for interface %s, and arp_interval/arp_ip_target module parameters not specified, thus bonding will not detect link failures! see bonding.txt for details\n",
1547 slave_dev->name);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001548 } else if (link_reporting == -1) {
1549 /* unable get link status using mii/ethtool */
Veaceslav Falico76444f52014-07-15 19:35:58 +02001550 netdev_warn(bond_dev, "can't get link status from interface %s; the network driver associated with this interface does not support MII or ETHTOOL link status reporting, thus miimon has no effect on this interface\n",
1551 slave_dev->name);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001552 }
1553 }
1554
1555 /* check for initial state */
Aviv Hellera30b0162016-07-05 12:09:47 +03001556 new_slave->link = BOND_LINK_NOCHANGE;
Michal Kubečekf31c7932012-04-17 02:02:06 +00001557 if (bond->params.miimon) {
1558 if (bond_check_dev_link(bond, slave_dev, 0) == BMSR_LSTATUS) {
1559 if (bond->params.updelay) {
Moni Shoua69a23382015-02-03 16:48:30 +02001560 bond_set_slave_link_state(new_slave,
Jiri Pirko5d397062015-12-03 12:12:19 +01001561 BOND_LINK_BACK,
1562 BOND_SLAVE_NOTIFY_NOW);
Michal Kubečekf31c7932012-04-17 02:02:06 +00001563 new_slave->delay = bond->params.updelay;
1564 } else {
Moni Shoua69a23382015-02-03 16:48:30 +02001565 bond_set_slave_link_state(new_slave,
Jiri Pirko5d397062015-12-03 12:12:19 +01001566 BOND_LINK_UP,
1567 BOND_SLAVE_NOTIFY_NOW);
Michal Kubečekf31c7932012-04-17 02:02:06 +00001568 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001569 } else {
Jiri Pirko5d397062015-12-03 12:12:19 +01001570 bond_set_slave_link_state(new_slave, BOND_LINK_DOWN,
1571 BOND_SLAVE_NOTIFY_NOW);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001572 }
Michal Kubečekf31c7932012-04-17 02:02:06 +00001573 } else if (bond->params.arp_interval) {
Moni Shoua69a23382015-02-03 16:48:30 +02001574 bond_set_slave_link_state(new_slave,
1575 (netif_carrier_ok(slave_dev) ?
Jiri Pirko5d397062015-12-03 12:12:19 +01001576 BOND_LINK_UP : BOND_LINK_DOWN),
1577 BOND_SLAVE_NOTIFY_NOW);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001578 } else {
Jiri Pirko5d397062015-12-03 12:12:19 +01001579 bond_set_slave_link_state(new_slave, BOND_LINK_UP,
1580 BOND_SLAVE_NOTIFY_NOW);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001581 }
1582
Michal Kubečekf31c7932012-04-17 02:02:06 +00001583 if (new_slave->link != BOND_LINK_DOWN)
Veaceslav Falico8e603462014-02-18 07:48:46 +01001584 new_slave->last_link_up = jiffies;
Veaceslav Falico76444f52014-07-15 19:35:58 +02001585 netdev_dbg(bond_dev, "Initial state of slave_dev is BOND_LINK_%s\n",
1586 new_slave->link == BOND_LINK_DOWN ? "DOWN" :
1587 (new_slave->link == BOND_LINK_UP ? "UP" : "BACK"));
Michal Kubečekf31c7932012-04-17 02:02:06 +00001588
Veaceslav Falicoec0865a2014-05-15 21:39:54 +02001589 if (bond_uses_primary(bond) && bond->params.primary[0]) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001590 /* if there is a primary slave, remember it */
Jiri Pirkoa5499522009-09-25 03:28:09 +00001591 if (strcmp(bond->params.primary, new_slave->dev->name) == 0) {
Nikolay Aleksandrov059b47e2014-09-09 23:17:00 +02001592 rcu_assign_pointer(bond->primary_slave, new_slave);
Jiri Pirkoa5499522009-09-25 03:28:09 +00001593 bond->force_primary = true;
1594 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001595 }
1596
Veaceslav Falico01844092014-05-15 21:39:55 +02001597 switch (BOND_MODE(bond)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001598 case BOND_MODE_ACTIVEBACKUP:
dingtianhong5e5b0662014-02-26 11:05:22 +08001599 bond_set_slave_inactive_flags(new_slave,
1600 BOND_SLAVE_NOTIFY_NOW);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001601 break;
1602 case BOND_MODE_8023AD:
1603 /* in 802.3ad mode, the internal mechanism
1604 * will activate the slaves in the selected
1605 * aggregator
1606 */
dingtianhong5e5b0662014-02-26 11:05:22 +08001607 bond_set_slave_inactive_flags(new_slave, BOND_SLAVE_NOTIFY_NOW);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001608 /* if this is the first slave */
Veaceslav Falico23c147e2013-09-27 15:10:57 +02001609 if (!prev_slave) {
dingtianhong3fdddd82014-05-12 15:08:43 +08001610 SLAVE_AD_INFO(new_slave)->id = 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001611 /* Initialize AD with the number of times that the AD timer is called in 1 second
1612 * can be called only after the mac address of the bond is set
1613 */
Peter Pan(潘卫平)56d00c672011-06-08 21:19:02 +00001614 bond_3ad_initialize(bond, 1000/AD_TIMER_INTERVAL);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001615 } else {
dingtianhong3fdddd82014-05-12 15:08:43 +08001616 SLAVE_AD_INFO(new_slave)->id =
1617 SLAVE_AD_INFO(prev_slave)->id + 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001618 }
1619
1620 bond_3ad_bind_slave(new_slave);
1621 break;
1622 case BOND_MODE_TLB:
1623 case BOND_MODE_ALB:
Jiri Pirkoe30bc062011-03-12 03:14:37 +00001624 bond_set_active_slave(new_slave);
dingtianhong5e5b0662014-02-26 11:05:22 +08001625 bond_set_slave_inactive_flags(new_slave, BOND_SLAVE_NOTIFY_NOW);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001626 break;
1627 default:
Veaceslav Falico76444f52014-07-15 19:35:58 +02001628 netdev_dbg(bond_dev, "This slave is always active in trunk mode\n");
Linus Torvalds1da177e2005-04-16 15:20:36 -07001629
1630 /* always active in trunk mode */
Jiri Pirkoe30bc062011-03-12 03:14:37 +00001631 bond_set_active_slave(new_slave);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001632
1633 /* In trunking mode there is little meaning to curr_active_slave
1634 * anyway (it holds no special properties of the bond device),
1635 * so we can change it without calling change_active_interface()
1636 */
Eric Dumazet4740d632014-07-15 06:56:55 -07001637 if (!rcu_access_pointer(bond->curr_active_slave) &&
1638 new_slave->link == BOND_LINK_UP)
nikolay@redhat.com278b2082013-08-01 16:54:51 +02001639 rcu_assign_pointer(bond->curr_active_slave, new_slave);
Stephen Hemminger3d632c32009-06-12 19:02:48 +00001640
Linus Torvalds1da177e2005-04-16 15:20:36 -07001641 break;
1642 } /* switch(bond_mode) */
1643
WANG Congf6dc31a2010-05-06 00:48:51 -07001644#ifdef CONFIG_NET_POLL_CONTROLLER
Xin Longa43e7cb2018-04-22 19:11:50 +08001645 if (bond->dev->npinfo) {
Amerigo Wang8a8efa22011-02-17 23:43:32 +00001646 if (slave_enable_netpoll(new_slave)) {
Veaceslav Falico76444f52014-07-15 19:35:58 +02001647 netdev_info(bond_dev, "master_dev is using netpoll, but new slave device does not support netpoll\n");
Amerigo Wang8a8efa22011-02-17 23:43:32 +00001648 res = -EBUSY;
stephen hemmingerf7d98212011-12-31 13:26:46 +00001649 goto err_detach;
Amerigo Wang8a8efa22011-02-17 23:43:32 +00001650 }
WANG Congf6dc31a2010-05-06 00:48:51 -07001651 }
1652#endif
Amerigo Wang8a8efa22011-02-17 23:43:32 +00001653
Michal Kubečekfbe168b2014-11-13 07:54:50 +01001654 if (!(bond_dev->features & NETIF_F_LRO))
1655 dev_disable_lro(slave_dev);
1656
Jiri Pirko35d48902011-03-22 02:38:12 +00001657 res = netdev_rx_handler_register(slave_dev, bond_handle_frame,
1658 new_slave);
1659 if (res) {
Veaceslav Falico76444f52014-07-15 19:35:58 +02001660 netdev_dbg(bond_dev, "Error %d calling netdev_rx_handler_register\n", res);
Veaceslav Falico5831d66e2013-09-25 09:20:32 +02001661 goto err_detach;
Jiri Pirko35d48902011-03-22 02:38:12 +00001662 }
1663
Jiri Pirko41f0b042015-12-03 12:12:14 +01001664 res = bond_master_upper_dev_link(bond, new_slave);
Veaceslav Falico1f718f02013-09-25 09:20:10 +02001665 if (res) {
Veaceslav Falico76444f52014-07-15 19:35:58 +02001666 netdev_dbg(bond_dev, "Error %d calling bond_master_upper_dev_link\n", res);
Veaceslav Falico1f718f02013-09-25 09:20:10 +02001667 goto err_unregister;
1668 }
1669
sfeldma@cumulusnetworks.com07699f92014-01-16 22:57:49 -08001670 res = bond_sysfs_slave_add(new_slave);
1671 if (res) {
Veaceslav Falico76444f52014-07-15 19:35:58 +02001672 netdev_dbg(bond_dev, "Error %d calling bond_sysfs_slave_add\n", res);
sfeldma@cumulusnetworks.com07699f92014-01-16 22:57:49 -08001673 goto err_upper_unlink;
1674 }
1675
Eric Dumazet7142fdb2018-07-31 06:30:54 -07001676 bond->nest_level = dev_get_nest_level(bond_dev) + 1;
1677
Xin Longd6d65b42018-03-26 01:16:46 +08001678 /* If the mode uses primary, then the following is handled by
1679 * bond_change_active_slave().
1680 */
1681 if (!bond_uses_primary(bond)) {
1682 /* set promiscuity level to new slave */
1683 if (bond_dev->flags & IFF_PROMISC) {
1684 res = dev_set_promiscuity(slave_dev, 1);
1685 if (res)
1686 goto err_sysfs_del;
1687 }
1688
1689 /* set allmulti level to new slave */
1690 if (bond_dev->flags & IFF_ALLMULTI) {
1691 res = dev_set_allmulti(slave_dev, 1);
Xin Long359d6e32018-03-26 01:16:47 +08001692 if (res) {
1693 if (bond_dev->flags & IFF_PROMISC)
1694 dev_set_promiscuity(slave_dev, -1);
Xin Longd6d65b42018-03-26 01:16:46 +08001695 goto err_sysfs_del;
Xin Long359d6e32018-03-26 01:16:47 +08001696 }
Xin Longd6d65b42018-03-26 01:16:46 +08001697 }
1698
1699 netif_addr_lock_bh(bond_dev);
1700 dev_mc_sync_multiple(slave_dev, bond_dev);
1701 dev_uc_sync_multiple(slave_dev, bond_dev);
1702 netif_addr_unlock_bh(bond_dev);
1703
1704 if (BOND_MODE(bond) == BOND_MODE_8023AD) {
1705 /* add lacpdu mc addr to mc list */
1706 u8 lacpdu_multicast[ETH_ALEN] = MULTICAST_LACPDU_ADDR;
1707
1708 dev_mc_add(slave_dev, lacpdu_multicast);
1709 }
1710 }
1711
Veaceslav Falico5378c2e2013-10-21 11:48:30 +02001712 bond->slave_cnt++;
1713 bond_compute_features(bond);
1714 bond_set_carrier(bond);
1715
Veaceslav Falicoec0865a2014-05-15 21:39:54 +02001716 if (bond_uses_primary(bond)) {
dingtianhongf80889a2014-02-12 12:06:40 +08001717 block_netpoll_tx();
Veaceslav Falico5378c2e2013-10-21 11:48:30 +02001718 bond_select_active_slave(bond);
dingtianhongf80889a2014-02-12 12:06:40 +08001719 unblock_netpoll_tx();
Veaceslav Falico5378c2e2013-10-21 11:48:30 +02001720 }
Veaceslav Falico1f718f02013-09-25 09:20:10 +02001721
Mahesh Bandewaree637712014-10-04 17:45:01 -07001722 if (bond_mode_uses_xmit_hash(bond))
1723 bond_update_slave_arr(bond, NULL);
1724
Debabrata Banerjeed7bfa992018-05-09 19:32:11 -04001725
Veaceslav Falico76444f52014-07-15 19:35:58 +02001726 netdev_info(bond_dev, "Enslaving %s as %s interface with %s link\n",
1727 slave_dev->name,
1728 bond_is_active_slave(new_slave) ? "an active" : "a backup",
1729 new_slave->link != BOND_LINK_DOWN ? "an up" : "a down");
Linus Torvalds1da177e2005-04-16 15:20:36 -07001730
1731 /* enslave is successful */
Moni Shoua69e61132015-02-03 16:48:31 +02001732 bond_queue_slave_event(new_slave);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001733 return 0;
1734
1735/* Undo stages on error */
Xin Longd6d65b42018-03-26 01:16:46 +08001736err_sysfs_del:
1737 bond_sysfs_slave_del(new_slave);
1738
sfeldma@cumulusnetworks.com07699f92014-01-16 22:57:49 -08001739err_upper_unlink:
Jiri Pirko41f0b042015-12-03 12:12:14 +01001740 bond_upper_dev_unlink(bond, new_slave);
sfeldma@cumulusnetworks.com07699f92014-01-16 22:57:49 -08001741
Veaceslav Falico1f718f02013-09-25 09:20:10 +02001742err_unregister:
1743 netdev_rx_handler_unregister(slave_dev);
1744
stephen hemmingerf7d98212011-12-31 13:26:46 +00001745err_detach:
nikolay@redhat.com1ff412a2013-08-06 12:40:15 +02001746 vlan_vids_del_by_dev(slave_dev, bond_dev);
Nikolay Aleksandrov059b47e2014-09-09 23:17:00 +02001747 if (rcu_access_pointer(bond->primary_slave) == new_slave)
1748 RCU_INIT_POINTER(bond->primary_slave, NULL);
Eric Dumazet4740d632014-07-15 06:56:55 -07001749 if (rcu_access_pointer(bond->curr_active_slave) == new_slave) {
dingtianhongf80889a2014-02-12 12:06:40 +08001750 block_netpoll_tx();
dingtianhongc8517032013-12-13 10:20:07 +08001751 bond_change_active_slave(bond, NULL);
nikolay@redhat.com3c5913b532013-04-18 07:33:36 +00001752 bond_select_active_slave(bond);
dingtianhongf80889a2014-02-12 12:06:40 +08001753 unblock_netpoll_tx();
nikolay@redhat.com3c5913b532013-04-18 07:33:36 +00001754 }
Nikolay Aleksandrov059b47e2014-09-09 23:17:00 +02001755 /* either primary_slave or curr_active_slave might've changed */
1756 synchronize_rcu();
nikolay@redhat.comfc7a72a2013-04-18 07:33:37 +00001757 slave_disable_netpoll(new_slave);
stephen hemmingerf7d98212011-12-31 13:26:46 +00001758
Linus Torvalds1da177e2005-04-16 15:20:36 -07001759err_close:
nikolay@redhat.comb6a5a7b2013-04-11 09:18:56 +00001760 slave_dev->priv_flags &= ~IFF_BONDING;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001761 dev_close(slave_dev);
1762
1763err_restore_mac:
Karl Heiss03d84a52016-01-11 08:28:43 -05001764 slave_dev->flags &= ~IFF_SLAVE;
dingtianhong00503b62014-01-25 13:00:29 +08001765 if (!bond->params.fail_over_mac ||
Veaceslav Falico01844092014-05-15 21:39:55 +02001766 BOND_MODE(bond) != BOND_MODE_ACTIVEBACKUP) {
Jay Vosburgh3915c1e82008-05-17 21:10:14 -07001767 /* XXX TODO - fom follow mode needs to change master's
1768 * MAC if this slave's MAC is in use by the bond, or at
1769 * least print a warning.
1770 */
Joe Perchesada0f862014-02-15 16:02:17 -08001771 ether_addr_copy(addr.sa_data, new_slave->perm_hwaddr);
Moni Shoua2ab82852007-10-09 19:43:39 -07001772 addr.sa_family = slave_dev->type;
1773 dev_set_mac_address(slave_dev, &addr);
1774 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001775
Jiri Pirkob15ba0f2010-05-18 05:42:40 +00001776err_restore_mtu:
1777 dev_set_mtu(slave_dev, new_slave->original_mtu);
1778
Linus Torvalds1da177e2005-04-16 15:20:36 -07001779err_free:
dingtianhong3fdddd82014-05-12 15:08:43 +08001780 bond_free_slave(new_slave);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001781
1782err_undo_flags:
Nikolay Aleksandrovb8fad452013-06-12 00:07:01 +02001783 /* Enslave of first slave has failed and we need to fix master's mac */
Nikolay Aleksandrov7d5cd2c2015-07-15 22:57:01 +02001784 if (!bond_has_slaves(bond)) {
1785 if (ether_addr_equal_64bits(bond_dev->dev_addr,
1786 slave_dev->dev_addr))
1787 eth_hw_addr_random(bond_dev);
1788 if (bond_dev->type != ARPHRD_ETHER) {
Jay Vosburgh40baec22015-11-06 17:23:23 -08001789 dev_close(bond_dev);
Nikolay Aleksandrov7d5cd2c2015-07-15 22:57:01 +02001790 ether_setup(bond_dev);
1791 bond_dev->flags |= IFF_MASTER;
1792 bond_dev->priv_flags &= ~IFF_TX_SKB_SHARING;
1793 }
1794 }
Stephen Hemminger3d632c32009-06-12 19:02:48 +00001795
Linus Torvalds1da177e2005-04-16 15:20:36 -07001796 return res;
1797}
1798
Nikolay Aleksandrov547942c2014-09-15 17:19:34 +02001799/* Try to release the slave device <slave> from the bond device <master>
Linus Torvalds1da177e2005-04-16 15:20:36 -07001800 * It is legal to access curr_active_slave without a lock because all the function
Nikolay Aleksandrov8c0bc552014-09-11 22:49:28 +02001801 * is RTNL-locked. If "all" is true it means that the function is being called
nikolay@redhat.com08963412013-02-18 14:09:42 +00001802 * while destroying a bond interface and all slaves are being released.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001803 *
1804 * The rules for slave state should be:
1805 * for Active/Backup:
1806 * Active stays on all backups go down
1807 * for Bonded connections:
1808 * The first up interface should be left on and all others downed.
1809 */
nikolay@redhat.com08963412013-02-18 14:09:42 +00001810static int __bond_release_one(struct net_device *bond_dev,
1811 struct net_device *slave_dev,
1812 bool all)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001813{
Wang Chen454d7c92008-11-12 23:37:49 -08001814 struct bonding *bond = netdev_priv(bond_dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001815 struct slave *slave, *oldcurrent;
1816 struct sockaddr addr;
Neil Horman5a0068d2013-09-27 12:22:15 -04001817 int old_flags = bond_dev->flags;
Michał Mirosławc8f44af2011-11-15 15:29:55 +00001818 netdev_features_t old_features = bond_dev->features;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001819
1820 /* slave is not a slave or master is not master of this slave */
1821 if (!(slave_dev->flags & IFF_SLAVE) ||
Jiri Pirko471cb5a2013-01-03 22:49:01 +00001822 !netdev_has_upper_dev(slave_dev, bond_dev)) {
Wengang Wanga22a9e42014-12-23 09:24:36 +08001823 netdev_dbg(bond_dev, "cannot release %s\n",
Veaceslav Falico76444f52014-07-15 19:35:58 +02001824 slave_dev->name);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001825 return -EINVAL;
1826 }
1827
Neil Hormane843fa52010-10-13 16:01:50 +00001828 block_netpoll_tx();
Linus Torvalds1da177e2005-04-16 15:20:36 -07001829
1830 slave = bond_get_slave_by_dev(bond, slave_dev);
1831 if (!slave) {
1832 /* not a slave of this bond */
Veaceslav Falico76444f52014-07-15 19:35:58 +02001833 netdev_info(bond_dev, "%s not enslaved\n",
1834 slave_dev->name);
Neil Hormane843fa52010-10-13 16:01:50 +00001835 unblock_netpoll_tx();
Linus Torvalds1da177e2005-04-16 15:20:36 -07001836 return -EINVAL;
1837 }
1838
Jiri Pirko57beaca2015-12-03 12:12:21 +01001839 bond_set_slave_inactive_flags(slave, BOND_SLAVE_NOTIFY_NOW);
1840
sfeldma@cumulusnetworks.com07699f92014-01-16 22:57:49 -08001841 bond_sysfs_slave_del(slave);
1842
Andy Gospodarek5f0c5f72014-09-28 22:34:37 -04001843 /* recompute stats just before removing the slave */
1844 bond_get_stats(bond->dev, &bond->bond_stats);
1845
Jiri Pirko41f0b042015-12-03 12:12:14 +01001846 bond_upper_dev_unlink(bond, slave);
Jiri Pirko35d48902011-03-22 02:38:12 +00001847 /* unregister rx_handler early so bond_handle_frame wouldn't be called
1848 * for this slave anymore.
1849 */
1850 netdev_rx_handler_unregister(slave_dev);
Jiri Pirko35d48902011-03-22 02:38:12 +00001851
Nikolay Aleksandrove4702592014-09-11 22:49:27 +02001852 if (BOND_MODE(bond) == BOND_MODE_8023AD)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001853 bond_3ad_unbind_slave(slave);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001854
Mahesh Bandewaree637712014-10-04 17:45:01 -07001855 if (bond_mode_uses_xmit_hash(bond))
1856 bond_update_slave_arr(bond, slave);
1857
Veaceslav Falico76444f52014-07-15 19:35:58 +02001858 netdev_info(bond_dev, "Releasing %s interface %s\n",
1859 bond_is_active_slave(slave) ? "active" : "backup",
1860 slave_dev->name);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001861
Eric Dumazet4740d632014-07-15 06:56:55 -07001862 oldcurrent = rcu_access_pointer(bond->curr_active_slave);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001863
Eric Dumazet857417182014-07-15 06:56:56 -07001864 RCU_INIT_POINTER(bond->current_arp_slave, NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001865
dingtianhong00503b62014-01-25 13:00:29 +08001866 if (!all && (!bond->params.fail_over_mac ||
Veaceslav Falico01844092014-05-15 21:39:55 +02001867 BOND_MODE(bond) != BOND_MODE_ACTIVEBACKUP)) {
dingtianhong844223a2014-01-02 09:13:16 +08001868 if (ether_addr_equal_64bits(bond_dev->dev_addr, slave->perm_hwaddr) &&
Veaceslav Falico0965a1f2013-09-25 09:20:21 +02001869 bond_has_slaves(bond))
Veaceslav Falico76444f52014-07-15 19:35:58 +02001870 netdev_warn(bond_dev, "the permanent HWaddr of %s - %pM - is still in use by %s - set the HWaddr of %s to a different address to avoid conflicts\n",
1871 slave_dev->name, slave->perm_hwaddr,
1872 bond_dev->name, slave_dev->name);
nikolay@redhat.comdec1e902013-08-01 16:54:47 +02001873 }
1874
Nikolay Aleksandrov059b47e2014-09-09 23:17:00 +02001875 if (rtnl_dereference(bond->primary_slave) == slave)
1876 RCU_INIT_POINTER(bond->primary_slave, NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001877
Nikolay Aleksandrov1c72cfdc92014-09-11 22:49:24 +02001878 if (oldcurrent == slave)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001879 bond_change_active_slave(bond, NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001880
Holger Eitzenberger58402052008-12-09 23:07:13 -08001881 if (bond_is_lb(bond)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001882 /* Must be called only after the slave has been
1883 * detached from the list and the curr_active_slave
1884 * has been cleared (if our_slave == old_current),
1885 * but before a new active slave is selected.
1886 */
1887 bond_alb_deinit_slave(bond, slave);
1888 }
1889
nikolay@redhat.com08963412013-02-18 14:09:42 +00001890 if (all) {
Paul E. McKenney36708b82013-12-09 15:19:53 -08001891 RCU_INIT_POINTER(bond->curr_active_slave, NULL);
nikolay@redhat.com08963412013-02-18 14:09:42 +00001892 } else if (oldcurrent == slave) {
Nikolay Aleksandrov547942c2014-09-15 17:19:34 +02001893 /* Note that we hold RTNL over this sequence, so there
Jay Vosburgh059fe7a2007-10-17 17:37:49 -07001894 * is no concern that another slave add/remove event
1895 * will interfere.
1896 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001897 bond_select_active_slave(bond);
Jay Vosburgh059fe7a2007-10-17 17:37:49 -07001898 }
1899
Veaceslav Falico0965a1f2013-09-25 09:20:21 +02001900 if (!bond_has_slaves(bond)) {
Jay Vosburghff59c452006-03-27 13:27:43 -08001901 bond_set_carrier(bond);
Jiri Pirko409cc1f2013-01-30 11:08:11 +01001902 eth_hw_addr_random(bond_dev);
Willem de Bruijn59a70d12019-01-08 12:32:42 -05001903 bond->nest_level = SINGLE_DEPTH_NESTING;
1904 } else {
1905 bond->nest_level = dev_get_nest_level(bond_dev) + 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001906 }
1907
Neil Hormane843fa52010-10-13 16:01:50 +00001908 unblock_netpoll_tx();
nikolay@redhat.com278b2082013-08-01 16:54:51 +02001909 synchronize_rcu();
Nikolay Aleksandrovee6154e2014-02-26 14:20:30 +01001910 bond->slave_cnt--;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001911
Veaceslav Falico0965a1f2013-09-25 09:20:21 +02001912 if (!bond_has_slaves(bond)) {
Shlomo Pongratz2af73d42012-04-03 22:56:19 +00001913 call_netdevice_notifiers(NETDEV_CHANGEADDR, bond->dev);
Veaceslav Falico80028ea2013-03-06 07:10:32 +00001914 call_netdevice_notifiers(NETDEV_RELEASE, bond->dev);
1915 }
Shlomo Pongratz2af73d42012-04-03 22:56:19 +00001916
Michał Mirosławb2a103e2011-05-07 03:22:17 +00001917 bond_compute_features(bond);
1918 if (!(bond_dev->features & NETIF_F_VLAN_CHALLENGED) &&
1919 (old_features & NETIF_F_VLAN_CHALLENGED))
Veaceslav Falico76444f52014-07-15 19:35:58 +02001920 netdev_info(bond_dev, "last VLAN challenged slave %s left bond %s - VLAN blocking is removed\n",
1921 slave_dev->name, bond_dev->name);
Michał Mirosławb2a103e2011-05-07 03:22:17 +00001922
nikolay@redhat.com1ff412a2013-08-06 12:40:15 +02001923 vlan_vids_del_by_dev(slave_dev, bond_dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001924
Nikolay Aleksandrov547942c2014-09-15 17:19:34 +02001925 /* If the mode uses primary, then this case was handled above by
Jay Vosburgh303d1cb2013-05-31 11:57:30 +00001926 * bond_change_active_slave(..., NULL)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001927 */
Veaceslav Falicoec0865a2014-05-15 21:39:54 +02001928 if (!bond_uses_primary(bond)) {
Neil Horman5a0068d2013-09-27 12:22:15 -04001929 /* unset promiscuity level from slave
1930 * NOTE: The NETDEV_CHANGEADDR call above may change the value
1931 * of the IFF_PROMISC flag in the bond_dev, but we need the
1932 * value of that flag before that change, as that was the value
1933 * when this slave was attached, so we cache at the start of the
1934 * function and use it here. Same goes for ALLMULTI below
1935 */
1936 if (old_flags & IFF_PROMISC)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001937 dev_set_promiscuity(slave_dev, -1);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001938
1939 /* unset allmulti level from slave */
Neil Horman5a0068d2013-09-27 12:22:15 -04001940 if (old_flags & IFF_ALLMULTI)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001941 dev_set_allmulti(slave_dev, -1);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001942
Jay Vosburgh303d1cb2013-05-31 11:57:30 +00001943 bond_hw_addr_flush(bond_dev, slave_dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001944 }
1945
Amerigo Wang8a8efa22011-02-17 23:43:32 +00001946 slave_disable_netpoll(slave);
WANG Congf6dc31a2010-05-06 00:48:51 -07001947
Linus Torvalds1da177e2005-04-16 15:20:36 -07001948 /* close slave before restoring its mac address */
1949 dev_close(slave_dev);
1950
dingtianhong00503b62014-01-25 13:00:29 +08001951 if (bond->params.fail_over_mac != BOND_FOM_ACTIVE ||
Veaceslav Falico01844092014-05-15 21:39:55 +02001952 BOND_MODE(bond) != BOND_MODE_ACTIVEBACKUP) {
Moni Shoua2ab82852007-10-09 19:43:39 -07001953 /* restore original ("permanent") mac address */
Joe Perchesada0f862014-02-15 16:02:17 -08001954 ether_addr_copy(addr.sa_data, slave->perm_hwaddr);
Moni Shoua2ab82852007-10-09 19:43:39 -07001955 addr.sa_family = slave_dev->type;
1956 dev_set_mac_address(slave_dev, &addr);
1957 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001958
Jiri Pirkob15ba0f2010-05-18 05:42:40 +00001959 dev_set_mtu(slave_dev, slave->original_mtu);
1960
Jiri Pirko2d7011c2011-03-16 08:46:43 +00001961 slave_dev->priv_flags &= ~IFF_BONDING;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001962
dingtianhong3fdddd82014-05-12 15:08:43 +08001963 bond_free_slave(slave);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001964
Nikolay Aleksandrov547942c2014-09-15 17:19:34 +02001965 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001966}
1967
nikolay@redhat.com08963412013-02-18 14:09:42 +00001968/* A wrapper used because of ndo_del_link */
1969int bond_release(struct net_device *bond_dev, struct net_device *slave_dev)
1970{
1971 return __bond_release_one(bond_dev, slave_dev, false);
1972}
1973
Nikolay Aleksandrov547942c2014-09-15 17:19:34 +02001974/* First release a slave and then destroy the bond if no more slaves are left.
1975 * Must be under rtnl_lock when this function is called.
1976 */
stephen hemminger26d8ee72010-10-15 05:09:34 +00001977static int bond_release_and_destroy(struct net_device *bond_dev,
1978 struct net_device *slave_dev)
Moni Shouad90a1622007-10-09 19:43:43 -07001979{
Wang Chen454d7c92008-11-12 23:37:49 -08001980 struct bonding *bond = netdev_priv(bond_dev);
Moni Shouad90a1622007-10-09 19:43:43 -07001981 int ret;
1982
1983 ret = bond_release(bond_dev, slave_dev);
Veaceslav Falico0965a1f2013-09-25 09:20:21 +02001984 if (ret == 0 && !bond_has_slaves(bond)) {
Amerigo Wang8a8efa22011-02-17 23:43:32 +00001985 bond_dev->priv_flags |= IFF_DISABLE_NETPOLL;
Veaceslav Falico76444f52014-07-15 19:35:58 +02001986 netdev_info(bond_dev, "Destroying bond %s\n",
1987 bond_dev->name);
Nikolay Aleksandrov06f6d102015-07-15 21:52:51 +02001988 bond_remove_proc_entry(bond);
Stephen Hemminger9e716262009-06-12 19:02:47 +00001989 unregister_netdevice(bond_dev);
Moni Shouad90a1622007-10-09 19:43:43 -07001990 }
1991 return ret;
1992}
1993
Linus Torvalds1da177e2005-04-16 15:20:36 -07001994static int bond_info_query(struct net_device *bond_dev, struct ifbond *info)
1995{
Wang Chen454d7c92008-11-12 23:37:49 -08001996 struct bonding *bond = netdev_priv(bond_dev);
Moni Shoua69a23382015-02-03 16:48:30 +02001997 bond_fill_ifbond(bond, info);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001998 return 0;
1999}
2000
2001static int bond_slave_info_query(struct net_device *bond_dev, struct ifslave *info)
2002{
Wang Chen454d7c92008-11-12 23:37:49 -08002003 struct bonding *bond = netdev_priv(bond_dev);
Veaceslav Falico9caff1e2013-09-25 09:20:14 +02002004 struct list_head *iter;
nikolay@redhat.comdec1e902013-08-01 16:54:47 +02002005 int i = 0, res = -ENODEV;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002006 struct slave *slave;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002007
Veaceslav Falico9caff1e2013-09-25 09:20:14 +02002008 bond_for_each_slave(bond, slave, iter) {
nikolay@redhat.comdec1e902013-08-01 16:54:47 +02002009 if (i++ == (int)info->slave_id) {
Eric Dumazet689c96c2009-04-23 03:39:04 +00002010 res = 0;
Moni Shoua69a23382015-02-03 16:48:30 +02002011 bond_fill_ifslave(slave, info);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002012 break;
2013 }
2014 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07002015
Eric Dumazet689c96c2009-04-23 03:39:04 +00002016 return res;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002017}
2018
2019/*-------------------------------- Monitoring -------------------------------*/
2020
Eric Dumazet4740d632014-07-15 06:56:55 -07002021/* called with rcu_read_lock() */
Jay Vosburghf0c76d62008-07-02 18:21:58 -07002022static int bond_miimon_inspect(struct bonding *bond)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002023{
nikolay@redhat.comdec1e902013-08-01 16:54:47 +02002024 int link_state, commit = 0;
Veaceslav Falico9caff1e2013-09-25 09:20:14 +02002025 struct list_head *iter;
Jay Vosburghf0c76d62008-07-02 18:21:58 -07002026 struct slave *slave;
Jiri Pirko41f89102009-04-24 03:57:29 +00002027 bool ignore_updelay;
2028
Eric Dumazet4740d632014-07-15 06:56:55 -07002029 ignore_updelay = !rcu_dereference(bond->curr_active_slave);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002030
dingtianhong4cb4f972013-12-13 10:19:39 +08002031 bond_for_each_slave_rcu(bond, slave, iter) {
Jay Vosburghf0c76d62008-07-02 18:21:58 -07002032 slave->new_link = BOND_LINK_NOCHANGE;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002033
Jay Vosburghf0c76d62008-07-02 18:21:58 -07002034 link_state = bond_check_dev_link(bond, slave->dev, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002035
2036 switch (slave->link) {
Jay Vosburghf0c76d62008-07-02 18:21:58 -07002037 case BOND_LINK_UP:
2038 if (link_state)
2039 continue;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002040
Jiri Pirko5d397062015-12-03 12:12:19 +01002041 bond_set_slave_link_state(slave, BOND_LINK_FAIL,
2042 BOND_SLAVE_NOTIFY_LATER);
Jay Vosburghf0c76d62008-07-02 18:21:58 -07002043 slave->delay = bond->params.downdelay;
2044 if (slave->delay) {
Veaceslav Falico76444f52014-07-15 19:35:58 +02002045 netdev_info(bond->dev, "link status down for %sinterface %s, disabling it in %d ms\n",
2046 (BOND_MODE(bond) ==
2047 BOND_MODE_ACTIVEBACKUP) ?
2048 (bond_is_active_slave(slave) ?
2049 "active " : "backup ") : "",
2050 slave->dev->name,
2051 bond->params.downdelay * bond->params.miimon);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002052 }
Jay Vosburghf0c76d62008-07-02 18:21:58 -07002053 /*FALLTHRU*/
2054 case BOND_LINK_FAIL:
2055 if (link_state) {
Nikolay Aleksandrov547942c2014-09-15 17:19:34 +02002056 /* recovered before downdelay expired */
Jiri Pirko5d397062015-12-03 12:12:19 +01002057 bond_set_slave_link_state(slave, BOND_LINK_UP,
2058 BOND_SLAVE_NOTIFY_LATER);
Veaceslav Falico8e603462014-02-18 07:48:46 +01002059 slave->last_link_up = jiffies;
Veaceslav Falico76444f52014-07-15 19:35:58 +02002060 netdev_info(bond->dev, "link status up again after %d ms for interface %s\n",
2061 (bond->params.downdelay - slave->delay) *
2062 bond->params.miimon,
2063 slave->dev->name);
Mahesh Bandewar86138a42017-04-11 22:36:00 -07002064 commit++;
Jay Vosburghf0c76d62008-07-02 18:21:58 -07002065 continue;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002066 }
Jay Vosburghf0c76d62008-07-02 18:21:58 -07002067
2068 if (slave->delay <= 0) {
2069 slave->new_link = BOND_LINK_DOWN;
2070 commit++;
2071 continue;
2072 }
2073
2074 slave->delay--;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002075 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002076
Jay Vosburghf0c76d62008-07-02 18:21:58 -07002077 case BOND_LINK_DOWN:
2078 if (!link_state)
2079 continue;
2080
Jiri Pirko5d397062015-12-03 12:12:19 +01002081 bond_set_slave_link_state(slave, BOND_LINK_BACK,
2082 BOND_SLAVE_NOTIFY_LATER);
Jay Vosburghf0c76d62008-07-02 18:21:58 -07002083 slave->delay = bond->params.updelay;
2084
2085 if (slave->delay) {
Veaceslav Falico76444f52014-07-15 19:35:58 +02002086 netdev_info(bond->dev, "link status up for interface %s, enabling it in %d ms\n",
2087 slave->dev->name,
2088 ignore_updelay ? 0 :
2089 bond->params.updelay *
2090 bond->params.miimon);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002091 }
Jay Vosburghf0c76d62008-07-02 18:21:58 -07002092 /*FALLTHRU*/
2093 case BOND_LINK_BACK:
2094 if (!link_state) {
Moni Shoua69a23382015-02-03 16:48:30 +02002095 bond_set_slave_link_state(slave,
Jiri Pirko5d397062015-12-03 12:12:19 +01002096 BOND_LINK_DOWN,
2097 BOND_SLAVE_NOTIFY_LATER);
Veaceslav Falico76444f52014-07-15 19:35:58 +02002098 netdev_info(bond->dev, "link status down again after %d ms for interface %s\n",
2099 (bond->params.updelay - slave->delay) *
2100 bond->params.miimon,
2101 slave->dev->name);
Mahesh Bandewar86138a42017-04-11 22:36:00 -07002102 commit++;
Jay Vosburghf0c76d62008-07-02 18:21:58 -07002103 continue;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002104 }
Jay Vosburghf0c76d62008-07-02 18:21:58 -07002105
Jiri Pirko41f89102009-04-24 03:57:29 +00002106 if (ignore_updelay)
2107 slave->delay = 0;
2108
Jay Vosburghf0c76d62008-07-02 18:21:58 -07002109 if (slave->delay <= 0) {
2110 slave->new_link = BOND_LINK_UP;
2111 commit++;
Jiri Pirko41f89102009-04-24 03:57:29 +00002112 ignore_updelay = false;
Jay Vosburghf0c76d62008-07-02 18:21:58 -07002113 continue;
2114 }
2115
2116 slave->delay--;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002117 break;
Jay Vosburghf0c76d62008-07-02 18:21:58 -07002118 }
2119 }
2120
2121 return commit;
2122}
2123
2124static void bond_miimon_commit(struct bonding *bond)
2125{
Veaceslav Falico9caff1e2013-09-25 09:20:14 +02002126 struct list_head *iter;
Nikolay Aleksandrov059b47e2014-09-09 23:17:00 +02002127 struct slave *slave, *primary;
Jay Vosburghf0c76d62008-07-02 18:21:58 -07002128
Veaceslav Falico9caff1e2013-09-25 09:20:14 +02002129 bond_for_each_slave(bond, slave, iter) {
Jay Vosburghf0c76d62008-07-02 18:21:58 -07002130 switch (slave->new_link) {
2131 case BOND_LINK_NOCHANGE:
2132 continue;
2133
2134 case BOND_LINK_UP:
Andreas Bornae0c8ee2017-08-10 06:41:44 +02002135 if (bond_update_speed_duplex(slave) &&
2136 bond_needs_speed_duplex(bond)) {
Mahesh Bandewarbc5ad402017-04-03 18:38:39 -07002137 slave->link = BOND_LINK_DOWN;
Andreas Bornba559a52017-08-12 00:36:55 +02002138 if (net_ratelimit())
2139 netdev_warn(bond->dev,
2140 "failed to get link speed/duplex for %s\n",
2141 slave->dev->name);
Mahesh Bandewarc5b9d362017-03-27 11:37:37 -07002142 continue;
2143 }
Jiri Pirko5d397062015-12-03 12:12:19 +01002144 bond_set_slave_link_state(slave, BOND_LINK_UP,
2145 BOND_SLAVE_NOTIFY_NOW);
Veaceslav Falico8e603462014-02-18 07:48:46 +01002146 slave->last_link_up = jiffies;
Jay Vosburghf0c76d62008-07-02 18:21:58 -07002147
Nikolay Aleksandrov059b47e2014-09-09 23:17:00 +02002148 primary = rtnl_dereference(bond->primary_slave);
Veaceslav Falico01844092014-05-15 21:39:55 +02002149 if (BOND_MODE(bond) == BOND_MODE_8023AD) {
Jay Vosburghf0c76d62008-07-02 18:21:58 -07002150 /* prevent it from being the active one */
Jiri Pirkoe30bc062011-03-12 03:14:37 +00002151 bond_set_backup_slave(slave);
Veaceslav Falico01844092014-05-15 21:39:55 +02002152 } else if (BOND_MODE(bond) != BOND_MODE_ACTIVEBACKUP) {
Jay Vosburghf0c76d62008-07-02 18:21:58 -07002153 /* make it immediately active */
Jiri Pirkoe30bc062011-03-12 03:14:37 +00002154 bond_set_active_slave(slave);
Nikolay Aleksandrov059b47e2014-09-09 23:17:00 +02002155 } else if (slave != primary) {
Jay Vosburghf0c76d62008-07-02 18:21:58 -07002156 /* prevent it from being the active one */
Jiri Pirkoe30bc062011-03-12 03:14:37 +00002157 bond_set_backup_slave(slave);
Jay Vosburghf0c76d62008-07-02 18:21:58 -07002158 }
2159
Veaceslav Falico76444f52014-07-15 19:35:58 +02002160 netdev_info(bond->dev, "link status definitely up for interface %s, %u Mbps %s duplex\n",
2161 slave->dev->name,
2162 slave->speed == SPEED_UNKNOWN ? 0 : slave->speed,
2163 slave->duplex ? "full" : "half");
Jay Vosburghf0c76d62008-07-02 18:21:58 -07002164
2165 /* notify ad that the link status has changed */
Veaceslav Falico01844092014-05-15 21:39:55 +02002166 if (BOND_MODE(bond) == BOND_MODE_8023AD)
Jay Vosburghf0c76d62008-07-02 18:21:58 -07002167 bond_3ad_handle_link_change(slave, BOND_LINK_UP);
2168
Holger Eitzenberger58402052008-12-09 23:07:13 -08002169 if (bond_is_lb(bond))
Jay Vosburghf0c76d62008-07-02 18:21:58 -07002170 bond_alb_handle_link_change(bond, slave,
2171 BOND_LINK_UP);
2172
Mahesh Bandewaree637712014-10-04 17:45:01 -07002173 if (BOND_MODE(bond) == BOND_MODE_XOR)
2174 bond_update_slave_arr(bond, NULL);
2175
Nikolay Aleksandrov059b47e2014-09-09 23:17:00 +02002176 if (!bond->curr_active_slave || slave == primary)
Jay Vosburghf0c76d62008-07-02 18:21:58 -07002177 goto do_failover;
2178
2179 continue;
2180
2181 case BOND_LINK_DOWN:
Jay Vosburghfba4acd2008-10-30 17:41:14 -07002182 if (slave->link_failure_count < UINT_MAX)
2183 slave->link_failure_count++;
2184
Jiri Pirko5d397062015-12-03 12:12:19 +01002185 bond_set_slave_link_state(slave, BOND_LINK_DOWN,
2186 BOND_SLAVE_NOTIFY_NOW);
Jay Vosburghf0c76d62008-07-02 18:21:58 -07002187
Veaceslav Falico01844092014-05-15 21:39:55 +02002188 if (BOND_MODE(bond) == BOND_MODE_ACTIVEBACKUP ||
2189 BOND_MODE(bond) == BOND_MODE_8023AD)
dingtianhong5e5b0662014-02-26 11:05:22 +08002190 bond_set_slave_inactive_flags(slave,
2191 BOND_SLAVE_NOTIFY_NOW);
Jay Vosburghf0c76d62008-07-02 18:21:58 -07002192
Veaceslav Falico76444f52014-07-15 19:35:58 +02002193 netdev_info(bond->dev, "link status definitely down for interface %s, disabling it\n",
2194 slave->dev->name);
Jay Vosburghf0c76d62008-07-02 18:21:58 -07002195
Veaceslav Falico01844092014-05-15 21:39:55 +02002196 if (BOND_MODE(bond) == BOND_MODE_8023AD)
Jay Vosburghf0c76d62008-07-02 18:21:58 -07002197 bond_3ad_handle_link_change(slave,
2198 BOND_LINK_DOWN);
2199
Jiri Pirkoae63e802009-05-27 05:42:36 +00002200 if (bond_is_lb(bond))
Jay Vosburghf0c76d62008-07-02 18:21:58 -07002201 bond_alb_handle_link_change(bond, slave,
2202 BOND_LINK_DOWN);
2203
Mahesh Bandewaree637712014-10-04 17:45:01 -07002204 if (BOND_MODE(bond) == BOND_MODE_XOR)
2205 bond_update_slave_arr(bond, NULL);
2206
Eric Dumazet4740d632014-07-15 06:56:55 -07002207 if (slave == rcu_access_pointer(bond->curr_active_slave))
Jay Vosburghf0c76d62008-07-02 18:21:58 -07002208 goto do_failover;
2209
2210 continue;
2211
Linus Torvalds1da177e2005-04-16 15:20:36 -07002212 default:
Veaceslav Falico76444f52014-07-15 19:35:58 +02002213 netdev_err(bond->dev, "invalid new link %d on slave %s\n",
2214 slave->new_link, slave->dev->name);
Jay Vosburghf0c76d62008-07-02 18:21:58 -07002215 slave->new_link = BOND_LINK_NOCHANGE;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002216
Jay Vosburghf0c76d62008-07-02 18:21:58 -07002217 continue;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002218 }
2219
Jay Vosburghf0c76d62008-07-02 18:21:58 -07002220do_failover:
Neil Hormane843fa52010-10-13 16:01:50 +00002221 block_netpoll_tx();
Linus Torvalds1da177e2005-04-16 15:20:36 -07002222 bond_select_active_slave(bond);
Neil Hormane843fa52010-10-13 16:01:50 +00002223 unblock_netpoll_tx();
Jay Vosburghf0c76d62008-07-02 18:21:58 -07002224 }
Jay Vosburgh059fe7a2007-10-17 17:37:49 -07002225
Jay Vosburghf0c76d62008-07-02 18:21:58 -07002226 bond_set_carrier(bond);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002227}
2228
Nikolay Aleksandrov547942c2014-09-15 17:19:34 +02002229/* bond_mii_monitor
Jay Vosburgh0b0eef62007-10-17 17:37:48 -07002230 *
2231 * Really a wrapper that splits the mii monitor into two phases: an
Jay Vosburghf0c76d62008-07-02 18:21:58 -07002232 * inspection, then (if inspection indicates something needs to be done)
2233 * an acquisition of appropriate locks followed by a commit phase to
2234 * implement whatever link state changes are indicated.
Jay Vosburgh0b0eef62007-10-17 17:37:48 -07002235 */
stephen hemminger6da67d22013-12-30 10:43:41 -08002236static void bond_mii_monitor(struct work_struct *work)
Jay Vosburgh0b0eef62007-10-17 17:37:48 -07002237{
2238 struct bonding *bond = container_of(work, struct bonding,
2239 mii_work.work);
Ben Hutchingsad246c92011-04-26 15:25:52 +00002240 bool should_notify_peers = false;
David S. Miller1f2cd842013-10-28 00:11:22 -04002241 unsigned long delay;
Jay Vosburgh0b0eef62007-10-17 17:37:48 -07002242
David S. Miller1f2cd842013-10-28 00:11:22 -04002243 delay = msecs_to_jiffies(bond->params.miimon);
2244
2245 if (!bond_has_slaves(bond))
dingtianhong6b6c5262013-10-24 11:09:03 +08002246 goto re_arm;
dingtianhong6b6c5262013-10-24 11:09:03 +08002247
dingtianhong4cb4f972013-12-13 10:19:39 +08002248 rcu_read_lock();
2249
Ben Hutchingsad246c92011-04-26 15:25:52 +00002250 should_notify_peers = bond_should_notify_peers(bond);
2251
David S. Miller1f2cd842013-10-28 00:11:22 -04002252 if (bond_miimon_inspect(bond)) {
dingtianhong4cb4f972013-12-13 10:19:39 +08002253 rcu_read_unlock();
David S. Miller1f2cd842013-10-28 00:11:22 -04002254
2255 /* Race avoidance with bond_close cancel of workqueue */
2256 if (!rtnl_trylock()) {
David S. Miller1f2cd842013-10-28 00:11:22 -04002257 delay = 1;
2258 should_notify_peers = false;
2259 goto re_arm;
2260 }
2261
Jay Vosburghf0c76d62008-07-02 18:21:58 -07002262 bond_miimon_commit(bond);
2263
David S. Miller1f2cd842013-10-28 00:11:22 -04002264 rtnl_unlock(); /* might sleep, hold no other locks */
dingtianhong4cb4f972013-12-13 10:19:39 +08002265 } else
2266 rcu_read_unlock();
Jay Vosburgh0b0eef62007-10-17 17:37:48 -07002267
Jay Vosburghf0c76d62008-07-02 18:21:58 -07002268re_arm:
Jay Vosburghe6d265e2011-10-28 15:42:50 +00002269 if (bond->params.miimon)
David S. Miller1f2cd842013-10-28 00:11:22 -04002270 queue_delayed_work(bond->wq, &bond->mii_work, delay);
2271
David S. Miller1f2cd842013-10-28 00:11:22 -04002272 if (should_notify_peers) {
2273 if (!rtnl_trylock())
2274 return;
2275 call_netdevice_notifiers(NETDEV_NOTIFY_PEERS, bond->dev);
2276 rtnl_unlock();
2277 }
Jay Vosburgh0b0eef62007-10-17 17:37:48 -07002278}
Jay Vosburghc3ade5c2005-06-26 17:52:20 -04002279
Veaceslav Falico50223ce2013-08-28 23:25:11 +02002280static bool bond_has_this_ip(struct bonding *bond, __be32 ip)
Jay Vosburghf5b2b962006-09-22 21:54:53 -07002281{
Veaceslav Falico50223ce2013-08-28 23:25:11 +02002282 struct net_device *upper;
2283 struct list_head *iter;
2284 bool ret = false;
Jay Vosburghf5b2b962006-09-22 21:54:53 -07002285
Andy Gospodarekeaddcd72012-03-22 16:14:29 +00002286 if (ip == bond_confirm_addr(bond->dev, 0, ip))
Veaceslav Falico50223ce2013-08-28 23:25:11 +02002287 return true;
Jay Vosburghf5b2b962006-09-22 21:54:53 -07002288
Veaceslav Falico50223ce2013-08-28 23:25:11 +02002289 rcu_read_lock();
Veaceslav Falico2f268f12013-09-25 09:20:07 +02002290 netdev_for_each_all_upper_dev_rcu(bond->dev, upper, iter) {
Veaceslav Falico50223ce2013-08-28 23:25:11 +02002291 if (ip == bond_confirm_addr(upper, 0, ip)) {
2292 ret = true;
2293 break;
2294 }
Jay Vosburghf5b2b962006-09-22 21:54:53 -07002295 }
Veaceslav Falico50223ce2013-08-28 23:25:11 +02002296 rcu_read_unlock();
Jay Vosburghf5b2b962006-09-22 21:54:53 -07002297
Veaceslav Falico50223ce2013-08-28 23:25:11 +02002298 return ret;
Jay Vosburghf5b2b962006-09-22 21:54:53 -07002299}
2300
Nikolay Aleksandrov547942c2014-09-15 17:19:34 +02002301/* We go to the (large) trouble of VLAN tagging ARP frames because
Jay Vosburghc3ade5c2005-06-26 17:52:20 -04002302 * switches in VLAN mode (especially if ports are configured as
2303 * "native" to a VLAN) might not pass non-tagged frames.
2304 */
dingtianhongfbd929f2014-03-25 17:44:43 +08002305static void bond_arp_send(struct net_device *slave_dev, int arp_op,
2306 __be32 dest_ip, __be32 src_ip,
Vlad Yasevich44a40852014-05-16 17:20:38 -04002307 struct bond_vlan_tag *tags)
Jay Vosburghc3ade5c2005-06-26 17:52:20 -04002308{
2309 struct sk_buff *skb;
Veaceslav Falico3e403a72014-07-17 17:02:23 +02002310 struct bond_vlan_tag *outer_tag = tags;
Jay Vosburghc3ade5c2005-06-26 17:52:20 -04002311
Veaceslav Falico76444f52014-07-15 19:35:58 +02002312 netdev_dbg(slave_dev, "arp %d on slave %s: dst %pI4 src %pI4\n",
2313 arp_op, slave_dev->name, &dest_ip, &src_ip);
Stephen Hemminger3d632c32009-06-12 19:02:48 +00002314
Jay Vosburghc3ade5c2005-06-26 17:52:20 -04002315 skb = arp_create(arp_op, ETH_P_ARP, dest_ip, slave_dev, src_ip,
2316 NULL, slave_dev->dev_addr, NULL);
2317
2318 if (!skb) {
dingtianhong4873ac32014-03-25 17:44:44 +08002319 net_err_ratelimited("ARP packet allocation failed\n");
Jay Vosburghc3ade5c2005-06-26 17:52:20 -04002320 return;
2321 }
dingtianhongfbd929f2014-03-25 17:44:43 +08002322
Veaceslav Falico3e403a72014-07-17 17:02:23 +02002323 if (!tags || tags->vlan_proto == VLAN_N_VID)
2324 goto xmit;
2325
2326 tags++;
2327
Vlad Yasevich44a40852014-05-16 17:20:38 -04002328 /* Go through all the tags backwards and add them to the packet */
Veaceslav Falico3e403a72014-07-17 17:02:23 +02002329 while (tags->vlan_proto != VLAN_N_VID) {
2330 if (!tags->vlan_id) {
2331 tags++;
Vlad Yasevich44a40852014-05-16 17:20:38 -04002332 continue;
Veaceslav Falico3e403a72014-07-17 17:02:23 +02002333 }
Vlad Yasevich44a40852014-05-16 17:20:38 -04002334
Veaceslav Falico76444f52014-07-15 19:35:58 +02002335 netdev_dbg(slave_dev, "inner tag: proto %X vid %X\n",
Veaceslav Falico3e403a72014-07-17 17:02:23 +02002336 ntohs(outer_tag->vlan_proto), tags->vlan_id);
Jiri Pirko62749e22014-11-19 14:04:58 +01002337 skb = vlan_insert_tag_set_proto(skb, tags->vlan_proto,
2338 tags->vlan_id);
Vlad Yasevich44a40852014-05-16 17:20:38 -04002339 if (!skb) {
2340 net_err_ratelimited("failed to insert inner VLAN tag\n");
2341 return;
2342 }
Veaceslav Falico3e403a72014-07-17 17:02:23 +02002343
2344 tags++;
Vlad Yasevich44a40852014-05-16 17:20:38 -04002345 }
2346 /* Set the outer tag */
Veaceslav Falico3e403a72014-07-17 17:02:23 +02002347 if (outer_tag->vlan_id) {
Veaceslav Falico76444f52014-07-15 19:35:58 +02002348 netdev_dbg(slave_dev, "outer tag: proto %X vid %X\n",
Veaceslav Falico3e403a72014-07-17 17:02:23 +02002349 ntohs(outer_tag->vlan_proto), outer_tag->vlan_id);
Jiri Pirkob4bef1b2014-11-19 14:04:57 +01002350 __vlan_hwaccel_put_tag(skb, outer_tag->vlan_proto,
2351 outer_tag->vlan_id);
Jay Vosburghc3ade5c2005-06-26 17:52:20 -04002352 }
Veaceslav Falico3e403a72014-07-17 17:02:23 +02002353
2354xmit:
Jay Vosburghc3ade5c2005-06-26 17:52:20 -04002355 arp_xmit(skb);
2356}
2357
Vlad Yasevich44a40852014-05-16 17:20:38 -04002358/* Validate the device path between the @start_dev and the @end_dev.
2359 * The path is valid if the @end_dev is reachable through device
2360 * stacking.
2361 * When the path is validated, collect any vlan information in the
2362 * path.
2363 */
Veaceslav Falico3e403a72014-07-17 17:02:23 +02002364struct bond_vlan_tag *bond_verify_device_path(struct net_device *start_dev,
2365 struct net_device *end_dev,
2366 int level)
Vlad Yasevich44a40852014-05-16 17:20:38 -04002367{
Veaceslav Falico3e403a72014-07-17 17:02:23 +02002368 struct bond_vlan_tag *tags;
Vlad Yasevich44a40852014-05-16 17:20:38 -04002369 struct net_device *upper;
2370 struct list_head *iter;
Vlad Yasevich44a40852014-05-16 17:20:38 -04002371
Veaceslav Falico3e403a72014-07-17 17:02:23 +02002372 if (start_dev == end_dev) {
2373 tags = kzalloc(sizeof(*tags) * (level + 1), GFP_ATOMIC);
2374 if (!tags)
2375 return ERR_PTR(-ENOMEM);
2376 tags[level].vlan_proto = VLAN_N_VID;
2377 return tags;
Vlad Yasevich44a40852014-05-16 17:20:38 -04002378 }
2379
Veaceslav Falico3e403a72014-07-17 17:02:23 +02002380 netdev_for_each_upper_dev_rcu(start_dev, upper, iter) {
2381 tags = bond_verify_device_path(upper, end_dev, level + 1);
2382 if (IS_ERR_OR_NULL(tags)) {
2383 if (IS_ERR(tags))
2384 return tags;
2385 continue;
2386 }
2387 if (is_vlan_dev(upper)) {
2388 tags[level].vlan_proto = vlan_dev_vlan_proto(upper);
2389 tags[level].vlan_id = vlan_dev_vlan_id(upper);
2390 }
2391
2392 return tags;
2393 }
2394
2395 return NULL;
Vlad Yasevich44a40852014-05-16 17:20:38 -04002396}
Jay Vosburghc3ade5c2005-06-26 17:52:20 -04002397
Linus Torvalds1da177e2005-04-16 15:20:36 -07002398static void bond_arp_send_all(struct bonding *bond, struct slave *slave)
2399{
Jay Vosburghc3ade5c2005-06-26 17:52:20 -04002400 struct rtable *rt;
Veaceslav Falico3e403a72014-07-17 17:02:23 +02002401 struct bond_vlan_tag *tags;
Veaceslav Falico27bc11e2013-08-28 23:25:10 +02002402 __be32 *targets = bond->params.arp_targets, addr;
dingtianhongfbd929f2014-03-25 17:44:43 +08002403 int i;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002404
Veaceslav Falico27bc11e2013-08-28 23:25:10 +02002405 for (i = 0; i < BOND_MAX_ARP_TARGETS && targets[i]; i++) {
Veaceslav Falico76444f52014-07-15 19:35:58 +02002406 netdev_dbg(bond->dev, "basa: target %pI4\n", &targets[i]);
Veaceslav Falico3e403a72014-07-17 17:02:23 +02002407 tags = NULL;
Jay Vosburghc3ade5c2005-06-26 17:52:20 -04002408
Veaceslav Falico27bc11e2013-08-28 23:25:10 +02002409 /* Find out through which dev should the packet go */
David S. Miller78fbfd82011-03-12 00:00:52 -05002410 rt = ip_route_output(dev_net(bond->dev), targets[i], 0,
2411 RTO_ONLINK, 0);
David S. Millerb23dd4f2011-03-02 14:31:35 -08002412 if (IS_ERR(rt)) {
Veaceslav Falico28572762014-02-28 12:39:19 +01002413 /* there's no route to target - try to send arp
2414 * probe to generate any traffic (arp_validate=0)
2415 */
dingtianhong4873ac32014-03-25 17:44:44 +08002416 if (bond->params.arp_validate)
2417 net_warn_ratelimited("%s: no route to arp_ip_target %pI4 and arp_validate is set\n",
2418 bond->dev->name,
2419 &targets[i]);
Vlad Yasevich44a40852014-05-16 17:20:38 -04002420 bond_arp_send(slave->dev, ARPOP_REQUEST, targets[i],
2421 0, tags);
Jay Vosburghc3ade5c2005-06-26 17:52:20 -04002422 continue;
2423 }
2424
Veaceslav Falico27bc11e2013-08-28 23:25:10 +02002425 /* bond device itself */
2426 if (rt->dst.dev == bond->dev)
2427 goto found;
2428
2429 rcu_read_lock();
Veaceslav Falico3e403a72014-07-17 17:02:23 +02002430 tags = bond_verify_device_path(bond->dev, rt->dst.dev, 0);
Veaceslav Falico27bc11e2013-08-28 23:25:10 +02002431 rcu_read_unlock();
2432
Veaceslav Falico3e403a72014-07-17 17:02:23 +02002433 if (!IS_ERR_OR_NULL(tags))
Vlad Yasevich44a40852014-05-16 17:20:38 -04002434 goto found;
2435
Veaceslav Falico27bc11e2013-08-28 23:25:10 +02002436 /* Not our device - skip */
Veaceslav Falico76444f52014-07-15 19:35:58 +02002437 netdev_dbg(bond->dev, "no path to arp_ip_target %pI4 via rt.dev %s\n",
2438 &targets[i], rt->dst.dev ? rt->dst.dev->name : "NULL");
Veaceslav Falico3e325822013-08-28 23:25:16 +02002439
Jay Vosburghed4b9f82005-09-14 14:52:09 -07002440 ip_rt_put(rt);
Veaceslav Falico27bc11e2013-08-28 23:25:10 +02002441 continue;
2442
2443found:
2444 addr = bond_confirm_addr(rt->dst.dev, targets[i], 0);
2445 ip_rt_put(rt);
2446 bond_arp_send(slave->dev, ARPOP_REQUEST, targets[i],
Vlad Yasevich44a40852014-05-16 17:20:38 -04002447 addr, tags);
Dan Carpentera67eed52014-07-25 15:21:21 +03002448 kfree(tags);
Jay Vosburghc3ade5c2005-06-26 17:52:20 -04002449 }
2450}
2451
Al Virod3bb52b2007-08-22 20:06:58 -04002452static void bond_validate_arp(struct bonding *bond, struct slave *slave, __be32 sip, __be32 tip)
Jay Vosburghf5b2b962006-09-22 21:54:53 -07002453{
Veaceslav Falico8599b522013-06-24 11:49:34 +02002454 int i;
2455
Veaceslav Falico87a7b842013-06-24 11:49:29 +02002456 if (!sip || !bond_has_this_ip(bond, tip)) {
Veaceslav Falico76444f52014-07-15 19:35:58 +02002457 netdev_dbg(bond->dev, "bva: sip %pI4 tip %pI4 not found\n",
2458 &sip, &tip);
Veaceslav Falico87a7b842013-06-24 11:49:29 +02002459 return;
Jay Vosburghf5b2b962006-09-22 21:54:53 -07002460 }
Veaceslav Falico87a7b842013-06-24 11:49:29 +02002461
Veaceslav Falico8599b522013-06-24 11:49:34 +02002462 i = bond_get_targets_ip(bond->params.arp_targets, sip);
2463 if (i == -1) {
Veaceslav Falico76444f52014-07-15 19:35:58 +02002464 netdev_dbg(bond->dev, "bva: sip %pI4 not found in targets\n",
2465 &sip);
Veaceslav Falico87a7b842013-06-24 11:49:29 +02002466 return;
2467 }
Veaceslav Falico49f17de2014-02-18 07:48:47 +01002468 slave->last_rx = jiffies;
Veaceslav Falico8599b522013-06-24 11:49:34 +02002469 slave->target_last_arp_rx[i] = jiffies;
Jay Vosburghf5b2b962006-09-22 21:54:53 -07002470}
2471
nikolay@redhat.com5bb9e0b2013-09-07 00:00:26 +02002472int bond_arp_rcv(const struct sk_buff *skb, struct bonding *bond,
2473 struct slave *slave)
Jay Vosburghf5b2b962006-09-22 21:54:53 -07002474{
Eric Dumazetde063b72012-06-11 19:23:07 +00002475 struct arphdr *arp = (struct arphdr *)skb->data;
Jay Vosburgh21a75f02016-02-02 13:35:56 -08002476 struct slave *curr_active_slave, *curr_arp_slave;
Jay Vosburghf5b2b962006-09-22 21:54:53 -07002477 unsigned char *arp_ptr;
Al Virod3bb52b2007-08-22 20:06:58 -04002478 __be32 sip, tip;
Veaceslav Falicof2cb6912014-02-18 07:48:42 +01002479 int alen, is_arp = skb->protocol == __cpu_to_be16(ETH_P_ARP);
Jay Vosburghf5b2b962006-09-22 21:54:53 -07002480
Veaceslav Falicof2cb6912014-02-18 07:48:42 +01002481 if (!slave_do_arp_validate(bond, slave)) {
dingtianhongbedabf92014-05-07 22:10:20 +08002482 if ((slave_do_arp_validate_only(bond) && is_arp) ||
2483 !slave_do_arp_validate_only(bond))
Veaceslav Falico49f17de2014-02-18 07:48:47 +01002484 slave->last_rx = jiffies;
Jiri Bohac13a8e0c2012-05-09 01:01:40 +00002485 return RX_HANDLER_ANOTHER;
Veaceslav Falicof2cb6912014-02-18 07:48:42 +01002486 } else if (!is_arp) {
2487 return RX_HANDLER_ANOTHER;
2488 }
Veaceslav Falico2c146102013-06-24 11:49:31 +02002489
Eric Dumazetde063b72012-06-11 19:23:07 +00002490 alen = arp_hdr_len(bond->dev);
Jay Vosburghf5b2b962006-09-22 21:54:53 -07002491
Veaceslav Falico76444f52014-07-15 19:35:58 +02002492 netdev_dbg(bond->dev, "bond_arp_rcv: skb->dev %s\n",
2493 skb->dev->name);
Jay Vosburghf5b2b962006-09-22 21:54:53 -07002494
Eric Dumazetde063b72012-06-11 19:23:07 +00002495 if (alen > skb_headlen(skb)) {
2496 arp = kmalloc(alen, GFP_ATOMIC);
2497 if (!arp)
2498 goto out_unlock;
2499 if (skb_copy_bits(skb, 0, arp, alen) < 0)
2500 goto out_unlock;
2501 }
Jay Vosburghf5b2b962006-09-22 21:54:53 -07002502
Jiri Pirko3aba8912011-04-19 03:48:16 +00002503 if (arp->ar_hln != bond->dev->addr_len ||
Jay Vosburghf5b2b962006-09-22 21:54:53 -07002504 skb->pkt_type == PACKET_OTHERHOST ||
2505 skb->pkt_type == PACKET_LOOPBACK ||
2506 arp->ar_hrd != htons(ARPHRD_ETHER) ||
2507 arp->ar_pro != htons(ETH_P_IP) ||
2508 arp->ar_pln != 4)
2509 goto out_unlock;
2510
2511 arp_ptr = (unsigned char *)(arp + 1);
Jiri Pirko3aba8912011-04-19 03:48:16 +00002512 arp_ptr += bond->dev->addr_len;
Jay Vosburghf5b2b962006-09-22 21:54:53 -07002513 memcpy(&sip, arp_ptr, 4);
Jiri Pirko3aba8912011-04-19 03:48:16 +00002514 arp_ptr += 4 + bond->dev->addr_len;
Jay Vosburghf5b2b962006-09-22 21:54:53 -07002515 memcpy(&tip, arp_ptr, 4);
2516
Veaceslav Falico76444f52014-07-15 19:35:58 +02002517 netdev_dbg(bond->dev, "bond_arp_rcv: %s/%d av %d sv %d sip %pI4 tip %pI4\n",
2518 slave->dev->name, bond_slave_state(slave),
2519 bond->params.arp_validate, slave_do_arp_validate(bond, slave),
2520 &sip, &tip);
Jay Vosburghf5b2b962006-09-22 21:54:53 -07002521
Veaceslav Falico010d3c32014-02-20 12:07:57 +01002522 curr_active_slave = rcu_dereference(bond->curr_active_slave);
Jay Vosburgh21a75f02016-02-02 13:35:56 -08002523 curr_arp_slave = rcu_dereference(bond->current_arp_slave);
Veaceslav Falico010d3c32014-02-20 12:07:57 +01002524
Jay Vosburgh21a75f02016-02-02 13:35:56 -08002525 /* We 'trust' the received ARP enough to validate it if:
Veaceslav Falicoaeea64a2013-06-24 11:49:32 +02002526 *
Jay Vosburgh21a75f02016-02-02 13:35:56 -08002527 * (a) the slave receiving the ARP is active (which includes the
2528 * current ARP slave, if any), or
2529 *
2530 * (b) the receiving slave isn't active, but there is a currently
2531 * active slave and it received valid arp reply(s) after it became
2532 * the currently active slave, or
2533 *
2534 * (c) there is an ARP slave that sent an ARP during the prior ARP
2535 * interval, and we receive an ARP reply on any slave. We accept
2536 * these because switch FDB update delays may deliver the ARP
2537 * reply to a slave other than the sender of the ARP request.
2538 *
2539 * Note: for (b), backup slaves are receiving the broadcast ARP
2540 * request, not a reply. This request passes from the sending
2541 * slave through the L2 switch(es) to the receiving slave. Since
2542 * this is checking the request, sip/tip are swapped for
2543 * validation.
2544 *
2545 * This is done to avoid endless looping when we can't reach the
Veaceslav Falicoaeea64a2013-06-24 11:49:32 +02002546 * arp_ip_target and fool ourselves with our own arp requests.
Jay Vosburghf5b2b962006-09-22 21:54:53 -07002547 */
Jiri Pirkoe30bc062011-03-12 03:14:37 +00002548 if (bond_is_active_slave(slave))
Jay Vosburghf5b2b962006-09-22 21:54:53 -07002549 bond_validate_arp(bond, slave, sip, tip);
Veaceslav Falico010d3c32014-02-20 12:07:57 +01002550 else if (curr_active_slave &&
2551 time_after(slave_last_rx(bond, curr_active_slave),
2552 curr_active_slave->last_link_up))
Jay Vosburghf5b2b962006-09-22 21:54:53 -07002553 bond_validate_arp(bond, slave, tip, sip);
Jay Vosburgh21a75f02016-02-02 13:35:56 -08002554 else if (curr_arp_slave && (arp->ar_op == htons(ARPOP_REPLY)) &&
2555 bond_time_in_interval(bond,
2556 dev_trans_start(curr_arp_slave->dev), 1))
2557 bond_validate_arp(bond, slave, sip, tip);
Jay Vosburghf5b2b962006-09-22 21:54:53 -07002558
2559out_unlock:
Eric Dumazetde063b72012-06-11 19:23:07 +00002560 if (arp != (struct arphdr *)skb->data)
2561 kfree(arp);
Jiri Bohac13a8e0c2012-05-09 01:01:40 +00002562 return RX_HANDLER_ANOTHER;
Jay Vosburghf5b2b962006-09-22 21:54:53 -07002563}
2564
Veaceslav Falicoe7f63f12013-08-03 03:50:36 +02002565/* function to verify if we're in the arp_interval timeslice, returns true if
2566 * (last_act - arp_interval) <= jiffies <= (last_act + mod * arp_interval +
2567 * arp_interval/2) . the arp_interval/2 is needed for really fast networks.
2568 */
2569static bool bond_time_in_interval(struct bonding *bond, unsigned long last_act,
2570 int mod)
2571{
2572 int delta_in_ticks = msecs_to_jiffies(bond->params.arp_interval);
2573
2574 return time_in_range(jiffies,
2575 last_act - delta_in_ticks,
2576 last_act + mod * delta_in_ticks + delta_in_ticks/2);
2577}
2578
Nikolay Aleksandrov547942c2014-09-15 17:19:34 +02002579/* This function is called regularly to monitor each slave's link
Linus Torvalds1da177e2005-04-16 15:20:36 -07002580 * ensuring that traffic is being sent and received when arp monitoring
2581 * is used in load-balancing mode. if the adapter has been dormant, then an
2582 * arp is transmitted to generate traffic. see activebackup_arp_monitor for
2583 * arp monitoring in active backup mode.
2584 */
stephen hemminger6da67d22013-12-30 10:43:41 -08002585static void bond_loadbalance_arp_mon(struct work_struct *work)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002586{
Jay Vosburgh1b76b312007-10-17 17:37:45 -07002587 struct bonding *bond = container_of(work, struct bonding,
2588 arp_work.work);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002589 struct slave *slave, *oldcurrent;
Veaceslav Falico9caff1e2013-09-25 09:20:14 +02002590 struct list_head *iter;
Ding Tianhong6fde8f02014-01-28 11:48:53 +08002591 int do_failover = 0, slave_state_changed = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002592
David S. Miller1f2cd842013-10-28 00:11:22 -04002593 if (!bond_has_slaves(bond))
dingtianhong7f1bb572013-10-24 11:09:17 +08002594 goto re_arm;
dingtianhong7f1bb572013-10-24 11:09:17 +08002595
dingtianhong2e52f4f2013-12-13 10:19:50 +08002596 rcu_read_lock();
2597
Eric Dumazet4740d632014-07-15 06:56:55 -07002598 oldcurrent = rcu_dereference(bond->curr_active_slave);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002599 /* see if any of the previous devices are up now (i.e. they have
2600 * xmt and rcv traffic). the curr_active_slave does not come into
Veaceslav Falico8e603462014-02-18 07:48:46 +01002601 * the picture unless it is null. also, slave->last_link_up is not
2602 * needed here because we send an arp on each slave and give a slave
2603 * as long as it needs to get the tx/rx within the delta.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002604 * TODO: what about up/down delay in arp mode? it wasn't here before
2605 * so it can wait
2606 */
dingtianhong2e52f4f2013-12-13 10:19:50 +08002607 bond_for_each_slave_rcu(bond, slave, iter) {
Jiri Bohaccb32f2a2010-09-02 05:45:54 +00002608 unsigned long trans_start = dev_trans_start(slave->dev);
2609
Nithin Sujir1a6906e2017-05-24 19:45:17 -07002610 slave->new_link = BOND_LINK_NOCHANGE;
2611
Linus Torvalds1da177e2005-04-16 15:20:36 -07002612 if (slave->link != BOND_LINK_UP) {
Veaceslav Falicoe7f63f12013-08-03 03:50:36 +02002613 if (bond_time_in_interval(bond, trans_start, 1) &&
Veaceslav Falico49f17de2014-02-18 07:48:47 +01002614 bond_time_in_interval(bond, slave->last_rx, 1)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002615
Nithin Sujir1a6906e2017-05-24 19:45:17 -07002616 slave->new_link = BOND_LINK_UP;
Ding Tianhong6fde8f02014-01-28 11:48:53 +08002617 slave_state_changed = 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002618
2619 /* primary_slave has no meaning in round-robin
2620 * mode. the window of a slave being up and
2621 * curr_active_slave being null after enslaving
2622 * is closed.
2623 */
2624 if (!oldcurrent) {
Veaceslav Falico76444f52014-07-15 19:35:58 +02002625 netdev_info(bond->dev, "link status definitely up for interface %s\n",
2626 slave->dev->name);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002627 do_failover = 1;
2628 } else {
Veaceslav Falico76444f52014-07-15 19:35:58 +02002629 netdev_info(bond->dev, "interface %s is now up\n",
2630 slave->dev->name);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002631 }
2632 }
2633 } else {
2634 /* slave->link == BOND_LINK_UP */
2635
2636 /* not all switches will respond to an arp request
2637 * when the source ip is 0, so don't take the link down
2638 * if we don't know our ip yet
2639 */
Veaceslav Falicoe7f63f12013-08-03 03:50:36 +02002640 if (!bond_time_in_interval(bond, trans_start, 2) ||
Veaceslav Falico49f17de2014-02-18 07:48:47 +01002641 !bond_time_in_interval(bond, slave->last_rx, 2)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002642
Nithin Sujir1a6906e2017-05-24 19:45:17 -07002643 slave->new_link = BOND_LINK_DOWN;
Ding Tianhong6fde8f02014-01-28 11:48:53 +08002644 slave_state_changed = 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002645
Stephen Hemminger3d632c32009-06-12 19:02:48 +00002646 if (slave->link_failure_count < UINT_MAX)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002647 slave->link_failure_count++;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002648
Veaceslav Falico76444f52014-07-15 19:35:58 +02002649 netdev_info(bond->dev, "interface %s is now down\n",
2650 slave->dev->name);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002651
Stephen Hemminger3d632c32009-06-12 19:02:48 +00002652 if (slave == oldcurrent)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002653 do_failover = 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002654 }
2655 }
2656
2657 /* note: if switch is in round-robin mode, all links
2658 * must tx arp to ensure all links rx an arp - otherwise
2659 * links may oscillate or not come up at all; if switch is
2660 * in something like xor mode, there is nothing we can
2661 * do - all replies will be rx'ed on same link causing slaves
2662 * to be unstable during low/no traffic periods
2663 */
Veaceslav Falicob6adc612014-05-15 21:39:57 +02002664 if (bond_slave_is_up(slave))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002665 bond_arp_send_all(bond, slave);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002666 }
2667
dingtianhong2e52f4f2013-12-13 10:19:50 +08002668 rcu_read_unlock();
2669
Ding Tianhong6fde8f02014-01-28 11:48:53 +08002670 if (do_failover || slave_state_changed) {
dingtianhong2e52f4f2013-12-13 10:19:50 +08002671 if (!rtnl_trylock())
2672 goto re_arm;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002673
Nithin Sujir1a6906e2017-05-24 19:45:17 -07002674 bond_for_each_slave(bond, slave, iter) {
2675 if (slave->new_link != BOND_LINK_NOCHANGE)
2676 slave->link = slave->new_link;
2677 }
2678
Ding Tianhong6fde8f02014-01-28 11:48:53 +08002679 if (slave_state_changed) {
2680 bond_slave_state_change(bond);
Mahesh Bandewaree637712014-10-04 17:45:01 -07002681 if (BOND_MODE(bond) == BOND_MODE_XOR)
2682 bond_update_slave_arr(bond, NULL);
Nikolay Aleksandrovb8e45002014-11-18 15:14:44 +01002683 }
2684 if (do_failover) {
Ding Tianhong6fde8f02014-01-28 11:48:53 +08002685 block_netpoll_tx();
Ding Tianhong6fde8f02014-01-28 11:48:53 +08002686 bond_select_active_slave(bond);
Ding Tianhong6fde8f02014-01-28 11:48:53 +08002687 unblock_netpoll_tx();
2688 }
dingtianhong2e52f4f2013-12-13 10:19:50 +08002689 rtnl_unlock();
Linus Torvalds1da177e2005-04-16 15:20:36 -07002690 }
2691
2692re_arm:
Jay Vosburghe6d265e2011-10-28 15:42:50 +00002693 if (bond->params.arp_interval)
Veaceslav Falicoe7f63f12013-08-03 03:50:36 +02002694 queue_delayed_work(bond->wq, &bond->arp_work,
2695 msecs_to_jiffies(bond->params.arp_interval));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002696}
2697
Nikolay Aleksandrov547942c2014-09-15 17:19:34 +02002698/* Called to inspect slaves for active-backup mode ARP monitor link state
Jay Vosburghb2220ca2008-05-17 21:10:13 -07002699 * changes. Sets new_link in slaves to specify what action should take
2700 * place for the slave. Returns 0 if no changes are found, >0 if changes
2701 * to link states must be committed.
2702 *
Nikolay Aleksandrov8c0bc552014-09-11 22:49:28 +02002703 * Called with rcu_read_lock held.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002704 */
Veaceslav Falicoe7f63f12013-08-03 03:50:36 +02002705static int bond_ab_arp_inspect(struct bonding *bond)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002706{
Veaceslav Falicodef44602013-08-03 03:50:35 +02002707 unsigned long trans_start, last_rx;
Veaceslav Falico9caff1e2013-09-25 09:20:14 +02002708 struct list_head *iter;
nikolay@redhat.comdec1e902013-08-01 16:54:47 +02002709 struct slave *slave;
nikolay@redhat.comdec1e902013-08-01 16:54:47 +02002710 int commit = 0;
Jiri Bohacda210f52012-08-30 12:02:47 +00002711
dingtianhongeb9fa4b2013-12-13 10:20:02 +08002712 bond_for_each_slave_rcu(bond, slave, iter) {
Jay Vosburghb2220ca2008-05-17 21:10:13 -07002713 slave->new_link = BOND_LINK_NOCHANGE;
Veaceslav Falicodef44602013-08-03 03:50:35 +02002714 last_rx = slave_last_rx(bond, slave);
Jay Vosburghb2220ca2008-05-17 21:10:13 -07002715
Linus Torvalds1da177e2005-04-16 15:20:36 -07002716 if (slave->link != BOND_LINK_UP) {
Veaceslav Falicoe7f63f12013-08-03 03:50:36 +02002717 if (bond_time_in_interval(bond, last_rx, 1)) {
Jay Vosburghb2220ca2008-05-17 21:10:13 -07002718 slave->new_link = BOND_LINK_UP;
2719 commit++;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002720 }
Jay Vosburghb2220ca2008-05-17 21:10:13 -07002721 continue;
2722 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07002723
Nikolay Aleksandrov547942c2014-09-15 17:19:34 +02002724 /* Give slaves 2*delta after being enslaved or made
Jay Vosburghb2220ca2008-05-17 21:10:13 -07002725 * active. This avoids bouncing, as the last receive
2726 * times need a full ARP monitor cycle to be updated.
2727 */
Veaceslav Falico8e603462014-02-18 07:48:46 +01002728 if (bond_time_in_interval(bond, slave->last_link_up, 2))
Jay Vosburghb2220ca2008-05-17 21:10:13 -07002729 continue;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002730
Nikolay Aleksandrov547942c2014-09-15 17:19:34 +02002731 /* Backup slave is down if:
Jay Vosburghb2220ca2008-05-17 21:10:13 -07002732 * - No current_arp_slave AND
2733 * - more than 3*delta since last receive AND
2734 * - the bond has an IP address
2735 *
2736 * Note: a non-null current_arp_slave indicates
2737 * the curr_active_slave went down and we are
2738 * searching for a new one; under this condition
2739 * we only take the curr_active_slave down - this
2740 * gives each slave a chance to tx/rx traffic
2741 * before being taken out
2742 */
Jiri Pirkoe30bc062011-03-12 03:14:37 +00002743 if (!bond_is_active_slave(slave) &&
Eric Dumazet857417182014-07-15 06:56:56 -07002744 !rcu_access_pointer(bond->current_arp_slave) &&
Veaceslav Falicoe7f63f12013-08-03 03:50:36 +02002745 !bond_time_in_interval(bond, last_rx, 3)) {
Jay Vosburghb2220ca2008-05-17 21:10:13 -07002746 slave->new_link = BOND_LINK_DOWN;
2747 commit++;
2748 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07002749
Nikolay Aleksandrov547942c2014-09-15 17:19:34 +02002750 /* Active slave is down if:
Jay Vosburghb2220ca2008-05-17 21:10:13 -07002751 * - more than 2*delta since transmitting OR
2752 * - (more than 2*delta since receive AND
2753 * the bond has an IP address)
2754 */
Jiri Bohaccb32f2a2010-09-02 05:45:54 +00002755 trans_start = dev_trans_start(slave->dev);
Jiri Pirkoe30bc062011-03-12 03:14:37 +00002756 if (bond_is_active_slave(slave) &&
Veaceslav Falicoe7f63f12013-08-03 03:50:36 +02002757 (!bond_time_in_interval(bond, trans_start, 2) ||
2758 !bond_time_in_interval(bond, last_rx, 2))) {
Jay Vosburghb2220ca2008-05-17 21:10:13 -07002759 slave->new_link = BOND_LINK_DOWN;
2760 commit++;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002761 }
2762 }
2763
Jay Vosburghb2220ca2008-05-17 21:10:13 -07002764 return commit;
2765}
Linus Torvalds1da177e2005-04-16 15:20:36 -07002766
Nikolay Aleksandrov547942c2014-09-15 17:19:34 +02002767/* Called to commit link state changes noted by inspection step of
Jay Vosburghb2220ca2008-05-17 21:10:13 -07002768 * active-backup mode ARP monitor.
2769 *
dingtianhongeb9fa4b2013-12-13 10:20:02 +08002770 * Called with RTNL hold.
Jay Vosburghb2220ca2008-05-17 21:10:13 -07002771 */
Veaceslav Falicoe7f63f12013-08-03 03:50:36 +02002772static void bond_ab_arp_commit(struct bonding *bond)
Jay Vosburghb2220ca2008-05-17 21:10:13 -07002773{
Jiri Bohaccb32f2a2010-09-02 05:45:54 +00002774 unsigned long trans_start;
Veaceslav Falico9caff1e2013-09-25 09:20:14 +02002775 struct list_head *iter;
nikolay@redhat.comdec1e902013-08-01 16:54:47 +02002776 struct slave *slave;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002777
Veaceslav Falico9caff1e2013-09-25 09:20:14 +02002778 bond_for_each_slave(bond, slave, iter) {
Jay Vosburghb2220ca2008-05-17 21:10:13 -07002779 switch (slave->new_link) {
2780 case BOND_LINK_NOCHANGE:
2781 continue;
2782
2783 case BOND_LINK_UP:
Jiri Bohaccb32f2a2010-09-02 05:45:54 +00002784 trans_start = dev_trans_start(slave->dev);
Eric Dumazet4740d632014-07-15 06:56:55 -07002785 if (rtnl_dereference(bond->curr_active_slave) != slave ||
2786 (!rtnl_dereference(bond->curr_active_slave) &&
Veaceslav Falicoe7f63f12013-08-03 03:50:36 +02002787 bond_time_in_interval(bond, trans_start, 1))) {
Eric Dumazet857417182014-07-15 06:56:56 -07002788 struct slave *current_arp_slave;
2789
2790 current_arp_slave = rtnl_dereference(bond->current_arp_slave);
Jiri Pirko5d397062015-12-03 12:12:19 +01002791 bond_set_slave_link_state(slave, BOND_LINK_UP,
2792 BOND_SLAVE_NOTIFY_NOW);
Eric Dumazet857417182014-07-15 06:56:56 -07002793 if (current_arp_slave) {
Veaceslav Falico5a430972012-04-05 03:47:43 +00002794 bond_set_slave_inactive_flags(
Eric Dumazet857417182014-07-15 06:56:56 -07002795 current_arp_slave,
dingtianhong5e5b0662014-02-26 11:05:22 +08002796 BOND_SLAVE_NOTIFY_NOW);
Eric Dumazet857417182014-07-15 06:56:56 -07002797 RCU_INIT_POINTER(bond->current_arp_slave, NULL);
Veaceslav Falico5a430972012-04-05 03:47:43 +00002798 }
Jay Vosburghb2220ca2008-05-17 21:10:13 -07002799
Veaceslav Falico76444f52014-07-15 19:35:58 +02002800 netdev_info(bond->dev, "link status definitely up for interface %s\n",
2801 slave->dev->name);
Jay Vosburghb2220ca2008-05-17 21:10:13 -07002802
Eric Dumazet4740d632014-07-15 06:56:55 -07002803 if (!rtnl_dereference(bond->curr_active_slave) ||
Nikolay Aleksandrov059b47e2014-09-09 23:17:00 +02002804 slave == rtnl_dereference(bond->primary_slave))
Jiri Pirkob9f60252009-08-31 11:09:38 +00002805 goto do_failover;
Jay Vosburghb2220ca2008-05-17 21:10:13 -07002806
Jay Vosburghb2220ca2008-05-17 21:10:13 -07002807 }
2808
Jiri Pirkob9f60252009-08-31 11:09:38 +00002809 continue;
Jay Vosburghb2220ca2008-05-17 21:10:13 -07002810
2811 case BOND_LINK_DOWN:
2812 if (slave->link_failure_count < UINT_MAX)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002813 slave->link_failure_count++;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002814
Jiri Pirko5d397062015-12-03 12:12:19 +01002815 bond_set_slave_link_state(slave, BOND_LINK_DOWN,
2816 BOND_SLAVE_NOTIFY_NOW);
dingtianhong5e5b0662014-02-26 11:05:22 +08002817 bond_set_slave_inactive_flags(slave,
2818 BOND_SLAVE_NOTIFY_NOW);
Jiri Pirkob9f60252009-08-31 11:09:38 +00002819
Veaceslav Falico76444f52014-07-15 19:35:58 +02002820 netdev_info(bond->dev, "link status definitely down for interface %s, disabling it\n",
2821 slave->dev->name);
Jay Vosburghb2220ca2008-05-17 21:10:13 -07002822
Eric Dumazet4740d632014-07-15 06:56:55 -07002823 if (slave == rtnl_dereference(bond->curr_active_slave)) {
Eric Dumazet857417182014-07-15 06:56:56 -07002824 RCU_INIT_POINTER(bond->current_arp_slave, NULL);
Jiri Pirkob9f60252009-08-31 11:09:38 +00002825 goto do_failover;
Jay Vosburghb2220ca2008-05-17 21:10:13 -07002826 }
Jiri Pirkob9f60252009-08-31 11:09:38 +00002827
2828 continue;
Jay Vosburghb2220ca2008-05-17 21:10:13 -07002829
2830 default:
Veaceslav Falico76444f52014-07-15 19:35:58 +02002831 netdev_err(bond->dev, "impossible: new_link %d on slave %s\n",
2832 slave->new_link, slave->dev->name);
Jiri Pirkob9f60252009-08-31 11:09:38 +00002833 continue;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002834 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07002835
Jiri Pirkob9f60252009-08-31 11:09:38 +00002836do_failover:
Neil Hormane843fa52010-10-13 16:01:50 +00002837 block_netpoll_tx();
Jiri Pirkob9f60252009-08-31 11:09:38 +00002838 bond_select_active_slave(bond);
Neil Hormane843fa52010-10-13 16:01:50 +00002839 unblock_netpoll_tx();
Jay Vosburghb2220ca2008-05-17 21:10:13 -07002840 }
2841
2842 bond_set_carrier(bond);
2843}
2844
Nikolay Aleksandrov547942c2014-09-15 17:19:34 +02002845/* Send ARP probes for active-backup mode ARP monitor.
dingtianhongb0929912014-02-26 11:05:23 +08002846 *
Nikolay Aleksandrov8c0bc552014-09-11 22:49:28 +02002847 * Called with rcu_read_lock held.
Jay Vosburghb2220ca2008-05-17 21:10:13 -07002848 */
Veaceslav Falicof2ebd472014-01-27 14:37:32 +01002849static bool bond_ab_arp_probe(struct bonding *bond)
Jay Vosburghb2220ca2008-05-17 21:10:13 -07002850{
dingtianhongeb9fa4b2013-12-13 10:20:02 +08002851 struct slave *slave, *before = NULL, *new_slave = NULL,
dingtianhongb0929912014-02-26 11:05:23 +08002852 *curr_arp_slave = rcu_dereference(bond->current_arp_slave),
2853 *curr_active_slave = rcu_dereference(bond->curr_active_slave);
Veaceslav Falico4087df872013-09-25 09:20:19 +02002854 struct list_head *iter;
2855 bool found = false;
dingtianhongb0929912014-02-26 11:05:23 +08002856 bool should_notify_rtnl = BOND_SLAVE_NOTIFY_LATER;
Veaceslav Falicof2ebd472014-01-27 14:37:32 +01002857
Veaceslav Falico98b90f22014-01-27 14:37:31 +01002858 if (curr_arp_slave && curr_active_slave)
Veaceslav Falico76444f52014-07-15 19:35:58 +02002859 netdev_info(bond->dev, "PROBE: c_arp %s && cas %s BAD\n",
2860 curr_arp_slave->dev->name,
2861 curr_active_slave->dev->name);
Jay Vosburghb2220ca2008-05-17 21:10:13 -07002862
Veaceslav Falico98b90f22014-01-27 14:37:31 +01002863 if (curr_active_slave) {
2864 bond_arp_send_all(bond, curr_active_slave);
dingtianhongb0929912014-02-26 11:05:23 +08002865 return should_notify_rtnl;
Jay Vosburghb2220ca2008-05-17 21:10:13 -07002866 }
2867
Linus Torvalds1da177e2005-04-16 15:20:36 -07002868 /* if we don't have a curr_active_slave, search for the next available
2869 * backup slave from the current_arp_slave and make it the candidate
2870 * for becoming the curr_active_slave
2871 */
Jay Vosburghb2220ca2008-05-17 21:10:13 -07002872
dingtianhongeb9fa4b2013-12-13 10:20:02 +08002873 if (!curr_arp_slave) {
dingtianhongb0929912014-02-26 11:05:23 +08002874 curr_arp_slave = bond_first_slave_rcu(bond);
2875 if (!curr_arp_slave)
2876 return should_notify_rtnl;
Jay Vosburghb2220ca2008-05-17 21:10:13 -07002877 }
2878
dingtianhongb0929912014-02-26 11:05:23 +08002879 bond_set_slave_inactive_flags(curr_arp_slave, BOND_SLAVE_NOTIFY_LATER);
Jay Vosburghb2220ca2008-05-17 21:10:13 -07002880
dingtianhongb0929912014-02-26 11:05:23 +08002881 bond_for_each_slave_rcu(bond, slave, iter) {
Veaceslav Falicob6adc612014-05-15 21:39:57 +02002882 if (!found && !before && bond_slave_is_up(slave))
Veaceslav Falico4087df872013-09-25 09:20:19 +02002883 before = slave;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002884
Veaceslav Falicob6adc612014-05-15 21:39:57 +02002885 if (found && !new_slave && bond_slave_is_up(slave))
Veaceslav Falico4087df872013-09-25 09:20:19 +02002886 new_slave = slave;
Jay Vosburghb2220ca2008-05-17 21:10:13 -07002887 /* if the link state is up at this point, we
2888 * mark it down - this can happen if we have
2889 * simultaneous link failures and
2890 * reselect_active_interface doesn't make this
2891 * one the current slave so it is still marked
2892 * up when it is actually down
2893 */
Veaceslav Falicob6adc612014-05-15 21:39:57 +02002894 if (!bond_slave_is_up(slave) && slave->link == BOND_LINK_UP) {
Jiri Pirko5d397062015-12-03 12:12:19 +01002895 bond_set_slave_link_state(slave, BOND_LINK_DOWN,
2896 BOND_SLAVE_NOTIFY_LATER);
Jay Vosburghb2220ca2008-05-17 21:10:13 -07002897 if (slave->link_failure_count < UINT_MAX)
2898 slave->link_failure_count++;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002899
dingtianhong5e5b0662014-02-26 11:05:22 +08002900 bond_set_slave_inactive_flags(slave,
dingtianhongb0929912014-02-26 11:05:23 +08002901 BOND_SLAVE_NOTIFY_LATER);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002902
Veaceslav Falico76444f52014-07-15 19:35:58 +02002903 netdev_info(bond->dev, "backup interface %s is now down\n",
2904 slave->dev->name);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002905 }
dingtianhongeb9fa4b2013-12-13 10:20:02 +08002906 if (slave == curr_arp_slave)
Veaceslav Falico4087df872013-09-25 09:20:19 +02002907 found = true;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002908 }
Veaceslav Falico4087df872013-09-25 09:20:19 +02002909
2910 if (!new_slave && before)
2911 new_slave = before;
2912
dingtianhongb0929912014-02-26 11:05:23 +08002913 if (!new_slave)
2914 goto check_state;
Veaceslav Falico4087df872013-09-25 09:20:19 +02002915
Jiri Pirko5d397062015-12-03 12:12:19 +01002916 bond_set_slave_link_state(new_slave, BOND_LINK_BACK,
2917 BOND_SLAVE_NOTIFY_LATER);
dingtianhongb0929912014-02-26 11:05:23 +08002918 bond_set_slave_active_flags(new_slave, BOND_SLAVE_NOTIFY_LATER);
Veaceslav Falico4087df872013-09-25 09:20:19 +02002919 bond_arp_send_all(bond, new_slave);
Veaceslav Falico8e603462014-02-18 07:48:46 +01002920 new_slave->last_link_up = jiffies;
dingtianhongeb9fa4b2013-12-13 10:20:02 +08002921 rcu_assign_pointer(bond->current_arp_slave, new_slave);
Veaceslav Falicof2ebd472014-01-27 14:37:32 +01002922
dingtianhongb0929912014-02-26 11:05:23 +08002923check_state:
2924 bond_for_each_slave_rcu(bond, slave, iter) {
Jiri Pirko5d397062015-12-03 12:12:19 +01002925 if (slave->should_notify || slave->should_notify_link) {
dingtianhongb0929912014-02-26 11:05:23 +08002926 should_notify_rtnl = BOND_SLAVE_NOTIFY_NOW;
2927 break;
2928 }
2929 }
2930 return should_notify_rtnl;
Jay Vosburghb2220ca2008-05-17 21:10:13 -07002931}
2932
stephen hemminger6da67d22013-12-30 10:43:41 -08002933static void bond_activebackup_arp_mon(struct work_struct *work)
Jay Vosburghb2220ca2008-05-17 21:10:13 -07002934{
2935 struct bonding *bond = container_of(work, struct bonding,
2936 arp_work.work);
dingtianhongb0929912014-02-26 11:05:23 +08002937 bool should_notify_peers = false;
2938 bool should_notify_rtnl = false;
David S. Miller1f2cd842013-10-28 00:11:22 -04002939 int delta_in_ticks;
Jay Vosburghb2220ca2008-05-17 21:10:13 -07002940
David S. Miller1f2cd842013-10-28 00:11:22 -04002941 delta_in_ticks = msecs_to_jiffies(bond->params.arp_interval);
2942
2943 if (!bond_has_slaves(bond))
dingtianhong80b9d232013-10-24 11:09:25 +08002944 goto re_arm;
dingtianhong80b9d232013-10-24 11:09:25 +08002945
dingtianhongeb9fa4b2013-12-13 10:20:02 +08002946 rcu_read_lock();
Ben Hutchingsad246c92011-04-26 15:25:52 +00002947
dingtianhongb0929912014-02-26 11:05:23 +08002948 should_notify_peers = bond_should_notify_peers(bond);
2949
2950 if (bond_ab_arp_inspect(bond)) {
2951 rcu_read_unlock();
2952
David S. Miller1f2cd842013-10-28 00:11:22 -04002953 /* Race avoidance with bond_close flush of workqueue */
2954 if (!rtnl_trylock()) {
David S. Miller1f2cd842013-10-28 00:11:22 -04002955 delta_in_ticks = 1;
2956 should_notify_peers = false;
2957 goto re_arm;
2958 }
2959
Veaceslav Falicoe7f63f12013-08-03 03:50:36 +02002960 bond_ab_arp_commit(bond);
dingtianhongb0929912014-02-26 11:05:23 +08002961
David S. Miller1f2cd842013-10-28 00:11:22 -04002962 rtnl_unlock();
dingtianhongb0929912014-02-26 11:05:23 +08002963 rcu_read_lock();
David S. Miller1f2cd842013-10-28 00:11:22 -04002964 }
2965
dingtianhongb0929912014-02-26 11:05:23 +08002966 should_notify_rtnl = bond_ab_arp_probe(bond);
2967 rcu_read_unlock();
Linus Torvalds1da177e2005-04-16 15:20:36 -07002968
2969re_arm:
Jay Vosburghe6d265e2011-10-28 15:42:50 +00002970 if (bond->params.arp_interval)
David S. Miller1f2cd842013-10-28 00:11:22 -04002971 queue_delayed_work(bond->wq, &bond->arp_work, delta_in_ticks);
2972
dingtianhongb0929912014-02-26 11:05:23 +08002973 if (should_notify_peers || should_notify_rtnl) {
David S. Miller1f2cd842013-10-28 00:11:22 -04002974 if (!rtnl_trylock())
2975 return;
dingtianhongb0929912014-02-26 11:05:23 +08002976
2977 if (should_notify_peers)
2978 call_netdevice_notifiers(NETDEV_NOTIFY_PEERS,
2979 bond->dev);
Jiri Pirko5d397062015-12-03 12:12:19 +01002980 if (should_notify_rtnl) {
dingtianhongb0929912014-02-26 11:05:23 +08002981 bond_slave_state_notify(bond);
Jiri Pirko5d397062015-12-03 12:12:19 +01002982 bond_slave_link_notify(bond);
2983 }
dingtianhongb0929912014-02-26 11:05:23 +08002984
David S. Miller1f2cd842013-10-28 00:11:22 -04002985 rtnl_unlock();
2986 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07002987}
2988
Linus Torvalds1da177e2005-04-16 15:20:36 -07002989/*-------------------------- netdev event handling --------------------------*/
2990
Nikolay Aleksandrov547942c2014-09-15 17:19:34 +02002991/* Change device name */
Linus Torvalds1da177e2005-04-16 15:20:36 -07002992static int bond_event_changename(struct bonding *bond)
2993{
Linus Torvalds1da177e2005-04-16 15:20:36 -07002994 bond_remove_proc_entry(bond);
2995 bond_create_proc_entry(bond);
Stephen Hemminger7e083842009-06-12 19:02:46 +00002996
Taku Izumif073c7c2010-12-09 15:17:13 +00002997 bond_debug_reregister(bond);
2998
Linus Torvalds1da177e2005-04-16 15:20:36 -07002999 return NOTIFY_DONE;
3000}
3001
Stephen Hemminger3d632c32009-06-12 19:02:48 +00003002static int bond_master_netdev_event(unsigned long event,
3003 struct net_device *bond_dev)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003004{
Wang Chen454d7c92008-11-12 23:37:49 -08003005 struct bonding *event_bond = netdev_priv(bond_dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003006
3007 switch (event) {
3008 case NETDEV_CHANGENAME:
3009 return bond_event_changename(event_bond);
Eric W. Biedermana64d49c2012-07-09 10:51:45 +00003010 case NETDEV_UNREGISTER:
3011 bond_remove_proc_entry(event_bond);
3012 break;
3013 case NETDEV_REGISTER:
3014 bond_create_proc_entry(event_bond);
3015 break;
nikolay@redhat.com6c8d23f2013-09-02 13:51:38 +02003016 case NETDEV_NOTIFY_PEERS:
3017 if (event_bond->send_peer_notif)
3018 event_bond->send_peer_notif--;
3019 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003020 default:
3021 break;
3022 }
3023
3024 return NOTIFY_DONE;
3025}
3026
Stephen Hemminger3d632c32009-06-12 19:02:48 +00003027static int bond_slave_netdev_event(unsigned long event,
3028 struct net_device *slave_dev)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003029{
Nikolay Aleksandrov059b47e2014-09-09 23:17:00 +02003030 struct slave *slave = bond_slave_get_rtnl(slave_dev), *primary;
nikolay@redhat.com61013912013-04-11 09:18:55 +00003031 struct bonding *bond;
3032 struct net_device *bond_dev;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003033
nikolay@redhat.com61013912013-04-11 09:18:55 +00003034 /* A netdev event can be generated while enslaving a device
3035 * before netdev_rx_handler_register is called in which case
3036 * slave will be NULL
3037 */
3038 if (!slave)
3039 return NOTIFY_DONE;
3040 bond_dev = slave->bond->dev;
3041 bond = slave->bond;
Nikolay Aleksandrov059b47e2014-09-09 23:17:00 +02003042 primary = rtnl_dereference(bond->primary_slave);
nikolay@redhat.com61013912013-04-11 09:18:55 +00003043
Linus Torvalds1da177e2005-04-16 15:20:36 -07003044 switch (event) {
3045 case NETDEV_UNREGISTER:
nikolay@redhat.com8d2ada72013-06-26 17:13:37 +02003046 if (bond_dev->type != ARPHRD_ETHER)
Jiri Pirko471cb5a2013-01-03 22:49:01 +00003047 bond_release_and_destroy(bond_dev, slave_dev);
3048 else
3049 bond_release(bond_dev, slave_dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003050 break;
Weiping Pan98f41f62011-10-31 17:20:48 +00003051 case NETDEV_UP:
Linus Torvalds1da177e2005-04-16 15:20:36 -07003052 case NETDEV_CHANGE:
Jiri Pirko471cb5a2013-01-03 22:49:01 +00003053 bond_update_speed_duplex(slave);
Mahesh Bandewar52bc6712015-10-31 12:45:11 -07003054 if (BOND_MODE(bond) == BOND_MODE_8023AD)
3055 bond_3ad_adapter_speed_duplex_changed(slave);
Mahesh Bandewar950ddcb2015-02-19 10:13:25 -08003056 /* Fallthrough */
3057 case NETDEV_DOWN:
Mahesh Bandewaree637712014-10-04 17:45:01 -07003058 /* Refresh slave-array if applicable!
3059 * If the setup does not use miimon or arpmon (mode-specific!),
3060 * then these events will not cause the slave-array to be
3061 * refreshed. This will cause xmit to use a slave that is not
3062 * usable. Avoid such situation by refeshing the array at these
3063 * events. If these (miimon/arpmon) parameters are configured
3064 * then array gets refreshed twice and that should be fine!
3065 */
3066 if (bond_mode_uses_xmit_hash(bond))
3067 bond_update_slave_arr(bond, NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003068 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003069 case NETDEV_CHANGEMTU:
Nikolay Aleksandrov547942c2014-09-15 17:19:34 +02003070 /* TODO: Should slaves be allowed to
Linus Torvalds1da177e2005-04-16 15:20:36 -07003071 * independently alter their MTU? For
3072 * an active-backup bond, slaves need
3073 * not be the same type of device, so
3074 * MTUs may vary. For other modes,
3075 * slaves arguably should have the
3076 * same MTUs. To do this, we'd need to
3077 * take over the slave's change_mtu
3078 * function for the duration of their
3079 * servitude.
3080 */
3081 break;
3082 case NETDEV_CHANGENAME:
Veaceslav Falico3ec775b2014-01-16 02:04:29 +01003083 /* we don't care if we don't have primary set */
Veaceslav Falicoec0865a2014-05-15 21:39:54 +02003084 if (!bond_uses_primary(bond) ||
Veaceslav Falico3ec775b2014-01-16 02:04:29 +01003085 !bond->params.primary[0])
3086 break;
3087
Nikolay Aleksandrov059b47e2014-09-09 23:17:00 +02003088 if (slave == primary) {
Veaceslav Falico3ec775b2014-01-16 02:04:29 +01003089 /* slave's name changed - he's no longer primary */
Nikolay Aleksandrov059b47e2014-09-09 23:17:00 +02003090 RCU_INIT_POINTER(bond->primary_slave, NULL);
Veaceslav Falico3ec775b2014-01-16 02:04:29 +01003091 } else if (!strcmp(slave_dev->name, bond->params.primary)) {
3092 /* we have a new primary slave */
Nikolay Aleksandrov059b47e2014-09-09 23:17:00 +02003093 rcu_assign_pointer(bond->primary_slave, slave);
Veaceslav Falico3ec775b2014-01-16 02:04:29 +01003094 } else { /* we didn't change primary - exit */
3095 break;
3096 }
3097
Veaceslav Falico76444f52014-07-15 19:35:58 +02003098 netdev_info(bond->dev, "Primary slave changed to %s, reselecting active slave\n",
Nikolay Aleksandrov059b47e2014-09-09 23:17:00 +02003099 primary ? slave_dev->name : "none");
dingtianhongf80889a2014-02-12 12:06:40 +08003100
3101 block_netpoll_tx();
Veaceslav Falico3ec775b2014-01-16 02:04:29 +01003102 bond_select_active_slave(bond);
dingtianhongf80889a2014-02-12 12:06:40 +08003103 unblock_netpoll_tx();
Linus Torvalds1da177e2005-04-16 15:20:36 -07003104 break;
Arthur Kepner8531c5f2005-08-23 01:34:53 -04003105 case NETDEV_FEAT_CHANGE:
3106 bond_compute_features(bond);
3107 break;
Jiri Pirko4aa5dee2013-07-20 12:13:53 +02003108 case NETDEV_RESEND_IGMP:
3109 /* Propagate to master device */
3110 call_netdevice_notifiers(event, slave->bond->dev);
3111 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003112 default:
3113 break;
3114 }
3115
3116 return NOTIFY_DONE;
3117}
3118
Nikolay Aleksandrov547942c2014-09-15 17:19:34 +02003119/* bond_netdev_event: handle netdev notifier chain events.
Linus Torvalds1da177e2005-04-16 15:20:36 -07003120 *
3121 * This function receives events for the netdev chain. The caller (an
Alan Sterne041c682006-03-27 01:16:30 -08003122 * ioctl handler calling blocking_notifier_call_chain) holds the necessary
Linus Torvalds1da177e2005-04-16 15:20:36 -07003123 * locks for us to safely manipulate the slave devices (RTNL lock,
3124 * dev_probe_lock).
3125 */
Stephen Hemminger3d632c32009-06-12 19:02:48 +00003126static int bond_netdev_event(struct notifier_block *this,
3127 unsigned long event, void *ptr)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003128{
Jiri Pirko351638e2013-05-28 01:30:21 +00003129 struct net_device *event_dev = netdev_notifier_info_to_dev(ptr);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003130
Veaceslav Falico76444f52014-07-15 19:35:58 +02003131 netdev_dbg(event_dev, "event: %lx\n", event);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003132
Jay Vosburgh0b680e72006-09-22 21:54:10 -07003133 if (!(event_dev->priv_flags & IFF_BONDING))
3134 return NOTIFY_DONE;
3135
Linus Torvalds1da177e2005-04-16 15:20:36 -07003136 if (event_dev->flags & IFF_MASTER) {
Sabrina Dubroca489b99b2019-04-12 15:04:10 +02003137 int ret;
3138
Veaceslav Falico76444f52014-07-15 19:35:58 +02003139 netdev_dbg(event_dev, "IFF_MASTER\n");
Sabrina Dubroca489b99b2019-04-12 15:04:10 +02003140 ret = bond_master_netdev_event(event, event_dev);
3141 if (ret != NOTIFY_DONE)
3142 return ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003143 }
3144
3145 if (event_dev->flags & IFF_SLAVE) {
Veaceslav Falico76444f52014-07-15 19:35:58 +02003146 netdev_dbg(event_dev, "IFF_SLAVE\n");
Linus Torvalds1da177e2005-04-16 15:20:36 -07003147 return bond_slave_netdev_event(event, event_dev);
3148 }
3149
3150 return NOTIFY_DONE;
3151}
3152
3153static struct notifier_block bond_netdev_notifier = {
3154 .notifier_call = bond_netdev_event,
3155};
3156
Jay Vosburgh169a3e62005-06-26 17:54:11 -04003157/*---------------------------- Hashing Policies -----------------------------*/
3158
Nikolay Aleksandrov32819dc2013-10-02 13:39:25 +02003159/* L2 hash helper */
3160static inline u32 bond_eth_hash(struct sk_buff *skb)
Jay Vosburgh169a3e62005-06-26 17:54:11 -04003161{
Jianhua Xiece04d632014-07-17 14:16:25 +08003162 struct ethhdr *ep, hdr_tmp;
Jay Vosburgh169a3e62005-06-26 17:54:11 -04003163
Jianhua Xiece04d632014-07-17 14:16:25 +08003164 ep = skb_header_pointer(skb, 0, sizeof(hdr_tmp), &hdr_tmp);
3165 if (ep)
3166 return ep->h_dest[5] ^ ep->h_source[5] ^ ep->h_proto;
John Eaglesham6b923cb2012-08-21 20:43:35 +00003167 return 0;
3168}
3169
Nikolay Aleksandrov32819dc2013-10-02 13:39:25 +02003170/* Extract the appropriate headers based on bond's xmit policy */
3171static bool bond_flow_dissect(struct bonding *bond, struct sk_buff *skb,
3172 struct flow_keys *fk)
John Eaglesham6b923cb2012-08-21 20:43:35 +00003173{
Nikolay Aleksandrov32819dc2013-10-02 13:39:25 +02003174 const struct ipv6hdr *iph6;
Eric Dumazet43945422013-04-15 17:03:24 +00003175 const struct iphdr *iph;
Nikolay Aleksandrov32819dc2013-10-02 13:39:25 +02003176 int noff, proto = -1;
John Eaglesham6b923cb2012-08-21 20:43:35 +00003177
Nikolay Aleksandrov32819dc2013-10-02 13:39:25 +02003178 if (bond->params.xmit_policy > BOND_XMIT_POLICY_LAYER23)
Tom Herbertcd79a232015-09-01 09:24:27 -07003179 return skb_flow_dissect_flow_keys(skb, fk, 0);
Nikolay Aleksandrov32819dc2013-10-02 13:39:25 +02003180
Jiri Pirko06635a32015-05-12 14:56:16 +02003181 fk->ports.ports = 0;
Nikolay Aleksandrov32819dc2013-10-02 13:39:25 +02003182 noff = skb_network_offset(skb);
3183 if (skb->protocol == htons(ETH_P_IP)) {
dingtianhong054bb882014-03-25 17:00:09 +08003184 if (unlikely(!pskb_may_pull(skb, noff + sizeof(*iph))))
Nikolay Aleksandrov32819dc2013-10-02 13:39:25 +02003185 return false;
John Eaglesham6b923cb2012-08-21 20:43:35 +00003186 iph = ip_hdr(skb);
Tom Herbertc3f83242015-06-04 09:16:40 -07003187 iph_to_flow_copy_v4addrs(fk, iph);
Nikolay Aleksandrov32819dc2013-10-02 13:39:25 +02003188 noff += iph->ihl << 2;
3189 if (!ip_is_fragment(iph))
3190 proto = iph->protocol;
3191 } else if (skb->protocol == htons(ETH_P_IPV6)) {
dingtianhong054bb882014-03-25 17:00:09 +08003192 if (unlikely(!pskb_may_pull(skb, noff + sizeof(*iph6))))
Nikolay Aleksandrov32819dc2013-10-02 13:39:25 +02003193 return false;
3194 iph6 = ipv6_hdr(skb);
Tom Herbertc3f83242015-06-04 09:16:40 -07003195 iph_to_flow_copy_v6addrs(fk, iph6);
Nikolay Aleksandrov32819dc2013-10-02 13:39:25 +02003196 noff += sizeof(*iph6);
3197 proto = iph6->nexthdr;
3198 } else {
3199 return false;
John Eaglesham6b923cb2012-08-21 20:43:35 +00003200 }
Nikolay Aleksandrov32819dc2013-10-02 13:39:25 +02003201 if (bond->params.xmit_policy == BOND_XMIT_POLICY_LAYER34 && proto >= 0)
Jiri Pirko06635a32015-05-12 14:56:16 +02003202 fk->ports.ports = skb_flow_get_ports(skb, noff, proto);
John Eaglesham6b923cb2012-08-21 20:43:35 +00003203
Nikolay Aleksandrov32819dc2013-10-02 13:39:25 +02003204 return true;
John Eaglesham6b923cb2012-08-21 20:43:35 +00003205}
3206
Nikolay Aleksandrov32819dc2013-10-02 13:39:25 +02003207/**
3208 * bond_xmit_hash - generate a hash value based on the xmit policy
3209 * @bond: bonding device
3210 * @skb: buffer to use for headers
Nikolay Aleksandrov32819dc2013-10-02 13:39:25 +02003211 *
3212 * This function will extract the necessary headers from the skb buffer and use
3213 * them to generate a hash based on the xmit_policy set in the bonding device
John Eaglesham6b923cb2012-08-21 20:43:35 +00003214 */
Mahesh Bandewaree62e862014-04-22 16:30:15 -07003215u32 bond_xmit_hash(struct bonding *bond, struct sk_buff *skb)
John Eaglesham6b923cb2012-08-21 20:43:35 +00003216{
Nikolay Aleksandrov32819dc2013-10-02 13:39:25 +02003217 struct flow_keys flow;
3218 u32 hash;
John Eaglesham6b923cb2012-08-21 20:43:35 +00003219
Eric Dumazet4b1b8652015-09-15 15:24:28 -07003220 if (bond->params.xmit_policy == BOND_XMIT_POLICY_ENCAP34 &&
3221 skb->l4_hash)
3222 return skb->hash;
3223
Nikolay Aleksandrov32819dc2013-10-02 13:39:25 +02003224 if (bond->params.xmit_policy == BOND_XMIT_POLICY_LAYER2 ||
3225 !bond_flow_dissect(bond, skb, &flow))
Mahesh Bandewaree62e862014-04-22 16:30:15 -07003226 return bond_eth_hash(skb);
Eric Dumazet43945422013-04-15 17:03:24 +00003227
Nikolay Aleksandrov32819dc2013-10-02 13:39:25 +02003228 if (bond->params.xmit_policy == BOND_XMIT_POLICY_LAYER23 ||
3229 bond->params.xmit_policy == BOND_XMIT_POLICY_ENCAP23)
3230 hash = bond_eth_hash(skb);
3231 else
Jiri Pirko06635a32015-05-12 14:56:16 +02003232 hash = (__force u32)flow.ports.ports;
Tom Herbertc3f83242015-06-04 09:16:40 -07003233 hash ^= (__force u32)flow_get_u32_dst(&flow) ^
3234 (__force u32)flow_get_u32_src(&flow);
Nikolay Aleksandrov32819dc2013-10-02 13:39:25 +02003235 hash ^= (hash >> 16);
3236 hash ^= (hash >> 8);
John Eaglesham6b923cb2012-08-21 20:43:35 +00003237
Hangbin Liu6f239c02017-11-06 09:01:57 +08003238 return hash >> 1;
Jay Vosburgh169a3e62005-06-26 17:54:11 -04003239}
3240
Linus Torvalds1da177e2005-04-16 15:20:36 -07003241/*-------------------------- Device entry points ----------------------------*/
3242
nikolay@redhat.comfbb0c412012-11-29 01:31:31 +00003243static void bond_work_init_all(struct bonding *bond)
3244{
3245 INIT_DELAYED_WORK(&bond->mcast_work,
3246 bond_resend_igmp_join_requests_delayed);
3247 INIT_DELAYED_WORK(&bond->alb_work, bond_alb_monitor);
3248 INIT_DELAYED_WORK(&bond->mii_work, bond_mii_monitor);
Veaceslav Falico01844092014-05-15 21:39:55 +02003249 if (BOND_MODE(bond) == BOND_MODE_ACTIVEBACKUP)
nikolay@redhat.comfbb0c412012-11-29 01:31:31 +00003250 INIT_DELAYED_WORK(&bond->arp_work, bond_activebackup_arp_mon);
3251 else
3252 INIT_DELAYED_WORK(&bond->arp_work, bond_loadbalance_arp_mon);
3253 INIT_DELAYED_WORK(&bond->ad_work, bond_3ad_state_machine_handler);
Mahesh Bandewaree637712014-10-04 17:45:01 -07003254 INIT_DELAYED_WORK(&bond->slave_arr_work, bond_slave_arr_handler);
nikolay@redhat.comfbb0c412012-11-29 01:31:31 +00003255}
3256
3257static void bond_work_cancel_all(struct bonding *bond)
3258{
3259 cancel_delayed_work_sync(&bond->mii_work);
3260 cancel_delayed_work_sync(&bond->arp_work);
3261 cancel_delayed_work_sync(&bond->alb_work);
3262 cancel_delayed_work_sync(&bond->ad_work);
3263 cancel_delayed_work_sync(&bond->mcast_work);
Mahesh Bandewaree637712014-10-04 17:45:01 -07003264 cancel_delayed_work_sync(&bond->slave_arr_work);
nikolay@redhat.comfbb0c412012-11-29 01:31:31 +00003265}
3266
Linus Torvalds1da177e2005-04-16 15:20:36 -07003267static int bond_open(struct net_device *bond_dev)
3268{
Wang Chen454d7c92008-11-12 23:37:49 -08003269 struct bonding *bond = netdev_priv(bond_dev);
Veaceslav Falico9caff1e2013-09-25 09:20:14 +02003270 struct list_head *iter;
Peter Pan(潘卫平)ba3211c2011-08-15 15:57:35 +00003271 struct slave *slave;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003272
Peter Pan(潘卫平)ba3211c2011-08-15 15:57:35 +00003273 /* reset slave->backup and slave->inactive */
Veaceslav Falico0965a1f2013-09-25 09:20:21 +02003274 if (bond_has_slaves(bond)) {
Veaceslav Falico9caff1e2013-09-25 09:20:14 +02003275 bond_for_each_slave(bond, slave, iter) {
Eric Dumazet4740d632014-07-15 06:56:55 -07003276 if (bond_uses_primary(bond) &&
3277 slave != rcu_access_pointer(bond->curr_active_slave)) {
dingtianhong5e5b0662014-02-26 11:05:22 +08003278 bond_set_slave_inactive_flags(slave,
3279 BOND_SLAVE_NOTIFY_NOW);
Wilson Kok8bbe71a2015-01-26 01:16:58 -05003280 } else if (BOND_MODE(bond) != BOND_MODE_8023AD) {
dingtianhong5e5b0662014-02-26 11:05:22 +08003281 bond_set_slave_active_flags(slave,
3282 BOND_SLAVE_NOTIFY_NOW);
Peter Pan(潘卫平)ba3211c2011-08-15 15:57:35 +00003283 }
3284 }
Peter Pan(潘卫平)ba3211c2011-08-15 15:57:35 +00003285 }
Peter Pan(潘卫平)ba3211c2011-08-15 15:57:35 +00003286
nikolay@redhat.comfbb0c412012-11-29 01:31:31 +00003287 bond_work_init_all(bond);
Flavio Leitner5a37e8c2010-10-05 14:23:57 +00003288
Holger Eitzenberger58402052008-12-09 23:07:13 -08003289 if (bond_is_lb(bond)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07003290 /* bond_alb_initialize must be called before the timer
3291 * is started.
3292 */
Veaceslav Falico01844092014-05-15 21:39:55 +02003293 if (bond_alb_initialize(bond, (BOND_MODE(bond) == BOND_MODE_ALB)))
stephen hemmingerb4739462010-01-25 23:34:15 +00003294 return -ENOMEM;
Mahesh Bandeware9f0fb82014-04-22 16:30:22 -07003295 if (bond->params.tlb_dynamic_lb)
3296 queue_delayed_work(bond->wq, &bond->alb_work, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003297 }
3298
nikolay@redhat.comfbb0c412012-11-29 01:31:31 +00003299 if (bond->params.miimon) /* link check interval, in milliseconds. */
Jay Vosburgh1b76b312007-10-17 17:37:45 -07003300 queue_delayed_work(bond->wq, &bond->mii_work, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003301
3302 if (bond->params.arp_interval) { /* arp interval, in milliseconds. */
Jay Vosburgh1b76b312007-10-17 17:37:45 -07003303 queue_delayed_work(bond->wq, &bond->arp_work, 0);
Veaceslav Falico3fe68df2014-02-18 07:48:39 +01003304 bond->recv_probe = bond_arp_rcv;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003305 }
3306
Veaceslav Falico01844092014-05-15 21:39:55 +02003307 if (BOND_MODE(bond) == BOND_MODE_8023AD) {
Jay Vosburgh1b76b312007-10-17 17:37:45 -07003308 queue_delayed_work(bond->wq, &bond->ad_work, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003309 /* register to receive LACPDUs */
Jiri Pirko3aba8912011-04-19 03:48:16 +00003310 bond->recv_probe = bond_3ad_lacpdu_recv;
Jay Vosburghfd989c82008-11-04 17:51:16 -08003311 bond_3ad_initiate_agg_selection(bond, 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003312 }
3313
Mahesh Bandewaree637712014-10-04 17:45:01 -07003314 if (bond_mode_uses_xmit_hash(bond))
3315 bond_update_slave_arr(bond, NULL);
3316
Linus Torvalds1da177e2005-04-16 15:20:36 -07003317 return 0;
3318}
3319
3320static int bond_close(struct net_device *bond_dev)
3321{
Wang Chen454d7c92008-11-12 23:37:49 -08003322 struct bonding *bond = netdev_priv(bond_dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003323
nikolay@redhat.comfbb0c412012-11-29 01:31:31 +00003324 bond_work_cancel_all(bond);
nikolay@redhat.com6c8d23f2013-09-02 13:51:38 +02003325 bond->send_peer_notif = 0;
nikolay@redhat.comee8487c2013-09-02 13:51:39 +02003326 if (bond_is_lb(bond))
Linus Torvalds1da177e2005-04-16 15:20:36 -07003327 bond_alb_deinitialize(bond);
Jiri Pirko3aba8912011-04-19 03:48:16 +00003328 bond->recv_probe = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003329
3330 return 0;
3331}
3332
Eric Dumazetfe309372016-03-17 17:23:36 -07003333/* fold stats, assuming all rtnl_link_stats64 fields are u64, but
3334 * that some drivers can provide 32bit values only.
3335 */
3336static void bond_fold_stats(struct rtnl_link_stats64 *_res,
3337 const struct rtnl_link_stats64 *_new,
3338 const struct rtnl_link_stats64 *_old)
3339{
3340 const u64 *new = (const u64 *)_new;
3341 const u64 *old = (const u64 *)_old;
3342 u64 *res = (u64 *)_res;
3343 int i;
3344
3345 for (i = 0; i < sizeof(*_res) / sizeof(u64); i++) {
3346 u64 nv = new[i];
3347 u64 ov = old[i];
Eric Dumazet8c989a62017-03-29 10:45:44 -07003348 s64 delta = nv - ov;
Eric Dumazetfe309372016-03-17 17:23:36 -07003349
3350 /* detects if this particular field is 32bit only */
3351 if (((nv | ov) >> 32) == 0)
Eric Dumazet8c989a62017-03-29 10:45:44 -07003352 delta = (s64)(s32)((u32)nv - (u32)ov);
3353
3354 /* filter anomalies, some drivers reset their stats
3355 * at down/up events.
3356 */
3357 if (delta > 0)
3358 res[i] += delta;
Eric Dumazetfe309372016-03-17 17:23:36 -07003359 }
3360}
3361
Eric Dumazet7142fdb2018-07-31 06:30:54 -07003362static int bond_get_nest_level(struct net_device *bond_dev)
3363{
3364 struct bonding *bond = netdev_priv(bond_dev);
3365
3366 return bond->nest_level;
3367}
3368
Eric Dumazet28172732010-07-07 14:58:56 -07003369static struct rtnl_link_stats64 *bond_get_stats(struct net_device *bond_dev,
3370 struct rtnl_link_stats64 *stats)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003371{
Wang Chen454d7c92008-11-12 23:37:49 -08003372 struct bonding *bond = netdev_priv(bond_dev);
Eric Dumazet28172732010-07-07 14:58:56 -07003373 struct rtnl_link_stats64 temp;
Veaceslav Falico9caff1e2013-09-25 09:20:14 +02003374 struct list_head *iter;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003375 struct slave *slave;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003376
Eric Dumazet7142fdb2018-07-31 06:30:54 -07003377 spin_lock_nested(&bond->stats_lock, bond_get_nest_level(bond_dev));
Andy Gospodarek5f0c5f72014-09-28 22:34:37 -04003378 memcpy(stats, &bond->bond_stats, sizeof(*stats));
Linus Torvalds1da177e2005-04-16 15:20:36 -07003379
Eric Dumazetfe309372016-03-17 17:23:36 -07003380 rcu_read_lock();
3381 bond_for_each_slave_rcu(bond, slave, iter) {
3382 const struct rtnl_link_stats64 *new =
Eric Dumazet28172732010-07-07 14:58:56 -07003383 dev_get_stats(slave->dev, &temp);
Stephen Hemmingereeda3fd2008-11-19 21:40:23 -08003384
Eric Dumazetfe309372016-03-17 17:23:36 -07003385 bond_fold_stats(stats, new, &slave->slave_stats);
Andy Gospodarek5f0c5f72014-09-28 22:34:37 -04003386
3387 /* save off the slave stats for the next run */
Eric Dumazetfe309372016-03-17 17:23:36 -07003388 memcpy(&slave->slave_stats, new, sizeof(*new));
Linus Torvalds1da177e2005-04-16 15:20:36 -07003389 }
Eric Dumazetfe309372016-03-17 17:23:36 -07003390 rcu_read_unlock();
3391
Andy Gospodarek5f0c5f72014-09-28 22:34:37 -04003392 memcpy(&bond->bond_stats, stats, sizeof(*stats));
Eric Dumazetfe309372016-03-17 17:23:36 -07003393 spin_unlock(&bond->stats_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003394
3395 return stats;
3396}
3397
3398static int bond_do_ioctl(struct net_device *bond_dev, struct ifreq *ifr, int cmd)
3399{
Jiri Pirko080a06e2013-10-18 17:43:36 +02003400 struct bonding *bond = netdev_priv(bond_dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003401 struct net_device *slave_dev = NULL;
3402 struct ifbond k_binfo;
3403 struct ifbond __user *u_binfo = NULL;
3404 struct ifslave k_sinfo;
3405 struct ifslave __user *u_sinfo = NULL;
3406 struct mii_ioctl_data *mii = NULL;
Nikolay Aleksandrovd1fbd3e2014-01-22 14:53:35 +01003407 struct bond_opt_value newval;
Gao feng387ff9112013-01-31 16:31:00 +00003408 struct net *net;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003409 int res = 0;
3410
Veaceslav Falico76444f52014-07-15 19:35:58 +02003411 netdev_dbg(bond_dev, "bond_ioctl: cmd=%d\n", cmd);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003412
3413 switch (cmd) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07003414 case SIOCGMIIPHY:
3415 mii = if_mii(ifr);
Stephen Hemminger3d632c32009-06-12 19:02:48 +00003416 if (!mii)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003417 return -EINVAL;
Stephen Hemminger3d632c32009-06-12 19:02:48 +00003418
Linus Torvalds1da177e2005-04-16 15:20:36 -07003419 mii->phy_id = 0;
3420 /* Fall Through */
3421 case SIOCGMIIREG:
Nikolay Aleksandrov547942c2014-09-15 17:19:34 +02003422 /* We do this again just in case we were called by SIOCGMIIREG
Linus Torvalds1da177e2005-04-16 15:20:36 -07003423 * instead of SIOCGMIIPHY.
3424 */
3425 mii = if_mii(ifr);
Stephen Hemminger3d632c32009-06-12 19:02:48 +00003426 if (!mii)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003427 return -EINVAL;
Stephen Hemminger3d632c32009-06-12 19:02:48 +00003428
Linus Torvalds1da177e2005-04-16 15:20:36 -07003429 if (mii->reg_num == 1) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07003430 mii->val_out = 0;
Stephen Hemminger3d632c32009-06-12 19:02:48 +00003431 if (netif_carrier_ok(bond->dev))
Linus Torvalds1da177e2005-04-16 15:20:36 -07003432 mii->val_out = BMSR_LSTATUS;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003433 }
3434
3435 return 0;
3436 case BOND_INFO_QUERY_OLD:
3437 case SIOCBONDINFOQUERY:
3438 u_binfo = (struct ifbond __user *)ifr->ifr_data;
3439
Stephen Hemminger3d632c32009-06-12 19:02:48 +00003440 if (copy_from_user(&k_binfo, u_binfo, sizeof(ifbond)))
Linus Torvalds1da177e2005-04-16 15:20:36 -07003441 return -EFAULT;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003442
3443 res = bond_info_query(bond_dev, &k_binfo);
Stephen Hemminger3d632c32009-06-12 19:02:48 +00003444 if (res == 0 &&
3445 copy_to_user(u_binfo, &k_binfo, sizeof(ifbond)))
3446 return -EFAULT;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003447
3448 return res;
3449 case BOND_SLAVE_INFO_QUERY_OLD:
3450 case SIOCBONDSLAVEINFOQUERY:
3451 u_sinfo = (struct ifslave __user *)ifr->ifr_data;
3452
Stephen Hemminger3d632c32009-06-12 19:02:48 +00003453 if (copy_from_user(&k_sinfo, u_sinfo, sizeof(ifslave)))
Linus Torvalds1da177e2005-04-16 15:20:36 -07003454 return -EFAULT;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003455
3456 res = bond_slave_info_query(bond_dev, &k_sinfo);
Stephen Hemminger3d632c32009-06-12 19:02:48 +00003457 if (res == 0 &&
3458 copy_to_user(u_sinfo, &k_sinfo, sizeof(ifslave)))
3459 return -EFAULT;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003460
3461 return res;
3462 default:
Linus Torvalds1da177e2005-04-16 15:20:36 -07003463 break;
3464 }
3465
Gao feng387ff9112013-01-31 16:31:00 +00003466 net = dev_net(bond_dev);
3467
3468 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
Linus Torvalds1da177e2005-04-16 15:20:36 -07003469 return -EPERM;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003470
Ying Xue0917b932014-01-15 10:23:37 +08003471 slave_dev = __dev_get_by_name(net, ifr->ifr_slave);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003472
Veaceslav Falico76444f52014-07-15 19:35:58 +02003473 netdev_dbg(bond_dev, "slave_dev=%p:\n", slave_dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003474
Stephen Hemminger3d632c32009-06-12 19:02:48 +00003475 if (!slave_dev)
Ying Xue0917b932014-01-15 10:23:37 +08003476 return -ENODEV;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003477
Veaceslav Falico76444f52014-07-15 19:35:58 +02003478 netdev_dbg(bond_dev, "slave_dev->name=%s:\n", slave_dev->name);
Ying Xue0917b932014-01-15 10:23:37 +08003479 switch (cmd) {
3480 case BOND_ENSLAVE_OLD:
3481 case SIOCBONDENSLAVE:
3482 res = bond_enslave(bond_dev, slave_dev);
3483 break;
3484 case BOND_RELEASE_OLD:
3485 case SIOCBONDRELEASE:
3486 res = bond_release(bond_dev, slave_dev);
3487 break;
3488 case BOND_SETHWADDR_OLD:
3489 case SIOCBONDSETHWADDR:
3490 bond_set_dev_addr(bond_dev, slave_dev);
3491 res = 0;
3492 break;
3493 case BOND_CHANGE_ACTIVE_OLD:
3494 case SIOCBONDCHANGEACTIVE:
Nikolay Aleksandrovd1fbd3e2014-01-22 14:53:35 +01003495 bond_opt_initstr(&newval, slave_dev->name);
3496 res = __bond_opt_set(bond, BOND_OPT_ACTIVE_SLAVE, &newval);
Ying Xue0917b932014-01-15 10:23:37 +08003497 break;
3498 default:
3499 res = -EOPNOTSUPP;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003500 }
3501
Linus Torvalds1da177e2005-04-16 15:20:36 -07003502 return res;
3503}
3504
Jiri Pirkod03462b2011-08-16 03:15:04 +00003505static void bond_change_rx_flags(struct net_device *bond_dev, int change)
3506{
3507 struct bonding *bond = netdev_priv(bond_dev);
3508
3509 if (change & IFF_PROMISC)
3510 bond_set_promiscuity(bond,
3511 bond_dev->flags & IFF_PROMISC ? 1 : -1);
3512
3513 if (change & IFF_ALLMULTI)
3514 bond_set_allmulti(bond,
3515 bond_dev->flags & IFF_ALLMULTI ? 1 : -1);
3516}
3517
Jay Vosburgh303d1cb2013-05-31 11:57:30 +00003518static void bond_set_rx_mode(struct net_device *bond_dev)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003519{
Wang Chen454d7c92008-11-12 23:37:49 -08003520 struct bonding *bond = netdev_priv(bond_dev);
Veaceslav Falico9caff1e2013-09-25 09:20:14 +02003521 struct list_head *iter;
Jay Vosburgh303d1cb2013-05-31 11:57:30 +00003522 struct slave *slave;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003523
Veaceslav Falicob3241872013-09-28 21:18:56 +02003524 rcu_read_lock();
Veaceslav Falicoec0865a2014-05-15 21:39:54 +02003525 if (bond_uses_primary(bond)) {
Veaceslav Falicob3241872013-09-28 21:18:56 +02003526 slave = rcu_dereference(bond->curr_active_slave);
Jay Vosburgh303d1cb2013-05-31 11:57:30 +00003527 if (slave) {
3528 dev_uc_sync(slave->dev, bond_dev);
3529 dev_mc_sync(slave->dev, bond_dev);
3530 }
Jay Vosburgh303d1cb2013-05-31 11:57:30 +00003531 } else {
Veaceslav Falicob3241872013-09-28 21:18:56 +02003532 bond_for_each_slave_rcu(bond, slave, iter) {
Jay Vosburgh303d1cb2013-05-31 11:57:30 +00003533 dev_uc_sync_multiple(slave->dev, bond_dev);
3534 dev_mc_sync_multiple(slave->dev, bond_dev);
3535 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07003536 }
Veaceslav Falicob3241872013-09-28 21:18:56 +02003537 rcu_read_unlock();
Linus Torvalds1da177e2005-04-16 15:20:36 -07003538}
3539
Shlomo Pongratz234bcf82012-04-03 22:56:20 +00003540static int bond_neigh_init(struct neighbour *n)
Stephen Hemminger00829822008-11-20 20:14:53 -08003541{
Shlomo Pongratz234bcf82012-04-03 22:56:20 +00003542 struct bonding *bond = netdev_priv(n->dev);
Shlomo Pongratz234bcf82012-04-03 22:56:20 +00003543 const struct net_device_ops *slave_ops;
3544 struct neigh_parms parms;
nikolay@redhat.comdec1e902013-08-01 16:54:47 +02003545 struct slave *slave;
Shlomo Pongratz234bcf82012-04-03 22:56:20 +00003546 int ret;
Stephen Hemminger00829822008-11-20 20:14:53 -08003547
nikolay@redhat.comdec1e902013-08-01 16:54:47 +02003548 slave = bond_first_slave(bond);
Shlomo Pongratz234bcf82012-04-03 22:56:20 +00003549 if (!slave)
3550 return 0;
Shlomo Pongratz234bcf82012-04-03 22:56:20 +00003551 slave_ops = slave->dev->netdev_ops;
Shlomo Pongratz234bcf82012-04-03 22:56:20 +00003552 if (!slave_ops->ndo_neigh_setup)
3553 return 0;
3554
3555 parms.neigh_setup = NULL;
3556 parms.neigh_cleanup = NULL;
3557 ret = slave_ops->ndo_neigh_setup(slave->dev, &parms);
3558 if (ret)
3559 return ret;
3560
Nikolay Aleksandrov547942c2014-09-15 17:19:34 +02003561 /* Assign slave's neigh_cleanup to neighbour in case cleanup is called
Shlomo Pongratz234bcf82012-04-03 22:56:20 +00003562 * after the last slave has been detached. Assumes that all slaves
3563 * utilize the same neigh_cleanup (true at this writing as only user
3564 * is ipoib).
3565 */
3566 n->parms->neigh_cleanup = parms.neigh_cleanup;
3567
3568 if (!parms.neigh_setup)
3569 return 0;
3570
3571 return parms.neigh_setup(n);
3572}
3573
Nikolay Aleksandrov547942c2014-09-15 17:19:34 +02003574/* The bonding ndo_neigh_setup is called at init time beofre any
Shlomo Pongratz234bcf82012-04-03 22:56:20 +00003575 * slave exists. So we must declare proxy setup function which will
3576 * be used at run time to resolve the actual slave neigh param setup.
Veaceslav Falico9918d5b2013-08-02 19:07:39 +02003577 *
3578 * It's also called by master devices (such as vlans) to setup their
3579 * underlying devices. In that case - do nothing, we're already set up from
3580 * our init.
Shlomo Pongratz234bcf82012-04-03 22:56:20 +00003581 */
3582static int bond_neigh_setup(struct net_device *dev,
3583 struct neigh_parms *parms)
3584{
Veaceslav Falico9918d5b2013-08-02 19:07:39 +02003585 /* modify only our neigh_parms */
3586 if (parms->dev == dev)
3587 parms->neigh_setup = bond_neigh_init;
Shlomo Pongratz234bcf82012-04-03 22:56:20 +00003588
Stephen Hemminger00829822008-11-20 20:14:53 -08003589 return 0;
3590}
3591
Nikolay Aleksandrov547942c2014-09-15 17:19:34 +02003592/* Change the MTU of all of a master's slaves to match the master */
Linus Torvalds1da177e2005-04-16 15:20:36 -07003593static int bond_change_mtu(struct net_device *bond_dev, int new_mtu)
3594{
Wang Chen454d7c92008-11-12 23:37:49 -08003595 struct bonding *bond = netdev_priv(bond_dev);
Veaceslav Falico81f23b12013-09-25 09:20:13 +02003596 struct slave *slave, *rollback_slave;
Veaceslav Falico9caff1e2013-09-25 09:20:14 +02003597 struct list_head *iter;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003598 int res = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003599
Veaceslav Falico76444f52014-07-15 19:35:58 +02003600 netdev_dbg(bond_dev, "bond=%p, new_mtu=%d\n", bond, new_mtu);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003601
Veaceslav Falico9caff1e2013-09-25 09:20:14 +02003602 bond_for_each_slave(bond, slave, iter) {
Veaceslav Falico76444f52014-07-15 19:35:58 +02003603 netdev_dbg(bond_dev, "s %p c_m %p\n",
3604 slave, slave->dev->netdev_ops->ndo_change_mtu);
Mitch Williamse944ef72005-11-09 10:36:50 -08003605
Linus Torvalds1da177e2005-04-16 15:20:36 -07003606 res = dev_set_mtu(slave->dev, new_mtu);
3607
3608 if (res) {
3609 /* If we failed to set the slave's mtu to the new value
3610 * we must abort the operation even in ACTIVE_BACKUP
3611 * mode, because if we allow the backup slaves to have
3612 * different mtu values than the active slave we'll
3613 * need to change their mtu when doing a failover. That
3614 * means changing their mtu from timer context, which
3615 * is probably not a good idea.
3616 */
Veaceslav Falico76444f52014-07-15 19:35:58 +02003617 netdev_dbg(bond_dev, "err %d %s\n", res,
3618 slave->dev->name);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003619 goto unwind;
3620 }
3621 }
3622
3623 bond_dev->mtu = new_mtu;
3624
3625 return 0;
3626
3627unwind:
3628 /* unwind from head to the slave that failed */
Veaceslav Falico9caff1e2013-09-25 09:20:14 +02003629 bond_for_each_slave(bond, rollback_slave, iter) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07003630 int tmp_res;
3631
Veaceslav Falico81f23b12013-09-25 09:20:13 +02003632 if (rollback_slave == slave)
3633 break;
3634
3635 tmp_res = dev_set_mtu(rollback_slave->dev, bond_dev->mtu);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003636 if (tmp_res) {
Veaceslav Falico76444f52014-07-15 19:35:58 +02003637 netdev_dbg(bond_dev, "unwind err %d dev %s\n",
3638 tmp_res, rollback_slave->dev->name);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003639 }
3640 }
3641
3642 return res;
3643}
3644
Nikolay Aleksandrov547942c2014-09-15 17:19:34 +02003645/* Change HW address
Linus Torvalds1da177e2005-04-16 15:20:36 -07003646 *
3647 * Note that many devices must be down to change the HW address, and
3648 * downing the master releases all slaves. We can make bonds full of
3649 * bonding devices to test this, however.
3650 */
3651static int bond_set_mac_address(struct net_device *bond_dev, void *addr)
3652{
Wang Chen454d7c92008-11-12 23:37:49 -08003653 struct bonding *bond = netdev_priv(bond_dev);
Veaceslav Falico81f23b12013-09-25 09:20:13 +02003654 struct slave *slave, *rollback_slave;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003655 struct sockaddr *sa = addr, tmp_sa;
Veaceslav Falico9caff1e2013-09-25 09:20:14 +02003656 struct list_head *iter;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003657 int res = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003658
Veaceslav Falico01844092014-05-15 21:39:55 +02003659 if (BOND_MODE(bond) == BOND_MODE_ALB)
Stephen Hemmingereb7cc592008-11-19 21:56:05 -08003660 return bond_alb_set_mac_address(bond_dev, addr);
3661
3662
Veaceslav Falico76444f52014-07-15 19:35:58 +02003663 netdev_dbg(bond_dev, "bond=%p\n", bond);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003664
Jay Vosburgh1b5acd22013-05-31 11:57:31 +00003665 /* If fail_over_mac is enabled, do nothing and return success.
3666 * Returning an error causes ifenslave to fail.
Jay Vosburghdd957c52007-10-09 19:57:24 -07003667 */
dingtianhongcc689aa2014-01-25 13:00:57 +08003668 if (bond->params.fail_over_mac &&
Veaceslav Falico01844092014-05-15 21:39:55 +02003669 BOND_MODE(bond) == BOND_MODE_ACTIVEBACKUP)
Jay Vosburghdd957c52007-10-09 19:57:24 -07003670 return 0;
Moni Shoua2ab82852007-10-09 19:43:39 -07003671
Stephen Hemminger3d632c32009-06-12 19:02:48 +00003672 if (!is_valid_ether_addr(sa->sa_data))
Linus Torvalds1da177e2005-04-16 15:20:36 -07003673 return -EADDRNOTAVAIL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003674
Veaceslav Falico9caff1e2013-09-25 09:20:14 +02003675 bond_for_each_slave(bond, slave, iter) {
Veaceslav Falico76444f52014-07-15 19:35:58 +02003676 netdev_dbg(bond_dev, "slave %p %s\n", slave, slave->dev->name);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003677 res = dev_set_mac_address(slave->dev, addr);
3678 if (res) {
3679 /* TODO: consider downing the slave
3680 * and retry ?
3681 * User should expect communications
3682 * breakage anyway until ARP finish
3683 * updating, so...
3684 */
Veaceslav Falico76444f52014-07-15 19:35:58 +02003685 netdev_dbg(bond_dev, "err %d %s\n", res, slave->dev->name);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003686 goto unwind;
3687 }
3688 }
3689
3690 /* success */
3691 memcpy(bond_dev->dev_addr, sa->sa_data, bond_dev->addr_len);
3692 return 0;
3693
3694unwind:
3695 memcpy(tmp_sa.sa_data, bond_dev->dev_addr, bond_dev->addr_len);
3696 tmp_sa.sa_family = bond_dev->type;
3697
3698 /* unwind from head to the slave that failed */
Veaceslav Falico9caff1e2013-09-25 09:20:14 +02003699 bond_for_each_slave(bond, rollback_slave, iter) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07003700 int tmp_res;
3701
Veaceslav Falico81f23b12013-09-25 09:20:13 +02003702 if (rollback_slave == slave)
3703 break;
3704
3705 tmp_res = dev_set_mac_address(rollback_slave->dev, &tmp_sa);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003706 if (tmp_res) {
Veaceslav Falico76444f52014-07-15 19:35:58 +02003707 netdev_dbg(bond_dev, "unwind err %d dev %s\n",
3708 tmp_res, rollback_slave->dev->name);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003709 }
3710 }
3711
3712 return res;
3713}
3714
Nikolay Aleksandrov15077222013-08-01 16:54:50 +02003715/**
3716 * bond_xmit_slave_id - transmit skb through slave with slave_id
3717 * @bond: bonding device that is transmitting
3718 * @skb: buffer to transmit
3719 * @slave_id: slave id up to slave_cnt-1 through which to transmit
3720 *
3721 * This function tries to transmit through slave with slave_id but in case
3722 * it fails, it tries to find the first available slave for transmission.
3723 * The skb is consumed in all cases, thus the function is void.
3724 */
stephen hemmingerda131dd2013-12-29 11:41:25 -08003725static void bond_xmit_slave_id(struct bonding *bond, struct sk_buff *skb, int slave_id)
Nikolay Aleksandrov15077222013-08-01 16:54:50 +02003726{
Veaceslav Falico9caff1e2013-09-25 09:20:14 +02003727 struct list_head *iter;
Nikolay Aleksandrov15077222013-08-01 16:54:50 +02003728 struct slave *slave;
3729 int i = slave_id;
3730
3731 /* Here we start from the slave with slave_id */
Veaceslav Falico9caff1e2013-09-25 09:20:14 +02003732 bond_for_each_slave_rcu(bond, slave, iter) {
Nikolay Aleksandrov15077222013-08-01 16:54:50 +02003733 if (--i < 0) {
Veaceslav Falico891ab542014-05-15 21:39:58 +02003734 if (bond_slave_can_tx(slave)) {
Nikolay Aleksandrov15077222013-08-01 16:54:50 +02003735 bond_dev_queue_xmit(bond, skb, slave->dev);
3736 return;
3737 }
3738 }
3739 }
3740
3741 /* Here we start from the first slave up to slave_id */
3742 i = slave_id;
Veaceslav Falico9caff1e2013-09-25 09:20:14 +02003743 bond_for_each_slave_rcu(bond, slave, iter) {
Nikolay Aleksandrov15077222013-08-01 16:54:50 +02003744 if (--i < 0)
3745 break;
Veaceslav Falico891ab542014-05-15 21:39:58 +02003746 if (bond_slave_can_tx(slave)) {
Nikolay Aleksandrov15077222013-08-01 16:54:50 +02003747 bond_dev_queue_xmit(bond, skb, slave->dev);
3748 return;
3749 }
3750 }
3751 /* no slave that can tx has been found */
Eric Dumazet31aa8602014-10-31 11:47:54 -07003752 bond_tx_drop(bond->dev, skb);
Nikolay Aleksandrov15077222013-08-01 16:54:50 +02003753}
3754
Nikolay Aleksandrov73958322013-11-05 13:51:41 +01003755/**
3756 * bond_rr_gen_slave_id - generate slave id based on packets_per_slave
3757 * @bond: bonding device to use
3758 *
3759 * Based on the value of the bonding device's packets_per_slave parameter
3760 * this function generates a slave id, which is usually used as the next
3761 * slave to transmit through.
3762 */
3763static u32 bond_rr_gen_slave_id(struct bonding *bond)
3764{
Nikolay Aleksandrov73958322013-11-05 13:51:41 +01003765 u32 slave_id;
Hannes Frederic Sowa809fa972014-01-22 02:29:41 +01003766 struct reciprocal_value reciprocal_packets_per_slave;
3767 int packets_per_slave = bond->params.packets_per_slave;
Nikolay Aleksandrov73958322013-11-05 13:51:41 +01003768
3769 switch (packets_per_slave) {
3770 case 0:
3771 slave_id = prandom_u32();
3772 break;
3773 case 1:
3774 slave_id = bond->rr_tx_counter;
3775 break;
3776 default:
Hannes Frederic Sowa809fa972014-01-22 02:29:41 +01003777 reciprocal_packets_per_slave =
3778 bond->params.reciprocal_packets_per_slave;
Nikolay Aleksandrov73958322013-11-05 13:51:41 +01003779 slave_id = reciprocal_divide(bond->rr_tx_counter,
Hannes Frederic Sowa809fa972014-01-22 02:29:41 +01003780 reciprocal_packets_per_slave);
Nikolay Aleksandrov73958322013-11-05 13:51:41 +01003781 break;
3782 }
3783 bond->rr_tx_counter++;
3784
3785 return slave_id;
3786}
3787
Linus Torvalds1da177e2005-04-16 15:20:36 -07003788static int bond_xmit_roundrobin(struct sk_buff *skb, struct net_device *bond_dev)
3789{
Wang Chen454d7c92008-11-12 23:37:49 -08003790 struct bonding *bond = netdev_priv(bond_dev);
Andy Gospodareka2fd9402010-03-25 14:49:05 +00003791 struct iphdr *iph = ip_hdr(skb);
Nikolay Aleksandrov15077222013-08-01 16:54:50 +02003792 struct slave *slave;
Nikolay Aleksandrov73958322013-11-05 13:51:41 +01003793 u32 slave_id;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003794
Nikolay Aleksandrov73958322013-11-05 13:51:41 +01003795 /* Start with the curr_active_slave that joined the bond as the
Andy Gospodareka2fd9402010-03-25 14:49:05 +00003796 * default for sending IGMP traffic. For failover purposes one
3797 * needs to maintain some consistency for the interface that will
3798 * send the join/membership reports. The curr_active_slave found
3799 * will send all of this type of traffic.
Jay Vosburghcf5f9042007-10-17 17:37:47 -07003800 */
Nikolay Aleksandrov15077222013-08-01 16:54:50 +02003801 if (iph->protocol == IPPROTO_IGMP && skb->protocol == htons(ETH_P_IP)) {
nikolay@redhat.com278b2082013-08-01 16:54:51 +02003802 slave = rcu_dereference(bond->curr_active_slave);
dingtianhonga64d0442014-09-19 21:04:57 +08003803 if (slave)
Nikolay Aleksandrov15077222013-08-01 16:54:50 +02003804 bond_dev_queue_xmit(bond, skb, slave->dev);
3805 else
3806 bond_xmit_slave_id(bond, skb, 0);
Andy Gospodareka2fd9402010-03-25 14:49:05 +00003807 } else {
Nikolay Aleksandrov9a72c2da2014-09-12 17:38:18 +02003808 int slave_cnt = ACCESS_ONCE(bond->slave_cnt);
3809
3810 if (likely(slave_cnt)) {
3811 slave_id = bond_rr_gen_slave_id(bond);
3812 bond_xmit_slave_id(bond, skb, slave_id % slave_cnt);
3813 } else {
Eric Dumazet31aa8602014-10-31 11:47:54 -07003814 bond_tx_drop(bond_dev, skb);
Nikolay Aleksandrov9a72c2da2014-09-12 17:38:18 +02003815 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07003816 }
Michał Mirosław0693e882011-05-07 01:48:02 +00003817
Patrick McHardyec634fe2009-07-05 19:23:38 -07003818 return NETDEV_TX_OK;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003819}
3820
Nikolay Aleksandrov547942c2014-09-15 17:19:34 +02003821/* In active-backup mode, we know that bond->curr_active_slave is always valid if
Linus Torvalds1da177e2005-04-16 15:20:36 -07003822 * the bond has a usable interface.
3823 */
3824static int bond_xmit_activebackup(struct sk_buff *skb, struct net_device *bond_dev)
3825{
Wang Chen454d7c92008-11-12 23:37:49 -08003826 struct bonding *bond = netdev_priv(bond_dev);
nikolay@redhat.com71bc3b22013-08-01 16:54:48 +02003827 struct slave *slave;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003828
nikolay@redhat.com278b2082013-08-01 16:54:51 +02003829 slave = rcu_dereference(bond->curr_active_slave);
nikolay@redhat.com71bc3b22013-08-01 16:54:48 +02003830 if (slave)
Nikolay Aleksandrov15077222013-08-01 16:54:50 +02003831 bond_dev_queue_xmit(bond, skb, slave->dev);
3832 else
Eric Dumazet31aa8602014-10-31 11:47:54 -07003833 bond_tx_drop(bond_dev, skb);
Michał Mirosław0693e882011-05-07 01:48:02 +00003834
Patrick McHardyec634fe2009-07-05 19:23:38 -07003835 return NETDEV_TX_OK;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003836}
3837
Mahesh Bandewaree637712014-10-04 17:45:01 -07003838/* Use this to update slave_array when (a) it's not appropriate to update
3839 * slave_array right away (note that update_slave_array() may sleep)
3840 * and / or (b) RTNL is not held.
Linus Torvalds1da177e2005-04-16 15:20:36 -07003841 */
Mahesh Bandewaree637712014-10-04 17:45:01 -07003842void bond_slave_arr_work_rearm(struct bonding *bond, unsigned long delay)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003843{
Mahesh Bandewaree637712014-10-04 17:45:01 -07003844 queue_delayed_work(bond->wq, &bond->slave_arr_work, delay);
3845}
Linus Torvalds1da177e2005-04-16 15:20:36 -07003846
Mahesh Bandewaree637712014-10-04 17:45:01 -07003847/* Slave array work handler. Holds only RTNL */
3848static void bond_slave_arr_handler(struct work_struct *work)
3849{
3850 struct bonding *bond = container_of(work, struct bonding,
3851 slave_arr_work.work);
3852 int ret;
3853
3854 if (!rtnl_trylock())
3855 goto err;
3856
3857 ret = bond_update_slave_arr(bond, NULL);
3858 rtnl_unlock();
3859 if (ret) {
3860 pr_warn_ratelimited("Failed to update slave array from WT\n");
3861 goto err;
3862 }
3863 return;
3864
3865err:
3866 bond_slave_arr_work_rearm(bond, 1);
3867}
3868
3869/* Build the usable slaves array in control path for modes that use xmit-hash
3870 * to determine the slave interface -
3871 * (a) BOND_MODE_8023AD
3872 * (b) BOND_MODE_XOR
3873 * (c) BOND_MODE_TLB && tlb_dynamic_lb == 0
3874 *
3875 * The caller is expected to hold RTNL only and NO other lock!
3876 */
3877int bond_update_slave_arr(struct bonding *bond, struct slave *skipslave)
3878{
3879 struct slave *slave;
3880 struct list_head *iter;
3881 struct bond_up_slave *new_arr, *old_arr;
Mahesh Bandewaree637712014-10-04 17:45:01 -07003882 int agg_id = 0;
3883 int ret = 0;
3884
3885#ifdef CONFIG_LOCKDEP
3886 WARN_ON(lockdep_is_held(&bond->mode_lock));
3887#endif
3888
3889 new_arr = kzalloc(offsetof(struct bond_up_slave, arr[bond->slave_cnt]),
3890 GFP_KERNEL);
3891 if (!new_arr) {
3892 ret = -ENOMEM;
3893 pr_err("Failed to build slave-array.\n");
3894 goto out;
3895 }
3896 if (BOND_MODE(bond) == BOND_MODE_8023AD) {
3897 struct ad_info ad_info;
3898
3899 if (bond_3ad_get_active_agg_info(bond, &ad_info)) {
3900 pr_debug("bond_3ad_get_active_agg_info failed\n");
3901 kfree_rcu(new_arr, rcu);
3902 /* No active aggragator means it's not safe to use
3903 * the previous array.
3904 */
3905 old_arr = rtnl_dereference(bond->slave_arr);
3906 if (old_arr) {
3907 RCU_INIT_POINTER(bond->slave_arr, NULL);
3908 kfree_rcu(old_arr, rcu);
3909 }
3910 goto out;
3911 }
Mahesh Bandewaree637712014-10-04 17:45:01 -07003912 agg_id = ad_info.aggregator_id;
3913 }
3914 bond_for_each_slave(bond, slave, iter) {
3915 if (BOND_MODE(bond) == BOND_MODE_8023AD) {
3916 struct aggregator *agg;
3917
3918 agg = SLAVE_AD_INFO(slave)->port.aggregator;
3919 if (!agg || agg->aggregator_identifier != agg_id)
3920 continue;
3921 }
3922 if (!bond_slave_can_tx(slave))
3923 continue;
3924 if (skipslave == slave)
3925 continue;
3926 new_arr->arr[new_arr->count++] = slave;
3927 }
3928
3929 old_arr = rtnl_dereference(bond->slave_arr);
3930 rcu_assign_pointer(bond->slave_arr, new_arr);
3931 if (old_arr)
3932 kfree_rcu(old_arr, rcu);
3933out:
3934 if (ret != 0 && skipslave) {
3935 int idx;
3936
3937 /* Rare situation where caller has asked to skip a specific
3938 * slave but allocation failed (most likely!). BTW this is
3939 * only possible when the call is initiated from
3940 * __bond_release_one(). In this situation; overwrite the
3941 * skipslave entry in the array with the last entry from the
3942 * array to avoid a situation where the xmit path may choose
3943 * this to-be-skipped slave to send a packet out.
3944 */
3945 old_arr = rtnl_dereference(bond->slave_arr);
3946 for (idx = 0; idx < old_arr->count; idx++) {
3947 if (skipslave == old_arr->arr[idx]) {
3948 old_arr->arr[idx] =
3949 old_arr->arr[old_arr->count-1];
3950 old_arr->count--;
3951 break;
3952 }
3953 }
3954 }
3955 return ret;
3956}
3957
3958/* Use this Xmit function for 3AD as well as XOR modes. The current
3959 * usable slave array is formed in the control path. The xmit function
3960 * just calculates hash and sends the packet out.
3961 */
Jonathan Toppins30369102015-01-26 01:17:01 -05003962static int bond_3ad_xor_xmit(struct sk_buff *skb, struct net_device *dev)
Mahesh Bandewaree637712014-10-04 17:45:01 -07003963{
3964 struct bonding *bond = netdev_priv(dev);
3965 struct slave *slave;
3966 struct bond_up_slave *slaves;
3967 unsigned int count;
3968
3969 slaves = rcu_dereference(bond->slave_arr);
3970 count = slaves ? ACCESS_ONCE(slaves->count) : 0;
3971 if (likely(count)) {
3972 slave = slaves->arr[bond_xmit_hash(bond, skb) % count];
3973 bond_dev_queue_xmit(bond, skb, slave->dev);
3974 } else {
Eric Dumazet31aa8602014-10-31 11:47:54 -07003975 bond_tx_drop(dev, skb);
Mahesh Bandewaree637712014-10-04 17:45:01 -07003976 }
Michał Mirosław0693e882011-05-07 01:48:02 +00003977
Patrick McHardyec634fe2009-07-05 19:23:38 -07003978 return NETDEV_TX_OK;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003979}
3980
Nikolay Aleksandrov78a646c2013-08-01 16:54:49 +02003981/* in broadcast mode, we send everything to all usable interfaces. */
Linus Torvalds1da177e2005-04-16 15:20:36 -07003982static int bond_xmit_broadcast(struct sk_buff *skb, struct net_device *bond_dev)
3983{
Wang Chen454d7c92008-11-12 23:37:49 -08003984 struct bonding *bond = netdev_priv(bond_dev);
Nikolay Aleksandrov78a646c2013-08-01 16:54:49 +02003985 struct slave *slave = NULL;
Veaceslav Falico9caff1e2013-09-25 09:20:14 +02003986 struct list_head *iter;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003987
Veaceslav Falico9caff1e2013-09-25 09:20:14 +02003988 bond_for_each_slave_rcu(bond, slave, iter) {
Nikolay Aleksandrov78a646c2013-08-01 16:54:49 +02003989 if (bond_is_last_slave(bond, slave))
3990 break;
Veaceslav Falicob6adc612014-05-15 21:39:57 +02003991 if (bond_slave_is_up(slave) && slave->link == BOND_LINK_UP) {
Nikolay Aleksandrov78a646c2013-08-01 16:54:49 +02003992 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003993
Nikolay Aleksandrov78a646c2013-08-01 16:54:49 +02003994 if (!skb2) {
dingtianhong9152e262014-03-25 17:00:10 +08003995 net_err_ratelimited("%s: Error: %s: skb_clone() failed\n",
3996 bond_dev->name, __func__);
Nikolay Aleksandrov78a646c2013-08-01 16:54:49 +02003997 continue;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003998 }
Nikolay Aleksandrov78a646c2013-08-01 16:54:49 +02003999 bond_dev_queue_xmit(bond, skb2, slave->dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -07004000 }
4001 }
Veaceslav Falicob6adc612014-05-15 21:39:57 +02004002 if (slave && bond_slave_is_up(slave) && slave->link == BOND_LINK_UP)
Nikolay Aleksandrov78a646c2013-08-01 16:54:49 +02004003 bond_dev_queue_xmit(bond, skb, slave->dev);
4004 else
Eric Dumazet31aa8602014-10-31 11:47:54 -07004005 bond_tx_drop(bond_dev, skb);
Stephen Hemminger3d632c32009-06-12 19:02:48 +00004006
Patrick McHardyec634fe2009-07-05 19:23:38 -07004007 return NETDEV_TX_OK;
Linus Torvalds1da177e2005-04-16 15:20:36 -07004008}
4009
4010/*------------------------- Device initialization ---------------------------*/
4011
Nikolay Aleksandrov547942c2014-09-15 17:19:34 +02004012/* Lookup the slave that corresponds to a qid */
Andy Gospodarekbb1d9122010-06-02 08:40:18 +00004013static inline int bond_slave_override(struct bonding *bond,
4014 struct sk_buff *skb)
4015{
Andy Gospodarekbb1d9122010-06-02 08:40:18 +00004016 struct slave *slave = NULL;
Veaceslav Falico9caff1e2013-09-25 09:20:14 +02004017 struct list_head *iter;
Andy Gospodarekbb1d9122010-06-02 08:40:18 +00004018
Michał Mirosław0693e882011-05-07 01:48:02 +00004019 if (!skb->queue_mapping)
4020 return 1;
Andy Gospodarekbb1d9122010-06-02 08:40:18 +00004021
4022 /* Find out if any slaves have the same mapping as this skb. */
dingtianhong3900f292014-01-02 09:13:06 +08004023 bond_for_each_slave_rcu(bond, slave, iter) {
4024 if (slave->queue_id == skb->queue_mapping) {
Anton Nayshtutf5e2dc52015-03-29 14:20:25 +03004025 if (bond_slave_is_up(slave) &&
4026 slave->link == BOND_LINK_UP) {
dingtianhong3900f292014-01-02 09:13:06 +08004027 bond_dev_queue_xmit(bond, skb, slave->dev);
4028 return 0;
4029 }
4030 /* If the slave isn't UP, use default transmit policy. */
Andy Gospodarekbb1d9122010-06-02 08:40:18 +00004031 break;
4032 }
4033 }
4034
dingtianhong3900f292014-01-02 09:13:06 +08004035 return 1;
Andy Gospodarekbb1d9122010-06-02 08:40:18 +00004036}
4037
Neil Horman374eeb52011-06-03 10:35:52 +00004038
Jason Wangf663dd92014-01-10 16:18:26 +08004039static u16 bond_select_queue(struct net_device *dev, struct sk_buff *skb,
Daniel Borkmann99932d42014-02-16 15:55:20 +01004040 void *accel_priv, select_queue_fallback_t fallback)
Andy Gospodarekbb1d9122010-06-02 08:40:18 +00004041{
Nikolay Aleksandrov547942c2014-09-15 17:19:34 +02004042 /* This helper function exists to help dev_pick_tx get the correct
Phil Oesterfd0e4352011-03-14 06:22:04 +00004043 * destination queue. Using a helper function skips a call to
Andy Gospodarekbb1d9122010-06-02 08:40:18 +00004044 * skb_tx_hash and will put the skbs in the queue we expect on their
4045 * way down to the bonding driver.
4046 */
Phil Oesterfd0e4352011-03-14 06:22:04 +00004047 u16 txq = skb_rx_queue_recorded(skb) ? skb_get_rx_queue(skb) : 0;
4048
Nikolay Aleksandrov547942c2014-09-15 17:19:34 +02004049 /* Save the original txq to restore before passing to the driver */
Jiri Pirkodf4ab5b2012-07-20 02:28:49 +00004050 qdisc_skb_cb(skb)->slave_dev_queue_mapping = skb->queue_mapping;
Neil Horman374eeb52011-06-03 10:35:52 +00004051
Phil Oesterfd0e4352011-03-14 06:22:04 +00004052 if (unlikely(txq >= dev->real_num_tx_queues)) {
David Decotignyd30ee672011-04-13 15:22:29 +00004053 do {
Phil Oesterfd0e4352011-03-14 06:22:04 +00004054 txq -= dev->real_num_tx_queues;
David Decotignyd30ee672011-04-13 15:22:29 +00004055 } while (txq >= dev->real_num_tx_queues);
Phil Oesterfd0e4352011-03-14 06:22:04 +00004056 }
4057 return txq;
Andy Gospodarekbb1d9122010-06-02 08:40:18 +00004058}
4059
Michał Mirosław0693e882011-05-07 01:48:02 +00004060static netdev_tx_t __bond_start_xmit(struct sk_buff *skb, struct net_device *dev)
Stephen Hemminger00829822008-11-20 20:14:53 -08004061{
Andy Gospodarekbb1d9122010-06-02 08:40:18 +00004062 struct bonding *bond = netdev_priv(dev);
4063
Veaceslav Falicod1e2e5c2014-05-15 21:39:52 +02004064 if (bond_should_override_tx_queue(bond) &&
4065 !bond_slave_override(bond, skb))
4066 return NETDEV_TX_OK;
Stephen Hemminger00829822008-11-20 20:14:53 -08004067
Veaceslav Falico01844092014-05-15 21:39:55 +02004068 switch (BOND_MODE(bond)) {
Stephen Hemminger00829822008-11-20 20:14:53 -08004069 case BOND_MODE_ROUNDROBIN:
4070 return bond_xmit_roundrobin(skb, dev);
4071 case BOND_MODE_ACTIVEBACKUP:
4072 return bond_xmit_activebackup(skb, dev);
Mahesh Bandewaree637712014-10-04 17:45:01 -07004073 case BOND_MODE_8023AD:
Stephen Hemminger00829822008-11-20 20:14:53 -08004074 case BOND_MODE_XOR:
Mahesh Bandewaree637712014-10-04 17:45:01 -07004075 return bond_3ad_xor_xmit(skb, dev);
Stephen Hemminger00829822008-11-20 20:14:53 -08004076 case BOND_MODE_BROADCAST:
4077 return bond_xmit_broadcast(skb, dev);
Stephen Hemminger00829822008-11-20 20:14:53 -08004078 case BOND_MODE_ALB:
Stephen Hemminger00829822008-11-20 20:14:53 -08004079 return bond_alb_xmit(skb, dev);
Mahesh Bandewarf05b42e2014-04-22 16:30:20 -07004080 case BOND_MODE_TLB:
4081 return bond_tlb_xmit(skb, dev);
Stephen Hemminger00829822008-11-20 20:14:53 -08004082 default:
4083 /* Should never happen, mode already checked */
Veaceslav Falico76444f52014-07-15 19:35:58 +02004084 netdev_err(dev, "Unknown bonding mode %d\n", BOND_MODE(bond));
Stephen Hemminger00829822008-11-20 20:14:53 -08004085 WARN_ON_ONCE(1);
Eric Dumazet31aa8602014-10-31 11:47:54 -07004086 bond_tx_drop(dev, skb);
Stephen Hemminger00829822008-11-20 20:14:53 -08004087 return NETDEV_TX_OK;
4088 }
4089}
4090
Michał Mirosław0693e882011-05-07 01:48:02 +00004091static netdev_tx_t bond_start_xmit(struct sk_buff *skb, struct net_device *dev)
4092{
4093 struct bonding *bond = netdev_priv(dev);
4094 netdev_tx_t ret = NETDEV_TX_OK;
4095
Nikolay Aleksandrov547942c2014-09-15 17:19:34 +02004096 /* If we risk deadlock from transmitting this in the
Michał Mirosław0693e882011-05-07 01:48:02 +00004097 * netpoll path, tell netpoll to queue the frame for later tx
4098 */
dingtianhong054bb882014-03-25 17:00:09 +08004099 if (unlikely(is_netpoll_tx_blocked(dev)))
Michał Mirosław0693e882011-05-07 01:48:02 +00004100 return NETDEV_TX_BUSY;
4101
nikolay@redhat.com278b2082013-08-01 16:54:51 +02004102 rcu_read_lock();
Veaceslav Falico0965a1f2013-09-25 09:20:21 +02004103 if (bond_has_slaves(bond))
Michał Mirosław0693e882011-05-07 01:48:02 +00004104 ret = __bond_start_xmit(skb, dev);
4105 else
Eric Dumazet31aa8602014-10-31 11:47:54 -07004106 bond_tx_drop(dev, skb);
nikolay@redhat.com278b2082013-08-01 16:54:51 +02004107 rcu_read_unlock();
Michał Mirosław0693e882011-05-07 01:48:02 +00004108
4109 return ret;
4110}
Stephen Hemminger00829822008-11-20 20:14:53 -08004111
Andy Gospodarekbb5b0522013-04-16 14:46:00 +00004112static int bond_ethtool_get_settings(struct net_device *bond_dev,
4113 struct ethtool_cmd *ecmd)
4114{
4115 struct bonding *bond = netdev_priv(bond_dev);
Andy Gospodarekbb5b0522013-04-16 14:46:00 +00004116 unsigned long speed = 0;
Veaceslav Falico9caff1e2013-09-25 09:20:14 +02004117 struct list_head *iter;
nikolay@redhat.comdec1e902013-08-01 16:54:47 +02004118 struct slave *slave;
Andy Gospodarekbb5b0522013-04-16 14:46:00 +00004119
4120 ecmd->duplex = DUPLEX_UNKNOWN;
4121 ecmd->port = PORT_OTHER;
4122
Veaceslav Falico8557cd72014-05-15 21:39:59 +02004123 /* Since bond_slave_can_tx returns false for all inactive or down slaves, we
Andy Gospodarekbb5b0522013-04-16 14:46:00 +00004124 * do not need to check mode. Though link speed might not represent
4125 * the true receive or transmit bandwidth (not all modes are symmetric)
4126 * this is an accurate maximum.
4127 */
Veaceslav Falico9caff1e2013-09-25 09:20:14 +02004128 bond_for_each_slave(bond, slave, iter) {
Veaceslav Falico8557cd72014-05-15 21:39:59 +02004129 if (bond_slave_can_tx(slave)) {
Andy Gospodarekbb5b0522013-04-16 14:46:00 +00004130 if (slave->speed != SPEED_UNKNOWN)
4131 speed += slave->speed;
4132 if (ecmd->duplex == DUPLEX_UNKNOWN &&
4133 slave->duplex != DUPLEX_UNKNOWN)
4134 ecmd->duplex = slave->duplex;
4135 }
4136 }
4137 ethtool_cmd_speed_set(ecmd, speed ? : SPEED_UNKNOWN);
nikolay@redhat.comdec1e902013-08-01 16:54:47 +02004138
Andy Gospodarekbb5b0522013-04-16 14:46:00 +00004139 return 0;
4140}
4141
Jay Vosburgh217df672005-09-26 16:11:50 -07004142static void bond_ethtool_get_drvinfo(struct net_device *bond_dev,
Jiri Pirko7826d432013-01-06 00:44:26 +00004143 struct ethtool_drvinfo *drvinfo)
Jay Vosburgh217df672005-09-26 16:11:50 -07004144{
Jiri Pirko7826d432013-01-06 00:44:26 +00004145 strlcpy(drvinfo->driver, DRV_NAME, sizeof(drvinfo->driver));
4146 strlcpy(drvinfo->version, DRV_VERSION, sizeof(drvinfo->version));
4147 snprintf(drvinfo->fw_version, sizeof(drvinfo->fw_version), "%d",
4148 BOND_ABI_VERSION);
Jay Vosburgh217df672005-09-26 16:11:50 -07004149}
4150
Jeff Garzik7282d492006-09-13 14:30:00 -04004151static const struct ethtool_ops bond_ethtool_ops = {
Jay Vosburgh217df672005-09-26 16:11:50 -07004152 .get_drvinfo = bond_ethtool_get_drvinfo,
Andy Gospodarekbb5b0522013-04-16 14:46:00 +00004153 .get_settings = bond_ethtool_get_settings,
Stephen Hemmingerfa53eba2008-09-13 21:17:09 -04004154 .get_link = ethtool_op_get_link,
Arthur Kepner8531c5f2005-08-23 01:34:53 -04004155};
4156
Stephen Hemmingereb7cc592008-11-19 21:56:05 -08004157static const struct net_device_ops bond_netdev_ops = {
Stephen Hemminger181470f2009-06-12 19:02:52 +00004158 .ndo_init = bond_init,
Stephen Hemminger9e716262009-06-12 19:02:47 +00004159 .ndo_uninit = bond_uninit,
Stephen Hemmingereb7cc592008-11-19 21:56:05 -08004160 .ndo_open = bond_open,
4161 .ndo_stop = bond_close,
Stephen Hemminger00829822008-11-20 20:14:53 -08004162 .ndo_start_xmit = bond_start_xmit,
Andy Gospodarekbb1d9122010-06-02 08:40:18 +00004163 .ndo_select_queue = bond_select_queue,
Ben Hutchingsbe1f3c22010-06-08 07:19:54 +00004164 .ndo_get_stats64 = bond_get_stats,
Stephen Hemmingereb7cc592008-11-19 21:56:05 -08004165 .ndo_do_ioctl = bond_do_ioctl,
Jiri Pirkod03462b2011-08-16 03:15:04 +00004166 .ndo_change_rx_flags = bond_change_rx_flags,
Jay Vosburgh303d1cb2013-05-31 11:57:30 +00004167 .ndo_set_rx_mode = bond_set_rx_mode,
Stephen Hemmingereb7cc592008-11-19 21:56:05 -08004168 .ndo_change_mtu = bond_change_mtu,
Jiri Pirkocc0e4072011-07-20 04:54:46 +00004169 .ndo_set_mac_address = bond_set_mac_address,
Stephen Hemminger00829822008-11-20 20:14:53 -08004170 .ndo_neigh_setup = bond_neigh_setup,
Jiri Pirkocc0e4072011-07-20 04:54:46 +00004171 .ndo_vlan_rx_add_vid = bond_vlan_rx_add_vid,
Stephen Hemmingereb7cc592008-11-19 21:56:05 -08004172 .ndo_vlan_rx_kill_vid = bond_vlan_rx_kill_vid,
Eric Dumazet7142fdb2018-07-31 06:30:54 -07004173 .ndo_get_lock_subclass = bond_get_nest_level,
WANG Congf6dc31a2010-05-06 00:48:51 -07004174#ifdef CONFIG_NET_POLL_CONTROLLER
Amerigo Wang8a8efa22011-02-17 23:43:32 +00004175 .ndo_netpoll_setup = bond_netpoll_setup,
WANG Congf6dc31a2010-05-06 00:48:51 -07004176 .ndo_netpoll_cleanup = bond_netpoll_cleanup,
4177 .ndo_poll_controller = bond_poll_controller,
4178#endif
Jiri Pirko9232ecc2011-02-13 09:33:01 +00004179 .ndo_add_slave = bond_enslave,
4180 .ndo_del_slave = bond_release,
Michał Mirosławb2a103e2011-05-07 03:22:17 +00004181 .ndo_fix_features = bond_fix_features,
Jiri Pirko18bfb922016-07-05 11:27:38 +02004182 .ndo_neigh_construct = netdev_default_l2upper_neigh_construct,
4183 .ndo_neigh_destroy = netdev_default_l2upper_neigh_destroy,
Scott Feldmanfc8f40d2015-05-10 09:47:57 -07004184 .ndo_bridge_setlink = switchdev_port_bridge_setlink,
Scott Feldman85fdb952015-05-10 09:48:05 -07004185 .ndo_bridge_getlink = switchdev_port_bridge_getlink,
Scott Feldman54ba5a02015-05-10 09:48:01 -07004186 .ndo_bridge_dellink = switchdev_port_bridge_dellink,
Samudrala, Sridhar45d41222015-05-13 21:55:43 -07004187 .ndo_fdb_add = switchdev_port_fdb_add,
4188 .ndo_fdb_del = switchdev_port_fdb_del,
4189 .ndo_fdb_dump = switchdev_port_fdb_dump,
Toshiaki Makita4847f042015-03-27 14:31:14 +09004190 .ndo_features_check = passthru_features_check,
Stephen Hemmingereb7cc592008-11-19 21:56:05 -08004191};
4192
Doug Goldsteinb3f92b62013-02-18 14:59:23 +00004193static const struct device_type bond_type = {
4194 .name = "bond",
4195};
4196
Amerigo Wang9e2e61f2010-03-31 21:30:52 +00004197static void bond_destructor(struct net_device *bond_dev)
4198{
4199 struct bonding *bond = netdev_priv(bond_dev);
4200 if (bond->wq)
4201 destroy_workqueue(bond->wq);
4202 free_netdev(bond_dev);
4203}
4204
Jiri Pirko0a2a78c2013-10-18 17:43:33 +02004205void bond_setup(struct net_device *bond_dev)
Linus Torvalds1da177e2005-04-16 15:20:36 -07004206{
Wang Chen454d7c92008-11-12 23:37:49 -08004207 struct bonding *bond = netdev_priv(bond_dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -07004208
Nikolay Aleksandrovb7435622014-09-11 22:49:25 +02004209 spin_lock_init(&bond->mode_lock);
Eric Dumazetfe309372016-03-17 17:23:36 -07004210 spin_lock_init(&bond->stats_lock);
Stephen Hemmingerd2991f72009-06-12 19:02:44 +00004211 bond->params = bonding_defaults;
Linus Torvalds1da177e2005-04-16 15:20:36 -07004212
4213 /* Initialize pointers */
Linus Torvalds1da177e2005-04-16 15:20:36 -07004214 bond->dev = bond_dev;
Linus Torvalds1da177e2005-04-16 15:20:36 -07004215
4216 /* Initialize the device entry points */
Stephen Hemminger181470f2009-06-12 19:02:52 +00004217 ether_setup(bond_dev);
Stephen Hemmingereb7cc592008-11-19 21:56:05 -08004218 bond_dev->netdev_ops = &bond_netdev_ops;
Arthur Kepner8531c5f2005-08-23 01:34:53 -04004219 bond_dev->ethtool_ops = &bond_ethtool_ops;
Linus Torvalds1da177e2005-04-16 15:20:36 -07004220
Amerigo Wang9e2e61f2010-03-31 21:30:52 +00004221 bond_dev->destructor = bond_destructor;
Linus Torvalds1da177e2005-04-16 15:20:36 -07004222
Doug Goldsteinb3f92b62013-02-18 14:59:23 +00004223 SET_NETDEV_DEVTYPE(bond_dev, &bond_type);
4224
Linus Torvalds1da177e2005-04-16 15:20:36 -07004225 /* Initialize the device options */
Zhang Shengju1098cee2016-03-16 09:59:15 +00004226 bond_dev->flags |= IFF_MASTER;
Phil Sutter1e6f20c2015-08-18 10:30:39 +02004227 bond_dev->priv_flags |= IFF_BONDING | IFF_UNICAST_FLT | IFF_NO_QUEUE;
Neil Horman550fd082011-07-26 06:05:38 +00004228 bond_dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_TX_SKB_SHARING);
Stephen Hemminger181470f2009-06-12 19:02:52 +00004229
Nikolay Aleksandrov547942c2014-09-15 17:19:34 +02004230 /* don't acquire bond device's netif_tx_lock when transmitting */
Linus Torvalds1da177e2005-04-16 15:20:36 -07004231 bond_dev->features |= NETIF_F_LLTX;
4232
4233 /* By default, we declare the bond to be fully
4234 * VLAN hardware accelerated capable. Special
4235 * care is taken in the various xmit functions
4236 * when there are slaves that are not hw accel
4237 * capable
4238 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07004239
Weilong Chenf9399812014-01-22 17:16:30 +08004240 /* Don't allow bond devices to change network namespaces. */
4241 bond_dev->features |= NETIF_F_NETNS_LOCAL;
4242
Michał Mirosławb2a103e2011-05-07 03:22:17 +00004243 bond_dev->hw_features = BOND_VLAN_FEATURES |
Patrick McHardyf6469682013-04-19 02:04:27 +00004244 NETIF_F_HW_VLAN_CTAG_RX |
4245 NETIF_F_HW_VLAN_CTAG_FILTER;
Michał Mirosławb2a103e2011-05-07 03:22:17 +00004246
Eric Dumazet24f87d42015-01-25 23:27:14 -08004247 bond_dev->hw_features |= NETIF_F_GSO_ENCAP_ALL;
Michał Mirosławb2a103e2011-05-07 03:22:17 +00004248 bond_dev->features |= bond_dev->hw_features;
YueHaibingb93a8c22019-06-26 16:08:44 +08004249 bond_dev->features |= NETIF_F_HW_VLAN_CTAG_TX;
Linus Torvalds1da177e2005-04-16 15:20:36 -07004250}
4251
Nikolay Aleksandrov547942c2014-09-15 17:19:34 +02004252/* Destroy a bonding device.
4253 * Must be under rtnl_lock when this function is called.
4254 */
Eric W. Biedermanc67dfb22009-10-29 14:18:24 +00004255static void bond_uninit(struct net_device *bond_dev)
Jay Vosburgha434e432008-10-30 17:41:15 -07004256{
Wang Chen454d7c92008-11-12 23:37:49 -08004257 struct bonding *bond = netdev_priv(bond_dev);
Veaceslav Falico544a0282013-09-25 09:20:15 +02004258 struct list_head *iter;
4259 struct slave *slave;
Mahesh Bandewaree637712014-10-04 17:45:01 -07004260 struct bond_up_slave *arr;
Jay Vosburgha434e432008-10-30 17:41:15 -07004261
WANG Congf6dc31a2010-05-06 00:48:51 -07004262 bond_netpoll_cleanup(bond_dev);
4263
Eric W. Biedermanc67dfb22009-10-29 14:18:24 +00004264 /* Release the bonded slaves */
Veaceslav Falico544a0282013-09-25 09:20:15 +02004265 bond_for_each_slave(bond, slave, iter)
nikolay@redhat.comdec1e902013-08-01 16:54:47 +02004266 __bond_release_one(bond_dev, slave->dev, true);
Veaceslav Falico76444f52014-07-15 19:35:58 +02004267 netdev_info(bond_dev, "Released all slaves\n");
Eric W. Biedermanc67dfb22009-10-29 14:18:24 +00004268
Mahesh Bandewaree637712014-10-04 17:45:01 -07004269 arr = rtnl_dereference(bond->slave_arr);
4270 if (arr) {
4271 RCU_INIT_POINTER(bond->slave_arr, NULL);
4272 kfree_rcu(arr, rcu);
4273 }
4274
Jay Vosburgha434e432008-10-30 17:41:15 -07004275 list_del(&bond->bond_list);
4276
Taku Izumif073c7c2010-12-09 15:17:13 +00004277 bond_debug_unregister(bond);
Jay Vosburgha434e432008-10-30 17:41:15 -07004278}
4279
Linus Torvalds1da177e2005-04-16 15:20:36 -07004280/*------------------------- Module initialization ---------------------------*/
4281
Linus Torvalds1da177e2005-04-16 15:20:36 -07004282static int bond_check_params(struct bond_params *params)
4283{
nikolay@redhat.com5a5c5fd2013-05-18 01:18:30 +00004284 int arp_validate_value, fail_over_mac_value, primary_reselect_value, i;
stephen hemmingerf3253332014-03-04 16:36:44 -08004285 struct bond_opt_value newval;
4286 const struct bond_opt_value *valptr;
Veaceslav Falico8599b522013-06-24 11:49:34 +02004287 int arp_all_targets_value;
Mahesh Bandewar6791e462015-05-09 00:01:55 -07004288 u16 ad_actor_sys_prio = 0;
Mahesh Bandeward22a5fc2015-05-09 00:01:57 -07004289 u16 ad_user_port_key = 0;
Jay Vosburghf5b2b962006-09-22 21:54:53 -07004290
Nikolay Aleksandrov547942c2014-09-15 17:19:34 +02004291 /* Convert string parameters. */
Linus Torvalds1da177e2005-04-16 15:20:36 -07004292 if (mode) {
Nikolay Aleksandrov2b3798d2014-01-22 14:53:17 +01004293 bond_opt_initstr(&newval, mode);
4294 valptr = bond_opt_parse(bond_opt_get(BOND_OPT_MODE), &newval);
4295 if (!valptr) {
4296 pr_err("Error: Invalid bonding mode \"%s\"\n", mode);
Linus Torvalds1da177e2005-04-16 15:20:36 -07004297 return -EINVAL;
4298 }
Nikolay Aleksandrov2b3798d2014-01-22 14:53:17 +01004299 bond_mode = valptr->value;
Linus Torvalds1da177e2005-04-16 15:20:36 -07004300 }
4301
Jay Vosburgh169a3e62005-06-26 17:54:11 -04004302 if (xmit_hash_policy) {
4303 if ((bond_mode != BOND_MODE_XOR) &&
Mahesh Bandewarf05b42e2014-04-22 16:30:20 -07004304 (bond_mode != BOND_MODE_8023AD) &&
4305 (bond_mode != BOND_MODE_TLB)) {
Joe Perchesa4aee5c2009-12-13 20:06:07 -08004306 pr_info("xmit_hash_policy param is irrelevant in mode %s\n",
Joe Perches90194262014-02-15 16:01:45 -08004307 bond_mode_name(bond_mode));
Jay Vosburgh169a3e62005-06-26 17:54:11 -04004308 } else {
Nikolay Aleksandrova4b32ce2014-01-22 14:53:19 +01004309 bond_opt_initstr(&newval, xmit_hash_policy);
4310 valptr = bond_opt_parse(bond_opt_get(BOND_OPT_XMIT_HASH),
4311 &newval);
4312 if (!valptr) {
Joe Perchesa4aee5c2009-12-13 20:06:07 -08004313 pr_err("Error: Invalid xmit_hash_policy \"%s\"\n",
Jay Vosburgh169a3e62005-06-26 17:54:11 -04004314 xmit_hash_policy);
4315 return -EINVAL;
4316 }
Nikolay Aleksandrova4b32ce2014-01-22 14:53:19 +01004317 xmit_hashtype = valptr->value;
Jay Vosburgh169a3e62005-06-26 17:54:11 -04004318 }
4319 }
4320
Linus Torvalds1da177e2005-04-16 15:20:36 -07004321 if (lacp_rate) {
4322 if (bond_mode != BOND_MODE_8023AD) {
Joe Perchesa4aee5c2009-12-13 20:06:07 -08004323 pr_info("lacp_rate param is irrelevant in mode %s\n",
4324 bond_mode_name(bond_mode));
Linus Torvalds1da177e2005-04-16 15:20:36 -07004325 } else {
Nikolay Aleksandrovd3131de2014-01-22 14:53:27 +01004326 bond_opt_initstr(&newval, lacp_rate);
4327 valptr = bond_opt_parse(bond_opt_get(BOND_OPT_LACP_RATE),
4328 &newval);
4329 if (!valptr) {
Joe Perchesa4aee5c2009-12-13 20:06:07 -08004330 pr_err("Error: Invalid lacp rate \"%s\"\n",
Nikolay Aleksandrovd3131de2014-01-22 14:53:27 +01004331 lacp_rate);
Linus Torvalds1da177e2005-04-16 15:20:36 -07004332 return -EINVAL;
4333 }
Nikolay Aleksandrovd3131de2014-01-22 14:53:27 +01004334 lacp_fast = valptr->value;
Linus Torvalds1da177e2005-04-16 15:20:36 -07004335 }
4336 }
4337
Jay Vosburghfd989c82008-11-04 17:51:16 -08004338 if (ad_select) {
Nikolay Aleksandrov548d28b2014-07-13 09:47:47 +02004339 bond_opt_initstr(&newval, ad_select);
Nikolay Aleksandrov9e5f5ee2014-01-22 14:53:29 +01004340 valptr = bond_opt_parse(bond_opt_get(BOND_OPT_AD_SELECT),
4341 &newval);
4342 if (!valptr) {
4343 pr_err("Error: Invalid ad_select \"%s\"\n", ad_select);
Jay Vosburghfd989c82008-11-04 17:51:16 -08004344 return -EINVAL;
4345 }
Nikolay Aleksandrov9e5f5ee2014-01-22 14:53:29 +01004346 params->ad_select = valptr->value;
4347 if (bond_mode != BOND_MODE_8023AD)
Joe Perches91565eb2014-02-15 15:57:04 -08004348 pr_warn("ad_select param only affects 802.3ad mode\n");
Jay Vosburghfd989c82008-11-04 17:51:16 -08004349 } else {
4350 params->ad_select = BOND_AD_STABLE;
4351 }
4352
Nicolas de Pesloüanf5841302009-08-28 13:18:34 +00004353 if (max_bonds < 0) {
Joe Perches91565eb2014-02-15 15:57:04 -08004354 pr_warn("Warning: max_bonds (%d) not in range %d-%d, so it was reset to BOND_DEFAULT_MAX_BONDS (%d)\n",
4355 max_bonds, 0, INT_MAX, BOND_DEFAULT_MAX_BONDS);
Linus Torvalds1da177e2005-04-16 15:20:36 -07004356 max_bonds = BOND_DEFAULT_MAX_BONDS;
4357 }
4358
4359 if (miimon < 0) {
Joe Perches91565eb2014-02-15 15:57:04 -08004360 pr_warn("Warning: miimon module parameter (%d), not in range 0-%d, so it was reset to 0\n",
4361 miimon, INT_MAX);
Nikolay Aleksandrovb98d9c62014-01-22 14:53:31 +01004362 miimon = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07004363 }
4364
4365 if (updelay < 0) {
Joe Perches91565eb2014-02-15 15:57:04 -08004366 pr_warn("Warning: updelay module parameter (%d), not in range 0-%d, so it was reset to 0\n",
4367 updelay, INT_MAX);
Linus Torvalds1da177e2005-04-16 15:20:36 -07004368 updelay = 0;
4369 }
4370
4371 if (downdelay < 0) {
Joe Perches91565eb2014-02-15 15:57:04 -08004372 pr_warn("Warning: downdelay module parameter (%d), not in range 0-%d, so it was reset to 0\n",
4373 downdelay, INT_MAX);
Linus Torvalds1da177e2005-04-16 15:20:36 -07004374 downdelay = 0;
4375 }
4376
4377 if ((use_carrier != 0) && (use_carrier != 1)) {
Joe Perches91565eb2014-02-15 15:57:04 -08004378 pr_warn("Warning: use_carrier module parameter (%d), not of valid value (0/1), so it was set to 1\n",
4379 use_carrier);
Linus Torvalds1da177e2005-04-16 15:20:36 -07004380 use_carrier = 1;
4381 }
4382
Ben Hutchingsad246c92011-04-26 15:25:52 +00004383 if (num_peer_notif < 0 || num_peer_notif > 255) {
Joe Perches91565eb2014-02-15 15:57:04 -08004384 pr_warn("Warning: num_grat_arp/num_unsol_na (%d) not in range 0-255 so it was reset to 1\n",
4385 num_peer_notif);
Ben Hutchingsad246c92011-04-26 15:25:52 +00004386 num_peer_notif = 1;
4387 }
4388
dingtianhong834db4b2013-12-21 14:40:17 +08004389 /* reset values for 802.3ad/TLB/ALB */
Veaceslav Falico267bed72014-05-15 21:39:53 +02004390 if (!bond_mode_uses_arp(bond_mode)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07004391 if (!miimon) {
Joe Perches91565eb2014-02-15 15:57:04 -08004392 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");
4393 pr_warn("Forcing miimon to 100msec\n");
dingtianhongfe9d04a2013-11-22 22:28:43 +08004394 miimon = BOND_DEFAULT_MIIMON;
Linus Torvalds1da177e2005-04-16 15:20:36 -07004395 }
4396 }
4397
Andy Gospodarekbb1d9122010-06-02 08:40:18 +00004398 if (tx_queues < 1 || tx_queues > 255) {
Joe Perches91565eb2014-02-15 15:57:04 -08004399 pr_warn("Warning: tx_queues (%d) should be between 1 and 255, resetting to %d\n",
4400 tx_queues, BOND_DEFAULT_TX_QUEUES);
Andy Gospodarekbb1d9122010-06-02 08:40:18 +00004401 tx_queues = BOND_DEFAULT_TX_QUEUES;
4402 }
4403
Andy Gospodarekebd8e492010-06-02 08:39:21 +00004404 if ((all_slaves_active != 0) && (all_slaves_active != 1)) {
Joe Perches91565eb2014-02-15 15:57:04 -08004405 pr_warn("Warning: all_slaves_active module parameter (%d), not of valid value (0/1), so it was set to 0\n",
4406 all_slaves_active);
Andy Gospodarekebd8e492010-06-02 08:39:21 +00004407 all_slaves_active = 0;
4408 }
4409
Flavio Leitnerc2952c32010-10-05 14:23:59 +00004410 if (resend_igmp < 0 || resend_igmp > 255) {
Joe Perches91565eb2014-02-15 15:57:04 -08004411 pr_warn("Warning: resend_igmp (%d) should be between 0 and 255, resetting to %d\n",
4412 resend_igmp, BOND_DEFAULT_RESEND_IGMP);
Flavio Leitnerc2952c32010-10-05 14:23:59 +00004413 resend_igmp = BOND_DEFAULT_RESEND_IGMP;
4414 }
4415
Nikolay Aleksandrovaa59d852014-01-22 14:53:18 +01004416 bond_opt_initval(&newval, packets_per_slave);
4417 if (!bond_opt_parse(bond_opt_get(BOND_OPT_PACKETS_PER_SLAVE), &newval)) {
Nikolay Aleksandrov73958322013-11-05 13:51:41 +01004418 pr_warn("Warning: packets_per_slave (%d) should be between 0 and %u resetting to 1\n",
4419 packets_per_slave, USHRT_MAX);
4420 packets_per_slave = 1;
4421 }
4422
Linus Torvalds1da177e2005-04-16 15:20:36 -07004423 if (bond_mode == BOND_MODE_ALB) {
Joe Perchesa4aee5c2009-12-13 20:06:07 -08004424 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",
4425 updelay);
Linus Torvalds1da177e2005-04-16 15:20:36 -07004426 }
4427
4428 if (!miimon) {
4429 if (updelay || downdelay) {
4430 /* just warn the user the up/down delay will have
4431 * no effect since miimon is zero...
4432 */
Joe Perches91565eb2014-02-15 15:57:04 -08004433 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",
4434 updelay, downdelay);
Linus Torvalds1da177e2005-04-16 15:20:36 -07004435 }
4436 } else {
4437 /* don't allow arp monitoring */
4438 if (arp_interval) {
Joe Perches91565eb2014-02-15 15:57:04 -08004439 pr_warn("Warning: miimon (%d) and arp_interval (%d) can't be used simultaneously, disabling ARP monitoring\n",
4440 miimon, arp_interval);
Linus Torvalds1da177e2005-04-16 15:20:36 -07004441 arp_interval = 0;
4442 }
4443
4444 if ((updelay % miimon) != 0) {
Joe Perches91565eb2014-02-15 15:57:04 -08004445 pr_warn("Warning: updelay (%d) is not a multiple of miimon (%d), updelay rounded to %d ms\n",
4446 updelay, miimon, (updelay / miimon) * miimon);
Linus Torvalds1da177e2005-04-16 15:20:36 -07004447 }
4448
4449 updelay /= miimon;
4450
4451 if ((downdelay % miimon) != 0) {
Joe Perches91565eb2014-02-15 15:57:04 -08004452 pr_warn("Warning: downdelay (%d) is not a multiple of miimon (%d), downdelay rounded to %d ms\n",
4453 downdelay, miimon,
4454 (downdelay / miimon) * miimon);
Linus Torvalds1da177e2005-04-16 15:20:36 -07004455 }
4456
4457 downdelay /= miimon;
4458 }
4459
4460 if (arp_interval < 0) {
Joe Perches91565eb2014-02-15 15:57:04 -08004461 pr_warn("Warning: arp_interval module parameter (%d), not in range 0-%d, so it was reset to 0\n",
4462 arp_interval, INT_MAX);
Nikolay Aleksandrov7bdb04e2014-01-22 14:53:23 +01004463 arp_interval = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07004464 }
4465
nikolay@redhat.com5a5c5fd2013-05-18 01:18:30 +00004466 for (arp_ip_count = 0, i = 0;
4467 (arp_ip_count < BOND_MAX_ARP_TARGETS) && arp_ip_target[i]; i++) {
dingtianhong89015c12013-12-04 18:59:31 +08004468 __be32 ip;
Nikolay Aleksandrov547942c2014-09-15 17:19:34 +02004469
4470 /* not a complete check, but good enough to catch mistakes */
dingtianhong89015c12013-12-04 18:59:31 +08004471 if (!in4_pton(arp_ip_target[i], -1, (u8 *)&ip, -1, NULL) ||
Veaceslav Falico2807a9f2014-05-15 21:39:56 +02004472 !bond_is_ip_target_ok(ip)) {
Joe Perches91565eb2014-02-15 15:57:04 -08004473 pr_warn("Warning: bad arp_ip_target module parameter (%s), ARP monitoring will not be performed\n",
4474 arp_ip_target[i]);
Linus Torvalds1da177e2005-04-16 15:20:36 -07004475 arp_interval = 0;
4476 } else {
Veaceslav Falico0afee4e2013-06-24 11:49:30 +02004477 if (bond_get_targets_ip(arp_target, ip) == -1)
4478 arp_target[arp_ip_count++] = ip;
4479 else
Joe Perches91565eb2014-02-15 15:57:04 -08004480 pr_warn("Warning: duplicate address %pI4 in arp_ip_target, skipping\n",
4481 &ip);
Linus Torvalds1da177e2005-04-16 15:20:36 -07004482 }
4483 }
4484
4485 if (arp_interval && !arp_ip_count) {
4486 /* don't allow arping if no arp_ip_target given... */
Joe Perches91565eb2014-02-15 15:57:04 -08004487 pr_warn("Warning: arp_interval module parameter (%d) specified without providing an arp_ip_target parameter, arp_interval was reset to 0\n",
4488 arp_interval);
Linus Torvalds1da177e2005-04-16 15:20:36 -07004489 arp_interval = 0;
4490 }
4491
Jay Vosburghf5b2b962006-09-22 21:54:53 -07004492 if (arp_validate) {
Jay Vosburghf5b2b962006-09-22 21:54:53 -07004493 if (!arp_interval) {
Joe Perchesa4aee5c2009-12-13 20:06:07 -08004494 pr_err("arp_validate requires arp_interval\n");
Jay Vosburghf5b2b962006-09-22 21:54:53 -07004495 return -EINVAL;
4496 }
4497
Nikolay Aleksandrov16228882014-01-22 14:53:20 +01004498 bond_opt_initstr(&newval, arp_validate);
4499 valptr = bond_opt_parse(bond_opt_get(BOND_OPT_ARP_VALIDATE),
4500 &newval);
4501 if (!valptr) {
Joe Perchesa4aee5c2009-12-13 20:06:07 -08004502 pr_err("Error: invalid arp_validate \"%s\"\n",
Nikolay Aleksandrov16228882014-01-22 14:53:20 +01004503 arp_validate);
Jay Vosburghf5b2b962006-09-22 21:54:53 -07004504 return -EINVAL;
4505 }
Nikolay Aleksandrov16228882014-01-22 14:53:20 +01004506 arp_validate_value = valptr->value;
4507 } else {
Jay Vosburghf5b2b962006-09-22 21:54:53 -07004508 arp_validate_value = 0;
Nikolay Aleksandrov16228882014-01-22 14:53:20 +01004509 }
Jay Vosburghf5b2b962006-09-22 21:54:53 -07004510
Veaceslav Falico8599b522013-06-24 11:49:34 +02004511 arp_all_targets_value = 0;
4512 if (arp_all_targets) {
Nikolay Aleksandrovedf36b22014-01-22 14:53:21 +01004513 bond_opt_initstr(&newval, arp_all_targets);
4514 valptr = bond_opt_parse(bond_opt_get(BOND_OPT_ARP_ALL_TARGETS),
4515 &newval);
4516 if (!valptr) {
Veaceslav Falico8599b522013-06-24 11:49:34 +02004517 pr_err("Error: invalid arp_all_targets_value \"%s\"\n",
4518 arp_all_targets);
4519 arp_all_targets_value = 0;
Nikolay Aleksandrovedf36b22014-01-22 14:53:21 +01004520 } else {
4521 arp_all_targets_value = valptr->value;
Veaceslav Falico8599b522013-06-24 11:49:34 +02004522 }
4523 }
4524
Linus Torvalds1da177e2005-04-16 15:20:36 -07004525 if (miimon) {
Joe Perchesa4aee5c2009-12-13 20:06:07 -08004526 pr_info("MII link monitoring set to %d ms\n", miimon);
Linus Torvalds1da177e2005-04-16 15:20:36 -07004527 } else if (arp_interval) {
Nikolay Aleksandrov16228882014-01-22 14:53:20 +01004528 valptr = bond_opt_get_val(BOND_OPT_ARP_VALIDATE,
4529 arp_validate_value);
Joe Perchesa4aee5c2009-12-13 20:06:07 -08004530 pr_info("ARP monitoring set to %d ms, validate %s, with %d target(s):",
Nikolay Aleksandrov16228882014-01-22 14:53:20 +01004531 arp_interval, valptr->string, arp_ip_count);
Linus Torvalds1da177e2005-04-16 15:20:36 -07004532
4533 for (i = 0; i < arp_ip_count; i++)
Joe Perches90194262014-02-15 16:01:45 -08004534 pr_cont(" %s", arp_ip_target[i]);
Linus Torvalds1da177e2005-04-16 15:20:36 -07004535
Joe Perches90194262014-02-15 16:01:45 -08004536 pr_cont("\n");
Linus Torvalds1da177e2005-04-16 15:20:36 -07004537
Jay Vosburghb8a97872008-06-13 18:12:04 -07004538 } else if (max_bonds) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07004539 /* miimon and arp_interval not set, we need one so things
4540 * work as expected, see bonding.txt for details
4541 */
Joe Perches90194262014-02-15 16:01:45 -08004542 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");
Linus Torvalds1da177e2005-04-16 15:20:36 -07004543 }
4544
Veaceslav Falicoec0865a2014-05-15 21:39:54 +02004545 if (primary && !bond_mode_uses_primary(bond_mode)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07004546 /* currently, using a primary only makes sense
4547 * in active backup, TLB or ALB modes
4548 */
Joe Perches91565eb2014-02-15 15:57:04 -08004549 pr_warn("Warning: %s primary device specified but has no effect in %s mode\n",
4550 primary, bond_mode_name(bond_mode));
Linus Torvalds1da177e2005-04-16 15:20:36 -07004551 primary = NULL;
4552 }
4553
Jiri Pirkoa5499522009-09-25 03:28:09 +00004554 if (primary && primary_reselect) {
Nikolay Aleksandrov388d3a62014-01-22 14:53:33 +01004555 bond_opt_initstr(&newval, primary_reselect);
4556 valptr = bond_opt_parse(bond_opt_get(BOND_OPT_PRIMARY_RESELECT),
4557 &newval);
4558 if (!valptr) {
Joe Perchesa4aee5c2009-12-13 20:06:07 -08004559 pr_err("Error: Invalid primary_reselect \"%s\"\n",
Nikolay Aleksandrov388d3a62014-01-22 14:53:33 +01004560 primary_reselect);
Jiri Pirkoa5499522009-09-25 03:28:09 +00004561 return -EINVAL;
4562 }
Nikolay Aleksandrov388d3a62014-01-22 14:53:33 +01004563 primary_reselect_value = valptr->value;
Jiri Pirkoa5499522009-09-25 03:28:09 +00004564 } else {
4565 primary_reselect_value = BOND_PRI_RESELECT_ALWAYS;
4566 }
4567
Jay Vosburgh3915c1e82008-05-17 21:10:14 -07004568 if (fail_over_mac) {
Nikolay Aleksandrov1df6b6a2014-01-22 14:53:22 +01004569 bond_opt_initstr(&newval, fail_over_mac);
4570 valptr = bond_opt_parse(bond_opt_get(BOND_OPT_FAIL_OVER_MAC),
4571 &newval);
4572 if (!valptr) {
Joe Perchesa4aee5c2009-12-13 20:06:07 -08004573 pr_err("Error: invalid fail_over_mac \"%s\"\n",
Nikolay Aleksandrov1df6b6a2014-01-22 14:53:22 +01004574 fail_over_mac);
Jay Vosburgh3915c1e82008-05-17 21:10:14 -07004575 return -EINVAL;
4576 }
Nikolay Aleksandrov1df6b6a2014-01-22 14:53:22 +01004577 fail_over_mac_value = valptr->value;
Jay Vosburgh3915c1e82008-05-17 21:10:14 -07004578 if (bond_mode != BOND_MODE_ACTIVEBACKUP)
Joe Perches91565eb2014-02-15 15:57:04 -08004579 pr_warn("Warning: fail_over_mac only affects active-backup mode\n");
Jay Vosburgh3915c1e82008-05-17 21:10:14 -07004580 } else {
4581 fail_over_mac_value = BOND_FOM_NONE;
4582 }
Jay Vosburghdd957c52007-10-09 19:57:24 -07004583
Mahesh Bandewar6791e462015-05-09 00:01:55 -07004584 bond_opt_initstr(&newval, "default");
4585 valptr = bond_opt_parse(
4586 bond_opt_get(BOND_OPT_AD_ACTOR_SYS_PRIO),
4587 &newval);
4588 if (!valptr) {
4589 pr_err("Error: No ad_actor_sys_prio default value");
4590 return -EINVAL;
4591 }
4592 ad_actor_sys_prio = valptr->value;
4593
Mahesh Bandeward22a5fc2015-05-09 00:01:57 -07004594 valptr = bond_opt_parse(bond_opt_get(BOND_OPT_AD_USER_PORT_KEY),
4595 &newval);
4596 if (!valptr) {
4597 pr_err("Error: No ad_user_port_key default value");
4598 return -EINVAL;
4599 }
4600 ad_user_port_key = valptr->value;
4601
dingtianhong3a7129e2013-12-21 14:40:12 +08004602 if (lp_interval == 0) {
Joe Perches91565eb2014-02-15 15:57:04 -08004603 pr_warn("Warning: ip_interval must be between 1 and %d, so it was reset to %d\n",
4604 INT_MAX, BOND_ALB_DEFAULT_LP_INTERVAL);
dingtianhong3a7129e2013-12-21 14:40:12 +08004605 lp_interval = BOND_ALB_DEFAULT_LP_INTERVAL;
4606 }
4607
Linus Torvalds1da177e2005-04-16 15:20:36 -07004608 /* fill params struct with the proper values */
4609 params->mode = bond_mode;
Jay Vosburgh169a3e62005-06-26 17:54:11 -04004610 params->xmit_policy = xmit_hashtype;
Linus Torvalds1da177e2005-04-16 15:20:36 -07004611 params->miimon = miimon;
Ben Hutchingsad246c92011-04-26 15:25:52 +00004612 params->num_peer_notif = num_peer_notif;
Linus Torvalds1da177e2005-04-16 15:20:36 -07004613 params->arp_interval = arp_interval;
Jay Vosburghf5b2b962006-09-22 21:54:53 -07004614 params->arp_validate = arp_validate_value;
Veaceslav Falico8599b522013-06-24 11:49:34 +02004615 params->arp_all_targets = arp_all_targets_value;
Linus Torvalds1da177e2005-04-16 15:20:36 -07004616 params->updelay = updelay;
4617 params->downdelay = downdelay;
4618 params->use_carrier = use_carrier;
4619 params->lacp_fast = lacp_fast;
4620 params->primary[0] = 0;
Jiri Pirkoa5499522009-09-25 03:28:09 +00004621 params->primary_reselect = primary_reselect_value;
Jay Vosburgh3915c1e82008-05-17 21:10:14 -07004622 params->fail_over_mac = fail_over_mac_value;
Andy Gospodarekbb1d9122010-06-02 08:40:18 +00004623 params->tx_queues = tx_queues;
Andy Gospodarekebd8e492010-06-02 08:39:21 +00004624 params->all_slaves_active = all_slaves_active;
Flavio Leitnerc2952c32010-10-05 14:23:59 +00004625 params->resend_igmp = resend_igmp;
stephen hemminger655f8912011-06-22 09:54:39 +00004626 params->min_links = min_links;
dingtianhong3a7129e2013-12-21 14:40:12 +08004627 params->lp_interval = lp_interval;
Hannes Frederic Sowa809fa972014-01-22 02:29:41 +01004628 params->packets_per_slave = packets_per_slave;
Mahesh Bandeware9f0fb82014-04-22 16:30:22 -07004629 params->tlb_dynamic_lb = 1; /* Default value */
Mahesh Bandewar6791e462015-05-09 00:01:55 -07004630 params->ad_actor_sys_prio = ad_actor_sys_prio;
Mahesh Bandewar74514952015-05-09 00:01:56 -07004631 eth_zero_addr(params->ad_actor_system);
Mahesh Bandeward22a5fc2015-05-09 00:01:57 -07004632 params->ad_user_port_key = ad_user_port_key;
Hannes Frederic Sowa809fa972014-01-22 02:29:41 +01004633 if (packets_per_slave > 0) {
4634 params->reciprocal_packets_per_slave =
4635 reciprocal_value(packets_per_slave);
4636 } else {
4637 /* reciprocal_packets_per_slave is unused if
4638 * packets_per_slave is 0 or 1, just initialize it
4639 */
4640 params->reciprocal_packets_per_slave =
4641 (struct reciprocal_value) { 0 };
4642 }
4643
Linus Torvalds1da177e2005-04-16 15:20:36 -07004644 if (primary) {
4645 strncpy(params->primary, primary, IFNAMSIZ);
4646 params->primary[IFNAMSIZ - 1] = 0;
4647 }
4648
4649 memcpy(params->arp_targets, arp_target, sizeof(arp_target));
4650
4651 return 0;
4652}
4653
Nikolay Aleksandrov547942c2014-09-15 17:19:34 +02004654/* Called from registration process */
Stephen Hemminger181470f2009-06-12 19:02:52 +00004655static int bond_init(struct net_device *bond_dev)
4656{
4657 struct bonding *bond = netdev_priv(bond_dev);
Eric W. Biedermanec87fd32009-10-29 14:18:26 +00004658 struct bond_net *bn = net_generic(dev_net(bond_dev), bond_net_id);
Stephen Hemminger181470f2009-06-12 19:02:52 +00004659
Veaceslav Falico76444f52014-07-15 19:35:58 +02004660 netdev_dbg(bond_dev, "Begin bond_init\n");
Stephen Hemminger181470f2009-06-12 19:02:52 +00004661
Bhaktipriya Shridharf9f225e2016-08-30 22:02:01 +05304662 bond->wq = alloc_ordered_workqueue(bond_dev->name, WQ_MEM_RECLAIM);
Stephen Hemminger181470f2009-06-12 19:02:52 +00004663 if (!bond->wq)
4664 return -ENOMEM;
4665
Eric Dumazet7142fdb2018-07-31 06:30:54 -07004666 bond->nest_level = SINGLE_DEPTH_NESTING;
Eric Dumazetd3fff6c2016-06-09 07:45:12 -07004667 netdev_lockdep_set_classes(bond_dev);
Stephen Hemminger181470f2009-06-12 19:02:52 +00004668
Eric W. Biedermanec87fd32009-10-29 14:18:26 +00004669 list_add_tail(&bond->bond_list, &bn->dev_list);
Stephen Hemminger181470f2009-06-12 19:02:52 +00004670
Eric W. Biederman6151b3d2009-10-29 14:18:22 +00004671 bond_prepare_sysfs_group(bond);
Jiri Pirko22bedad32010-04-01 21:22:57 +00004672
Taku Izumif073c7c2010-12-09 15:17:13 +00004673 bond_debug_register(bond);
4674
Jiri Pirko409cc1f2013-01-30 11:08:11 +01004675 /* Ensure valid dev_addr */
4676 if (is_zero_ether_addr(bond_dev->dev_addr) &&
nikolay@redhat.com97a1e632013-06-26 17:13:38 +02004677 bond_dev->addr_assign_type == NET_ADDR_PERM)
Jiri Pirko409cc1f2013-01-30 11:08:11 +01004678 eth_hw_addr_random(bond_dev);
Jiri Pirko409cc1f2013-01-30 11:08:11 +01004679
Stephen Hemminger181470f2009-06-12 19:02:52 +00004680 return 0;
4681}
4682
Jiri Pirko0a2a78c2013-10-18 17:43:33 +02004683unsigned int bond_get_num_tx_queues(void)
Jiri Pirkod5da4512011-08-10 06:09:44 +00004684{
stephen hemmingerefacb302012-04-10 18:34:43 +00004685 return tx_queues;
Jiri Pirkod5da4512011-08-10 06:09:44 +00004686}
4687
Mitch Williamsdfe60392005-11-09 10:36:04 -08004688/* Create a new bond based on the specified name and bonding parameters.
Jay Vosburghe4b91c42007-01-19 18:15:31 -08004689 * If name is NULL, obtain a suitable "bond%d" name for us.
Mitch Williamsdfe60392005-11-09 10:36:04 -08004690 * Caller must NOT hold rtnl_lock; we need to release it here before we
4691 * set up our sysfs entries.
4692 */
Eric W. Biedermanec87fd32009-10-29 14:18:26 +00004693int bond_create(struct net *net, const char *name)
Mitch Williamsdfe60392005-11-09 10:36:04 -08004694{
4695 struct net_device *bond_dev;
Paie913fb22015-04-29 14:24:23 -04004696 struct bonding *bond;
4697 struct alb_bond_info *bond_info;
Mitch Williamsdfe60392005-11-09 10:36:04 -08004698 int res;
4699
4700 rtnl_lock();
Jay Vosburgh027ea042008-01-17 16:25:02 -08004701
Jiri Pirko1c5cae82011-04-30 01:21:32 +00004702 bond_dev = alloc_netdev_mq(sizeof(struct bonding),
Tom Gundersenc835a672014-07-14 16:37:24 +02004703 name ? name : "bond%d", NET_NAME_UNKNOWN,
Jiri Pirko1c5cae82011-04-30 01:21:32 +00004704 bond_setup, tx_queues);
Mitch Williamsdfe60392005-11-09 10:36:04 -08004705 if (!bond_dev) {
Joe Perchesa4aee5c2009-12-13 20:06:07 -08004706 pr_err("%s: eek! can't alloc netdev!\n", name);
Amerigo Wang9e2e61f2010-03-31 21:30:52 +00004707 rtnl_unlock();
4708 return -ENOMEM;
Mitch Williamsdfe60392005-11-09 10:36:04 -08004709 }
4710
Paie913fb22015-04-29 14:24:23 -04004711 /*
4712 * Initialize rx_hashtbl_used_head to RLB_NULL_INDEX.
4713 * It is set to 0 by default which is wrong.
4714 */
4715 bond = netdev_priv(bond_dev);
4716 bond_info = &(BOND_ALB_INFO(bond));
4717 bond_info->rx_hashtbl_used_head = RLB_NULL_INDEX;
4718
Eric W. Biedermanec87fd32009-10-29 14:18:26 +00004719 dev_net_set(bond_dev, net);
Eric W. Biederman88ead972009-10-29 14:18:25 +00004720 bond_dev->rtnl_link_ops = &bond_link_ops;
4721
Mitch Williamsdfe60392005-11-09 10:36:04 -08004722 res = register_netdevice(bond_dev);
Peter Zijlstra0daa23032006-11-08 19:51:01 -08004723
Phil Oestere826eaf2011-03-14 06:22:05 +00004724 netif_carrier_off(bond_dev);
4725
Mitch Williamsdfe60392005-11-09 10:36:04 -08004726 rtnl_unlock();
Amerigo Wang9e2e61f2010-03-31 21:30:52 +00004727 if (res < 0)
4728 bond_destructor(bond_dev);
Mitch Williamsdfe60392005-11-09 10:36:04 -08004729 return res;
4730}
4731
Alexey Dobriyan2c8c1e72010-01-17 03:35:32 +00004732static int __net_init bond_net_init(struct net *net)
Eric W. Biedermanec87fd32009-10-29 14:18:26 +00004733{
Eric W. Biederman15449742009-11-29 15:46:04 +00004734 struct bond_net *bn = net_generic(net, bond_net_id);
Eric W. Biedermanec87fd32009-10-29 14:18:26 +00004735
4736 bn->net = net;
4737 INIT_LIST_HEAD(&bn->dev_list);
4738
Eric W. Biedermanec87fd32009-10-29 14:18:26 +00004739 bond_create_proc_dir(bn);
Eric W. Biederman4c224002011-10-12 21:56:25 +00004740 bond_create_sysfs(bn);
Veaceslav Falico87a7b842013-06-24 11:49:29 +02004741
Eric W. Biederman15449742009-11-29 15:46:04 +00004742 return 0;
Eric W. Biedermanec87fd32009-10-29 14:18:26 +00004743}
4744
Alexey Dobriyan2c8c1e72010-01-17 03:35:32 +00004745static void __net_exit bond_net_exit(struct net *net)
Eric W. Biedermanec87fd32009-10-29 14:18:26 +00004746{
Eric W. Biederman15449742009-11-29 15:46:04 +00004747 struct bond_net *bn = net_generic(net, bond_net_id);
nikolay@redhat.com69b02162013-04-06 00:54:38 +00004748 struct bonding *bond, *tmp_bond;
4749 LIST_HEAD(list);
Eric W. Biedermanec87fd32009-10-29 14:18:26 +00004750
Eric W. Biederman4c224002011-10-12 21:56:25 +00004751 bond_destroy_sysfs(bn);
nikolay@redhat.com69b02162013-04-06 00:54:38 +00004752
4753 /* Kill off any bonds created after unregistering bond rtnl ops */
4754 rtnl_lock();
4755 list_for_each_entry_safe(bond, tmp_bond, &bn->dev_list, bond_list)
4756 unregister_netdevice_queue(bond->dev, &list);
4757 unregister_netdevice_many(&list);
4758 rtnl_unlock();
Veaceslav Falico23fa5c22014-07-17 12:04:08 +02004759
4760 bond_destroy_proc_dir(bn);
Eric W. Biedermanec87fd32009-10-29 14:18:26 +00004761}
4762
4763static struct pernet_operations bond_net_ops = {
4764 .init = bond_net_init,
4765 .exit = bond_net_exit,
Eric W. Biederman15449742009-11-29 15:46:04 +00004766 .id = &bond_net_id,
4767 .size = sizeof(struct bond_net),
Eric W. Biedermanec87fd32009-10-29 14:18:26 +00004768};
4769
Linus Torvalds1da177e2005-04-16 15:20:36 -07004770static int __init bonding_init(void)
4771{
Linus Torvalds1da177e2005-04-16 15:20:36 -07004772 int i;
4773 int res;
4774
Amerigo Wangbd33acc2011-03-06 21:58:46 +00004775 pr_info("%s", bond_version);
Linus Torvalds1da177e2005-04-16 15:20:36 -07004776
Mitch Williamsdfe60392005-11-09 10:36:04 -08004777 res = bond_check_params(&bonding_defaults);
Stephen Hemminger3d632c32009-06-12 19:02:48 +00004778 if (res)
Mitch Williamsdfe60392005-11-09 10:36:04 -08004779 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07004780
Eric W. Biederman15449742009-11-29 15:46:04 +00004781 res = register_pernet_subsys(&bond_net_ops);
Eric W. Biedermanec87fd32009-10-29 14:18:26 +00004782 if (res)
4783 goto out;
Jay Vosburgh027ea042008-01-17 16:25:02 -08004784
Jiri Pirko0a2a78c2013-10-18 17:43:33 +02004785 res = bond_netlink_init();
Eric W. Biederman88ead972009-10-29 14:18:25 +00004786 if (res)
Eric W. Biederman66391042009-10-29 23:58:54 +00004787 goto err_link;
Eric W. Biederman88ead972009-10-29 14:18:25 +00004788
Taku Izumif073c7c2010-12-09 15:17:13 +00004789 bond_create_debugfs();
4790
Linus Torvalds1da177e2005-04-16 15:20:36 -07004791 for (i = 0; i < max_bonds; i++) {
Eric W. Biedermanec87fd32009-10-29 14:18:26 +00004792 res = bond_create(&init_net, NULL);
Mitch Williamsdfe60392005-11-09 10:36:04 -08004793 if (res)
4794 goto err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07004795 }
4796
Linus Torvalds1da177e2005-04-16 15:20:36 -07004797 register_netdevice_notifier(&bond_netdev_notifier);
Mitch Williamsdfe60392005-11-09 10:36:04 -08004798out:
Linus Torvalds1da177e2005-04-16 15:20:36 -07004799 return res;
Eric W. Biederman88ead972009-10-29 14:18:25 +00004800err:
Thomas Richterdb298682014-04-09 12:52:59 +02004801 bond_destroy_debugfs();
Jiri Pirko0a2a78c2013-10-18 17:43:33 +02004802 bond_netlink_fini();
Eric W. Biederman66391042009-10-29 23:58:54 +00004803err_link:
Eric W. Biederman15449742009-11-29 15:46:04 +00004804 unregister_pernet_subsys(&bond_net_ops);
Eric W. Biederman88ead972009-10-29 14:18:25 +00004805 goto out;
Mitch Williamsdfe60392005-11-09 10:36:04 -08004806
Linus Torvalds1da177e2005-04-16 15:20:36 -07004807}
4808
4809static void __exit bonding_exit(void)
4810{
4811 unregister_netdevice_notifier(&bond_netdev_notifier);
Linus Torvalds1da177e2005-04-16 15:20:36 -07004812
Taku Izumif073c7c2010-12-09 15:17:13 +00004813 bond_destroy_debugfs();
Pavel Emelyanovae68c392008-05-02 17:49:39 -07004814
Jiri Pirko0a2a78c2013-10-18 17:43:33 +02004815 bond_netlink_fini();
nikolay@redhat.comffcdedb2013-04-06 00:54:37 +00004816 unregister_pernet_subsys(&bond_net_ops);
Neil Hormane843fa52010-10-13 16:01:50 +00004817
4818#ifdef CONFIG_NET_POLL_CONTROLLER
Nikolay Aleksandrov547942c2014-09-15 17:19:34 +02004819 /* Make sure we don't have an imbalance on our netpoll blocking */
Neil Hormanfb4fa762010-12-06 09:05:50 +00004820 WARN_ON(atomic_read(&netpoll_block_tx));
Neil Hormane843fa52010-10-13 16:01:50 +00004821#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07004822}
4823
4824module_init(bonding_init);
4825module_exit(bonding_exit);
4826MODULE_LICENSE("GPL");
4827MODULE_VERSION(DRV_VERSION);
4828MODULE_DESCRIPTION(DRV_DESCRIPTION ", v" DRV_VERSION);
4829MODULE_AUTHOR("Thomas Davis, tadavis@lbl.gov and many others");