2 * Bond several ethernet interfaces into a Cisco, running 'Etherchannel'.
4 * Portions are (c) Copyright 1995 Simon "Guru Aleph-Null" Janes
5 * NCM: Network and Communications Management, Inc.
7 * BUT, I'm the one who modified it for ethernet, so:
8 * (c) Copyright 1999, Thomas Davis, tadavis@lbl.gov
10 * This software may be used and distributed according to the terms
11 * of the GNU Public License, incorporated herein by reference.
15 #ifndef _LINUX_BONDING_H
16 #define _LINUX_BONDING_H
18 #include <linux/timer.h>
19 #include <linux/proc_fs.h>
20 #include <linux/if_bonding.h>
21 #include <linux/cpumask.h>
22 #include <linux/in6.h>
23 #include <linux/netpoll.h>
24 #include <linux/inetdevice.h>
25 #include <linux/etherdevice.h>
26 #include <linux/reciprocal_div.h>
30 #include "bond_options.h"
32 #define DRV_VERSION "3.7.1"
33 #define DRV_RELDATE "April 27, 2011"
34 #define DRV_NAME "bonding"
35 #define DRV_DESCRIPTION "Ethernet Channel Bonding Driver"
37 #define bond_version DRV_DESCRIPTION ": v" DRV_VERSION " (" DRV_RELDATE ")\n"
39 #define BOND_MAX_ARP_TARGETS 16
41 #define BOND_DEFAULT_MIIMON 100
44 * Less bad way to call ioctl from within the kernel; this needs to be
45 * done some other way to get the call out of interrupt context.
46 * Needs "ioctl" variable to be supplied by calling context.
48 #define IOCTL(dev, arg, cmd) ({ \
50 mm_segment_t fs = get_fs(); \
52 res = ioctl(dev, arg, cmd); \
56 #define BOND_MODE(bond) ((bond)->params.mode)
58 /* slave list primitives */
59 #define bond_slave_list(bond) (&(bond)->dev->adj_list.lower)
61 #define bond_has_slaves(bond) !list_empty(bond_slave_list(bond))
63 /* IMPORTANT: bond_first/last_slave can return NULL in case of an empty list */
64 #define bond_first_slave(bond) \
65 (bond_has_slaves(bond) ? \
66 netdev_adjacent_get_private(bond_slave_list(bond)->next) : \
68 #define bond_last_slave(bond) \
69 (bond_has_slaves(bond) ? \
70 netdev_adjacent_get_private(bond_slave_list(bond)->prev) : \
73 /* Caller must have rcu_read_lock */
74 #define bond_first_slave_rcu(bond) \
75 netdev_lower_get_first_private_rcu(bond->dev)
77 #define bond_is_first_slave(bond, pos) (pos == bond_first_slave(bond))
78 #define bond_is_last_slave(bond, pos) (pos == bond_last_slave(bond))
81 * bond_for_each_slave - iterate over all slaves
82 * @bond: the bond holding this list
84 * @iter: list_head * iterator
86 * Caller must hold bond->lock
88 #define bond_for_each_slave(bond, pos, iter) \
89 netdev_for_each_lower_private((bond)->dev, pos, iter)
91 /* Caller must have rcu_read_lock */
92 #define bond_for_each_slave_rcu(bond, pos, iter) \
93 netdev_for_each_lower_private_rcu((bond)->dev, pos, iter)
95 #ifdef CONFIG_NET_POLL_CONTROLLER
96 extern atomic_t netpoll_block_tx;
98 static inline void block_netpoll_tx(void)
100 atomic_inc(&netpoll_block_tx);
103 static inline void unblock_netpoll_tx(void)
105 atomic_dec(&netpoll_block_tx);
108 static inline int is_netpoll_tx_blocked(struct net_device *dev)
110 if (unlikely(netpoll_tx_running(dev)))
111 return atomic_read(&netpoll_block_tx);
115 #define block_netpoll_tx()
116 #define unblock_netpoll_tx()
117 #define is_netpoll_tx_blocked(dev) (0)
133 unsigned int min_links;
135 char primary[IFNAMSIZ];
136 int primary_reselect;
137 __be32 arp_targets[BOND_MAX_ARP_TARGETS];
139 int all_slaves_active;
142 int packets_per_slave;
144 struct reciprocal_value reciprocal_packets_per_slave;
147 struct bond_parm_tbl {
153 struct net_device *dev; /* first - useful for panic debug */
154 struct bonding *bond; /* our master */
156 /* all three in jiffies */
157 unsigned long last_link_up;
158 unsigned long last_rx;
159 unsigned long target_last_arp_rx[BOND_MAX_ARP_TARGETS];
160 s8 link; /* one of BOND_LINK_XXXX */
162 u8 backup:1, /* indicates backup slave. Value corresponds with
163 BOND_STATE_ACTIVE and BOND_STATE_BACKUP */
164 inactive:1, /* indicates inactive slave */
165 should_notify:1; /* indicateds whether the state changed */
168 u32 link_failure_count;
171 u8 perm_hwaddr[ETH_ALEN];
172 struct ad_slave_info *ad_info;
173 struct tlb_slave_info tlb_info;
174 #ifdef CONFIG_NET_POLL_CONTROLLER
181 * Link pseudo-state only used internally by monitors
183 #define BOND_LINK_NOCHANGE -1
186 * Here are the locking policies for the two bonding locks:
188 * 1) Get bond->lock when reading/writing slave list.
189 * 2) Get bond->curr_slave_lock when reading/writing bond->curr_active_slave.
190 * (It is unnecessary when the write-lock is put with bond->lock.)
191 * 3) When we lock with bond->curr_slave_lock, we must lock with bond->lock
195 struct net_device *dev; /* first - useful for panic debug */
196 struct slave __rcu *curr_active_slave;
197 struct slave __rcu *current_arp_slave;
198 struct slave *primary_slave;
200 s32 slave_cnt; /* never change this value outside the attach/detach wrappers */
201 int (*recv_probe)(const struct sk_buff *, struct bonding *,
204 rwlock_t curr_slave_lock;
207 #ifdef CONFIG_PROC_FS
208 struct proc_dir_entry *proc_entry;
209 char proc_file_name[IFNAMSIZ];
210 #endif /* CONFIG_PROC_FS */
211 struct list_head bond_list;
213 struct ad_bond_info ad_info;
214 struct alb_bond_info alb_info;
215 struct bond_params params;
216 struct workqueue_struct *wq;
217 struct delayed_work mii_work;
218 struct delayed_work arp_work;
219 struct delayed_work alb_work;
220 struct delayed_work ad_work;
221 struct delayed_work mcast_work;
222 #ifdef CONFIG_DEBUG_FS
223 /* debugging support via debugfs */
224 struct dentry *debug_dir;
225 #endif /* CONFIG_DEBUG_FS */
228 #define bond_slave_get_rcu(dev) \
229 ((struct slave *) rcu_dereference(dev->rx_handler_data))
231 #define bond_slave_get_rtnl(dev) \
232 ((struct slave *) rtnl_dereference(dev->rx_handler_data))
234 #define bond_deref_active_protected(bond) \
235 rcu_dereference_protected(bond->curr_active_slave, \
236 lockdep_is_held(&bond->curr_slave_lock))
238 struct bond_vlan_tag {
240 unsigned short vlan_id;
244 * Returns NULL if the net_device does not belong to any of the bond's slaves
246 * Caller must hold bond lock for read
248 static inline struct slave *bond_get_slave_by_dev(struct bonding *bond,
249 struct net_device *slave_dev)
251 return netdev_lower_dev_get_private(bond->dev, slave_dev);
254 static inline struct bonding *bond_get_bond_by_slave(struct slave *slave)
259 static inline bool bond_should_override_tx_queue(struct bonding *bond)
261 return BOND_MODE(bond) == BOND_MODE_ACTIVEBACKUP ||
262 BOND_MODE(bond) == BOND_MODE_ROUNDROBIN;
265 static inline bool bond_is_lb(const struct bonding *bond)
267 return BOND_MODE(bond) == BOND_MODE_TLB ||
268 BOND_MODE(bond) == BOND_MODE_ALB;
271 static inline bool bond_is_nondyn_tlb(const struct bonding *bond)
273 return (BOND_MODE(bond) == BOND_MODE_TLB) &&
274 (bond->params.tlb_dynamic_lb == 0);
277 static inline bool bond_mode_uses_arp(int mode)
279 return mode != BOND_MODE_8023AD && mode != BOND_MODE_TLB &&
280 mode != BOND_MODE_ALB;
283 static inline bool bond_mode_uses_primary(int mode)
285 return mode == BOND_MODE_ACTIVEBACKUP || mode == BOND_MODE_TLB ||
286 mode == BOND_MODE_ALB;
289 static inline bool bond_uses_primary(struct bonding *bond)
291 return bond_mode_uses_primary(BOND_MODE(bond));
294 static inline bool bond_slave_is_up(struct slave *slave)
296 return netif_running(slave->dev) && netif_carrier_ok(slave->dev);
299 static inline void bond_set_active_slave(struct slave *slave)
303 rtmsg_ifinfo(RTM_NEWLINK, slave->dev, 0, GFP_ATOMIC);
307 static inline void bond_set_backup_slave(struct slave *slave)
309 if (!slave->backup) {
311 rtmsg_ifinfo(RTM_NEWLINK, slave->dev, 0, GFP_ATOMIC);
315 static inline void bond_set_slave_state(struct slave *slave,
316 int slave_state, bool notify)
318 if (slave->backup == slave_state)
321 slave->backup = slave_state;
323 rtmsg_ifinfo(RTM_NEWLINK, slave->dev, 0, GFP_ATOMIC);
324 slave->should_notify = 0;
326 if (slave->should_notify)
327 slave->should_notify = 0;
329 slave->should_notify = 1;
333 static inline void bond_slave_state_change(struct bonding *bond)
335 struct list_head *iter;
338 bond_for_each_slave(bond, tmp, iter) {
339 if (tmp->link == BOND_LINK_UP)
340 bond_set_active_slave(tmp);
341 else if (tmp->link == BOND_LINK_DOWN)
342 bond_set_backup_slave(tmp);
346 static inline void bond_slave_state_notify(struct bonding *bond)
348 struct list_head *iter;
351 bond_for_each_slave(bond, tmp, iter) {
352 if (tmp->should_notify) {
353 rtmsg_ifinfo(RTM_NEWLINK, tmp->dev, 0, GFP_ATOMIC);
354 tmp->should_notify = 0;
359 static inline int bond_slave_state(struct slave *slave)
361 return slave->backup;
364 static inline bool bond_is_active_slave(struct slave *slave)
366 return !bond_slave_state(slave);
369 static inline bool bond_slave_can_tx(struct slave *slave)
371 return bond_slave_is_up(slave) && slave->link == BOND_LINK_UP &&
372 bond_is_active_slave(slave);
375 #define BOND_PRI_RESELECT_ALWAYS 0
376 #define BOND_PRI_RESELECT_BETTER 1
377 #define BOND_PRI_RESELECT_FAILURE 2
379 #define BOND_FOM_NONE 0
380 #define BOND_FOM_ACTIVE 1
381 #define BOND_FOM_FOLLOW 2
383 #define BOND_ARP_TARGETS_ANY 0
384 #define BOND_ARP_TARGETS_ALL 1
386 #define BOND_ARP_VALIDATE_NONE 0
387 #define BOND_ARP_VALIDATE_ACTIVE (1 << BOND_STATE_ACTIVE)
388 #define BOND_ARP_VALIDATE_BACKUP (1 << BOND_STATE_BACKUP)
389 #define BOND_ARP_VALIDATE_ALL (BOND_ARP_VALIDATE_ACTIVE | \
390 BOND_ARP_VALIDATE_BACKUP)
391 #define BOND_ARP_FILTER (BOND_ARP_VALIDATE_ALL + 1)
392 #define BOND_ARP_FILTER_ACTIVE (BOND_ARP_VALIDATE_ACTIVE | \
394 #define BOND_ARP_FILTER_BACKUP (BOND_ARP_VALIDATE_BACKUP | \
397 #define BOND_SLAVE_NOTIFY_NOW true
398 #define BOND_SLAVE_NOTIFY_LATER false
400 static inline int slave_do_arp_validate(struct bonding *bond,
403 return bond->params.arp_validate & (1 << bond_slave_state(slave));
406 static inline int slave_do_arp_validate_only(struct bonding *bond)
408 return bond->params.arp_validate & BOND_ARP_FILTER;
411 static inline int bond_is_ip_target_ok(__be32 addr)
413 return !ipv4_is_lbcast(addr) && !ipv4_is_zeronet(addr);
416 /* Get the oldest arp which we've received on this slave for bond's
419 static inline unsigned long slave_oldest_target_arp_rx(struct bonding *bond,
423 unsigned long ret = slave->target_last_arp_rx[0];
425 for (; (i < BOND_MAX_ARP_TARGETS) && bond->params.arp_targets[i]; i++)
426 if (time_before(slave->target_last_arp_rx[i], ret))
427 ret = slave->target_last_arp_rx[i];
432 static inline unsigned long slave_last_rx(struct bonding *bond,
435 if (bond->params.arp_all_targets == BOND_ARP_TARGETS_ALL)
436 return slave_oldest_target_arp_rx(bond, slave);
438 return slave->last_rx;
441 #ifdef CONFIG_NET_POLL_CONTROLLER
442 static inline void bond_netpoll_send_skb(const struct slave *slave,
445 struct netpoll *np = slave->np;
448 netpoll_send_skb(np, skb);
451 static inline void bond_netpoll_send_skb(const struct slave *slave,
457 static inline void bond_set_slave_inactive_flags(struct slave *slave,
460 if (!bond_is_lb(slave->bond))
461 bond_set_slave_state(slave, BOND_STATE_BACKUP, notify);
462 if (!slave->bond->params.all_slaves_active)
466 static inline void bond_set_slave_active_flags(struct slave *slave,
469 bond_set_slave_state(slave, BOND_STATE_ACTIVE, notify);
473 static inline bool bond_is_slave_inactive(struct slave *slave)
475 return slave->inactive;
478 static inline __be32 bond_confirm_addr(struct net_device *dev, __be32 dst, __be32 local)
480 struct in_device *in_dev;
484 in_dev = __in_dev_get_rcu(dev);
487 addr = inet_confirm_addr(dev_net(dev), in_dev, dst, local,
494 struct net *net; /* Associated network namespace */
495 struct list_head dev_list;
496 #ifdef CONFIG_PROC_FS
497 struct proc_dir_entry *proc_dir;
499 struct class_attribute class_attr_bonding_masters;
502 int bond_arp_rcv(const struct sk_buff *skb, struct bonding *bond, struct slave *slave);
503 void bond_dev_queue_xmit(struct bonding *bond, struct sk_buff *skb, struct net_device *slave_dev);
504 int bond_create(struct net *net, const char *name);
505 int bond_create_sysfs(struct bond_net *net);
506 void bond_destroy_sysfs(struct bond_net *net);
507 void bond_prepare_sysfs_group(struct bonding *bond);
508 int bond_sysfs_slave_add(struct slave *slave);
509 void bond_sysfs_slave_del(struct slave *slave);
510 int bond_enslave(struct net_device *bond_dev, struct net_device *slave_dev);
511 int bond_release(struct net_device *bond_dev, struct net_device *slave_dev);
512 u32 bond_xmit_hash(struct bonding *bond, struct sk_buff *skb);
513 void bond_select_active_slave(struct bonding *bond);
514 void bond_change_active_slave(struct bonding *bond, struct slave *new_active);
515 void bond_create_debugfs(void);
516 void bond_destroy_debugfs(void);
517 void bond_debug_register(struct bonding *bond);
518 void bond_debug_unregister(struct bonding *bond);
519 void bond_debug_reregister(struct bonding *bond);
520 const char *bond_mode_name(int mode);
521 void bond_setup(struct net_device *bond_dev);
522 unsigned int bond_get_num_tx_queues(void);
523 int bond_netlink_init(void);
524 void bond_netlink_fini(void);
525 struct net_device *bond_option_active_slave_get_rcu(struct bonding *bond);
526 const char *bond_slave_link_status(s8 link);
527 struct bond_vlan_tag *bond_verify_device_path(struct net_device *start_dev,
528 struct net_device *end_dev,
531 #ifdef CONFIG_PROC_FS
532 void bond_create_proc_entry(struct bonding *bond);
533 void bond_remove_proc_entry(struct bonding *bond);
534 void bond_create_proc_dir(struct bond_net *bn);
535 void bond_destroy_proc_dir(struct bond_net *bn);
537 static inline void bond_create_proc_entry(struct bonding *bond)
541 static inline void bond_remove_proc_entry(struct bonding *bond)
545 static inline void bond_create_proc_dir(struct bond_net *bn)
549 static inline void bond_destroy_proc_dir(struct bond_net *bn)
554 static inline struct slave *bond_slave_has_mac(struct bonding *bond,
557 struct list_head *iter;
560 bond_for_each_slave(bond, tmp, iter)
561 if (ether_addr_equal_64bits(mac, tmp->dev->dev_addr))
567 /* Caller must hold rcu_read_lock() for read */
568 static inline struct slave *bond_slave_has_mac_rcu(struct bonding *bond,
571 struct list_head *iter;
574 bond_for_each_slave_rcu(bond, tmp, iter)
575 if (ether_addr_equal_64bits(mac, tmp->dev->dev_addr))
581 /* Caller must hold rcu_read_lock() for read */
582 static inline bool bond_slave_has_mac_rx(struct bonding *bond, const u8 *mac)
584 struct list_head *iter;
586 struct netdev_hw_addr *ha;
588 bond_for_each_slave_rcu(bond, tmp, iter)
589 if (ether_addr_equal_64bits(mac, tmp->dev->dev_addr))
592 if (netdev_uc_empty(bond->dev))
595 netdev_for_each_uc_addr(ha, bond->dev)
596 if (ether_addr_equal_64bits(mac, ha->addr))
602 /* Check if the ip is present in arp ip list, or first free slot if ip == 0
603 * Returns -1 if not found, index if found
605 static inline int bond_get_targets_ip(__be32 *targets, __be32 ip)
609 for (i = 0; i < BOND_MAX_ARP_TARGETS; i++)
610 if (targets[i] == ip)
612 else if (targets[i] == 0)
618 /* exported from bond_main.c */
619 extern int bond_net_id;
620 extern const struct bond_parm_tbl bond_lacp_tbl[];
621 extern const struct bond_parm_tbl xmit_hashtype_tbl[];
622 extern const struct bond_parm_tbl arp_validate_tbl[];
623 extern const struct bond_parm_tbl arp_all_targets_tbl[];
624 extern const struct bond_parm_tbl fail_over_mac_tbl[];
625 extern const struct bond_parm_tbl pri_reselect_tbl[];
626 extern struct bond_parm_tbl ad_select_tbl[];
628 /* exported from bond_netlink.c */
629 extern struct rtnl_link_ops bond_link_ops;
631 #endif /* _LINUX_BONDING_H */