]> git.proxmox.com Git - mirror_ubuntu-bionic-kernel.git/blob - drivers/net/bonding/bonding.h
bonding: convert first/last slave logic to use neighbours
[mirror_ubuntu-bionic-kernel.git] / drivers / net / bonding / bonding.h
1 /*
2 * Bond several ethernet interfaces into a Cisco, running 'Etherchannel'.
3 *
4 * Portions are (c) Copyright 1995 Simon "Guru Aleph-Null" Janes
5 * NCM: Network and Communications Management, Inc.
6 *
7 * BUT, I'm the one who modified it for ethernet, so:
8 * (c) Copyright 1999, Thomas Davis, tadavis@lbl.gov
9 *
10 * This software may be used and distributed according to the terms
11 * of the GNU Public License, incorporated herein by reference.
12 *
13 */
14
15 #ifndef _LINUX_BONDING_H
16 #define _LINUX_BONDING_H
17
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 "bond_3ad.h"
27 #include "bond_alb.h"
28
29 #define DRV_VERSION "3.7.1"
30 #define DRV_RELDATE "April 27, 2011"
31 #define DRV_NAME "bonding"
32 #define DRV_DESCRIPTION "Ethernet Channel Bonding Driver"
33
34 #define bond_version DRV_DESCRIPTION ": v" DRV_VERSION " (" DRV_RELDATE ")\n"
35
36 #define BOND_MAX_ARP_TARGETS 16
37
38 #define IS_UP(dev) \
39 ((((dev)->flags & IFF_UP) == IFF_UP) && \
40 netif_running(dev) && \
41 netif_carrier_ok(dev))
42
43 /*
44 * Checks whether slave is ready for transmit.
45 */
46 #define SLAVE_IS_OK(slave) \
47 (((slave)->dev->flags & IFF_UP) && \
48 netif_running((slave)->dev) && \
49 ((slave)->link == BOND_LINK_UP) && \
50 bond_is_active_slave(slave))
51
52
53 #define USES_PRIMARY(mode) \
54 (((mode) == BOND_MODE_ACTIVEBACKUP) || \
55 ((mode) == BOND_MODE_TLB) || \
56 ((mode) == BOND_MODE_ALB))
57
58 #define TX_QUEUE_OVERRIDE(mode) \
59 (((mode) == BOND_MODE_ACTIVEBACKUP) || \
60 ((mode) == BOND_MODE_ROUNDROBIN))
61 /*
62 * Less bad way to call ioctl from within the kernel; this needs to be
63 * done some other way to get the call out of interrupt context.
64 * Needs "ioctl" variable to be supplied by calling context.
65 */
66 #define IOCTL(dev, arg, cmd) ({ \
67 int res = 0; \
68 mm_segment_t fs = get_fs(); \
69 set_fs(get_ds()); \
70 res = ioctl(dev, arg, cmd); \
71 set_fs(fs); \
72 res; })
73
74 /* slave list primitives */
75 #define bond_slave_list(bond) (&(bond)->dev->adj_list.lower)
76
77 #define bond_has_slaves(bond) !list_empty(bond_slave_list(bond))
78
79 #define bond_to_slave(ptr) list_entry(ptr, struct slave, list)
80
81 /* IMPORTANT: bond_first/last_slave can return NULL in case of an empty list */
82 #define bond_first_slave(bond) \
83 (bond_has_slaves(bond) ? \
84 netdev_adjacent_get_private(bond_slave_list(bond)->next) : \
85 NULL)
86 #define bond_last_slave(bond) \
87 (bond_has_slaves(bond) ? \
88 netdev_adjacent_get_private(bond_slave_list(bond)->prev) : \
89 NULL)
90
91 #define bond_is_first_slave(bond, pos) (pos == bond_first_slave(bond))
92 #define bond_is_last_slave(bond, pos) (pos == bond_last_slave(bond))
93
94 /* Since bond_first/last_slave can return NULL, these can return NULL too */
95 #define bond_next_slave(bond, pos) \
96 (bond_is_last_slave(bond, pos) ? bond_first_slave(bond) : \
97 bond_to_slave((pos)->list.next))
98
99 #define bond_prev_slave(bond, pos) \
100 (bond_is_first_slave(bond, pos) ? bond_last_slave(bond) : \
101 bond_to_slave((pos)->list.prev))
102
103 /**
104 * bond_for_each_slave - iterate over all slaves
105 * @bond: the bond holding this list
106 * @pos: current slave
107 * @iter: list_head * iterator
108 *
109 * Caller must hold bond->lock
110 */
111 #define bond_for_each_slave(bond, pos, iter) \
112 netdev_for_each_lower_private((bond)->dev, pos, iter)
113
114 /* Caller must have rcu_read_lock */
115 #define bond_for_each_slave_rcu(bond, pos, iter) \
116 netdev_for_each_lower_private_rcu((bond)->dev, pos, iter)
117
118 #ifdef CONFIG_NET_POLL_CONTROLLER
119 extern atomic_t netpoll_block_tx;
120
121 static inline void block_netpoll_tx(void)
122 {
123 atomic_inc(&netpoll_block_tx);
124 }
125
126 static inline void unblock_netpoll_tx(void)
127 {
128 atomic_dec(&netpoll_block_tx);
129 }
130
131 static inline int is_netpoll_tx_blocked(struct net_device *dev)
132 {
133 if (unlikely(netpoll_tx_running(dev)))
134 return atomic_read(&netpoll_block_tx);
135 return 0;
136 }
137 #else
138 #define block_netpoll_tx()
139 #define unblock_netpoll_tx()
140 #define is_netpoll_tx_blocked(dev) (0)
141 #endif
142
143 struct bond_params {
144 int mode;
145 int xmit_policy;
146 int miimon;
147 u8 num_peer_notif;
148 int arp_interval;
149 int arp_validate;
150 int arp_all_targets;
151 int use_carrier;
152 int fail_over_mac;
153 int updelay;
154 int downdelay;
155 int lacp_fast;
156 unsigned int min_links;
157 int ad_select;
158 char primary[IFNAMSIZ];
159 int primary_reselect;
160 __be32 arp_targets[BOND_MAX_ARP_TARGETS];
161 int tx_queues;
162 int all_slaves_active;
163 int resend_igmp;
164 int lp_interval;
165 };
166
167 struct bond_parm_tbl {
168 char *modename;
169 int mode;
170 };
171
172 #define BOND_MAX_MODENAME_LEN 20
173
174 struct slave {
175 struct net_device *dev; /* first - useful for panic debug */
176 struct list_head list;
177 struct bonding *bond; /* our master */
178 int delay;
179 unsigned long jiffies;
180 unsigned long last_arp_rx;
181 unsigned long target_last_arp_rx[BOND_MAX_ARP_TARGETS];
182 s8 link; /* one of BOND_LINK_XXXX */
183 s8 new_link;
184 u8 backup:1, /* indicates backup slave. Value corresponds with
185 BOND_STATE_ACTIVE and BOND_STATE_BACKUP */
186 inactive:1; /* indicates inactive slave */
187 u8 duplex;
188 u32 original_mtu;
189 u32 link_failure_count;
190 u32 speed;
191 u16 queue_id;
192 u8 perm_hwaddr[ETH_ALEN];
193 struct ad_slave_info ad_info; /* HUGE - better to dynamically alloc */
194 struct tlb_slave_info tlb_info;
195 #ifdef CONFIG_NET_POLL_CONTROLLER
196 struct netpoll *np;
197 #endif
198 };
199
200 /*
201 * Link pseudo-state only used internally by monitors
202 */
203 #define BOND_LINK_NOCHANGE -1
204
205 /*
206 * Here are the locking policies for the two bonding locks:
207 *
208 * 1) Get bond->lock when reading/writing slave list.
209 * 2) Get bond->curr_slave_lock when reading/writing bond->curr_active_slave.
210 * (It is unnecessary when the write-lock is put with bond->lock.)
211 * 3) When we lock with bond->curr_slave_lock, we must lock with bond->lock
212 * beforehand.
213 */
214 struct bonding {
215 struct net_device *dev; /* first - useful for panic debug */
216 struct list_head slave_list;
217 struct slave *curr_active_slave;
218 struct slave *current_arp_slave;
219 struct slave *primary_slave;
220 bool force_primary;
221 s32 slave_cnt; /* never change this value outside the attach/detach wrappers */
222 int (*recv_probe)(const struct sk_buff *, struct bonding *,
223 struct slave *);
224 rwlock_t lock;
225 rwlock_t curr_slave_lock;
226 u8 send_peer_notif;
227 u8 igmp_retrans;
228 #ifdef CONFIG_PROC_FS
229 struct proc_dir_entry *proc_entry;
230 char proc_file_name[IFNAMSIZ];
231 #endif /* CONFIG_PROC_FS */
232 struct list_head bond_list;
233 int (*xmit_hash_policy)(struct sk_buff *, int);
234 u16 rr_tx_counter;
235 struct ad_bond_info ad_info;
236 struct alb_bond_info alb_info;
237 struct bond_params params;
238 struct workqueue_struct *wq;
239 struct delayed_work mii_work;
240 struct delayed_work arp_work;
241 struct delayed_work alb_work;
242 struct delayed_work ad_work;
243 struct delayed_work mcast_work;
244 #ifdef CONFIG_DEBUG_FS
245 /* debugging support via debugfs */
246 struct dentry *debug_dir;
247 #endif /* CONFIG_DEBUG_FS */
248 };
249
250 #define bond_slave_get_rcu(dev) \
251 ((struct slave *) rcu_dereference(dev->rx_handler_data))
252
253 #define bond_slave_get_rtnl(dev) \
254 ((struct slave *) rtnl_dereference(dev->rx_handler_data))
255
256 /**
257 * Returns NULL if the net_device does not belong to any of the bond's slaves
258 *
259 * Caller must hold bond lock for read
260 */
261 static inline struct slave *bond_get_slave_by_dev(struct bonding *bond,
262 struct net_device *slave_dev)
263 {
264 return netdev_lower_dev_get_private(bond->dev, slave_dev);
265 }
266
267 static inline struct bonding *bond_get_bond_by_slave(struct slave *slave)
268 {
269 if (!slave || !slave->bond)
270 return NULL;
271 return slave->bond;
272 }
273
274 static inline bool bond_is_lb(const struct bonding *bond)
275 {
276 return (bond->params.mode == BOND_MODE_TLB ||
277 bond->params.mode == BOND_MODE_ALB);
278 }
279
280 static inline void bond_set_active_slave(struct slave *slave)
281 {
282 slave->backup = 0;
283 }
284
285 static inline void bond_set_backup_slave(struct slave *slave)
286 {
287 slave->backup = 1;
288 }
289
290 static inline int bond_slave_state(struct slave *slave)
291 {
292 return slave->backup;
293 }
294
295 static inline bool bond_is_active_slave(struct slave *slave)
296 {
297 return !bond_slave_state(slave);
298 }
299
300 #define BOND_PRI_RESELECT_ALWAYS 0
301 #define BOND_PRI_RESELECT_BETTER 1
302 #define BOND_PRI_RESELECT_FAILURE 2
303
304 #define BOND_FOM_NONE 0
305 #define BOND_FOM_ACTIVE 1
306 #define BOND_FOM_FOLLOW 2
307
308 #define BOND_ARP_TARGETS_ANY 0
309 #define BOND_ARP_TARGETS_ALL 1
310
311 #define BOND_ARP_VALIDATE_NONE 0
312 #define BOND_ARP_VALIDATE_ACTIVE (1 << BOND_STATE_ACTIVE)
313 #define BOND_ARP_VALIDATE_BACKUP (1 << BOND_STATE_BACKUP)
314 #define BOND_ARP_VALIDATE_ALL (BOND_ARP_VALIDATE_ACTIVE | \
315 BOND_ARP_VALIDATE_BACKUP)
316
317 static inline int slave_do_arp_validate(struct bonding *bond,
318 struct slave *slave)
319 {
320 return bond->params.arp_validate & (1 << bond_slave_state(slave));
321 }
322
323 /* Get the oldest arp which we've received on this slave for bond's
324 * arp_targets.
325 */
326 static inline unsigned long slave_oldest_target_arp_rx(struct bonding *bond,
327 struct slave *slave)
328 {
329 int i = 1;
330 unsigned long ret = slave->target_last_arp_rx[0];
331
332 for (; (i < BOND_MAX_ARP_TARGETS) && bond->params.arp_targets[i]; i++)
333 if (time_before(slave->target_last_arp_rx[i], ret))
334 ret = slave->target_last_arp_rx[i];
335
336 return ret;
337 }
338
339 static inline unsigned long slave_last_rx(struct bonding *bond,
340 struct slave *slave)
341 {
342 if (slave_do_arp_validate(bond, slave)) {
343 if (bond->params.arp_all_targets == BOND_ARP_TARGETS_ALL)
344 return slave_oldest_target_arp_rx(bond, slave);
345 else
346 return slave->last_arp_rx;
347 }
348
349 return slave->dev->last_rx;
350 }
351
352 #ifdef CONFIG_NET_POLL_CONTROLLER
353 static inline void bond_netpoll_send_skb(const struct slave *slave,
354 struct sk_buff *skb)
355 {
356 struct netpoll *np = slave->np;
357
358 if (np)
359 netpoll_send_skb(np, skb);
360 }
361 #else
362 static inline void bond_netpoll_send_skb(const struct slave *slave,
363 struct sk_buff *skb)
364 {
365 }
366 #endif
367
368 static inline void bond_set_slave_inactive_flags(struct slave *slave)
369 {
370 if (!bond_is_lb(slave->bond))
371 bond_set_backup_slave(slave);
372 if (!slave->bond->params.all_slaves_active)
373 slave->inactive = 1;
374 }
375
376 static inline void bond_set_slave_active_flags(struct slave *slave)
377 {
378 bond_set_active_slave(slave);
379 slave->inactive = 0;
380 }
381
382 static inline bool bond_is_slave_inactive(struct slave *slave)
383 {
384 return slave->inactive;
385 }
386
387 static inline __be32 bond_confirm_addr(struct net_device *dev, __be32 dst, __be32 local)
388 {
389 struct in_device *in_dev;
390 __be32 addr = 0;
391
392 rcu_read_lock();
393 in_dev = __in_dev_get_rcu(dev);
394
395 if (in_dev)
396 addr = inet_confirm_addr(in_dev, dst, local, RT_SCOPE_HOST);
397
398 rcu_read_unlock();
399 return addr;
400 }
401
402 static inline bool slave_can_tx(struct slave *slave)
403 {
404 if (IS_UP(slave->dev) && slave->link == BOND_LINK_UP &&
405 bond_is_active_slave(slave))
406 return true;
407 else
408 return false;
409 }
410
411 struct bond_net;
412
413 int bond_arp_rcv(const struct sk_buff *skb, struct bonding *bond, struct slave *slave);
414 struct vlan_entry *bond_next_vlan(struct bonding *bond, struct vlan_entry *curr);
415 int bond_dev_queue_xmit(struct bonding *bond, struct sk_buff *skb, struct net_device *slave_dev);
416 void bond_xmit_slave_id(struct bonding *bond, struct sk_buff *skb, int slave_id);
417 int bond_create(struct net *net, const char *name);
418 int bond_create_sysfs(struct bond_net *net);
419 void bond_destroy_sysfs(struct bond_net *net);
420 void bond_prepare_sysfs_group(struct bonding *bond);
421 int bond_create_slave_symlinks(struct net_device *master, struct net_device *slave);
422 void bond_destroy_slave_symlinks(struct net_device *master, struct net_device *slave);
423 int bond_enslave(struct net_device *bond_dev, struct net_device *slave_dev);
424 int bond_release(struct net_device *bond_dev, struct net_device *slave_dev);
425 void bond_mii_monitor(struct work_struct *);
426 void bond_loadbalance_arp_mon(struct work_struct *);
427 void bond_activebackup_arp_mon(struct work_struct *);
428 void bond_set_mode_ops(struct bonding *bond, int mode);
429 int bond_parse_parm(const char *mode_arg, const struct bond_parm_tbl *tbl);
430 void bond_select_active_slave(struct bonding *bond);
431 void bond_change_active_slave(struct bonding *bond, struct slave *new_active);
432 void bond_create_debugfs(void);
433 void bond_destroy_debugfs(void);
434 void bond_debug_register(struct bonding *bond);
435 void bond_debug_unregister(struct bonding *bond);
436 void bond_debug_reregister(struct bonding *bond);
437 const char *bond_mode_name(int mode);
438
439 struct bond_net {
440 struct net * net; /* Associated network namespace */
441 struct list_head dev_list;
442 #ifdef CONFIG_PROC_FS
443 struct proc_dir_entry * proc_dir;
444 #endif
445 struct class_attribute class_attr_bonding_masters;
446 };
447
448 #ifdef CONFIG_PROC_FS
449 void bond_create_proc_entry(struct bonding *bond);
450 void bond_remove_proc_entry(struct bonding *bond);
451 void bond_create_proc_dir(struct bond_net *bn);
452 void bond_destroy_proc_dir(struct bond_net *bn);
453 #else
454 static inline void bond_create_proc_entry(struct bonding *bond)
455 {
456 }
457
458 static inline void bond_remove_proc_entry(struct bonding *bond)
459 {
460 }
461
462 static inline void bond_create_proc_dir(struct bond_net *bn)
463 {
464 }
465
466 static inline void bond_destroy_proc_dir(struct bond_net *bn)
467 {
468 }
469 #endif
470
471 static inline struct slave *bond_slave_has_mac(struct bonding *bond,
472 const u8 *mac)
473 {
474 struct list_head *iter;
475 struct slave *tmp;
476
477 bond_for_each_slave(bond, tmp, iter)
478 if (ether_addr_equal_64bits(mac, tmp->dev->dev_addr))
479 return tmp;
480
481 return NULL;
482 }
483
484 /* Check if the ip is present in arp ip list, or first free slot if ip == 0
485 * Returns -1 if not found, index if found
486 */
487 static inline int bond_get_targets_ip(__be32 *targets, __be32 ip)
488 {
489 int i;
490
491 for (i = 0; i < BOND_MAX_ARP_TARGETS; i++)
492 if (targets[i] == ip)
493 return i;
494 else if (targets[i] == 0)
495 break;
496
497 return -1;
498 }
499
500 /* exported from bond_main.c */
501 extern int bond_net_id;
502 extern const struct bond_parm_tbl bond_lacp_tbl[];
503 extern const struct bond_parm_tbl bond_mode_tbl[];
504 extern const struct bond_parm_tbl xmit_hashtype_tbl[];
505 extern const struct bond_parm_tbl arp_validate_tbl[];
506 extern const struct bond_parm_tbl arp_all_targets_tbl[];
507 extern const struct bond_parm_tbl fail_over_mac_tbl[];
508 extern const struct bond_parm_tbl pri_reselect_tbl[];
509 extern struct bond_parm_tbl ad_select_tbl[];
510
511 #endif /* _LINUX_BONDING_H */