bonding.h 14 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292293294295296297298299300301302303304305306307308309310311312313314315316317318319320321322323324325326327328329330331332333334335336337338339340341342343344345346347348349350351352353354355356357358359360361362363364365366367368369370371372373374375376377378379380381382383384385386387388389390391392393394395396397398399400401402403404405406407408409410411412413414415416417418419420421422423424425426427428429430431432433434435436437438439440441442443444445446447448449450451452453454455456457458459460461462463464465466467468469470471472473474475476477478479480481482483484485486487488489490491492493494495496497498499500501502503504505506507508509510511512513514515516517518519520521522523524525526527
  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. #ifndef _LINUX_BONDING_H
  15. #define _LINUX_BONDING_H
  16. #include <linux/timer.h>
  17. #include <linux/proc_fs.h>
  18. #include <linux/if_bonding.h>
  19. #include <linux/cpumask.h>
  20. #include <linux/in6.h>
  21. #include <linux/netpoll.h>
  22. #include <linux/inetdevice.h>
  23. #include <linux/etherdevice.h>
  24. #include "bond_3ad.h"
  25. #include "bond_alb.h"
  26. #define DRV_VERSION "3.7.1"
  27. #define DRV_RELDATE "April 27, 2011"
  28. #define DRV_NAME "bonding"
  29. #define DRV_DESCRIPTION "Ethernet Channel Bonding Driver"
  30. #define bond_version DRV_DESCRIPTION ": v" DRV_VERSION " (" DRV_RELDATE ")\n"
  31. #define BOND_MAX_ARP_TARGETS 16
  32. #define IS_UP(dev) \
  33. ((((dev)->flags & IFF_UP) == IFF_UP) && \
  34. netif_running(dev) && \
  35. netif_carrier_ok(dev))
  36. /*
  37. * Checks whether slave is ready for transmit.
  38. */
  39. #define SLAVE_IS_OK(slave) \
  40. (((slave)->dev->flags & IFF_UP) && \
  41. netif_running((slave)->dev) && \
  42. ((slave)->link == BOND_LINK_UP) && \
  43. bond_is_active_slave(slave))
  44. #define USES_PRIMARY(mode) \
  45. (((mode) == BOND_MODE_ACTIVEBACKUP) || \
  46. ((mode) == BOND_MODE_TLB) || \
  47. ((mode) == BOND_MODE_ALB))
  48. #define TX_QUEUE_OVERRIDE(mode) \
  49. (((mode) == BOND_MODE_ACTIVEBACKUP) || \
  50. ((mode) == BOND_MODE_ROUNDROBIN))
  51. #define BOND_MODE_IS_LB(mode) \
  52. (((mode) == BOND_MODE_TLB) || \
  53. ((mode) == BOND_MODE_ALB))
  54. #define IS_IP_TARGET_UNUSABLE_ADDRESS(a) \
  55. ((htonl(INADDR_BROADCAST) == a) || \
  56. ipv4_is_zeronet(a))
  57. /*
  58. * Less bad way to call ioctl from within the kernel; this needs to be
  59. * done some other way to get the call out of interrupt context.
  60. * Needs "ioctl" variable to be supplied by calling context.
  61. */
  62. #define IOCTL(dev, arg, cmd) ({ \
  63. int res = 0; \
  64. mm_segment_t fs = get_fs(); \
  65. set_fs(get_ds()); \
  66. res = ioctl(dev, arg, cmd); \
  67. set_fs(fs); \
  68. res; })
  69. /* slave list primitives */
  70. #define bond_slave_list(bond) (&(bond)->dev->adj_list.lower)
  71. #define bond_has_slaves(bond) !list_empty(bond_slave_list(bond))
  72. /* IMPORTANT: bond_first/last_slave can return NULL in case of an empty list */
  73. #define bond_first_slave(bond) \
  74. (bond_has_slaves(bond) ? \
  75. netdev_adjacent_get_private(bond_slave_list(bond)->next) : \
  76. NULL)
  77. #define bond_last_slave(bond) \
  78. (bond_has_slaves(bond) ? \
  79. netdev_adjacent_get_private(bond_slave_list(bond)->prev) : \
  80. NULL)
  81. #define bond_is_first_slave(bond, pos) (pos == bond_first_slave(bond))
  82. #define bond_is_last_slave(bond, pos) (pos == bond_last_slave(bond))
  83. /**
  84. * bond_for_each_slave - iterate over all slaves
  85. * @bond: the bond holding this list
  86. * @pos: current slave
  87. * @iter: list_head * iterator
  88. *
  89. * Caller must hold bond->lock
  90. */
  91. #define bond_for_each_slave(bond, pos, iter) \
  92. netdev_for_each_lower_private((bond)->dev, pos, iter)
  93. /* Caller must have rcu_read_lock */
  94. #define bond_for_each_slave_rcu(bond, pos, iter) \
  95. netdev_for_each_lower_private_rcu((bond)->dev, pos, iter)
  96. #ifdef CONFIG_NET_POLL_CONTROLLER
  97. extern atomic_t netpoll_block_tx;
  98. static inline void block_netpoll_tx(void)
  99. {
  100. atomic_inc(&netpoll_block_tx);
  101. }
  102. static inline void unblock_netpoll_tx(void)
  103. {
  104. atomic_dec(&netpoll_block_tx);
  105. }
  106. static inline int is_netpoll_tx_blocked(struct net_device *dev)
  107. {
  108. if (unlikely(netpoll_tx_running(dev)))
  109. return atomic_read(&netpoll_block_tx);
  110. return 0;
  111. }
  112. #else
  113. #define block_netpoll_tx()
  114. #define unblock_netpoll_tx()
  115. #define is_netpoll_tx_blocked(dev) (0)
  116. #endif
  117. struct bond_params {
  118. int mode;
  119. int xmit_policy;
  120. int miimon;
  121. u8 num_peer_notif;
  122. int arp_interval;
  123. int arp_validate;
  124. int arp_all_targets;
  125. int use_carrier;
  126. int fail_over_mac;
  127. int updelay;
  128. int downdelay;
  129. int lacp_fast;
  130. unsigned int min_links;
  131. int ad_select;
  132. char primary[IFNAMSIZ];
  133. int primary_reselect;
  134. __be32 arp_targets[BOND_MAX_ARP_TARGETS];
  135. int tx_queues;
  136. int all_slaves_active;
  137. int resend_igmp;
  138. int lp_interval;
  139. int packets_per_slave;
  140. };
  141. struct bond_parm_tbl {
  142. char *modename;
  143. int mode;
  144. };
  145. #define BOND_MAX_MODENAME_LEN 20
  146. struct slave {
  147. struct net_device *dev; /* first - useful for panic debug */
  148. struct bonding *bond; /* our master */
  149. int delay;
  150. unsigned long jiffies;
  151. unsigned long last_arp_rx;
  152. unsigned long target_last_arp_rx[BOND_MAX_ARP_TARGETS];
  153. s8 link; /* one of BOND_LINK_XXXX */
  154. s8 new_link;
  155. u8 backup:1, /* indicates backup slave. Value corresponds with
  156. BOND_STATE_ACTIVE and BOND_STATE_BACKUP */
  157. inactive:1; /* indicates inactive slave */
  158. u8 duplex;
  159. u32 original_mtu;
  160. u32 link_failure_count;
  161. u32 speed;
  162. u16 queue_id;
  163. u8 perm_hwaddr[ETH_ALEN];
  164. struct ad_slave_info ad_info; /* HUGE - better to dynamically alloc */
  165. struct tlb_slave_info tlb_info;
  166. #ifdef CONFIG_NET_POLL_CONTROLLER
  167. struct netpoll *np;
  168. #endif
  169. };
  170. /*
  171. * Link pseudo-state only used internally by monitors
  172. */
  173. #define BOND_LINK_NOCHANGE -1
  174. /*
  175. * Here are the locking policies for the two bonding locks:
  176. *
  177. * 1) Get bond->lock when reading/writing slave list.
  178. * 2) Get bond->curr_slave_lock when reading/writing bond->curr_active_slave.
  179. * (It is unnecessary when the write-lock is put with bond->lock.)
  180. * 3) When we lock with bond->curr_slave_lock, we must lock with bond->lock
  181. * beforehand.
  182. */
  183. struct bonding {
  184. struct net_device *dev; /* first - useful for panic debug */
  185. struct slave *curr_active_slave;
  186. struct slave *current_arp_slave;
  187. struct slave *primary_slave;
  188. bool force_primary;
  189. s32 slave_cnt; /* never change this value outside the attach/detach wrappers */
  190. int (*recv_probe)(const struct sk_buff *, struct bonding *,
  191. struct slave *);
  192. rwlock_t lock;
  193. rwlock_t curr_slave_lock;
  194. u8 send_peer_notif;
  195. u8 igmp_retrans;
  196. #ifdef CONFIG_PROC_FS
  197. struct proc_dir_entry *proc_entry;
  198. char proc_file_name[IFNAMSIZ];
  199. #endif /* CONFIG_PROC_FS */
  200. struct list_head bond_list;
  201. u32 rr_tx_counter;
  202. struct ad_bond_info ad_info;
  203. struct alb_bond_info alb_info;
  204. struct bond_params params;
  205. struct workqueue_struct *wq;
  206. struct delayed_work mii_work;
  207. struct delayed_work arp_work;
  208. struct delayed_work alb_work;
  209. struct delayed_work ad_work;
  210. struct delayed_work mcast_work;
  211. #ifdef CONFIG_DEBUG_FS
  212. /* debugging support via debugfs */
  213. struct dentry *debug_dir;
  214. #endif /* CONFIG_DEBUG_FS */
  215. };
  216. #define bond_slave_get_rcu(dev) \
  217. ((struct slave *) rcu_dereference(dev->rx_handler_data))
  218. #define bond_slave_get_rtnl(dev) \
  219. ((struct slave *) rtnl_dereference(dev->rx_handler_data))
  220. /**
  221. * Returns NULL if the net_device does not belong to any of the bond's slaves
  222. *
  223. * Caller must hold bond lock for read
  224. */
  225. static inline struct slave *bond_get_slave_by_dev(struct bonding *bond,
  226. struct net_device *slave_dev)
  227. {
  228. return netdev_lower_dev_get_private(bond->dev, slave_dev);
  229. }
  230. static inline struct bonding *bond_get_bond_by_slave(struct slave *slave)
  231. {
  232. if (!slave || !slave->bond)
  233. return NULL;
  234. return slave->bond;
  235. }
  236. static inline bool bond_is_lb(const struct bonding *bond)
  237. {
  238. return BOND_MODE_IS_LB(bond->params.mode);
  239. }
  240. static inline void bond_set_active_slave(struct slave *slave)
  241. {
  242. slave->backup = 0;
  243. }
  244. static inline void bond_set_backup_slave(struct slave *slave)
  245. {
  246. slave->backup = 1;
  247. }
  248. static inline int bond_slave_state(struct slave *slave)
  249. {
  250. return slave->backup;
  251. }
  252. static inline bool bond_is_active_slave(struct slave *slave)
  253. {
  254. return !bond_slave_state(slave);
  255. }
  256. #define BOND_PRI_RESELECT_ALWAYS 0
  257. #define BOND_PRI_RESELECT_BETTER 1
  258. #define BOND_PRI_RESELECT_FAILURE 2
  259. #define BOND_FOM_NONE 0
  260. #define BOND_FOM_ACTIVE 1
  261. #define BOND_FOM_FOLLOW 2
  262. #define BOND_ARP_TARGETS_ANY 0
  263. #define BOND_ARP_TARGETS_ALL 1
  264. #define BOND_ARP_VALIDATE_NONE 0
  265. #define BOND_ARP_VALIDATE_ACTIVE (1 << BOND_STATE_ACTIVE)
  266. #define BOND_ARP_VALIDATE_BACKUP (1 << BOND_STATE_BACKUP)
  267. #define BOND_ARP_VALIDATE_ALL (BOND_ARP_VALIDATE_ACTIVE | \
  268. BOND_ARP_VALIDATE_BACKUP)
  269. static inline int slave_do_arp_validate(struct bonding *bond,
  270. struct slave *slave)
  271. {
  272. return bond->params.arp_validate & (1 << bond_slave_state(slave));
  273. }
  274. /* Get the oldest arp which we've received on this slave for bond's
  275. * arp_targets.
  276. */
  277. static inline unsigned long slave_oldest_target_arp_rx(struct bonding *bond,
  278. struct slave *slave)
  279. {
  280. int i = 1;
  281. unsigned long ret = slave->target_last_arp_rx[0];
  282. for (; (i < BOND_MAX_ARP_TARGETS) && bond->params.arp_targets[i]; i++)
  283. if (time_before(slave->target_last_arp_rx[i], ret))
  284. ret = slave->target_last_arp_rx[i];
  285. return ret;
  286. }
  287. static inline unsigned long slave_last_rx(struct bonding *bond,
  288. struct slave *slave)
  289. {
  290. if (slave_do_arp_validate(bond, slave)) {
  291. if (bond->params.arp_all_targets == BOND_ARP_TARGETS_ALL)
  292. return slave_oldest_target_arp_rx(bond, slave);
  293. else
  294. return slave->last_arp_rx;
  295. }
  296. return slave->dev->last_rx;
  297. }
  298. #ifdef CONFIG_NET_POLL_CONTROLLER
  299. static inline void bond_netpoll_send_skb(const struct slave *slave,
  300. struct sk_buff *skb)
  301. {
  302. struct netpoll *np = slave->np;
  303. if (np)
  304. netpoll_send_skb(np, skb);
  305. }
  306. #else
  307. static inline void bond_netpoll_send_skb(const struct slave *slave,
  308. struct sk_buff *skb)
  309. {
  310. }
  311. #endif
  312. static inline void bond_set_slave_inactive_flags(struct slave *slave)
  313. {
  314. if (!bond_is_lb(slave->bond))
  315. bond_set_backup_slave(slave);
  316. if (!slave->bond->params.all_slaves_active)
  317. slave->inactive = 1;
  318. }
  319. static inline void bond_set_slave_active_flags(struct slave *slave)
  320. {
  321. bond_set_active_slave(slave);
  322. slave->inactive = 0;
  323. }
  324. static inline bool bond_is_slave_inactive(struct slave *slave)
  325. {
  326. return slave->inactive;
  327. }
  328. static inline __be32 bond_confirm_addr(struct net_device *dev, __be32 dst, __be32 local)
  329. {
  330. struct in_device *in_dev;
  331. __be32 addr = 0;
  332. rcu_read_lock();
  333. in_dev = __in_dev_get_rcu(dev);
  334. if (in_dev)
  335. addr = inet_confirm_addr(in_dev, dst, local, RT_SCOPE_HOST);
  336. rcu_read_unlock();
  337. return addr;
  338. }
  339. static inline bool slave_can_tx(struct slave *slave)
  340. {
  341. if (IS_UP(slave->dev) && slave->link == BOND_LINK_UP &&
  342. bond_is_active_slave(slave))
  343. return true;
  344. else
  345. return false;
  346. }
  347. struct bond_net;
  348. int bond_arp_rcv(const struct sk_buff *skb, struct bonding *bond, struct slave *slave);
  349. int bond_dev_queue_xmit(struct bonding *bond, struct sk_buff *skb, struct net_device *slave_dev);
  350. void bond_xmit_slave_id(struct bonding *bond, struct sk_buff *skb, int slave_id);
  351. int bond_create(struct net *net, const char *name);
  352. int bond_create_sysfs(struct bond_net *net);
  353. void bond_destroy_sysfs(struct bond_net *net);
  354. void bond_prepare_sysfs_group(struct bonding *bond);
  355. int bond_enslave(struct net_device *bond_dev, struct net_device *slave_dev);
  356. int bond_release(struct net_device *bond_dev, struct net_device *slave_dev);
  357. void bond_mii_monitor(struct work_struct *);
  358. void bond_loadbalance_arp_mon(struct work_struct *);
  359. void bond_activebackup_arp_mon(struct work_struct *);
  360. int bond_xmit_hash(struct bonding *bond, struct sk_buff *skb, int count);
  361. int bond_parse_parm(const char *mode_arg, const struct bond_parm_tbl *tbl);
  362. void bond_select_active_slave(struct bonding *bond);
  363. void bond_change_active_slave(struct bonding *bond, struct slave *new_active);
  364. void bond_create_debugfs(void);
  365. void bond_destroy_debugfs(void);
  366. void bond_debug_register(struct bonding *bond);
  367. void bond_debug_unregister(struct bonding *bond);
  368. void bond_debug_reregister(struct bonding *bond);
  369. const char *bond_mode_name(int mode);
  370. void bond_setup(struct net_device *bond_dev);
  371. unsigned int bond_get_num_tx_queues(void);
  372. int bond_netlink_init(void);
  373. void bond_netlink_fini(void);
  374. int bond_option_mode_set(struct bonding *bond, int mode);
  375. int bond_option_active_slave_set(struct bonding *bond, struct net_device *slave_dev);
  376. struct net_device *bond_option_active_slave_get_rcu(struct bonding *bond);
  377. struct net_device *bond_option_active_slave_get(struct bonding *bond);
  378. struct bond_net {
  379. struct net * net; /* Associated network namespace */
  380. struct list_head dev_list;
  381. #ifdef CONFIG_PROC_FS
  382. struct proc_dir_entry * proc_dir;
  383. #endif
  384. struct class_attribute class_attr_bonding_masters;
  385. };
  386. #ifdef CONFIG_PROC_FS
  387. void bond_create_proc_entry(struct bonding *bond);
  388. void bond_remove_proc_entry(struct bonding *bond);
  389. void bond_create_proc_dir(struct bond_net *bn);
  390. void bond_destroy_proc_dir(struct bond_net *bn);
  391. #else
  392. static inline void bond_create_proc_entry(struct bonding *bond)
  393. {
  394. }
  395. static inline void bond_remove_proc_entry(struct bonding *bond)
  396. {
  397. }
  398. static inline void bond_create_proc_dir(struct bond_net *bn)
  399. {
  400. }
  401. static inline void bond_destroy_proc_dir(struct bond_net *bn)
  402. {
  403. }
  404. #endif
  405. static inline struct slave *bond_slave_has_mac(struct bonding *bond,
  406. const u8 *mac)
  407. {
  408. struct list_head *iter;
  409. struct slave *tmp;
  410. bond_for_each_slave(bond, tmp, iter)
  411. if (ether_addr_equal_64bits(mac, tmp->dev->dev_addr))
  412. return tmp;
  413. return NULL;
  414. }
  415. /* Caller must hold rcu_read_lock() for read */
  416. static inline struct slave *bond_slave_has_mac_rcu(struct bonding *bond,
  417. const u8 *mac)
  418. {
  419. struct list_head *iter;
  420. struct slave *tmp;
  421. bond_for_each_slave_rcu(bond, tmp, iter)
  422. if (ether_addr_equal_64bits(mac, tmp->dev->dev_addr))
  423. return tmp;
  424. return NULL;
  425. }
  426. /* Check if the ip is present in arp ip list, or first free slot if ip == 0
  427. * Returns -1 if not found, index if found
  428. */
  429. static inline int bond_get_targets_ip(__be32 *targets, __be32 ip)
  430. {
  431. int i;
  432. for (i = 0; i < BOND_MAX_ARP_TARGETS; i++)
  433. if (targets[i] == ip)
  434. return i;
  435. else if (targets[i] == 0)
  436. break;
  437. return -1;
  438. }
  439. /* exported from bond_main.c */
  440. extern int bond_net_id;
  441. extern const struct bond_parm_tbl bond_lacp_tbl[];
  442. extern const struct bond_parm_tbl bond_mode_tbl[];
  443. extern const struct bond_parm_tbl xmit_hashtype_tbl[];
  444. extern const struct bond_parm_tbl arp_validate_tbl[];
  445. extern const struct bond_parm_tbl arp_all_targets_tbl[];
  446. extern const struct bond_parm_tbl fail_over_mac_tbl[];
  447. extern const struct bond_parm_tbl pri_reselect_tbl[];
  448. extern struct bond_parm_tbl ad_select_tbl[];
  449. /* exported from bond_netlink.c */
  450. extern struct rtnl_link_ops bond_link_ops;
  451. #endif /* _LINUX_BONDING_H */