bonding.h revision 28c719260da032c999ecb4b73ba56311c635ef4e
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/* IMPORTANT: bond_first/last_slave can return NULL in case of an empty list */ 80#define bond_first_slave(bond) \ 81 (bond_has_slaves(bond) ? \ 82 netdev_adjacent_get_private(bond_slave_list(bond)->next) : \ 83 NULL) 84#define bond_last_slave(bond) \ 85 (bond_has_slaves(bond) ? \ 86 netdev_adjacent_get_private(bond_slave_list(bond)->prev) : \ 87 NULL) 88 89#define bond_is_first_slave(bond, pos) (pos == bond_first_slave(bond)) 90#define bond_is_last_slave(bond, pos) (pos == bond_last_slave(bond)) 91 92/** 93 * bond_for_each_slave - iterate over all slaves 94 * @bond: the bond holding this list 95 * @pos: current slave 96 * @iter: list_head * iterator 97 * 98 * Caller must hold bond->lock 99 */ 100#define bond_for_each_slave(bond, pos, iter) \ 101 netdev_for_each_lower_private((bond)->dev, pos, iter) 102 103/* Caller must have rcu_read_lock */ 104#define bond_for_each_slave_rcu(bond, pos, iter) \ 105 netdev_for_each_lower_private_rcu((bond)->dev, pos, iter) 106 107#ifdef CONFIG_NET_POLL_CONTROLLER 108extern atomic_t netpoll_block_tx; 109 110static inline void block_netpoll_tx(void) 111{ 112 atomic_inc(&netpoll_block_tx); 113} 114 115static inline void unblock_netpoll_tx(void) 116{ 117 atomic_dec(&netpoll_block_tx); 118} 119 120static inline int is_netpoll_tx_blocked(struct net_device *dev) 121{ 122 if (unlikely(netpoll_tx_running(dev))) 123 return atomic_read(&netpoll_block_tx); 124 return 0; 125} 126#else 127#define block_netpoll_tx() 128#define unblock_netpoll_tx() 129#define is_netpoll_tx_blocked(dev) (0) 130#endif 131 132struct bond_params { 133 int mode; 134 int xmit_policy; 135 int miimon; 136 u8 num_peer_notif; 137 int arp_interval; 138 int arp_validate; 139 int arp_all_targets; 140 int use_carrier; 141 int fail_over_mac; 142 int updelay; 143 int downdelay; 144 int lacp_fast; 145 unsigned int min_links; 146 int ad_select; 147 char primary[IFNAMSIZ]; 148 int primary_reselect; 149 __be32 arp_targets[BOND_MAX_ARP_TARGETS]; 150 int tx_queues; 151 int all_slaves_active; 152 int resend_igmp; 153 int lp_interval; 154}; 155 156struct bond_parm_tbl { 157 char *modename; 158 int mode; 159}; 160 161#define BOND_MAX_MODENAME_LEN 20 162 163struct slave { 164 struct net_device *dev; /* first - useful for panic debug */ 165 struct bonding *bond; /* our master */ 166 int delay; 167 unsigned long jiffies; 168 unsigned long last_arp_rx; 169 unsigned long target_last_arp_rx[BOND_MAX_ARP_TARGETS]; 170 s8 link; /* one of BOND_LINK_XXXX */ 171 s8 new_link; 172 u8 backup:1, /* indicates backup slave. Value corresponds with 173 BOND_STATE_ACTIVE and BOND_STATE_BACKUP */ 174 inactive:1; /* indicates inactive slave */ 175 u8 duplex; 176 u32 original_mtu; 177 u32 link_failure_count; 178 u32 speed; 179 u16 queue_id; 180 u8 perm_hwaddr[ETH_ALEN]; 181 struct ad_slave_info ad_info; /* HUGE - better to dynamically alloc */ 182 struct tlb_slave_info tlb_info; 183#ifdef CONFIG_NET_POLL_CONTROLLER 184 struct netpoll *np; 185#endif 186}; 187 188/* 189 * Link pseudo-state only used internally by monitors 190 */ 191#define BOND_LINK_NOCHANGE -1 192 193/* 194 * Here are the locking policies for the two bonding locks: 195 * 196 * 1) Get bond->lock when reading/writing slave list. 197 * 2) Get bond->curr_slave_lock when reading/writing bond->curr_active_slave. 198 * (It is unnecessary when the write-lock is put with bond->lock.) 199 * 3) When we lock with bond->curr_slave_lock, we must lock with bond->lock 200 * beforehand. 201 */ 202struct bonding { 203 struct net_device *dev; /* first - useful for panic debug */ 204 struct slave *curr_active_slave; 205 struct slave *current_arp_slave; 206 struct slave *primary_slave; 207 bool force_primary; 208 s32 slave_cnt; /* never change this value outside the attach/detach wrappers */ 209 int (*recv_probe)(const struct sk_buff *, struct bonding *, 210 struct slave *); 211 rwlock_t lock; 212 rwlock_t curr_slave_lock; 213 u8 send_peer_notif; 214 u8 igmp_retrans; 215#ifdef CONFIG_PROC_FS 216 struct proc_dir_entry *proc_entry; 217 char proc_file_name[IFNAMSIZ]; 218#endif /* CONFIG_PROC_FS */ 219 struct list_head bond_list; 220 u16 rr_tx_counter; 221 struct ad_bond_info ad_info; 222 struct alb_bond_info alb_info; 223 struct bond_params params; 224 struct workqueue_struct *wq; 225 struct delayed_work mii_work; 226 struct delayed_work arp_work; 227 struct delayed_work alb_work; 228 struct delayed_work ad_work; 229 struct delayed_work mcast_work; 230#ifdef CONFIG_DEBUG_FS 231 /* debugging support via debugfs */ 232 struct dentry *debug_dir; 233#endif /* CONFIG_DEBUG_FS */ 234}; 235 236#define bond_slave_get_rcu(dev) \ 237 ((struct slave *) rcu_dereference(dev->rx_handler_data)) 238 239#define bond_slave_get_rtnl(dev) \ 240 ((struct slave *) rtnl_dereference(dev->rx_handler_data)) 241 242/** 243 * Returns NULL if the net_device does not belong to any of the bond's slaves 244 * 245 * Caller must hold bond lock for read 246 */ 247static inline struct slave *bond_get_slave_by_dev(struct bonding *bond, 248 struct net_device *slave_dev) 249{ 250 return netdev_lower_dev_get_private(bond->dev, slave_dev); 251} 252 253static inline struct bonding *bond_get_bond_by_slave(struct slave *slave) 254{ 255 if (!slave || !slave->bond) 256 return NULL; 257 return slave->bond; 258} 259 260static inline bool bond_is_lb(const struct bonding *bond) 261{ 262 return (bond->params.mode == BOND_MODE_TLB || 263 bond->params.mode == BOND_MODE_ALB); 264} 265 266static inline void bond_set_active_slave(struct slave *slave) 267{ 268 slave->backup = 0; 269} 270 271static inline void bond_set_backup_slave(struct slave *slave) 272{ 273 slave->backup = 1; 274} 275 276static inline int bond_slave_state(struct slave *slave) 277{ 278 return slave->backup; 279} 280 281static inline bool bond_is_active_slave(struct slave *slave) 282{ 283 return !bond_slave_state(slave); 284} 285 286#define BOND_PRI_RESELECT_ALWAYS 0 287#define BOND_PRI_RESELECT_BETTER 1 288#define BOND_PRI_RESELECT_FAILURE 2 289 290#define BOND_FOM_NONE 0 291#define BOND_FOM_ACTIVE 1 292#define BOND_FOM_FOLLOW 2 293 294#define BOND_ARP_TARGETS_ANY 0 295#define BOND_ARP_TARGETS_ALL 1 296 297#define BOND_ARP_VALIDATE_NONE 0 298#define BOND_ARP_VALIDATE_ACTIVE (1 << BOND_STATE_ACTIVE) 299#define BOND_ARP_VALIDATE_BACKUP (1 << BOND_STATE_BACKUP) 300#define BOND_ARP_VALIDATE_ALL (BOND_ARP_VALIDATE_ACTIVE | \ 301 BOND_ARP_VALIDATE_BACKUP) 302 303static inline int slave_do_arp_validate(struct bonding *bond, 304 struct slave *slave) 305{ 306 return bond->params.arp_validate & (1 << bond_slave_state(slave)); 307} 308 309/* Get the oldest arp which we've received on this slave for bond's 310 * arp_targets. 311 */ 312static inline unsigned long slave_oldest_target_arp_rx(struct bonding *bond, 313 struct slave *slave) 314{ 315 int i = 1; 316 unsigned long ret = slave->target_last_arp_rx[0]; 317 318 for (; (i < BOND_MAX_ARP_TARGETS) && bond->params.arp_targets[i]; i++) 319 if (time_before(slave->target_last_arp_rx[i], ret)) 320 ret = slave->target_last_arp_rx[i]; 321 322 return ret; 323} 324 325static inline unsigned long slave_last_rx(struct bonding *bond, 326 struct slave *slave) 327{ 328 if (slave_do_arp_validate(bond, slave)) { 329 if (bond->params.arp_all_targets == BOND_ARP_TARGETS_ALL) 330 return slave_oldest_target_arp_rx(bond, slave); 331 else 332 return slave->last_arp_rx; 333 } 334 335 return slave->dev->last_rx; 336} 337 338#ifdef CONFIG_NET_POLL_CONTROLLER 339static inline void bond_netpoll_send_skb(const struct slave *slave, 340 struct sk_buff *skb) 341{ 342 struct netpoll *np = slave->np; 343 344 if (np) 345 netpoll_send_skb(np, skb); 346} 347#else 348static inline void bond_netpoll_send_skb(const struct slave *slave, 349 struct sk_buff *skb) 350{ 351} 352#endif 353 354static inline void bond_set_slave_inactive_flags(struct slave *slave) 355{ 356 if (!bond_is_lb(slave->bond)) 357 bond_set_backup_slave(slave); 358 if (!slave->bond->params.all_slaves_active) 359 slave->inactive = 1; 360} 361 362static inline void bond_set_slave_active_flags(struct slave *slave) 363{ 364 bond_set_active_slave(slave); 365 slave->inactive = 0; 366} 367 368static inline bool bond_is_slave_inactive(struct slave *slave) 369{ 370 return slave->inactive; 371} 372 373static inline __be32 bond_confirm_addr(struct net_device *dev, __be32 dst, __be32 local) 374{ 375 struct in_device *in_dev; 376 __be32 addr = 0; 377 378 rcu_read_lock(); 379 in_dev = __in_dev_get_rcu(dev); 380 381 if (in_dev) 382 addr = inet_confirm_addr(in_dev, dst, local, RT_SCOPE_HOST); 383 384 rcu_read_unlock(); 385 return addr; 386} 387 388static inline bool slave_can_tx(struct slave *slave) 389{ 390 if (IS_UP(slave->dev) && slave->link == BOND_LINK_UP && 391 bond_is_active_slave(slave)) 392 return true; 393 else 394 return false; 395} 396 397struct bond_net; 398 399int bond_arp_rcv(const struct sk_buff *skb, struct bonding *bond, struct slave *slave); 400int bond_dev_queue_xmit(struct bonding *bond, struct sk_buff *skb, struct net_device *slave_dev); 401void bond_xmit_slave_id(struct bonding *bond, struct sk_buff *skb, int slave_id); 402int bond_create(struct net *net, const char *name); 403int bond_create_sysfs(struct bond_net *net); 404void bond_destroy_sysfs(struct bond_net *net); 405void bond_prepare_sysfs_group(struct bonding *bond); 406int bond_enslave(struct net_device *bond_dev, struct net_device *slave_dev); 407int bond_release(struct net_device *bond_dev, struct net_device *slave_dev); 408void bond_mii_monitor(struct work_struct *); 409void bond_loadbalance_arp_mon(struct work_struct *); 410void bond_activebackup_arp_mon(struct work_struct *); 411int bond_xmit_hash(struct bonding *bond, struct sk_buff *skb, int count); 412int bond_parse_parm(const char *mode_arg, const struct bond_parm_tbl *tbl); 413void bond_select_active_slave(struct bonding *bond); 414void bond_change_active_slave(struct bonding *bond, struct slave *new_active); 415void bond_create_debugfs(void); 416void bond_destroy_debugfs(void); 417void bond_debug_register(struct bonding *bond); 418void bond_debug_unregister(struct bonding *bond); 419void bond_debug_reregister(struct bonding *bond); 420const char *bond_mode_name(int mode); 421 422struct bond_net { 423 struct net * net; /* Associated network namespace */ 424 struct list_head dev_list; 425#ifdef CONFIG_PROC_FS 426 struct proc_dir_entry * proc_dir; 427#endif 428 struct class_attribute class_attr_bonding_masters; 429}; 430 431#ifdef CONFIG_PROC_FS 432void bond_create_proc_entry(struct bonding *bond); 433void bond_remove_proc_entry(struct bonding *bond); 434void bond_create_proc_dir(struct bond_net *bn); 435void bond_destroy_proc_dir(struct bond_net *bn); 436#else 437static inline void bond_create_proc_entry(struct bonding *bond) 438{ 439} 440 441static inline void bond_remove_proc_entry(struct bonding *bond) 442{ 443} 444 445static inline void bond_create_proc_dir(struct bond_net *bn) 446{ 447} 448 449static inline void bond_destroy_proc_dir(struct bond_net *bn) 450{ 451} 452#endif 453 454static inline struct slave *bond_slave_has_mac(struct bonding *bond, 455 const u8 *mac) 456{ 457 struct list_head *iter; 458 struct slave *tmp; 459 460 bond_for_each_slave(bond, tmp, iter) 461 if (ether_addr_equal_64bits(mac, tmp->dev->dev_addr)) 462 return tmp; 463 464 return NULL; 465} 466 467/* Caller must hold rcu_read_lock() for read */ 468static inline struct slave *bond_slave_has_mac_rcu(struct bonding *bond, 469 const u8 *mac) 470{ 471 struct list_head *iter; 472 struct slave *tmp; 473 474 bond_for_each_slave_rcu(bond, tmp, iter) 475 if (ether_addr_equal_64bits(mac, tmp->dev->dev_addr)) 476 return tmp; 477 478 return NULL; 479} 480 481/* Check if the ip is present in arp ip list, or first free slot if ip == 0 482 * Returns -1 if not found, index if found 483 */ 484static inline int bond_get_targets_ip(__be32 *targets, __be32 ip) 485{ 486 int i; 487 488 for (i = 0; i < BOND_MAX_ARP_TARGETS; i++) 489 if (targets[i] == ip) 490 return i; 491 else if (targets[i] == 0) 492 break; 493 494 return -1; 495} 496 497/* exported from bond_main.c */ 498extern int bond_net_id; 499extern const struct bond_parm_tbl bond_lacp_tbl[]; 500extern const struct bond_parm_tbl bond_mode_tbl[]; 501extern const struct bond_parm_tbl xmit_hashtype_tbl[]; 502extern const struct bond_parm_tbl arp_validate_tbl[]; 503extern const struct bond_parm_tbl arp_all_targets_tbl[]; 504extern const struct bond_parm_tbl fail_over_mac_tbl[]; 505extern const struct bond_parm_tbl pri_reselect_tbl[]; 506extern struct bond_parm_tbl ad_select_tbl[]; 507 508#endif /* _LINUX_BONDING_H */ 509