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 _NET_BONDING_H 16 #define _NET_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 <linux/reciprocal_div.h> 27 #include <linux/if_link.h> 28 29 #include <net/bond_3ad.h> 30 #include <net/bond_alb.h> 31 #include <net/bond_options.h> 32 33 #define DRV_VERSION "3.7.1" 34 #define DRV_RELDATE "April 27, 2011" 35 #define DRV_NAME "bonding" 36 #define DRV_DESCRIPTION "Ethernet Channel Bonding Driver" 37 38 #define bond_version DRV_DESCRIPTION ": v" DRV_VERSION " (" DRV_RELDATE ")\n" 39 40 #define BOND_MAX_ARP_TARGETS 16 41 42 #define BOND_DEFAULT_MIIMON 100 43 44 /* 45 * Less bad way to call ioctl from within the kernel; this needs to be 46 * done some other way to get the call out of interrupt context. 47 * Needs "ioctl" variable to be supplied by calling context. 48 */ 49 #define IOCTL(dev, arg, cmd) ({ \ 50 int res = 0; \ 51 mm_segment_t fs = get_fs(); \ 52 set_fs(get_ds()); \ 53 res = ioctl(dev, arg, cmd); \ 54 set_fs(fs); \ 55 res; }) 56 57 #define BOND_MODE(bond) ((bond)->params.mode) 58 59 /* slave list primitives */ 60 #define bond_slave_list(bond) (&(bond)->dev->adj_list.lower) 61 62 #define bond_has_slaves(bond) !list_empty(bond_slave_list(bond)) 63 64 /* IMPORTANT: bond_first/last_slave can return NULL in case of an empty list */ 65 #define bond_first_slave(bond) \ 66 (bond_has_slaves(bond) ? \ 67 netdev_adjacent_get_private(bond_slave_list(bond)->next) : \ 68 NULL) 69 #define bond_last_slave(bond) \ 70 (bond_has_slaves(bond) ? \ 71 netdev_adjacent_get_private(bond_slave_list(bond)->prev) : \ 72 NULL) 73 74 /* Caller must have rcu_read_lock */ 75 #define bond_first_slave_rcu(bond) \ 76 netdev_lower_get_first_private_rcu(bond->dev) 77 78 #define bond_is_first_slave(bond, pos) (pos == bond_first_slave(bond)) 79 #define bond_is_last_slave(bond, pos) (pos == bond_last_slave(bond)) 80 81 /** 82 * bond_for_each_slave - iterate over all slaves 83 * @bond: the bond holding this list 84 * @pos: current slave 85 * @iter: list_head * iterator 86 * 87 * Caller must hold RTNL 88 */ 89 #define bond_for_each_slave(bond, pos, iter) \ 90 netdev_for_each_lower_private((bond)->dev, pos, iter) 91 92 /* Caller must have rcu_read_lock */ 93 #define bond_for_each_slave_rcu(bond, pos, iter) \ 94 netdev_for_each_lower_private_rcu((bond)->dev, pos, iter) 95 96 #ifdef CONFIG_NET_POLL_CONTROLLER 97 extern atomic_t netpoll_block_tx; 98 99 static inline void block_netpoll_tx(void) 100 { 101 atomic_inc(&netpoll_block_tx); 102 } 103 104 static inline void unblock_netpoll_tx(void) 105 { 106 atomic_dec(&netpoll_block_tx); 107 } 108 109 static inline int is_netpoll_tx_blocked(struct net_device *dev) 110 { 111 if (unlikely(netpoll_tx_running(dev))) 112 return atomic_read(&netpoll_block_tx); 113 return 0; 114 } 115 #else 116 #define block_netpoll_tx() 117 #define unblock_netpoll_tx() 118 #define is_netpoll_tx_blocked(dev) (0) 119 #endif 120 121 struct bond_params { 122 int mode; 123 int xmit_policy; 124 int miimon; 125 u8 num_peer_notif; 126 int arp_interval; 127 int arp_validate; 128 int arp_all_targets; 129 int use_carrier; 130 int fail_over_mac; 131 int updelay; 132 int downdelay; 133 int lacp_fast; 134 unsigned int min_links; 135 int ad_select; 136 char primary[IFNAMSIZ]; 137 int primary_reselect; 138 __be32 arp_targets[BOND_MAX_ARP_TARGETS]; 139 int tx_queues; 140 int all_slaves_active; 141 int resend_igmp; 142 int lp_interval; 143 int packets_per_slave; 144 int tlb_dynamic_lb; 145 struct reciprocal_value reciprocal_packets_per_slave; 146 }; 147 148 struct bond_parm_tbl { 149 char *modename; 150 int mode; 151 }; 152 153 struct slave { 154 struct net_device *dev; /* first - useful for panic debug */ 155 struct bonding *bond; /* our master */ 156 int delay; 157 /* all three in jiffies */ 158 unsigned long last_link_up; 159 unsigned long last_rx; 160 unsigned long target_last_arp_rx[BOND_MAX_ARP_TARGETS]; 161 s8 link; /* one of BOND_LINK_XXXX */ 162 s8 new_link; 163 u8 backup:1, /* indicates backup slave. Value corresponds with 164 BOND_STATE_ACTIVE and BOND_STATE_BACKUP */ 165 inactive:1, /* indicates inactive slave */ 166 should_notify:1; /* indicateds whether the state changed */ 167 u8 duplex; 168 u32 original_mtu; 169 u32 link_failure_count; 170 u32 speed; 171 u16 queue_id; 172 u8 perm_hwaddr[ETH_ALEN]; 173 struct ad_slave_info *ad_info; 174 struct tlb_slave_info tlb_info; 175 #ifdef CONFIG_NET_POLL_CONTROLLER 176 struct netpoll *np; 177 #endif 178 struct kobject kobj; 179 struct rtnl_link_stats64 slave_stats; 180 }; 181 182 struct bond_up_slave { 183 unsigned int count; 184 struct rcu_head rcu; 185 struct slave *arr[0]; 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 * Get rcu_read_lock when reading or RTNL when writing slave list. 196 */ 197 struct bonding { 198 struct net_device *dev; /* first - useful for panic debug */ 199 struct slave __rcu *curr_active_slave; 200 struct slave __rcu *current_arp_slave; 201 struct slave __rcu *primary_slave; 202 struct bond_up_slave __rcu *slave_arr; /* Array of usable slaves */ 203 bool force_primary; 204 s32 slave_cnt; /* never change this value outside the attach/detach wrappers */ 205 int (*recv_probe)(const struct sk_buff *, struct bonding *, 206 struct slave *); 207 /* mode_lock is used for mode-specific locking needs, currently used by: 208 * 3ad mode (4) - protect against running bond_3ad_unbind_slave() and 209 * bond_3ad_state_machine_handler() concurrently and also 210 * the access to the state machine shared variables. 211 * TLB mode (5) - to sync the use and modifications of its hash table 212 * ALB mode (6) - to sync the use and modifications of its hash table 213 */ 214 spinlock_t mode_lock; 215 u8 send_peer_notif; 216 u8 igmp_retrans; 217 #ifdef CONFIG_PROC_FS 218 struct proc_dir_entry *proc_entry; 219 char proc_file_name[IFNAMSIZ]; 220 #endif /* CONFIG_PROC_FS */ 221 struct list_head bond_list; 222 u32 rr_tx_counter; 223 struct ad_bond_info ad_info; 224 struct alb_bond_info alb_info; 225 struct bond_params params; 226 struct workqueue_struct *wq; 227 struct delayed_work mii_work; 228 struct delayed_work arp_work; 229 struct delayed_work alb_work; 230 struct delayed_work ad_work; 231 struct delayed_work mcast_work; 232 struct delayed_work slave_arr_work; 233 #ifdef CONFIG_DEBUG_FS 234 /* debugging support via debugfs */ 235 struct dentry *debug_dir; 236 #endif /* CONFIG_DEBUG_FS */ 237 struct rtnl_link_stats64 bond_stats; 238 }; 239 240 #define bond_slave_get_rcu(dev) \ 241 ((struct slave *) rcu_dereference(dev->rx_handler_data)) 242 243 #define bond_slave_get_rtnl(dev) \ 244 ((struct slave *) rtnl_dereference(dev->rx_handler_data)) 245 246 struct bond_vlan_tag { 247 __be16 vlan_proto; 248 unsigned short vlan_id; 249 }; 250 251 /** 252 * Returns NULL if the net_device does not belong to any of the bond's slaves 253 * 254 * Caller must hold bond lock for read 255 */ 256 static inline struct slave *bond_get_slave_by_dev(struct bonding *bond, 257 struct net_device *slave_dev) 258 { 259 return netdev_lower_dev_get_private(bond->dev, slave_dev); 260 } 261 262 static inline struct bonding *bond_get_bond_by_slave(struct slave *slave) 263 { 264 return slave->bond; 265 } 266 267 static inline bool bond_should_override_tx_queue(struct bonding *bond) 268 { 269 return BOND_MODE(bond) == BOND_MODE_ACTIVEBACKUP || 270 BOND_MODE(bond) == BOND_MODE_ROUNDROBIN; 271 } 272 273 static inline bool bond_is_lb(const struct bonding *bond) 274 { 275 return BOND_MODE(bond) == BOND_MODE_TLB || 276 BOND_MODE(bond) == BOND_MODE_ALB; 277 } 278 279 static inline bool bond_is_nondyn_tlb(const struct bonding *bond) 280 { 281 return (BOND_MODE(bond) == BOND_MODE_TLB) && 282 (bond->params.tlb_dynamic_lb == 0); 283 } 284 285 static inline bool bond_mode_uses_xmit_hash(const struct bonding *bond) 286 { 287 return (BOND_MODE(bond) == BOND_MODE_8023AD || 288 BOND_MODE(bond) == BOND_MODE_XOR || 289 bond_is_nondyn_tlb(bond)); 290 } 291 292 static inline bool bond_mode_uses_arp(int mode) 293 { 294 return mode != BOND_MODE_8023AD && mode != BOND_MODE_TLB && 295 mode != BOND_MODE_ALB; 296 } 297 298 static inline bool bond_mode_uses_primary(int mode) 299 { 300 return mode == BOND_MODE_ACTIVEBACKUP || mode == BOND_MODE_TLB || 301 mode == BOND_MODE_ALB; 302 } 303 304 static inline bool bond_uses_primary(struct bonding *bond) 305 { 306 return bond_mode_uses_primary(BOND_MODE(bond)); 307 } 308 309 static inline bool bond_slave_is_up(struct slave *slave) 310 { 311 return netif_running(slave->dev) && netif_carrier_ok(slave->dev); 312 } 313 314 static inline void bond_set_active_slave(struct slave *slave) 315 { 316 if (slave->backup) { 317 slave->backup = 0; 318 rtmsg_ifinfo(RTM_NEWLINK, slave->dev, 0, GFP_ATOMIC); 319 } 320 } 321 322 static inline void bond_set_backup_slave(struct slave *slave) 323 { 324 if (!slave->backup) { 325 slave->backup = 1; 326 rtmsg_ifinfo(RTM_NEWLINK, slave->dev, 0, GFP_ATOMIC); 327 } 328 } 329 330 static inline void bond_set_slave_state(struct slave *slave, 331 int slave_state, bool notify) 332 { 333 if (slave->backup == slave_state) 334 return; 335 336 slave->backup = slave_state; 337 if (notify) { 338 rtmsg_ifinfo(RTM_NEWLINK, slave->dev, 0, GFP_ATOMIC); 339 slave->should_notify = 0; 340 } else { 341 if (slave->should_notify) 342 slave->should_notify = 0; 343 else 344 slave->should_notify = 1; 345 } 346 } 347 348 static inline void bond_slave_state_change(struct bonding *bond) 349 { 350 struct list_head *iter; 351 struct slave *tmp; 352 353 bond_for_each_slave(bond, tmp, iter) { 354 if (tmp->link == BOND_LINK_UP) 355 bond_set_active_slave(tmp); 356 else if (tmp->link == BOND_LINK_DOWN) 357 bond_set_backup_slave(tmp); 358 } 359 } 360 361 static inline void bond_slave_state_notify(struct bonding *bond) 362 { 363 struct list_head *iter; 364 struct slave *tmp; 365 366 bond_for_each_slave(bond, tmp, iter) { 367 if (tmp->should_notify) { 368 rtmsg_ifinfo(RTM_NEWLINK, tmp->dev, 0, GFP_ATOMIC); 369 tmp->should_notify = 0; 370 } 371 } 372 } 373 374 static inline int bond_slave_state(struct slave *slave) 375 { 376 return slave->backup; 377 } 378 379 static inline bool bond_is_active_slave(struct slave *slave) 380 { 381 return !bond_slave_state(slave); 382 } 383 384 static inline bool bond_slave_can_tx(struct slave *slave) 385 { 386 return bond_slave_is_up(slave) && slave->link == BOND_LINK_UP && 387 bond_is_active_slave(slave); 388 } 389 390 #define BOND_PRI_RESELECT_ALWAYS 0 391 #define BOND_PRI_RESELECT_BETTER 1 392 #define BOND_PRI_RESELECT_FAILURE 2 393 394 #define BOND_FOM_NONE 0 395 #define BOND_FOM_ACTIVE 1 396 #define BOND_FOM_FOLLOW 2 397 398 #define BOND_ARP_TARGETS_ANY 0 399 #define BOND_ARP_TARGETS_ALL 1 400 401 #define BOND_ARP_VALIDATE_NONE 0 402 #define BOND_ARP_VALIDATE_ACTIVE (1 << BOND_STATE_ACTIVE) 403 #define BOND_ARP_VALIDATE_BACKUP (1 << BOND_STATE_BACKUP) 404 #define BOND_ARP_VALIDATE_ALL (BOND_ARP_VALIDATE_ACTIVE | \ 405 BOND_ARP_VALIDATE_BACKUP) 406 #define BOND_ARP_FILTER (BOND_ARP_VALIDATE_ALL + 1) 407 #define BOND_ARP_FILTER_ACTIVE (BOND_ARP_VALIDATE_ACTIVE | \ 408 BOND_ARP_FILTER) 409 #define BOND_ARP_FILTER_BACKUP (BOND_ARP_VALIDATE_BACKUP | \ 410 BOND_ARP_FILTER) 411 412 #define BOND_SLAVE_NOTIFY_NOW true 413 #define BOND_SLAVE_NOTIFY_LATER false 414 415 static inline int slave_do_arp_validate(struct bonding *bond, 416 struct slave *slave) 417 { 418 return bond->params.arp_validate & (1 << bond_slave_state(slave)); 419 } 420 421 static inline int slave_do_arp_validate_only(struct bonding *bond) 422 { 423 return bond->params.arp_validate & BOND_ARP_FILTER; 424 } 425 426 static inline int bond_is_ip_target_ok(__be32 addr) 427 { 428 return !ipv4_is_lbcast(addr) && !ipv4_is_zeronet(addr); 429 } 430 431 /* Get the oldest arp which we've received on this slave for bond's 432 * arp_targets. 433 */ 434 static inline unsigned long slave_oldest_target_arp_rx(struct bonding *bond, 435 struct slave *slave) 436 { 437 int i = 1; 438 unsigned long ret = slave->target_last_arp_rx[0]; 439 440 for (; (i < BOND_MAX_ARP_TARGETS) && bond->params.arp_targets[i]; i++) 441 if (time_before(slave->target_last_arp_rx[i], ret)) 442 ret = slave->target_last_arp_rx[i]; 443 444 return ret; 445 } 446 447 static inline unsigned long slave_last_rx(struct bonding *bond, 448 struct slave *slave) 449 { 450 if (bond->params.arp_all_targets == BOND_ARP_TARGETS_ALL) 451 return slave_oldest_target_arp_rx(bond, slave); 452 453 return slave->last_rx; 454 } 455 456 #ifdef CONFIG_NET_POLL_CONTROLLER 457 static inline void bond_netpoll_send_skb(const struct slave *slave, 458 struct sk_buff *skb) 459 { 460 struct netpoll *np = slave->np; 461 462 if (np) 463 netpoll_send_skb(np, skb); 464 } 465 #else 466 static inline void bond_netpoll_send_skb(const struct slave *slave, 467 struct sk_buff *skb) 468 { 469 } 470 #endif 471 472 static inline void bond_set_slave_inactive_flags(struct slave *slave, 473 bool notify) 474 { 475 if (!bond_is_lb(slave->bond)) 476 bond_set_slave_state(slave, BOND_STATE_BACKUP, notify); 477 if (!slave->bond->params.all_slaves_active) 478 slave->inactive = 1; 479 } 480 481 static inline void bond_set_slave_active_flags(struct slave *slave, 482 bool notify) 483 { 484 bond_set_slave_state(slave, BOND_STATE_ACTIVE, notify); 485 slave->inactive = 0; 486 } 487 488 static inline bool bond_is_slave_inactive(struct slave *slave) 489 { 490 return slave->inactive; 491 } 492 493 static inline __be32 bond_confirm_addr(struct net_device *dev, __be32 dst, __be32 local) 494 { 495 struct in_device *in_dev; 496 __be32 addr = 0; 497 498 rcu_read_lock(); 499 in_dev = __in_dev_get_rcu(dev); 500 501 if (in_dev) 502 addr = inet_confirm_addr(dev_net(dev), in_dev, dst, local, 503 RT_SCOPE_HOST); 504 rcu_read_unlock(); 505 return addr; 506 } 507 508 struct bond_net { 509 struct net *net; /* Associated network namespace */ 510 struct list_head dev_list; 511 #ifdef CONFIG_PROC_FS 512 struct proc_dir_entry *proc_dir; 513 #endif 514 struct class_attribute class_attr_bonding_masters; 515 }; 516 517 int bond_arp_rcv(const struct sk_buff *skb, struct bonding *bond, struct slave *slave); 518 void bond_dev_queue_xmit(struct bonding *bond, struct sk_buff *skb, struct net_device *slave_dev); 519 int bond_create(struct net *net, const char *name); 520 int bond_create_sysfs(struct bond_net *net); 521 void bond_destroy_sysfs(struct bond_net *net); 522 void bond_prepare_sysfs_group(struct bonding *bond); 523 int bond_sysfs_slave_add(struct slave *slave); 524 void bond_sysfs_slave_del(struct slave *slave); 525 int bond_enslave(struct net_device *bond_dev, struct net_device *slave_dev); 526 int bond_release(struct net_device *bond_dev, struct net_device *slave_dev); 527 u32 bond_xmit_hash(struct bonding *bond, struct sk_buff *skb); 528 void bond_select_active_slave(struct bonding *bond); 529 void bond_change_active_slave(struct bonding *bond, struct slave *new_active); 530 void bond_create_debugfs(void); 531 void bond_destroy_debugfs(void); 532 void bond_debug_register(struct bonding *bond); 533 void bond_debug_unregister(struct bonding *bond); 534 void bond_debug_reregister(struct bonding *bond); 535 const char *bond_mode_name(int mode); 536 void bond_setup(struct net_device *bond_dev); 537 unsigned int bond_get_num_tx_queues(void); 538 int bond_netlink_init(void); 539 void bond_netlink_fini(void); 540 struct net_device *bond_option_active_slave_get_rcu(struct bonding *bond); 541 const char *bond_slave_link_status(s8 link); 542 struct bond_vlan_tag *bond_verify_device_path(struct net_device *start_dev, 543 struct net_device *end_dev, 544 int level); 545 int bond_update_slave_arr(struct bonding *bond, struct slave *skipslave); 546 void bond_slave_arr_work_rearm(struct bonding *bond, unsigned long delay); 547 548 #ifdef CONFIG_PROC_FS 549 void bond_create_proc_entry(struct bonding *bond); 550 void bond_remove_proc_entry(struct bonding *bond); 551 void bond_create_proc_dir(struct bond_net *bn); 552 void bond_destroy_proc_dir(struct bond_net *bn); 553 #else 554 static inline void bond_create_proc_entry(struct bonding *bond) 555 { 556 } 557 558 static inline void bond_remove_proc_entry(struct bonding *bond) 559 { 560 } 561 562 static inline void bond_create_proc_dir(struct bond_net *bn) 563 { 564 } 565 566 static inline void bond_destroy_proc_dir(struct bond_net *bn) 567 { 568 } 569 #endif 570 571 static inline struct slave *bond_slave_has_mac(struct bonding *bond, 572 const u8 *mac) 573 { 574 struct list_head *iter; 575 struct slave *tmp; 576 577 bond_for_each_slave(bond, tmp, iter) 578 if (ether_addr_equal_64bits(mac, tmp->dev->dev_addr)) 579 return tmp; 580 581 return NULL; 582 } 583 584 /* Caller must hold rcu_read_lock() for read */ 585 static inline struct slave *bond_slave_has_mac_rcu(struct bonding *bond, 586 const u8 *mac) 587 { 588 struct list_head *iter; 589 struct slave *tmp; 590 591 bond_for_each_slave_rcu(bond, tmp, iter) 592 if (ether_addr_equal_64bits(mac, tmp->dev->dev_addr)) 593 return tmp; 594 595 return NULL; 596 } 597 598 /* Caller must hold rcu_read_lock() for read */ 599 static inline bool bond_slave_has_mac_rx(struct bonding *bond, const u8 *mac) 600 { 601 struct list_head *iter; 602 struct slave *tmp; 603 struct netdev_hw_addr *ha; 604 605 bond_for_each_slave_rcu(bond, tmp, iter) 606 if (ether_addr_equal_64bits(mac, tmp->dev->dev_addr)) 607 return true; 608 609 if (netdev_uc_empty(bond->dev)) 610 return false; 611 612 netdev_for_each_uc_addr(ha, bond->dev) 613 if (ether_addr_equal_64bits(mac, ha->addr)) 614 return true; 615 616 return false; 617 } 618 619 /* Check if the ip is present in arp ip list, or first free slot if ip == 0 620 * Returns -1 if not found, index if found 621 */ 622 static inline int bond_get_targets_ip(__be32 *targets, __be32 ip) 623 { 624 int i; 625 626 for (i = 0; i < BOND_MAX_ARP_TARGETS; i++) 627 if (targets[i] == ip) 628 return i; 629 else if (targets[i] == 0) 630 break; 631 632 return -1; 633 } 634 635 /* exported from bond_main.c */ 636 extern int bond_net_id; 637 extern const struct bond_parm_tbl bond_lacp_tbl[]; 638 extern const struct bond_parm_tbl xmit_hashtype_tbl[]; 639 extern const struct bond_parm_tbl arp_validate_tbl[]; 640 extern const struct bond_parm_tbl arp_all_targets_tbl[]; 641 extern const struct bond_parm_tbl fail_over_mac_tbl[]; 642 extern const struct bond_parm_tbl pri_reselect_tbl[]; 643 extern struct bond_parm_tbl ad_select_tbl[]; 644 645 /* exported from bond_netlink.c */ 646 extern struct rtnl_link_ops bond_link_ops; 647 648 static inline void bond_tx_drop(struct net_device *dev, struct sk_buff *skb) 649 { 650 atomic_long_inc(&dev->tx_dropped); 651 dev_kfree_skb_any(skb); 652 } 653 654 #endif /* _NET_BONDING_H */ 655