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_to_slave(ptr) list_entry(ptr, struct slave, list)
  76
  77 /* IMPORTANT: bond_first/last_slave can return NULL in case of an empty list */
  78 #define bond_first_slave(bond) \
  79         list_first_entry_or_null(&(bond)->slave_list, struct slave, list)
  80 #define bond_last_slave(bond) \
  81         (list_empty(&(bond)->slave_list) ? NULL : \
  82                                            bond_to_slave((bond)->slave_list.prev))
  83
  84 #define bond_is_first_slave(bond, pos) ((pos)->list.prev == &(bond)->slave_list)
  85 #define bond_is_last_slave(bond, pos) ((pos)->list.next == &(bond)->slave_list)
  86
  87 /* Since bond_first/last_slave can return NULL, these can return NULL too */
  88 #define bond_next_slave(bond, pos) \
  89         (bond_is_last_slave(bond, pos) ? bond_first_slave(bond) : \
  90                                          bond_to_slave((pos)->list.next))
  91
  92 #define bond_prev_slave(bond, pos) \
  93         (bond_is_first_slave(bond, pos) ? bond_last_slave(bond) : \
  94                                           bond_to_slave((pos)->list.prev))
  95
  96 /**
  97  * bond_for_each_slave_from - iterate the slaves list from a starting point
  98  * @bond:       the bond holding this list.
  99  * @pos:        current slave.
 100  * @cnt:        counter for max number of moves
 101  * @start:      starting point.
 102  *
 103  * Caller must hold bond->lock
 104  */
 105 #define bond_for_each_slave_from(bond, pos, cnt, start) \
 106         for (cnt = 0, pos = start; pos && cnt < (bond)->slave_cnt; \
 107              cnt++, pos = bond_next_slave(bond, pos))
 108
 109 /**
 110  * bond_for_each_slave - iterate over all slaves
 111  * @bond:       the bond holding this list
 112  * @pos:        current slave
 113  *
 114  * Caller must hold bond->lock
 115  */
 116 #define bond_for_each_slave(bond, pos) \
 117         list_for_each_entry(pos, &(bond)->slave_list, list)
 118
 119 /* Caller must have rcu_read_lock */
 120 #define bond_for_each_slave_rcu(bond, pos) \
 121         list_for_each_entry_rcu(pos, &(bond)->slave_list, list)
 122
 123 /**
 124  * bond_for_each_slave_reverse - iterate in reverse from a given position
 125  * @bond:       the bond holding this list
 126  * @pos:        slave to continue from
 127  *
 128  * Caller must hold bond->lock
 129  */
 130 #define bond_for_each_slave_continue_reverse(bond, pos) \
 131         list_for_each_entry_continue_reverse(pos, &(bond)->slave_list, list)
 132
 133 #ifdef CONFIG_NET_POLL_CONTROLLER
 134 extern atomic_t netpoll_block_tx;
 135
 136 static inline void block_netpoll_tx(void)
 137 {
 138         atomic_inc(&netpoll_block_tx);
 139 }
 140
 141 static inline void unblock_netpoll_tx(void)
 142 {
 143         atomic_dec(&netpoll_block_tx);
 144 }
 145
 146 static inline int is_netpoll_tx_blocked(struct net_device *dev)
 147 {
 148         if (unlikely(netpoll_tx_running(dev)))
 149                 return atomic_read(&netpoll_block_tx);
 150         return 0;
 151 }
 152 #else
 153 #define block_netpoll_tx()
 154 #define unblock_netpoll_tx()
 155 #define is_netpoll_tx_blocked(dev) (0)
 156 #endif
 157
 158 struct bond_params {
 159         int mode;
 160         int xmit_policy;
 161         int miimon;
 162         u8 num_peer_notif;
 163         int arp_interval;
 164         int arp_validate;
 165         int arp_all_targets;
 166         int use_carrier;
 167         int fail_over_mac;
 168         int updelay;
 169         int downdelay;
 170         int lacp_fast;
 171         unsigned int min_links;
 172         int ad_select;
 173         char primary[IFNAMSIZ];
 174         int primary_reselect;
 175         __be32 arp_targets[BOND_MAX_ARP_TARGETS];
 176         int tx_queues;
 177         int all_slaves_active;
 178         int resend_igmp;
 179 };
 180
 181 struct bond_parm_tbl {
 182         char *modename;
 183         int mode;
 184 };
 185
 186 #define BOND_MAX_MODENAME_LEN 20
 187
 188 struct slave {
 189         struct net_device *dev; /* first - useful for panic debug */
 190         struct list_head list;
 191         struct bonding *bond; /* our master */
 192         int    delay;
 193         unsigned long jiffies;
 194         unsigned long last_arp_rx;
 195         unsigned long target_last_arp_rx[BOND_MAX_ARP_TARGETS];
 196         s8     link;    /* one of BOND_LINK_XXXX */
 197         s8     new_link;
 198         u8     backup:1,   /* indicates backup slave. Value corresponds with
 199                               BOND_STATE_ACTIVE and BOND_STATE_BACKUP */
 200                inactive:1; /* indicates inactive slave */
 201         u8     duplex;
 202         u32    original_mtu;
 203         u32    link_failure_count;
 204         u32    speed;
 205         u16    queue_id;
 206         u8     perm_hwaddr[ETH_ALEN];
 207         struct ad_slave_info ad_info; /* HUGE - better to dynamically alloc */
 208         struct tlb_slave_info tlb_info;
 209 #ifdef CONFIG_NET_POLL_CONTROLLER
 210         struct netpoll *np;
 211 #endif
 212 };
 213
 214 /*
 215  * Link pseudo-state only used internally by monitors
 216  */
 217 #define BOND_LINK_NOCHANGE -1
 218
 219 /*
 220  * Here are the locking policies for the two bonding locks:
 221  *
 222  * 1) Get bond->lock when reading/writing slave list.
 223  * 2) Get bond->curr_slave_lock when reading/writing bond->curr_active_slave.
 224  *    (It is unnecessary when the write-lock is put with bond->lock.)
 225  * 3) When we lock with bond->curr_slave_lock, we must lock with bond->lock
 226  *    beforehand.
 227  */
 228 struct bonding {
 229         struct   net_device *dev; /* first - useful for panic debug */
 230         struct   list_head slave_list;
 231         struct   slave *curr_active_slave;
 232         struct   slave *current_arp_slave;
 233         struct   slave *primary_slave;
 234         bool     force_primary;
 235         s32      slave_cnt; /* never change this value outside the attach/detach wrappers */
 236         int     (*recv_probe)(const struct sk_buff *, struct bonding *,
 237                               struct slave *);
 238         rwlock_t lock;
 239         rwlock_t curr_slave_lock;
 240         u8       send_peer_notif;
 241         u8       igmp_retrans;
 242 #ifdef CONFIG_PROC_FS
 243         struct   proc_dir_entry *proc_entry;
 244         char     proc_file_name[IFNAMSIZ];
 245 #endif /* CONFIG_PROC_FS */
 246         struct   list_head bond_list;
 247         int      (*xmit_hash_policy)(struct sk_buff *, int);
 248         u16      rr_tx_counter;
 249         struct   ad_bond_info ad_info;
 250         struct   alb_bond_info alb_info;
 251         struct   bond_params params;
 252         struct   workqueue_struct *wq;
 253         struct   delayed_work mii_work;
 254         struct   delayed_work arp_work;
 255         struct   delayed_work alb_work;
 256         struct   delayed_work ad_work;
 257         struct   delayed_work mcast_work;
 258 #ifdef CONFIG_DEBUG_FS
 259         /* debugging support via debugfs */
 260         struct   dentry *debug_dir;
 261 #endif /* CONFIG_DEBUG_FS */
 262 };
 263
 264 #define bond_slave_get_rcu(dev) \
 265         ((struct slave *) rcu_dereference(dev->rx_handler_data))
 266
 267 #define bond_slave_get_rtnl(dev) \
 268         ((struct slave *) rtnl_dereference(dev->rx_handler_data))
 269
 270 /**
 271  * Returns NULL if the net_device does not belong to any of the bond's slaves
 272  *
 273  * Caller must hold bond lock for read
 274  */
 275 static inline struct slave *bond_get_slave_by_dev(struct bonding *bond,
 276                                                   struct net_device *slave_dev)
 277 {
 278         struct slave *slave = NULL;
 279
 280         bond_for_each_slave(bond, slave)
 281                 if (slave->dev == slave_dev)
 282                         return slave;
 283
 284         return NULL;
 285 }
 286
 287 static inline struct bonding *bond_get_bond_by_slave(struct slave *slave)
 288 {
 289         if (!slave || !slave->bond)
 290                 return NULL;
 291         return slave->bond;
 292 }
 293
 294 static inline bool bond_is_lb(const struct bonding *bond)
 295 {
 296         return (bond->params.mode == BOND_MODE_TLB ||
 297                 bond->params.mode == BOND_MODE_ALB);
 298 }
 299
 300 static inline void bond_set_active_slave(struct slave *slave)
 301 {
 302         slave->backup = 0;
 303 }
 304
 305 static inline void bond_set_backup_slave(struct slave *slave)
 306 {
 307         slave->backup = 1;
 308 }
 309
 310 static inline int bond_slave_state(struct slave *slave)
 311 {
 312         return slave->backup;
 313 }
 314
 315 static inline bool bond_is_active_slave(struct slave *slave)
 316 {
 317         return !bond_slave_state(slave);
 318 }
 319
 320 #define BOND_PRI_RESELECT_ALWAYS        0
 321 #define BOND_PRI_RESELECT_BETTER        1
 322 #define BOND_PRI_RESELECT_FAILURE       2
 323
 324 #define BOND_FOM_NONE                   0
 325 #define BOND_FOM_ACTIVE                 1
 326 #define BOND_FOM_FOLLOW                 2
 327
 328 #define BOND_ARP_TARGETS_ANY            0
 329 #define BOND_ARP_TARGETS_ALL            1
 330
 331 #define BOND_ARP_VALIDATE_NONE          0
 332 #define BOND_ARP_VALIDATE_ACTIVE        (1 << BOND_STATE_ACTIVE)
 333 #define BOND_ARP_VALIDATE_BACKUP        (1 << BOND_STATE_BACKUP)
 334 #define BOND_ARP_VALIDATE_ALL           (BOND_ARP_VALIDATE_ACTIVE | \
 335                                          BOND_ARP_VALIDATE_BACKUP)
 336
 337 static inline int slave_do_arp_validate(struct bonding *bond,
 338                                         struct slave *slave)
 339 {
 340         return bond->params.arp_validate & (1 << bond_slave_state(slave));
 341 }
 342
 343 /* Get the oldest arp which we've received on this slave for bond's
 344  * arp_targets.
 345  */
 346 static inline unsigned long slave_oldest_target_arp_rx(struct bonding *bond,
 347                                                        struct slave *slave)
 348 {
 349         int i = 1;
 350         unsigned long ret = slave->target_last_arp_rx[0];
 351
 352         for (; (i < BOND_MAX_ARP_TARGETS) && bond->params.arp_targets[i]; i++)
 353                 if (time_before(slave->target_last_arp_rx[i], ret))
 354                         ret = slave->target_last_arp_rx[i];
 355
 356         return ret;
 357 }
 358
 359 static inline unsigned long slave_last_rx(struct bonding *bond,
 360                                         struct slave *slave)
 361 {
 362         if (slave_do_arp_validate(bond, slave)) {
 363                 if (bond->params.arp_all_targets == BOND_ARP_TARGETS_ALL)
 364                         return slave_oldest_target_arp_rx(bond, slave);
 365                 else
 366                         return slave->last_arp_rx;
 367         }
 368
 369         return slave->dev->last_rx;
 370 }
 371
 372 #ifdef CONFIG_NET_POLL_CONTROLLER
 373 static inline void bond_netpoll_send_skb(const struct slave *slave,
 374                                          struct sk_buff *skb)
 375 {
 376         struct netpoll *np = slave->np;
 377
 378         if (np)
 379                 netpoll_send_skb(np, skb);
 380 }
 381 #else
 382 static inline void bond_netpoll_send_skb(const struct slave *slave,
 383                                          struct sk_buff *skb)
 384 {
 385 }
 386 #endif
 387
 388 static inline void bond_set_slave_inactive_flags(struct slave *slave)
 389 {
 390         if (!bond_is_lb(slave->bond))
 391                 bond_set_backup_slave(slave);
 392         if (!slave->bond->params.all_slaves_active)
 393                 slave->inactive = 1;
 394 }
 395
 396 static inline void bond_set_slave_active_flags(struct slave *slave)
 397 {
 398         bond_set_active_slave(slave);
 399         slave->inactive = 0;
 400 }
 401
 402 static inline bool bond_is_slave_inactive(struct slave *slave)
 403 {
 404         return slave->inactive;
 405 }
 406
 407 static inline __be32 bond_confirm_addr(struct net_device *dev, __be32 dst, __be32 local)
 408 {
 409         struct in_device *in_dev;
 410         __be32 addr = 0;
 411
 412         rcu_read_lock();
 413         in_dev = __in_dev_get_rcu(dev);
 414
 415         if (in_dev)
 416                 addr = inet_confirm_addr(in_dev, dst, local, RT_SCOPE_HOST);
 417
 418         rcu_read_unlock();
 419         return addr;
 420 }
 421
 422 static inline bool slave_can_tx(struct slave *slave)
 423 {
 424         if (IS_UP(slave->dev) && slave->link == BOND_LINK_UP &&
 425             bond_is_active_slave(slave))
 426                 return true;
 427         else
 428                 return false;
 429 }
 430
 431 struct bond_net;
 432
 433 struct vlan_entry *bond_next_vlan(struct bonding *bond, struct vlan_entry *curr);
 434 int bond_dev_queue_xmit(struct bonding *bond, struct sk_buff *skb, struct net_device *slave_dev);
 435 void bond_xmit_slave_id(struct bonding *bond, struct sk_buff *skb, int slave_id);
 436 int bond_create(struct net *net, const char *name);
 437 int bond_create_sysfs(struct bond_net *net);
 438 void bond_destroy_sysfs(struct bond_net *net);
 439 void bond_prepare_sysfs_group(struct bonding *bond);
 440 int bond_create_slave_symlinks(struct net_device *master, struct net_device *slave);
 441 void bond_destroy_slave_symlinks(struct net_device *master, struct net_device *slave);
 442 int bond_enslave(struct net_device *bond_dev, struct net_device *slave_dev);
 443 int bond_release(struct net_device *bond_dev, struct net_device *slave_dev);
 444 void bond_mii_monitor(struct work_struct *);
 445 void bond_loadbalance_arp_mon(struct work_struct *);
 446 void bond_activebackup_arp_mon(struct work_struct *);
 447 void bond_set_mode_ops(struct bonding *bond, int mode);
 448 int bond_parse_parm(const char *mode_arg, const struct bond_parm_tbl *tbl);
 449 void bond_select_active_slave(struct bonding *bond);
 450 void bond_change_active_slave(struct bonding *bond, struct slave *new_active);
 451 void bond_create_debugfs(void);
 452 void bond_destroy_debugfs(void);
 453 void bond_debug_register(struct bonding *bond);
 454 void bond_debug_unregister(struct bonding *bond);
 455 void bond_debug_reregister(struct bonding *bond);
 456 const char *bond_mode_name(int mode);
 457
 458 struct bond_net {
 459         struct net *            net;    /* Associated network namespace */
 460         struct list_head        dev_list;
 461 #ifdef CONFIG_PROC_FS
 462         struct proc_dir_entry * proc_dir;
 463 #endif
 464         struct class_attribute  class_attr_bonding_masters;
 465 };
 466
 467 #ifdef CONFIG_PROC_FS
 468 void bond_create_proc_entry(struct bonding *bond);
 469 void bond_remove_proc_entry(struct bonding *bond);
 470 void bond_create_proc_dir(struct bond_net *bn);
 471 void bond_destroy_proc_dir(struct bond_net *bn);
 472 #else
 473 static inline void bond_create_proc_entry(struct bonding *bond)
 474 {
 475 }
 476
 477 static inline void bond_remove_proc_entry(struct bonding *bond)
 478 {
 479 }
 480
 481 static inline void bond_create_proc_dir(struct bond_net *bn)
 482 {
 483 }
 484
 485 static inline void bond_destroy_proc_dir(struct bond_net *bn)
 486 {
 487 }
 488 #endif
 489
 490 static inline struct slave *bond_slave_has_mac(struct bonding *bond,
 491                                                const u8 *mac)
 492 {
 493         struct slave *tmp;
 494
 495         bond_for_each_slave(bond, tmp)
 496                 if (ether_addr_equal_64bits(mac, tmp->dev->dev_addr))
 497                         return tmp;
 498
 499         return NULL;
 500 }
 501
 502 /* Check if the ip is present in arp ip list, or first free slot if ip == 0
 503  * Returns -1 if not found, index if found
 504  */
 505 static inline int bond_get_targets_ip(__be32 *targets, __be32 ip)
 506 {
 507         int i;
 508
 509         for (i = 0; i < BOND_MAX_ARP_TARGETS; i++)
 510                 if (targets[i] == ip)
 511                         return i;
 512                 else if (targets[i] == 0)
 513                         break;
 514
 515         return -1;
 516 }
 517
 518 /* exported from bond_main.c */
 519 extern int bond_net_id;
 520 extern const struct bond_parm_tbl bond_lacp_tbl[];
 521 extern const struct bond_parm_tbl bond_mode_tbl[];
 522 extern const struct bond_parm_tbl xmit_hashtype_tbl[];
 523 extern const struct bond_parm_tbl arp_validate_tbl[];
 524 extern const struct bond_parm_tbl arp_all_targets_tbl[];
 525 extern const struct bond_parm_tbl fail_over_mac_tbl[];
 526 extern const struct bond_parm_tbl pri_reselect_tbl[];
 527 extern struct bond_parm_tbl ad_select_tbl[];
 528
 529 #endif /* _LINUX_BONDING_H */