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 /**
 120  * bond_for_each_slave_reverse - iterate in reverse from a given position
 121  * @bond:       the bond holding this list
 122  * @pos:        slave to continue from
 123  *
 124  * Caller must hold bond->lock
 125  */
 126 #define bond_for_each_slave_continue_reverse(bond, pos) \
 127         list_for_each_entry_continue_reverse(pos, &(bond)->slave_list, list)
 128
 129 #ifdef CONFIG_NET_POLL_CONTROLLER
 130 extern atomic_t netpoll_block_tx;
 131
 132 static inline void block_netpoll_tx(void)
 133 {
 134         atomic_inc(&netpoll_block_tx);
 135 }
 136
 137 static inline void unblock_netpoll_tx(void)
 138 {
 139         atomic_dec(&netpoll_block_tx);
 140 }
 141
 142 static inline int is_netpoll_tx_blocked(struct net_device *dev)
 143 {
 144         if (unlikely(netpoll_tx_running(dev)))
 145                 return atomic_read(&netpoll_block_tx);
 146         return 0;
 147 }
 148 #else
 149 #define block_netpoll_tx()
 150 #define unblock_netpoll_tx()
 151 #define is_netpoll_tx_blocked(dev) (0)
 152 #endif
 153
 154 struct bond_params {
 155         int mode;
 156         int xmit_policy;
 157         int miimon;
 158         u8 num_peer_notif;
 159         int arp_interval;
 160         int arp_validate;
 161         int arp_all_targets;
 162         int use_carrier;
 163         int fail_over_mac;
 164         int updelay;
 165         int downdelay;
 166         int lacp_fast;
 167         unsigned int min_links;
 168         int ad_select;
 169         char primary[IFNAMSIZ];
 170         int primary_reselect;
 171         __be32 arp_targets[BOND_MAX_ARP_TARGETS];
 172         int tx_queues;
 173         int all_slaves_active;
 174         int resend_igmp;
 175 };
 176
 177 struct bond_parm_tbl {
 178         char *modename;
 179         int mode;
 180 };
 181
 182 #define BOND_MAX_MODENAME_LEN 20
 183
 184 struct vlan_entry {
 185         struct list_head vlan_list;
 186         unsigned short vlan_id;
 187 };
 188
 189 struct slave {
 190         struct net_device *dev; /* first - useful for panic debug */
 191         struct list_head list;
 192         struct bonding *bond; /* our master */
 193         int    delay;
 194         unsigned long jiffies;
 195         unsigned long last_arp_rx;
 196         unsigned long target_last_arp_rx[BOND_MAX_ARP_TARGETS];
 197         s8     link;    /* one of BOND_LINK_XXXX */
 198         s8     new_link;
 199         u8     backup:1,   /* indicates backup slave. Value corresponds with
 200                               BOND_STATE_ACTIVE and BOND_STATE_BACKUP */
 201                inactive:1; /* indicates inactive slave */
 202         u8     duplex;
 203         u32    original_mtu;
 204         u32    link_failure_count;
 205         u32    speed;
 206         u16    queue_id;
 207         u8     perm_hwaddr[ETH_ALEN];
 208         struct ad_slave_info ad_info; /* HUGE - better to dynamically alloc */
 209         struct tlb_slave_info tlb_info;
 210 #ifdef CONFIG_NET_POLL_CONTROLLER
 211         struct netpoll *np;
 212 #endif
 213 };
 214
 215 /*
 216  * Link pseudo-state only used internally by monitors
 217  */
 218 #define BOND_LINK_NOCHANGE -1
 219
 220 /*
 221  * Here are the locking policies for the two bonding locks:
 222  *
 223  * 1) Get bond->lock when reading/writing slave list.
 224  * 2) Get bond->curr_slave_lock when reading/writing bond->curr_active_slave.
 225  *    (It is unnecessary when the write-lock is put with bond->lock.)
 226  * 3) When we lock with bond->curr_slave_lock, we must lock with bond->lock
 227  *    beforehand.
 228  */
 229 struct bonding {
 230         struct   net_device *dev; /* first - useful for panic debug */
 231         struct   list_head slave_list;
 232         struct   slave *curr_active_slave;
 233         struct   slave *current_arp_slave;
 234         struct   slave *primary_slave;
 235         bool     force_primary;
 236         s32      slave_cnt; /* never change this value outside the attach/detach wrappers */
 237         int     (*recv_probe)(const struct sk_buff *, struct bonding *,
 238                               struct slave *);
 239         rwlock_t lock;
 240         rwlock_t curr_slave_lock;
 241         u8       send_peer_notif;
 242         u8       igmp_retrans;
 243 #ifdef CONFIG_PROC_FS
 244         struct   proc_dir_entry *proc_entry;
 245         char     proc_file_name[IFNAMSIZ];
 246 #endif /* CONFIG_PROC_FS */
 247         struct   list_head bond_list;
 248         int      (*xmit_hash_policy)(struct sk_buff *, int);
 249         u16      rr_tx_counter;
 250         struct   ad_bond_info ad_info;
 251         struct   alb_bond_info alb_info;
 252         struct   bond_params params;
 253         struct   list_head vlan_list;
 254         struct   workqueue_struct *wq;
 255         struct   delayed_work mii_work;
 256         struct   delayed_work arp_work;
 257         struct   delayed_work alb_work;
 258         struct   delayed_work ad_work;
 259         struct   delayed_work mcast_work;
 260 #ifdef CONFIG_DEBUG_FS
 261         /* debugging support via debugfs */
 262         struct   dentry *debug_dir;
 263 #endif /* CONFIG_DEBUG_FS */
 264 };
 265
 266 static inline bool bond_vlan_used(struct bonding *bond)
 267 {
 268         return !list_empty(&bond->vlan_list);
 269 }
 270
 271 #define bond_slave_get_rcu(dev) \
 272         ((struct slave *) rcu_dereference(dev->rx_handler_data))
 273
 274 #define bond_slave_get_rtnl(dev) \
 275         ((struct slave *) rtnl_dereference(dev->rx_handler_data))
 276
 277 /**
 278  * Returns NULL if the net_device does not belong to any of the bond's slaves
 279  *
 280  * Caller must hold bond lock for read
 281  */
 282 static inline struct slave *bond_get_slave_by_dev(struct bonding *bond,
 283                                                   struct net_device *slave_dev)
 284 {
 285         struct slave *slave = NULL;
 286
 287         bond_for_each_slave(bond, slave)
 288                 if (slave->dev == slave_dev)
 289                         return slave;
 290
 291         return NULL;
 292 }
 293
 294 static inline struct bonding *bond_get_bond_by_slave(struct slave *slave)
 295 {
 296         if (!slave || !slave->bond)
 297                 return NULL;
 298         return slave->bond;
 299 }
 300
 301 static inline bool bond_is_lb(const struct bonding *bond)
 302 {
 303         return (bond->params.mode == BOND_MODE_TLB ||
 304                 bond->params.mode == BOND_MODE_ALB);
 305 }
 306
 307 static inline void bond_set_active_slave(struct slave *slave)
 308 {
 309         slave->backup = 0;
 310 }
 311
 312 static inline void bond_set_backup_slave(struct slave *slave)
 313 {
 314         slave->backup = 1;
 315 }
 316
 317 static inline int bond_slave_state(struct slave *slave)
 318 {
 319         return slave->backup;
 320 }
 321
 322 static inline bool bond_is_active_slave(struct slave *slave)
 323 {
 324         return !bond_slave_state(slave);
 325 }
 326
 327 #define BOND_PRI_RESELECT_ALWAYS        0
 328 #define BOND_PRI_RESELECT_BETTER        1
 329 #define BOND_PRI_RESELECT_FAILURE       2
 330
 331 #define BOND_FOM_NONE                   0
 332 #define BOND_FOM_ACTIVE                 1
 333 #define BOND_FOM_FOLLOW                 2
 334
 335 #define BOND_ARP_TARGETS_ANY            0
 336 #define BOND_ARP_TARGETS_ALL            1
 337
 338 #define BOND_ARP_VALIDATE_NONE          0
 339 #define BOND_ARP_VALIDATE_ACTIVE        (1 << BOND_STATE_ACTIVE)
 340 #define BOND_ARP_VALIDATE_BACKUP        (1 << BOND_STATE_BACKUP)
 341 #define BOND_ARP_VALIDATE_ALL           (BOND_ARP_VALIDATE_ACTIVE | \
 342                                          BOND_ARP_VALIDATE_BACKUP)
 343
 344 static inline int slave_do_arp_validate(struct bonding *bond,
 345                                         struct slave *slave)
 346 {
 347         return bond->params.arp_validate & (1 << bond_slave_state(slave));
 348 }
 349
 350 /* Get the oldest arp which we've received on this slave for bond's
 351  * arp_targets.
 352  */
 353 static inline unsigned long slave_oldest_target_arp_rx(struct bonding *bond,
 354                                                        struct slave *slave)
 355 {
 356         int i = 1;
 357         unsigned long ret = slave->target_last_arp_rx[0];
 358
 359         for (; (i < BOND_MAX_ARP_TARGETS) && bond->params.arp_targets[i]; i++)
 360                 if (time_before(slave->target_last_arp_rx[i], ret))
 361                         ret = slave->target_last_arp_rx[i];
 362
 363         return ret;
 364 }
 365
 366 static inline unsigned long slave_last_rx(struct bonding *bond,
 367                                         struct slave *slave)
 368 {
 369         if (slave_do_arp_validate(bond, slave)) {
 370                 if (bond->params.arp_all_targets == BOND_ARP_TARGETS_ALL)
 371                         return slave_oldest_target_arp_rx(bond, slave);
 372                 else
 373                         return slave->last_arp_rx;
 374         }
 375
 376         return slave->dev->last_rx;
 377 }
 378
 379 #ifdef CONFIG_NET_POLL_CONTROLLER
 380 static inline void bond_netpoll_send_skb(const struct slave *slave,
 381                                          struct sk_buff *skb)
 382 {
 383         struct netpoll *np = slave->np;
 384
 385         if (np)
 386                 netpoll_send_skb(np, skb);
 387 }
 388 #else
 389 static inline void bond_netpoll_send_skb(const struct slave *slave,
 390                                          struct sk_buff *skb)
 391 {
 392 }
 393 #endif
 394
 395 static inline void bond_set_slave_inactive_flags(struct slave *slave)
 396 {
 397         if (!bond_is_lb(slave->bond))
 398                 bond_set_backup_slave(slave);
 399         if (!slave->bond->params.all_slaves_active)
 400                 slave->inactive = 1;
 401 }
 402
 403 static inline void bond_set_slave_active_flags(struct slave *slave)
 404 {
 405         bond_set_active_slave(slave);
 406         slave->inactive = 0;
 407 }
 408
 409 static inline bool bond_is_slave_inactive(struct slave *slave)
 410 {
 411         return slave->inactive;
 412 }
 413
 414 static inline __be32 bond_confirm_addr(struct net_device *dev, __be32 dst, __be32 local)
 415 {
 416         struct in_device *in_dev;
 417         __be32 addr = 0;
 418
 419         rcu_read_lock();
 420         in_dev = __in_dev_get_rcu(dev);
 421
 422         if (in_dev)
 423                 addr = inet_confirm_addr(in_dev, dst, local, RT_SCOPE_HOST);
 424
 425         rcu_read_unlock();
 426         return addr;
 427 }
 428
 429 static inline bool slave_can_tx(struct slave *slave)
 430 {
 431         if (IS_UP(slave->dev) && slave->link == BOND_LINK_UP &&
 432             bond_is_active_slave(slave))
 433                 return true;
 434         else
 435                 return false;
 436 }
 437
 438 struct bond_net;
 439
 440 struct vlan_entry *bond_next_vlan(struct bonding *bond, struct vlan_entry *curr);
 441 int bond_dev_queue_xmit(struct bonding *bond, struct sk_buff *skb, struct net_device *slave_dev);
 442 void bond_xmit_slave_id(struct bonding *bond, struct sk_buff *skb, int slave_id);
 443 int bond_create(struct net *net, const char *name);
 444 int bond_create_sysfs(struct bond_net *net);
 445 void bond_destroy_sysfs(struct bond_net *net);
 446 void bond_prepare_sysfs_group(struct bonding *bond);
 447 int bond_create_slave_symlinks(struct net_device *master, struct net_device *slave);
 448 void bond_destroy_slave_symlinks(struct net_device *master, struct net_device *slave);
 449 int bond_enslave(struct net_device *bond_dev, struct net_device *slave_dev);
 450 int bond_release(struct net_device *bond_dev, struct net_device *slave_dev);
 451 void bond_mii_monitor(struct work_struct *);
 452 void bond_loadbalance_arp_mon(struct work_struct *);
 453 void bond_activebackup_arp_mon(struct work_struct *);
 454 void bond_set_mode_ops(struct bonding *bond, int mode);
 455 int bond_parse_parm(const char *mode_arg, const struct bond_parm_tbl *tbl);
 456 void bond_select_active_slave(struct bonding *bond);
 457 void bond_change_active_slave(struct bonding *bond, struct slave *new_active);
 458 void bond_create_debugfs(void);
 459 void bond_destroy_debugfs(void);
 460 void bond_debug_register(struct bonding *bond);
 461 void bond_debug_unregister(struct bonding *bond);
 462 void bond_debug_reregister(struct bonding *bond);
 463 const char *bond_mode_name(int mode);
 464
 465 struct bond_net {
 466         struct net *            net;    /* Associated network namespace */
 467         struct list_head        dev_list;
 468 #ifdef CONFIG_PROC_FS
 469         struct proc_dir_entry * proc_dir;
 470 #endif
 471         struct class_attribute  class_attr_bonding_masters;
 472 };
 473
 474 #ifdef CONFIG_PROC_FS
 475 void bond_create_proc_entry(struct bonding *bond);
 476 void bond_remove_proc_entry(struct bonding *bond);
 477 void bond_create_proc_dir(struct bond_net *bn);
 478 void bond_destroy_proc_dir(struct bond_net *bn);
 479 #else
 480 static inline void bond_create_proc_entry(struct bonding *bond)
 481 {
 482 }
 483
 484 static inline void bond_remove_proc_entry(struct bonding *bond)
 485 {
 486 }
 487
 488 static inline void bond_create_proc_dir(struct bond_net *bn)
 489 {
 490 }
 491
 492 static inline void bond_destroy_proc_dir(struct bond_net *bn)
 493 {
 494 }
 495 #endif
 496
 497 static inline struct slave *bond_slave_has_mac(struct bonding *bond,
 498                                                const u8 *mac)
 499 {
 500         struct slave *tmp;
 501
 502         bond_for_each_slave(bond, tmp)
 503                 if (ether_addr_equal_64bits(mac, tmp->dev->dev_addr))
 504                         return tmp;
 505
 506         return NULL;
 507 }
 508
 509 /* Check if the ip is present in arp ip list, or first free slot if ip == 0
 510  * Returns -1 if not found, index if found
 511  */
 512 static inline int bond_get_targets_ip(__be32 *targets, __be32 ip)
 513 {
 514         int i;
 515
 516         for (i = 0; i < BOND_MAX_ARP_TARGETS; i++)
 517                 if (targets[i] == ip)
 518                         return i;
 519                 else if (targets[i] == 0)
 520                         break;
 521
 522         return -1;
 523 }
 524
 525 /* exported from bond_main.c */
 526 extern int bond_net_id;
 527 extern const struct bond_parm_tbl bond_lacp_tbl[];
 528 extern const struct bond_parm_tbl bond_mode_tbl[];
 529 extern const struct bond_parm_tbl xmit_hashtype_tbl[];
 530 extern const struct bond_parm_tbl arp_validate_tbl[];
 531 extern const struct bond_parm_tbl arp_all_targets_tbl[];
 532 extern const struct bond_parm_tbl fail_over_mac_tbl[];
 533 extern const struct bond_parm_tbl pri_reselect_tbl[];
 534 extern struct bond_parm_tbl ad_select_tbl[];
 535
 536 #endif /* _LINUX_BONDING_H */