tjh.dev
Linux kernel mirror (for testing)
git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git
kernel
os
linux
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 BOND_DEFAULT_MIIMON 100
39
40#define IS_UP(dev) \
41 ((((dev)->flags & IFF_UP) == IFF_UP) && \
42 netif_running(dev) && \
43 netif_carrier_ok(dev))
44
45/*
46 * Checks whether slave is ready for transmit.
47 */
48#define SLAVE_IS_OK(slave) \
49 (((slave)->dev->flags & IFF_UP) && \
50 netif_running((slave)->dev) && \
51 ((slave)->link == BOND_LINK_UP) && \
52 bond_is_active_slave(slave))
53
54
55#define USES_PRIMARY(mode) \
56 (((mode) == BOND_MODE_ACTIVEBACKUP) || \
57 ((mode) == BOND_MODE_TLB) || \
58 ((mode) == BOND_MODE_ALB))
59
60#define BOND_NO_USES_ARP(mode) \
61 (((mode) == BOND_MODE_8023AD) || \
62 ((mode) == BOND_MODE_TLB) || \
63 ((mode) == BOND_MODE_ALB))
64
65#define TX_QUEUE_OVERRIDE(mode) \
66 (((mode) == BOND_MODE_ACTIVEBACKUP) || \
67 ((mode) == BOND_MODE_ROUNDROBIN))
68
69#define BOND_MODE_IS_LB(mode) \
70 (((mode) == BOND_MODE_TLB) || \
71 ((mode) == BOND_MODE_ALB))
72
73#define IS_IP_TARGET_UNUSABLE_ADDRESS(a) \
74 ((htonl(INADDR_BROADCAST) == a) || \
75 ipv4_is_zeronet(a))
76/*
77 * Less bad way to call ioctl from within the kernel; this needs to be
78 * done some other way to get the call out of interrupt context.
79 * Needs "ioctl" variable to be supplied by calling context.
80 */
81#define IOCTL(dev, arg, cmd) ({ \
82 int res = 0; \
83 mm_segment_t fs = get_fs(); \
84 set_fs(get_ds()); \
85 res = ioctl(dev, arg, cmd); \
86 set_fs(fs); \
87 res; })
88
89/* slave list primitives */
90#define bond_slave_list(bond) (&(bond)->dev->adj_list.lower)
91
92#define bond_has_slaves(bond) !list_empty(bond_slave_list(bond))
93
94/* IMPORTANT: bond_first/last_slave can return NULL in case of an empty list */
95#define bond_first_slave(bond) \
96 (bond_has_slaves(bond) ? \
97 netdev_adjacent_get_private(bond_slave_list(bond)->next) : \
98 NULL)
99#define bond_last_slave(bond) \
100 (bond_has_slaves(bond) ? \
101 netdev_adjacent_get_private(bond_slave_list(bond)->prev) : \
102 NULL)
103
104#define bond_is_first_slave(bond, pos) (pos == bond_first_slave(bond))
105#define bond_is_last_slave(bond, pos) (pos == bond_last_slave(bond))
106
107/**
108 * bond_for_each_slave - iterate over all slaves
109 * @bond: the bond holding this list
110 * @pos: current slave
111 * @iter: list_head * iterator
112 *
113 * Caller must hold bond->lock
114 */
115#define bond_for_each_slave(bond, pos, iter) \
116 netdev_for_each_lower_private((bond)->dev, pos, iter)
117
118/* Caller must have rcu_read_lock */
119#define bond_for_each_slave_rcu(bond, pos, iter) \
120 netdev_for_each_lower_private_rcu((bond)->dev, pos, iter)
121
122#ifdef CONFIG_NET_POLL_CONTROLLER
123extern atomic_t netpoll_block_tx;
124
125static inline void block_netpoll_tx(void)
126{
127 atomic_inc(&netpoll_block_tx);
128}
129
130static inline void unblock_netpoll_tx(void)
131{
132 atomic_dec(&netpoll_block_tx);
133}
134
135static inline int is_netpoll_tx_blocked(struct net_device *dev)
136{
137 if (unlikely(netpoll_tx_running(dev)))
138 return atomic_read(&netpoll_block_tx);
139 return 0;
140}
141#else
142#define block_netpoll_tx()
143#define unblock_netpoll_tx()
144#define is_netpoll_tx_blocked(dev) (0)
145#endif
146
147struct bond_params {
148 int mode;
149 int xmit_policy;
150 int miimon;
151 u8 num_peer_notif;
152 int arp_interval;
153 int arp_validate;
154 int arp_all_targets;
155 int use_carrier;
156 int fail_over_mac;
157 int updelay;
158 int downdelay;
159 int lacp_fast;
160 unsigned int min_links;
161 int ad_select;
162 char primary[IFNAMSIZ];
163 int primary_reselect;
164 __be32 arp_targets[BOND_MAX_ARP_TARGETS];
165 int tx_queues;
166 int all_slaves_active;
167 int resend_igmp;
168 int lp_interval;
169 int packets_per_slave;
170};
171
172struct bond_parm_tbl {
173 char *modename;
174 int mode;
175};
176
177#define BOND_MAX_MODENAME_LEN 20
178
179struct slave {
180 struct net_device *dev; /* first - useful for panic debug */
181 struct bonding *bond; /* our master */
182 int delay;
183 unsigned long jiffies;
184 unsigned long last_arp_rx;
185 unsigned long target_last_arp_rx[BOND_MAX_ARP_TARGETS];
186 s8 link; /* one of BOND_LINK_XXXX */
187 s8 new_link;
188 u8 backup:1, /* indicates backup slave. Value corresponds with
189 BOND_STATE_ACTIVE and BOND_STATE_BACKUP */
190 inactive:1; /* indicates inactive slave */
191 u8 duplex;
192 u32 original_mtu;
193 u32 link_failure_count;
194 u32 speed;
195 u16 queue_id;
196 u8 perm_hwaddr[ETH_ALEN];
197 struct ad_slave_info ad_info; /* HUGE - better to dynamically alloc */
198 struct tlb_slave_info tlb_info;
199#ifdef CONFIG_NET_POLL_CONTROLLER
200 struct netpoll *np;
201#endif
202};
203
204/*
205 * Link pseudo-state only used internally by monitors
206 */
207#define BOND_LINK_NOCHANGE -1
208
209/*
210 * Here are the locking policies for the two bonding locks:
211 *
212 * 1) Get bond->lock when reading/writing slave list.
213 * 2) Get bond->curr_slave_lock when reading/writing bond->curr_active_slave.
214 * (It is unnecessary when the write-lock is put with bond->lock.)
215 * 3) When we lock with bond->curr_slave_lock, we must lock with bond->lock
216 * beforehand.
217 */
218struct bonding {
219 struct net_device *dev; /* first - useful for panic debug */
220 struct slave *curr_active_slave;
221 struct slave *current_arp_slave;
222 struct slave *primary_slave;
223 bool force_primary;
224 s32 slave_cnt; /* never change this value outside the attach/detach wrappers */
225 int (*recv_probe)(const struct sk_buff *, struct bonding *,
226 struct slave *);
227 rwlock_t lock;
228 rwlock_t curr_slave_lock;
229 u8 send_peer_notif;
230 u8 igmp_retrans;
231#ifdef CONFIG_PROC_FS
232 struct proc_dir_entry *proc_entry;
233 char proc_file_name[IFNAMSIZ];
234#endif /* CONFIG_PROC_FS */
235 struct list_head bond_list;
236 u32 rr_tx_counter;
237 struct ad_bond_info ad_info;
238 struct alb_bond_info alb_info;
239 struct bond_params params;
240 struct workqueue_struct *wq;
241 struct delayed_work mii_work;
242 struct delayed_work arp_work;
243 struct delayed_work alb_work;
244 struct delayed_work ad_work;
245 struct delayed_work mcast_work;
246#ifdef CONFIG_DEBUG_FS
247 /* debugging support via debugfs */
248 struct dentry *debug_dir;
249#endif /* CONFIG_DEBUG_FS */
250};
251
252#define bond_slave_get_rcu(dev) \
253 ((struct slave *) rcu_dereference(dev->rx_handler_data))
254
255#define bond_slave_get_rtnl(dev) \
256 ((struct slave *) rtnl_dereference(dev->rx_handler_data))
257
258/**
259 * Returns NULL if the net_device does not belong to any of the bond's slaves
260 *
261 * Caller must hold bond lock for read
262 */
263static inline struct slave *bond_get_slave_by_dev(struct bonding *bond,
264 struct net_device *slave_dev)
265{
266 return netdev_lower_dev_get_private(bond->dev, slave_dev);
267}
268
269static inline struct bonding *bond_get_bond_by_slave(struct slave *slave)
270{
271 if (!slave || !slave->bond)
272 return NULL;
273 return slave->bond;
274}
275
276static inline bool bond_is_lb(const struct bonding *bond)
277{
278 return BOND_MODE_IS_LB(bond->params.mode);
279}
280
281static inline void bond_set_active_slave(struct slave *slave)
282{
283 slave->backup = 0;
284}
285
286static inline void bond_set_backup_slave(struct slave *slave)
287{
288 slave->backup = 1;
289}
290
291static inline int bond_slave_state(struct slave *slave)
292{
293 return slave->backup;
294}
295
296static inline bool bond_is_active_slave(struct slave *slave)
297{
298 return !bond_slave_state(slave);
299}
300
301#define BOND_PRI_RESELECT_ALWAYS 0
302#define BOND_PRI_RESELECT_BETTER 1
303#define BOND_PRI_RESELECT_FAILURE 2
304
305#define BOND_FOM_NONE 0
306#define BOND_FOM_ACTIVE 1
307#define BOND_FOM_FOLLOW 2
308
309#define BOND_ARP_TARGETS_ANY 0
310#define BOND_ARP_TARGETS_ALL 1
311
312#define BOND_ARP_VALIDATE_NONE 0
313#define BOND_ARP_VALIDATE_ACTIVE (1 << BOND_STATE_ACTIVE)
314#define BOND_ARP_VALIDATE_BACKUP (1 << BOND_STATE_BACKUP)
315#define BOND_ARP_VALIDATE_ALL (BOND_ARP_VALIDATE_ACTIVE | \
316 BOND_ARP_VALIDATE_BACKUP)
317
318static inline int slave_do_arp_validate(struct bonding *bond,
319 struct slave *slave)
320{
321 return bond->params.arp_validate & (1 << bond_slave_state(slave));
322}
323
324/* Get the oldest arp which we've received on this slave for bond's
325 * arp_targets.
326 */
327static inline unsigned long slave_oldest_target_arp_rx(struct bonding *bond,
328 struct slave *slave)
329{
330 int i = 1;
331 unsigned long ret = slave->target_last_arp_rx[0];
332
333 for (; (i < BOND_MAX_ARP_TARGETS) && bond->params.arp_targets[i]; i++)
334 if (time_before(slave->target_last_arp_rx[i], ret))
335 ret = slave->target_last_arp_rx[i];
336
337 return ret;
338}
339
340static inline unsigned long slave_last_rx(struct bonding *bond,
341 struct slave *slave)
342{
343 if (slave_do_arp_validate(bond, slave)) {
344 if (bond->params.arp_all_targets == BOND_ARP_TARGETS_ALL)
345 return slave_oldest_target_arp_rx(bond, slave);
346 else
347 return slave->last_arp_rx;
348 }
349
350 return slave->dev->last_rx;
351}
352
353#ifdef CONFIG_NET_POLL_CONTROLLER
354static inline void bond_netpoll_send_skb(const struct slave *slave,
355 struct sk_buff *skb)
356{
357 struct netpoll *np = slave->np;
358
359 if (np)
360 netpoll_send_skb(np, skb);
361}
362#else
363static inline void bond_netpoll_send_skb(const struct slave *slave,
364 struct sk_buff *skb)
365{
366}
367#endif
368
369static inline void bond_set_slave_inactive_flags(struct slave *slave)
370{
371 if (!bond_is_lb(slave->bond))
372 bond_set_backup_slave(slave);
373 if (!slave->bond->params.all_slaves_active)
374 slave->inactive = 1;
375}
376
377static inline void bond_set_slave_active_flags(struct slave *slave)
378{
379 bond_set_active_slave(slave);
380 slave->inactive = 0;
381}
382
383static inline bool bond_is_slave_inactive(struct slave *slave)
384{
385 return slave->inactive;
386}
387
388static inline __be32 bond_confirm_addr(struct net_device *dev, __be32 dst, __be32 local)
389{
390 struct in_device *in_dev;
391 __be32 addr = 0;
392
393 rcu_read_lock();
394 in_dev = __in_dev_get_rcu(dev);
395
396 if (in_dev)
397 addr = inet_confirm_addr(in_dev, dst, local, RT_SCOPE_HOST);
398
399 rcu_read_unlock();
400 return addr;
401}
402
403static inline bool slave_can_tx(struct slave *slave)
404{
405 if (IS_UP(slave->dev) && slave->link == BOND_LINK_UP &&
406 bond_is_active_slave(slave))
407 return true;
408 else
409 return false;
410}
411
412struct bond_net;
413
414int bond_arp_rcv(const struct sk_buff *skb, struct bonding *bond, struct slave *slave);
415int bond_dev_queue_xmit(struct bonding *bond, struct sk_buff *skb, struct net_device *slave_dev);
416void bond_xmit_slave_id(struct bonding *bond, struct sk_buff *skb, int slave_id);
417int bond_create(struct net *net, const char *name);
418int bond_create_sysfs(struct bond_net *net);
419void bond_destroy_sysfs(struct bond_net *net);
420void bond_prepare_sysfs_group(struct bonding *bond);
421int bond_enslave(struct net_device *bond_dev, struct net_device *slave_dev);
422int bond_release(struct net_device *bond_dev, struct net_device *slave_dev);
423void bond_mii_monitor(struct work_struct *);
424void bond_loadbalance_arp_mon(struct work_struct *);
425void bond_activebackup_arp_mon(struct work_struct *);
426int bond_xmit_hash(struct bonding *bond, struct sk_buff *skb, int count);
427int bond_parse_parm(const char *mode_arg, const struct bond_parm_tbl *tbl);
428void bond_select_active_slave(struct bonding *bond);
429void bond_change_active_slave(struct bonding *bond, struct slave *new_active);
430void bond_create_debugfs(void);
431void bond_destroy_debugfs(void);
432void bond_debug_register(struct bonding *bond);
433void bond_debug_unregister(struct bonding *bond);
434void bond_debug_reregister(struct bonding *bond);
435const char *bond_mode_name(int mode);
436void bond_setup(struct net_device *bond_dev);
437unsigned int bond_get_num_tx_queues(void);
438int bond_netlink_init(void);
439void bond_netlink_fini(void);
440int bond_option_mode_set(struct bonding *bond, int mode);
441int bond_option_active_slave_set(struct bonding *bond, struct net_device *slave_dev);
442struct net_device *bond_option_active_slave_get_rcu(struct bonding *bond);
443struct net_device *bond_option_active_slave_get(struct bonding *bond);
444
445struct bond_net {
446 struct net * net; /* Associated network namespace */
447 struct list_head dev_list;
448#ifdef CONFIG_PROC_FS
449 struct proc_dir_entry * proc_dir;
450#endif
451 struct class_attribute class_attr_bonding_masters;
452};
453
454#ifdef CONFIG_PROC_FS
455void bond_create_proc_entry(struct bonding *bond);
456void bond_remove_proc_entry(struct bonding *bond);
457void bond_create_proc_dir(struct bond_net *bn);
458void bond_destroy_proc_dir(struct bond_net *bn);
459#else
460static inline void bond_create_proc_entry(struct bonding *bond)
461{
462}
463
464static inline void bond_remove_proc_entry(struct bonding *bond)
465{
466}
467
468static inline void bond_create_proc_dir(struct bond_net *bn)
469{
470}
471
472static inline void bond_destroy_proc_dir(struct bond_net *bn)
473{
474}
475#endif
476
477static inline struct slave *bond_slave_has_mac(struct bonding *bond,
478 const u8 *mac)
479{
480 struct list_head *iter;
481 struct slave *tmp;
482
483 bond_for_each_slave(bond, tmp, iter)
484 if (ether_addr_equal_64bits(mac, tmp->dev->dev_addr))
485 return tmp;
486
487 return NULL;
488}
489
490/* Caller must hold rcu_read_lock() for read */
491static inline struct slave *bond_slave_has_mac_rcu(struct bonding *bond,
492 const u8 *mac)
493{
494 struct list_head *iter;
495 struct slave *tmp;
496
497 bond_for_each_slave_rcu(bond, tmp, iter)
498 if (ether_addr_equal_64bits(mac, tmp->dev->dev_addr))
499 return tmp;
500
501 return NULL;
502}
503
504/* Check if the ip is present in arp ip list, or first free slot if ip == 0
505 * Returns -1 if not found, index if found
506 */
507static inline int bond_get_targets_ip(__be32 *targets, __be32 ip)
508{
509 int i;
510
511 for (i = 0; i < BOND_MAX_ARP_TARGETS; i++)
512 if (targets[i] == ip)
513 return i;
514 else if (targets[i] == 0)
515 break;
516
517 return -1;
518}
519
520/* exported from bond_main.c */
521extern int bond_net_id;
522extern const struct bond_parm_tbl bond_lacp_tbl[];
523extern const struct bond_parm_tbl bond_mode_tbl[];
524extern const struct bond_parm_tbl xmit_hashtype_tbl[];
525extern const struct bond_parm_tbl arp_validate_tbl[];
526extern const struct bond_parm_tbl arp_all_targets_tbl[];
527extern const struct bond_parm_tbl fail_over_mac_tbl[];
528extern const struct bond_parm_tbl pri_reselect_tbl[];
529extern struct bond_parm_tbl ad_select_tbl[];
530
531/* exported from bond_netlink.c */
532extern struct rtnl_link_ops bond_link_ops;
533
534#endif /* _LINUX_BONDING_H */