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