]> bbs.cooldavid.org Git - net-next-2.6.git/blame - include/net/sch_generic.h
net: cleanup include/net
[net-next-2.6.git] / include / net / sch_generic.h
CommitLineData
1da177e4
LT
1#ifndef __NET_SCHED_GENERIC_H
2#define __NET_SCHED_GENERIC_H
3
1da177e4
LT
4#include <linux/netdevice.h>
5#include <linux/types.h>
6#include <linux/rcupdate.h>
7#include <linux/module.h>
1da177e4
LT
8#include <linux/pkt_sched.h>
9#include <linux/pkt_cls.h>
10#include <net/gen_stats.h>
be577ddc 11#include <net/rtnetlink.h>
1da177e4
LT
12
13struct Qdisc_ops;
14struct qdisc_walker;
15struct tcf_walker;
16struct module;
17
fd2c3ef7 18struct qdisc_rate_table {
1da177e4
LT
19 struct tc_ratespec rate;
20 u32 data[256];
21 struct qdisc_rate_table *next;
22 int refcnt;
23};
24
fd2c3ef7 25enum qdisc_state_t {
e2627c8c 26 __QDISC_STATE_RUNNING,
37437bb2 27 __QDISC_STATE_SCHED,
a9312ae8 28 __QDISC_STATE_DEACTIVATED,
e2627c8c
DM
29};
30
175f9c1b
JK
31struct qdisc_size_table {
32 struct list_head list;
33 struct tc_sizespec szopts;
34 int refcnt;
35 u16 data[];
36};
37
fd2c3ef7 38struct Qdisc {
1da177e4
LT
39 int (*enqueue)(struct sk_buff *skb, struct Qdisc *dev);
40 struct sk_buff * (*dequeue)(struct Qdisc *dev);
41 unsigned flags;
b00355db
JP
42#define TCQ_F_BUILTIN 1
43#define TCQ_F_THROTTLED 2
44#define TCQ_F_INGRESS 4
bbd8a0d3 45#define TCQ_F_CAN_BYPASS 8
23bcf634 46#define TCQ_F_MQROOT 16
b00355db 47#define TCQ_F_WARN_NONWC (1 << 16)
1da177e4
LT
48 int padded;
49 struct Qdisc_ops *ops;
175f9c1b 50 struct qdisc_size_table *stab;
5e140dfc 51 struct list_head list;
1da177e4
LT
52 u32 handle;
53 u32 parent;
54 atomic_t refcnt;
1da177e4 55 struct gnet_stats_rate_est rate_est;
1da177e4
LT
56 int (*reshape_fail)(struct sk_buff *skb,
57 struct Qdisc *q);
58
72b25a91
DM
59 void *u32_node;
60
1da177e4
LT
61 /* This field is deprecated, but it is still used by CBQ
62 * and it will live until better solution will be invented.
63 */
64 struct Qdisc *__parent;
5e140dfc
ED
65 struct netdev_queue *dev_queue;
66 struct Qdisc *next_sched;
67
68 struct sk_buff *gso_skb;
69 /*
70 * For performance sake on SMP, we put highly modified fields at the end
71 */
72 unsigned long state;
73 struct sk_buff_head q;
c1a8f1f1 74 struct gnet_stats_basic_packed bstats;
5e140dfc 75 struct gnet_stats_queue qstats;
1da177e4
LT
76};
77
fd2c3ef7 78struct Qdisc_class_ops {
1da177e4 79 /* Child qdisc manipulation */
926e61b7 80 struct netdev_queue * (*select_queue)(struct Qdisc *, struct tcmsg *);
1da177e4
LT
81 int (*graft)(struct Qdisc *, unsigned long cl,
82 struct Qdisc *, struct Qdisc **);
83 struct Qdisc * (*leaf)(struct Qdisc *, unsigned long cl);
43effa1e 84 void (*qlen_notify)(struct Qdisc *, unsigned long);
1da177e4
LT
85
86 /* Class manipulation routines */
87 unsigned long (*get)(struct Qdisc *, u32 classid);
88 void (*put)(struct Qdisc *, unsigned long);
89 int (*change)(struct Qdisc *, u32, u32,
1e90474c 90 struct nlattr **, unsigned long *);
1da177e4
LT
91 int (*delete)(struct Qdisc *, unsigned long);
92 void (*walk)(struct Qdisc *, struct qdisc_walker * arg);
93
94 /* Filter manipulation */
95 struct tcf_proto ** (*tcf_chain)(struct Qdisc *, unsigned long);
96 unsigned long (*bind_tcf)(struct Qdisc *, unsigned long,
97 u32 classid);
98 void (*unbind_tcf)(struct Qdisc *, unsigned long);
99
100 /* rtnetlink specific */
101 int (*dump)(struct Qdisc *, unsigned long,
102 struct sk_buff *skb, struct tcmsg*);
103 int (*dump_stats)(struct Qdisc *, unsigned long,
104 struct gnet_dump *);
105};
106
fd2c3ef7 107struct Qdisc_ops {
1da177e4 108 struct Qdisc_ops *next;
20fea08b 109 const struct Qdisc_class_ops *cl_ops;
1da177e4
LT
110 char id[IFNAMSIZ];
111 int priv_size;
112
113 int (*enqueue)(struct sk_buff *, struct Qdisc *);
114 struct sk_buff * (*dequeue)(struct Qdisc *);
90d841fd 115 struct sk_buff * (*peek)(struct Qdisc *);
1da177e4
LT
116 unsigned int (*drop)(struct Qdisc *);
117
1e90474c 118 int (*init)(struct Qdisc *, struct nlattr *arg);
1da177e4
LT
119 void (*reset)(struct Qdisc *);
120 void (*destroy)(struct Qdisc *);
1e90474c 121 int (*change)(struct Qdisc *, struct nlattr *arg);
6ec1c69a 122 void (*attach)(struct Qdisc *);
1da177e4
LT
123
124 int (*dump)(struct Qdisc *, struct sk_buff *);
125 int (*dump_stats)(struct Qdisc *, struct gnet_dump *);
126
127 struct module *owner;
128};
129
130
fd2c3ef7 131struct tcf_result {
1da177e4
LT
132 unsigned long class;
133 u32 classid;
134};
135
fd2c3ef7 136struct tcf_proto_ops {
1da177e4
LT
137 struct tcf_proto_ops *next;
138 char kind[IFNAMSIZ];
139
140 int (*classify)(struct sk_buff*, struct tcf_proto*,
141 struct tcf_result *);
142 int (*init)(struct tcf_proto*);
143 void (*destroy)(struct tcf_proto*);
144
145 unsigned long (*get)(struct tcf_proto*, u32 handle);
146 void (*put)(struct tcf_proto*, unsigned long);
147 int (*change)(struct tcf_proto*, unsigned long,
add93b61 148 u32 handle, struct nlattr **,
1da177e4
LT
149 unsigned long *);
150 int (*delete)(struct tcf_proto*, unsigned long);
151 void (*walk)(struct tcf_proto*, struct tcf_walker *arg);
152
153 /* rtnetlink specific */
154 int (*dump)(struct tcf_proto*, unsigned long,
155 struct sk_buff *skb, struct tcmsg*);
156
157 struct module *owner;
158};
159
fd2c3ef7 160struct tcf_proto {
1da177e4
LT
161 /* Fast access part */
162 struct tcf_proto *next;
163 void *root;
164 int (*classify)(struct sk_buff*, struct tcf_proto*,
165 struct tcf_result *);
66c6f529 166 __be16 protocol;
1da177e4
LT
167
168 /* All the rest */
169 u32 prio;
170 u32 classid;
171 struct Qdisc *q;
172 void *data;
173 struct tcf_proto_ops *ops;
174};
175
175f9c1b
JK
176struct qdisc_skb_cb {
177 unsigned int pkt_len;
178 char data[];
179};
180
bbd8a0d3
KK
181static inline int qdisc_qlen(struct Qdisc *q)
182{
183 return q->q.qlen;
184}
185
175f9c1b
JK
186static inline struct qdisc_skb_cb *qdisc_skb_cb(struct sk_buff *skb)
187{
188 return (struct qdisc_skb_cb *)skb->cb;
189}
190
83874000
DM
191static inline spinlock_t *qdisc_lock(struct Qdisc *qdisc)
192{
193 return &qdisc->q.lock;
194}
195
7698b4fc
DM
196static inline struct Qdisc *qdisc_root(struct Qdisc *qdisc)
197{
198 return qdisc->dev_queue->qdisc;
199}
200
2540e051
JP
201static inline struct Qdisc *qdisc_root_sleeping(struct Qdisc *qdisc)
202{
203 return qdisc->dev_queue->qdisc_sleeping;
204}
205
7e43f112
DM
206/* The qdisc root lock is a mechanism by which to top level
207 * of a qdisc tree can be locked from any qdisc node in the
208 * forest. This allows changing the configuration of some
209 * aspect of the qdisc tree while blocking out asynchronous
210 * qdisc access in the packet processing paths.
211 *
212 * It is only legal to do this when the root will not change
213 * on us. Otherwise we'll potentially lock the wrong qdisc
214 * root. This is enforced by holding the RTNL semaphore, which
215 * all users of this lock accessor must do.
216 */
7698b4fc
DM
217static inline spinlock_t *qdisc_root_lock(struct Qdisc *qdisc)
218{
219 struct Qdisc *root = qdisc_root(qdisc);
220
7e43f112 221 ASSERT_RTNL();
83874000 222 return qdisc_lock(root);
7698b4fc
DM
223}
224
f6f9b93f
JP
225static inline spinlock_t *qdisc_root_sleeping_lock(struct Qdisc *qdisc)
226{
227 struct Qdisc *root = qdisc_root_sleeping(qdisc);
228
229 ASSERT_RTNL();
230 return qdisc_lock(root);
231}
232
5ce2d488
DM
233static inline struct net_device *qdisc_dev(struct Qdisc *qdisc)
234{
235 return qdisc->dev_queue->dev;
236}
1da177e4 237
78a5b30b
DM
238static inline void sch_tree_lock(struct Qdisc *q)
239{
fe439dd0 240 spin_lock_bh(qdisc_root_sleeping_lock(q));
78a5b30b
DM
241}
242
243static inline void sch_tree_unlock(struct Qdisc *q)
244{
fe439dd0 245 spin_unlock_bh(qdisc_root_sleeping_lock(q));
78a5b30b
DM
246}
247
248#define tcf_tree_lock(tp) sch_tree_lock((tp)->q)
249#define tcf_tree_unlock(tp) sch_tree_unlock((tp)->q)
1da177e4 250
e41a33e6
TG
251extern struct Qdisc noop_qdisc;
252extern struct Qdisc_ops noop_qdisc_ops;
6ec1c69a
DM
253extern struct Qdisc_ops pfifo_fast_ops;
254extern struct Qdisc_ops mq_qdisc_ops;
e41a33e6 255
fd2c3ef7 256struct Qdisc_class_common {
6fe1c7a5
PM
257 u32 classid;
258 struct hlist_node hnode;
259};
260
fd2c3ef7 261struct Qdisc_class_hash {
6fe1c7a5
PM
262 struct hlist_head *hash;
263 unsigned int hashsize;
264 unsigned int hashmask;
265 unsigned int hashelems;
266};
267
268static inline unsigned int qdisc_class_hash(u32 id, u32 mask)
269{
270 id ^= id >> 8;
271 id ^= id >> 4;
272 return id & mask;
273}
274
275static inline struct Qdisc_class_common *
276qdisc_class_find(struct Qdisc_class_hash *hash, u32 id)
277{
278 struct Qdisc_class_common *cl;
279 struct hlist_node *n;
280 unsigned int h;
281
282 h = qdisc_class_hash(id, hash->hashmask);
283 hlist_for_each_entry(cl, n, &hash->hash[h], hnode) {
284 if (cl->classid == id)
285 return cl;
286 }
287 return NULL;
288}
289
290extern int qdisc_class_hash_init(struct Qdisc_class_hash *);
291extern void qdisc_class_hash_insert(struct Qdisc_class_hash *, struct Qdisc_class_common *);
292extern void qdisc_class_hash_remove(struct Qdisc_class_hash *, struct Qdisc_class_common *);
293extern void qdisc_class_hash_grow(struct Qdisc *, struct Qdisc_class_hash *);
294extern void qdisc_class_hash_destroy(struct Qdisc_class_hash *);
295
e41a33e6
TG
296extern void dev_init_scheduler(struct net_device *dev);
297extern void dev_shutdown(struct net_device *dev);
298extern void dev_activate(struct net_device *dev);
299extern void dev_deactivate(struct net_device *dev);
589983cd
PM
300extern struct Qdisc *dev_graft_qdisc(struct netdev_queue *dev_queue,
301 struct Qdisc *qdisc);
e41a33e6
TG
302extern void qdisc_reset(struct Qdisc *qdisc);
303extern void qdisc_destroy(struct Qdisc *qdisc);
43effa1e 304extern void qdisc_tree_decrease_qlen(struct Qdisc *qdisc, unsigned int n);
5ce2d488 305extern struct Qdisc *qdisc_alloc(struct netdev_queue *dev_queue,
bb949fbd 306 struct Qdisc_ops *ops);
e41a33e6 307extern struct Qdisc *qdisc_create_dflt(struct net_device *dev,
bb949fbd 308 struct netdev_queue *dev_queue,
9f9afec4 309 struct Qdisc_ops *ops, u32 parentid);
175f9c1b
JK
310extern void qdisc_calculate_pkt_len(struct sk_buff *skb,
311 struct qdisc_size_table *stab);
a48b5a61 312extern void tcf_destroy(struct tcf_proto *tp);
ff31ab56 313extern void tcf_destroy_chain(struct tcf_proto **fl);
1da177e4 314
5aa70995
DM
315/* Reset all TX qdiscs of a device. */
316static inline void qdisc_reset_all_tx(struct net_device *dev)
317{
e8a0464c
DM
318 unsigned int i;
319 for (i = 0; i < dev->num_tx_queues; i++)
320 qdisc_reset(netdev_get_tx_queue(dev, i)->qdisc);
5aa70995
DM
321}
322
3e745dd6
DM
323/* Are all TX queues of the device empty? */
324static inline bool qdisc_all_tx_empty(const struct net_device *dev)
325{
e8a0464c
DM
326 unsigned int i;
327 for (i = 0; i < dev->num_tx_queues; i++) {
328 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
329 const struct Qdisc *q = txq->qdisc;
3e745dd6 330
e8a0464c
DM
331 if (q->q.qlen)
332 return false;
333 }
334 return true;
3e745dd6
DM
335}
336
6fa9864b
DM
337/* Are any of the TX qdiscs changing? */
338static inline bool qdisc_tx_changing(struct net_device *dev)
339{
e8a0464c
DM
340 unsigned int i;
341 for (i = 0; i < dev->num_tx_queues; i++) {
342 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
343 if (txq->qdisc != txq->qdisc_sleeping)
344 return true;
345 }
346 return false;
6fa9864b
DM
347}
348
e8a0464c 349/* Is the device using the noop qdisc on all queues? */
05297949
DM
350static inline bool qdisc_tx_is_noop(const struct net_device *dev)
351{
e8a0464c
DM
352 unsigned int i;
353 for (i = 0; i < dev->num_tx_queues; i++) {
354 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
355 if (txq->qdisc != &noop_qdisc)
356 return false;
357 }
358 return true;
05297949
DM
359}
360
0abf77e5
JK
361static inline unsigned int qdisc_pkt_len(struct sk_buff *skb)
362{
175f9c1b 363 return qdisc_skb_cb(skb)->pkt_len;
0abf77e5
JK
364}
365
c27f339a 366/* additional qdisc xmit flags (NET_XMIT_MASK in linux/netdevice.h) */
378a2f09
JP
367enum net_xmit_qdisc_t {
368 __NET_XMIT_STOLEN = 0x00010000,
c27f339a 369 __NET_XMIT_BYPASS = 0x00020000,
378a2f09
JP
370};
371
c27f339a 372#ifdef CONFIG_NET_CLS_ACT
378a2f09 373#define net_xmit_drop_count(e) ((e) & __NET_XMIT_STOLEN ? 0 : 1)
378a2f09
JP
374#else
375#define net_xmit_drop_count(e) (1)
376#endif
377
5f86173b
JK
378static inline int qdisc_enqueue(struct sk_buff *skb, struct Qdisc *sch)
379{
3a682fbd 380#ifdef CONFIG_NET_SCHED
175f9c1b
JK
381 if (sch->stab)
382 qdisc_calculate_pkt_len(skb, sch->stab);
3a682fbd 383#endif
5f86173b
JK
384 return sch->enqueue(skb, sch);
385}
386
387static inline int qdisc_enqueue_root(struct sk_buff *skb, struct Qdisc *sch)
388{
175f9c1b 389 qdisc_skb_cb(skb)->pkt_len = skb->len;
378a2f09 390 return qdisc_enqueue(skb, sch) & NET_XMIT_MASK;
5f86173b
JK
391}
392
bbd8a0d3
KK
393static inline void __qdisc_update_bstats(struct Qdisc *sch, unsigned int len)
394{
395 sch->bstats.bytes += len;
396 sch->bstats.packets++;
397}
398
9972b25d
TG
399static inline int __qdisc_enqueue_tail(struct sk_buff *skb, struct Qdisc *sch,
400 struct sk_buff_head *list)
401{
402 __skb_queue_tail(list, skb);
0abf77e5 403 sch->qstats.backlog += qdisc_pkt_len(skb);
bbd8a0d3 404 __qdisc_update_bstats(sch, qdisc_pkt_len(skb));
9972b25d
TG
405
406 return NET_XMIT_SUCCESS;
407}
408
409static inline int qdisc_enqueue_tail(struct sk_buff *skb, struct Qdisc *sch)
410{
411 return __qdisc_enqueue_tail(skb, sch, &sch->q);
412}
413
414static inline struct sk_buff *__qdisc_dequeue_head(struct Qdisc *sch,
415 struct sk_buff_head *list)
416{
417 struct sk_buff *skb = __skb_dequeue(list);
418
419 if (likely(skb != NULL))
0abf77e5 420 sch->qstats.backlog -= qdisc_pkt_len(skb);
9972b25d
TG
421
422 return skb;
423}
424
425static inline struct sk_buff *qdisc_dequeue_head(struct Qdisc *sch)
426{
427 return __qdisc_dequeue_head(sch, &sch->q);
428}
429
430static inline struct sk_buff *__qdisc_dequeue_tail(struct Qdisc *sch,
431 struct sk_buff_head *list)
432{
433 struct sk_buff *skb = __skb_dequeue_tail(list);
434
435 if (likely(skb != NULL))
0abf77e5 436 sch->qstats.backlog -= qdisc_pkt_len(skb);
9972b25d
TG
437
438 return skb;
439}
440
441static inline struct sk_buff *qdisc_dequeue_tail(struct Qdisc *sch)
442{
443 return __qdisc_dequeue_tail(sch, &sch->q);
444}
445
48a8f519
PM
446static inline struct sk_buff *qdisc_peek_head(struct Qdisc *sch)
447{
448 return skb_peek(&sch->q);
449}
450
77be155c
JP
451/* generic pseudo peek method for non-work-conserving qdisc */
452static inline struct sk_buff *qdisc_peek_dequeued(struct Qdisc *sch)
453{
454 /* we can reuse ->gso_skb because peek isn't called for root qdiscs */
61c9eaf9 455 if (!sch->gso_skb) {
77be155c 456 sch->gso_skb = sch->dequeue(sch);
61c9eaf9
JP
457 if (sch->gso_skb)
458 /* it's still part of the queue */
459 sch->q.qlen++;
460 }
77be155c
JP
461
462 return sch->gso_skb;
463}
464
465/* use instead of qdisc->dequeue() for all qdiscs queried with ->peek() */
466static inline struct sk_buff *qdisc_dequeue_peeked(struct Qdisc *sch)
467{
468 struct sk_buff *skb = sch->gso_skb;
469
61c9eaf9 470 if (skb) {
77be155c 471 sch->gso_skb = NULL;
61c9eaf9
JP
472 sch->q.qlen--;
473 } else {
77be155c 474 skb = sch->dequeue(sch);
61c9eaf9 475 }
77be155c
JP
476
477 return skb;
478}
479
9972b25d
TG
480static inline void __qdisc_reset_queue(struct Qdisc *sch,
481 struct sk_buff_head *list)
482{
483 /*
484 * We do not know the backlog in bytes of this list, it
485 * is up to the caller to correct it
486 */
93245dd6 487 __skb_queue_purge(list);
9972b25d
TG
488}
489
490static inline void qdisc_reset_queue(struct Qdisc *sch)
491{
492 __qdisc_reset_queue(sch, &sch->q);
493 sch->qstats.backlog = 0;
494}
495
496static inline unsigned int __qdisc_queue_drop(struct Qdisc *sch,
497 struct sk_buff_head *list)
498{
499 struct sk_buff *skb = __qdisc_dequeue_tail(sch, list);
500
501 if (likely(skb != NULL)) {
0abf77e5 502 unsigned int len = qdisc_pkt_len(skb);
9972b25d
TG
503 kfree_skb(skb);
504 return len;
505 }
506
507 return 0;
508}
509
510static inline unsigned int qdisc_queue_drop(struct Qdisc *sch)
511{
512 return __qdisc_queue_drop(sch, &sch->q);
513}
514
515static inline int qdisc_drop(struct sk_buff *skb, struct Qdisc *sch)
516{
517 kfree_skb(skb);
518 sch->qstats.drops++;
519
520 return NET_XMIT_DROP;
521}
522
523static inline int qdisc_reshape_fail(struct sk_buff *skb, struct Qdisc *sch)
524{
525 sch->qstats.drops++;
526
c3bc7cff 527#ifdef CONFIG_NET_CLS_ACT
9972b25d
TG
528 if (sch->reshape_fail == NULL || sch->reshape_fail(skb, sch))
529 goto drop;
530
531 return NET_XMIT_SUCCESS;
532
533drop:
534#endif
535 kfree_skb(skb);
536 return NET_XMIT_DROP;
537}
538
e9bef55d
JDB
539/* Length to Time (L2T) lookup in a qdisc_rate_table, to determine how
540 long it will take to send a packet given its size.
541 */
542static inline u32 qdisc_l2t(struct qdisc_rate_table* rtab, unsigned int pktlen)
543{
e08b0998
JDB
544 int slot = pktlen + rtab->rate.cell_align + rtab->rate.overhead;
545 if (slot < 0)
546 slot = 0;
e9bef55d
JDB
547 slot >>= rtab->rate.cell_log;
548 if (slot > 255)
549 return (rtab->data[255]*(slot >> 8) + rtab->data[slot & 0xFF]);
550 return rtab->data[slot];
551}
552
12da81d1
JHS
553#ifdef CONFIG_NET_CLS_ACT
554static inline struct sk_buff *skb_act_clone(struct sk_buff *skb, gfp_t gfp_mask)
555{
556 struct sk_buff *n = skb_clone(skb, gfp_mask);
557
558 if (n) {
559 n->tc_verd = SET_TC_VERD(n->tc_verd, 0);
560 n->tc_verd = CLR_TC_OK2MUNGE(n->tc_verd);
561 n->tc_verd = CLR_TC_MUNGED(n->tc_verd);
12da81d1
JHS
562 }
563 return n;
564}
565#endif
566
1da177e4 567#endif