?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
cgroup, cpuset: remove cgroup_subsys->post_clone()
[linux-2.6/libata-dev.git] / include / net / sch_generic.h
blob4616f468d5995f9a0b46d7351c5c8f7dc3a869c8
1 #ifndef __NET_SCHED_GENERIC_H
2 #define __NET_SCHED_GENERIC_H
3
4 #include <linux/netdevice.h>
5 #include <linux/types.h>
6 #include <linux/rcupdate.h>
7 #include <linux/pkt_sched.h>
8 #include <linux/pkt_cls.h>
9 #include <net/gen_stats.h>
10 #include <net/rtnetlink.h>
11
12 struct Qdisc_ops;
13 struct qdisc_walker;
14 struct tcf_walker;
15 struct module;
16
17 struct qdisc_rate_table {
18 struct tc_ratespec rate;
19 u32 data[256];
20 struct qdisc_rate_table *next;
21 int refcnt;
22 };
23
24 enum qdisc_state_t {
25 __QDISC_STATE_SCHED,
26 __QDISC_STATE_DEACTIVATED,
27 __QDISC_STATE_THROTTLED,
28 };
29
30 /*
31 * following bits are only changed while qdisc lock is held
32 */
33 enum qdisc___state_t {
34 __QDISC___STATE_RUNNING = 1,
35 };
36
37 struct qdisc_size_table {
38 struct rcu_head rcu;
39 struct list_head list;
40 struct tc_sizespec szopts;
41 int refcnt;
42 u16 data[];
43 };
44
45 struct Qdisc {
46 int (*enqueue)(struct sk_buff *skb, struct Qdisc *dev);
47 struct sk_buff * (*dequeue)(struct Qdisc *dev);
48 unsigned int flags;
49 #define TCQ_F_BUILTIN 1
50 #define TCQ_F_INGRESS 2
51 #define TCQ_F_CAN_BYPASS 4
52 #define TCQ_F_MQROOT 8
53 #define TCQ_F_WARN_NONWC (1 << 16)
54 int padded;
55 const struct Qdisc_ops *ops;
56 struct qdisc_size_table __rcu *stab;
57 struct list_head list;
58 u32 handle;
59 u32 parent;
60 atomic_t refcnt;
61 struct gnet_stats_rate_est rate_est;
62 int (*reshape_fail)(struct sk_buff *skb,
63 struct Qdisc *q);
64
65 void *u32_node;
66
67 /* This field is deprecated, but it is still used by CBQ
68 * and it will live until better solution will be invented.
69 */
70 struct Qdisc *__parent;
71 struct netdev_queue *dev_queue;
72 struct Qdisc *next_sched;
73
74 struct sk_buff *gso_skb;
75 /*
76 * For performance sake on SMP, we put highly modified fields at the end
77 */
78 unsigned long state;
79 struct sk_buff_head q;
80 struct gnet_stats_basic_packed bstats;
81 unsigned int __state;
82 struct gnet_stats_queue qstats;
83 struct rcu_head rcu_head;
84 spinlock_t busylock;
85 u32 limit;
86 };
87
88 static inline bool qdisc_is_running(const struct Qdisc *qdisc)
89 {
90 return (qdisc->__state & __QDISC___STATE_RUNNING) ? true : false;
91 }
92
93 static inline bool qdisc_run_begin(struct Qdisc *qdisc)
94 {
95 if (qdisc_is_running(qdisc))
96 return false;
97 qdisc->__state |= __QDISC___STATE_RUNNING;
98 return true;
99 }
100
101 static inline void qdisc_run_end(struct Qdisc *qdisc)
102 {
103 qdisc->__state &= ~__QDISC___STATE_RUNNING;
104 }
105
106 static inline bool qdisc_is_throttled(const struct Qdisc *qdisc)
107 {
108 return test_bit(__QDISC_STATE_THROTTLED, &qdisc->state) ? true : false;
109 }
110
111 static inline void qdisc_throttled(struct Qdisc *qdisc)
112 {
113 set_bit(__QDISC_STATE_THROTTLED, &qdisc->state);
114 }
115
116 static inline void qdisc_unthrottled(struct Qdisc *qdisc)
117 {
118 clear_bit(__QDISC_STATE_THROTTLED, &qdisc->state);
119 }
120
121 struct Qdisc_class_ops {
122 /* Child qdisc manipulation */
123 struct netdev_queue * (*select_queue)(struct Qdisc *, struct tcmsg *);
124 int (*graft)(struct Qdisc *, unsigned long cl,
125 struct Qdisc *, struct Qdisc **);
126 struct Qdisc * (*leaf)(struct Qdisc *, unsigned long cl);
127 void (*qlen_notify)(struct Qdisc *, unsigned long);
128
129 /* Class manipulation routines */
130 unsigned long (*get)(struct Qdisc *, u32 classid);
131 void (*put)(struct Qdisc *, unsigned long);
132 int (*change)(struct Qdisc *, u32, u32,
133 struct nlattr **, unsigned long *);
134 int (*delete)(struct Qdisc *, unsigned long);
135 void (*walk)(struct Qdisc *, struct qdisc_walker * arg);
136
137 /* Filter manipulation */
138 struct tcf_proto ** (*tcf_chain)(struct Qdisc *, unsigned long);
139 unsigned long (*bind_tcf)(struct Qdisc *, unsigned long,
140 u32 classid);
141 void (*unbind_tcf)(struct Qdisc *, unsigned long);
142
143 /* rtnetlink specific */
144 int (*dump)(struct Qdisc *, unsigned long,
145 struct sk_buff *skb, struct tcmsg*);
146 int (*dump_stats)(struct Qdisc *, unsigned long,
147 struct gnet_dump *);
148 };
149
150 struct Qdisc_ops {
151 struct Qdisc_ops *next;
152 const struct Qdisc_class_ops *cl_ops;
153 char id[IFNAMSIZ];
154 int priv_size;
155
156 int (*enqueue)(struct sk_buff *, struct Qdisc *);
157 struct sk_buff * (*dequeue)(struct Qdisc *);
158 struct sk_buff * (*peek)(struct Qdisc *);
159 unsigned int (*drop)(struct Qdisc *);
160
161 int (*init)(struct Qdisc *, struct nlattr *arg);
162 void (*reset)(struct Qdisc *);
163 void (*destroy)(struct Qdisc *);
164 int (*change)(struct Qdisc *, struct nlattr *arg);
165 void (*attach)(struct Qdisc *);
166
167 int (*dump)(struct Qdisc *, struct sk_buff *);
168 int (*dump_stats)(struct Qdisc *, struct gnet_dump *);
169
170 struct module *owner;
171 };
172
173
174 struct tcf_result {
175 unsigned long class;
176 u32 classid;
177 };
178
179 struct tcf_proto_ops {
180 struct tcf_proto_ops *next;
181 char kind[IFNAMSIZ];
182
183 int (*classify)(struct sk_buff *,
184 const struct tcf_proto *,
185 struct tcf_result *);
186 int (*init)(struct tcf_proto*);
187 void (*destroy)(struct tcf_proto*);
188
189 unsigned long (*get)(struct tcf_proto*, u32 handle);
190 void (*put)(struct tcf_proto*, unsigned long);
191 int (*change)(struct sk_buff *,
192 struct tcf_proto*, unsigned long,
193 u32 handle, struct nlattr **,
194 unsigned long *);
195 int (*delete)(struct tcf_proto*, unsigned long);
196 void (*walk)(struct tcf_proto*, struct tcf_walker *arg);
197
198 /* rtnetlink specific */
199 int (*dump)(struct tcf_proto*, unsigned long,
200 struct sk_buff *skb, struct tcmsg*);
201
202 struct module *owner;
203 };
204
205 struct tcf_proto {
206 /* Fast access part */
207 struct tcf_proto *next;
208 void *root;
209 int (*classify)(struct sk_buff *,
210 const struct tcf_proto *,
211 struct tcf_result *);
212 __be16 protocol;
213
214 /* All the rest */
215 u32 prio;
216 u32 classid;
217 struct Qdisc *q;
218 void *data;
219 const struct tcf_proto_ops *ops;
220 };
221
222 struct qdisc_skb_cb {
223 unsigned int pkt_len;
224 u16 slave_dev_queue_mapping;
225 u16 _pad;
226 unsigned char data[20];
227 };
228
229 static inline void qdisc_cb_private_validate(const struct sk_buff *skb, int sz)
230 {
231 struct qdisc_skb_cb *qcb;
232
233 BUILD_BUG_ON(sizeof(skb->cb) < offsetof(struct qdisc_skb_cb, data) + sz);
234 BUILD_BUG_ON(sizeof(qcb->data) < sz);
235 }
236
237 static inline int qdisc_qlen(const struct Qdisc *q)
238 {
239 return q->q.qlen;
240 }
241
242 static inline struct qdisc_skb_cb *qdisc_skb_cb(const struct sk_buff *skb)
243 {
244 return (struct qdisc_skb_cb *)skb->cb;
245 }
246
247 static inline spinlock_t *qdisc_lock(struct Qdisc *qdisc)
248 {
249 return &qdisc->q.lock;
250 }
251
252 static inline struct Qdisc *qdisc_root(const struct Qdisc *qdisc)
253 {
254 return qdisc->dev_queue->qdisc;
255 }
256
257 static inline struct Qdisc *qdisc_root_sleeping(const struct Qdisc *qdisc)
258 {
259 return qdisc->dev_queue->qdisc_sleeping;
260 }
261
262 /* The qdisc root lock is a mechanism by which to top level
263 * of a qdisc tree can be locked from any qdisc node in the
264 * forest. This allows changing the configuration of some
265 * aspect of the qdisc tree while blocking out asynchronous
266 * qdisc access in the packet processing paths.
267 *
268 * It is only legal to do this when the root will not change
269 * on us. Otherwise we'll potentially lock the wrong qdisc
270 * root. This is enforced by holding the RTNL semaphore, which
271 * all users of this lock accessor must do.
272 */
273 static inline spinlock_t *qdisc_root_lock(const struct Qdisc *qdisc)
274 {
275 struct Qdisc *root = qdisc_root(qdisc);
276
277 ASSERT_RTNL();
278 return qdisc_lock(root);
279 }
280
281 static inline spinlock_t *qdisc_root_sleeping_lock(const struct Qdisc *qdisc)
282 {
283 struct Qdisc *root = qdisc_root_sleeping(qdisc);
284
285 ASSERT_RTNL();
286 return qdisc_lock(root);
287 }
288
289 static inline struct net_device *qdisc_dev(const struct Qdisc *qdisc)
290 {
291 return qdisc->dev_queue->dev;
292 }
293
294 static inline void sch_tree_lock(const struct Qdisc *q)
295 {
296 spin_lock_bh(qdisc_root_sleeping_lock(q));
297 }
298
299 static inline void sch_tree_unlock(const struct Qdisc *q)
300 {
301 spin_unlock_bh(qdisc_root_sleeping_lock(q));
302 }
303
304 #define tcf_tree_lock(tp) sch_tree_lock((tp)->q)
305 #define tcf_tree_unlock(tp) sch_tree_unlock((tp)->q)
306
307 extern struct Qdisc noop_qdisc;
308 extern struct Qdisc_ops noop_qdisc_ops;
309 extern struct Qdisc_ops pfifo_fast_ops;
310 extern struct Qdisc_ops mq_qdisc_ops;
311
312 struct Qdisc_class_common {
313 u32 classid;
314 struct hlist_node hnode;
315 };
316
317 struct Qdisc_class_hash {
318 struct hlist_head *hash;
319 unsigned int hashsize;
320 unsigned int hashmask;
321 unsigned int hashelems;
322 };
323
324 static inline unsigned int qdisc_class_hash(u32 id, u32 mask)
325 {
326 id ^= id >> 8;
327 id ^= id >> 4;
328 return id & mask;
329 }
330
331 static inline struct Qdisc_class_common *
332 qdisc_class_find(const struct Qdisc_class_hash *hash, u32 id)
333 {
334 struct Qdisc_class_common *cl;
335 struct hlist_node *n;
336 unsigned int h;
337
338 h = qdisc_class_hash(id, hash->hashmask);
339 hlist_for_each_entry(cl, n, &hash->hash[h], hnode) {
340 if (cl->classid == id)
341 return cl;
342 }
343 return NULL;
344 }
345
346 extern int qdisc_class_hash_init(struct Qdisc_class_hash *);
347 extern void qdisc_class_hash_insert(struct Qdisc_class_hash *, struct Qdisc_class_common *);
348 extern void qdisc_class_hash_remove(struct Qdisc_class_hash *, struct Qdisc_class_common *);
349 extern void qdisc_class_hash_grow(struct Qdisc *, struct Qdisc_class_hash *);
350 extern void qdisc_class_hash_destroy(struct Qdisc_class_hash *);
351
352 extern void dev_init_scheduler(struct net_device *dev);
353 extern void dev_shutdown(struct net_device *dev);
354 extern void dev_activate(struct net_device *dev);
355 extern void dev_deactivate(struct net_device *dev);
356 extern void dev_deactivate_many(struct list_head *head);
357 extern struct Qdisc *dev_graft_qdisc(struct netdev_queue *dev_queue,
358 struct Qdisc *qdisc);
359 extern void qdisc_reset(struct Qdisc *qdisc);
360 extern void qdisc_destroy(struct Qdisc *qdisc);
361 extern void qdisc_tree_decrease_qlen(struct Qdisc *qdisc, unsigned int n);
362 extern struct Qdisc *qdisc_alloc(struct netdev_queue *dev_queue,
363 struct Qdisc_ops *ops);
364 extern struct Qdisc *qdisc_create_dflt(struct netdev_queue *dev_queue,
365 struct Qdisc_ops *ops, u32 parentid);
366 extern void __qdisc_calculate_pkt_len(struct sk_buff *skb,
367 const struct qdisc_size_table *stab);
368 extern void tcf_destroy(struct tcf_proto *tp);
369 extern void tcf_destroy_chain(struct tcf_proto **fl);
370
371 /* Reset all TX qdiscs greater then index of a device. */
372 static inline void qdisc_reset_all_tx_gt(struct net_device *dev, unsigned int i)
373 {
374 struct Qdisc *qdisc;
375
376 for (; i < dev->num_tx_queues; i++) {
377 qdisc = netdev_get_tx_queue(dev, i)->qdisc;
378 if (qdisc) {
379 spin_lock_bh(qdisc_lock(qdisc));
380 qdisc_reset(qdisc);
381 spin_unlock_bh(qdisc_lock(qdisc));
382 }
383 }
384 }
385
386 static inline void qdisc_reset_all_tx(struct net_device *dev)
387 {
388 qdisc_reset_all_tx_gt(dev, 0);
389 }
390
391 /* Are all TX queues of the device empty? */
392 static inline bool qdisc_all_tx_empty(const struct net_device *dev)
393 {
394 unsigned int i;
395 for (i = 0; i < dev->num_tx_queues; i++) {
396 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
397 const struct Qdisc *q = txq->qdisc;
398
399 if (q->q.qlen)
400 return false;
401 }
402 return true;
403 }
404
405 /* Are any of the TX qdiscs changing? */
406 static inline bool qdisc_tx_changing(const struct net_device *dev)
407 {
408 unsigned int i;
409 for (i = 0; i < dev->num_tx_queues; i++) {
410 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
411 if (txq->qdisc != txq->qdisc_sleeping)
412 return true;
413 }
414 return false;
415 }
416
417 /* Is the device using the noop qdisc on all queues? */
418 static inline bool qdisc_tx_is_noop(const struct net_device *dev)
419 {
420 unsigned int i;
421 for (i = 0; i < dev->num_tx_queues; i++) {
422 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
423 if (txq->qdisc != &noop_qdisc)
424 return false;
425 }
426 return true;
427 }
428
429 static inline unsigned int qdisc_pkt_len(const struct sk_buff *skb)
430 {
431 return qdisc_skb_cb(skb)->pkt_len;
432 }
433
434 /* additional qdisc xmit flags (NET_XMIT_MASK in linux/netdevice.h) */
435 enum net_xmit_qdisc_t {
436 __NET_XMIT_STOLEN = 0x00010000,
437 __NET_XMIT_BYPASS = 0x00020000,
438 };
439
440 #ifdef CONFIG_NET_CLS_ACT
441 #define net_xmit_drop_count(e) ((e) & __NET_XMIT_STOLEN ? 0 : 1)
442 #else
443 #define net_xmit_drop_count(e) (1)
444 #endif
445
446 static inline void qdisc_calculate_pkt_len(struct sk_buff *skb,
447 const struct Qdisc *sch)
448 {
449 #ifdef CONFIG_NET_SCHED
450 struct qdisc_size_table *stab = rcu_dereference_bh(sch->stab);
451
452 if (stab)
453 __qdisc_calculate_pkt_len(skb, stab);
454 #endif
455 }
456
457 static inline int qdisc_enqueue(struct sk_buff *skb, struct Qdisc *sch)
458 {
459 qdisc_calculate_pkt_len(skb, sch);
460 return sch->enqueue(skb, sch);
461 }
462
463 static inline int qdisc_enqueue_root(struct sk_buff *skb, struct Qdisc *sch)
464 {
465 qdisc_skb_cb(skb)->pkt_len = skb->len;
466 return qdisc_enqueue(skb, sch) & NET_XMIT_MASK;
467 }
468
469
470 static inline void bstats_update(struct gnet_stats_basic_packed *bstats,
471 const struct sk_buff *skb)
472 {
473 bstats->bytes += qdisc_pkt_len(skb);
474 bstats->packets += skb_is_gso(skb) ? skb_shinfo(skb)->gso_segs : 1;
475 }
476
477 static inline void qdisc_bstats_update(struct Qdisc *sch,
478 const struct sk_buff *skb)
479 {
480 bstats_update(&sch->bstats, skb);
481 }
482
483 static inline int __qdisc_enqueue_tail(struct sk_buff *skb, struct Qdisc *sch,
484 struct sk_buff_head *list)
485 {
486 __skb_queue_tail(list, skb);
487 sch->qstats.backlog += qdisc_pkt_len(skb);
488
489 return NET_XMIT_SUCCESS;
490 }
491
492 static inline int qdisc_enqueue_tail(struct sk_buff *skb, struct Qdisc *sch)
493 {
494 return __qdisc_enqueue_tail(skb, sch, &sch->q);
495 }
496
497 static inline struct sk_buff *__qdisc_dequeue_head(struct Qdisc *sch,
498 struct sk_buff_head *list)
499 {
500 struct sk_buff *skb = __skb_dequeue(list);
501
502 if (likely(skb != NULL)) {
503 sch->qstats.backlog -= qdisc_pkt_len(skb);
504 qdisc_bstats_update(sch, skb);
505 }
506
507 return skb;
508 }
509
510 static inline struct sk_buff *qdisc_dequeue_head(struct Qdisc *sch)
511 {
512 return __qdisc_dequeue_head(sch, &sch->q);
513 }
514
515 static inline unsigned int __qdisc_queue_drop_head(struct Qdisc *sch,
516 struct sk_buff_head *list)
517 {
518 struct sk_buff *skb = __skb_dequeue(list);
519
520 if (likely(skb != NULL)) {
521 unsigned int len = qdisc_pkt_len(skb);
522 sch->qstats.backlog -= len;
523 kfree_skb(skb);
524 return len;
525 }
526
527 return 0;
528 }
529
530 static inline unsigned int qdisc_queue_drop_head(struct Qdisc *sch)
531 {
532 return __qdisc_queue_drop_head(sch, &sch->q);
533 }
534
535 static inline struct sk_buff *__qdisc_dequeue_tail(struct Qdisc *sch,
536 struct sk_buff_head *list)
537 {
538 struct sk_buff *skb = __skb_dequeue_tail(list);
539
540 if (likely(skb != NULL))
541 sch->qstats.backlog -= qdisc_pkt_len(skb);
542
543 return skb;
544 }
545
546 static inline struct sk_buff *qdisc_dequeue_tail(struct Qdisc *sch)
547 {
548 return __qdisc_dequeue_tail(sch, &sch->q);
549 }
550
551 static inline struct sk_buff *qdisc_peek_head(struct Qdisc *sch)
552 {
553 return skb_peek(&sch->q);
554 }
555
556 /* generic pseudo peek method for non-work-conserving qdisc */
557 static inline struct sk_buff *qdisc_peek_dequeued(struct Qdisc *sch)
558 {
559 /* we can reuse ->gso_skb because peek isn't called for root qdiscs */
560 if (!sch->gso_skb) {
561 sch->gso_skb = sch->dequeue(sch);
562 if (sch->gso_skb)
563 /* it's still part of the queue */
564 sch->q.qlen++;
565 }
566
567 return sch->gso_skb;
568 }
569
570 /* use instead of qdisc->dequeue() for all qdiscs queried with ->peek() */
571 static inline struct sk_buff *qdisc_dequeue_peeked(struct Qdisc *sch)
572 {
573 struct sk_buff *skb = sch->gso_skb;
574
575 if (skb) {
576 sch->gso_skb = NULL;
577 sch->q.qlen--;
578 } else {
579 skb = sch->dequeue(sch);
580 }
581
582 return skb;
583 }
584
585 static inline void __qdisc_reset_queue(struct Qdisc *sch,
586 struct sk_buff_head *list)
587 {
588 /*
589 * We do not know the backlog in bytes of this list, it
590 * is up to the caller to correct it
591 */
592 __skb_queue_purge(list);
593 }
594
595 static inline void qdisc_reset_queue(struct Qdisc *sch)
596 {
597 __qdisc_reset_queue(sch, &sch->q);
598 sch->qstats.backlog = 0;
599 }
600
601 static inline unsigned int __qdisc_queue_drop(struct Qdisc *sch,
602 struct sk_buff_head *list)
603 {
604 struct sk_buff *skb = __qdisc_dequeue_tail(sch, list);
605
606 if (likely(skb != NULL)) {
607 unsigned int len = qdisc_pkt_len(skb);
608 kfree_skb(skb);
609 return len;
610 }
611
612 return 0;
613 }
614
615 static inline unsigned int qdisc_queue_drop(struct Qdisc *sch)
616 {
617 return __qdisc_queue_drop(sch, &sch->q);
618 }
619
620 static inline int qdisc_drop(struct sk_buff *skb, struct Qdisc *sch)
621 {
622 kfree_skb(skb);
623 sch->qstats.drops++;
624
625 return NET_XMIT_DROP;
626 }
627
628 static inline int qdisc_reshape_fail(struct sk_buff *skb, struct Qdisc *sch)
629 {
630 sch->qstats.drops++;
631
632 #ifdef CONFIG_NET_CLS_ACT
633 if (sch->reshape_fail == NULL || sch->reshape_fail(skb, sch))
634 goto drop;
635
636 return NET_XMIT_SUCCESS;
637
638 drop:
639 #endif
640 kfree_skb(skb);
641 return NET_XMIT_DROP;
642 }
643
644 /* Length to Time (L2T) lookup in a qdisc_rate_table, to determine how
645 long it will take to send a packet given its size.
646 */
647 static inline u32 qdisc_l2t(struct qdisc_rate_table* rtab, unsigned int pktlen)
648 {
649 int slot = pktlen + rtab->rate.cell_align + rtab->rate.overhead;
650 if (slot < 0)
651 slot = 0;
652 slot >>= rtab->rate.cell_log;
653 if (slot > 255)
654 return rtab->data[255]*(slot >> 8) + rtab->data[slot & 0xFF];
655 return rtab->data[slot];
656 }
657
658 #ifdef CONFIG_NET_CLS_ACT
659 static inline struct sk_buff *skb_act_clone(struct sk_buff *skb, gfp_t gfp_mask,
660 int action)
661 {
662 struct sk_buff *n;
663
664 n = skb_clone(skb, gfp_mask);
665
666 if (n) {
667 n->tc_verd = SET_TC_VERD(n->tc_verd, 0);
668 n->tc_verd = CLR_TC_OK2MUNGE(n->tc_verd);
669 n->tc_verd = CLR_TC_MUNGED(n->tc_verd);
670 }
671 return n;
672 }
673 #endif
674
675 #endif
Linux 2.6 libata-dev (mirror)