include/net/sch_generic.h

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