pkt_sched.h revision a08d2590a042f8173713cb3b1bf7bb7aeb3ce584
1#ifndef __LINUX_PKT_SCHED_H 2#define __LINUX_PKT_SCHED_H 3 4#include <linux/types.h> 5 6/* Logical priority bands not depending on specific packet scheduler. 7 Every scheduler will map them to real traffic classes, if it has 8 no more precise mechanism to classify packets. 9 10 These numbers have no special meaning, though their coincidence 11 with obsolete IPv6 values is not occasional :-). New IPv6 drafts 12 preferred full anarchy inspired by diffserv group. 13 14 Note: TC_PRIO_BESTEFFORT does not mean that it is the most unhappy 15 class, actually, as rule it will be handled with more care than 16 filler or even bulk. 17 */ 18 19#define TC_PRIO_BESTEFFORT 0 20#define TC_PRIO_FILLER 1 21#define TC_PRIO_BULK 2 22#define TC_PRIO_INTERACTIVE_BULK 4 23#define TC_PRIO_INTERACTIVE 6 24#define TC_PRIO_CONTROL 7 25 26#define TC_PRIO_MAX 15 27 28/* Generic queue statistics, available for all the elements. 29 Particular schedulers may have also their private records. 30 */ 31 32struct tc_stats { 33 __u64 bytes; /* Number of enqueued bytes */ 34 __u32 packets; /* Number of enqueued packets */ 35 __u32 drops; /* Packets dropped because of lack of resources */ 36 __u32 overlimits; /* Number of throttle events when this 37 * flow goes out of allocated bandwidth */ 38 __u32 bps; /* Current flow byte rate */ 39 __u32 pps; /* Current flow packet rate */ 40 __u32 qlen; 41 __u32 backlog; 42}; 43 44struct tc_estimator { 45 signed char interval; 46 unsigned char ewma_log; 47}; 48 49/* "Handles" 50 --------- 51 52 All the traffic control objects have 32bit identifiers, or "handles". 53 54 They can be considered as opaque numbers from user API viewpoint, 55 but actually they always consist of two fields: major and 56 minor numbers, which are interpreted by kernel specially, 57 that may be used by applications, though not recommended. 58 59 F.e. qdisc handles always have minor number equal to zero, 60 classes (or flows) have major equal to parent qdisc major, and 61 minor uniquely identifying class inside qdisc. 62 63 Macros to manipulate handles: 64 */ 65 66#define TC_H_MAJ_MASK (0xFFFF0000U) 67#define TC_H_MIN_MASK (0x0000FFFFU) 68#define TC_H_MAJ(h) ((h)&TC_H_MAJ_MASK) 69#define TC_H_MIN(h) ((h)&TC_H_MIN_MASK) 70#define TC_H_MAKE(maj,min) (((maj)&TC_H_MAJ_MASK)|((min)&TC_H_MIN_MASK)) 71 72#define TC_H_UNSPEC (0U) 73#define TC_H_ROOT (0xFFFFFFFFU) 74#define TC_H_INGRESS (0xFFFFFFF1U) 75 76struct tc_ratespec { 77 unsigned char cell_log; 78 unsigned char __reserved; 79 unsigned short overhead; 80 short cell_align; 81 unsigned short mpu; 82 __u32 rate; 83}; 84 85#define TC_RTAB_SIZE 1024 86 87struct tc_sizespec { 88 unsigned char cell_log; 89 unsigned char size_log; 90 short cell_align; 91 int overhead; 92 unsigned int linklayer; 93 unsigned int mpu; 94 unsigned int mtu; 95 unsigned int tsize; 96}; 97 98enum { 99 TCA_STAB_UNSPEC, 100 TCA_STAB_BASE, 101 TCA_STAB_DATA, 102 __TCA_STAB_MAX 103}; 104 105#define TCA_STAB_MAX (__TCA_STAB_MAX - 1) 106 107/* FIFO section */ 108 109struct tc_fifo_qopt { 110 __u32 limit; /* Queue length: bytes for bfifo, packets for pfifo */ 111}; 112 113/* PRIO section */ 114 115#define TCQ_PRIO_BANDS 16 116#define TCQ_MIN_PRIO_BANDS 2 117 118struct tc_prio_qopt { 119 int bands; /* Number of bands */ 120 __u8 priomap[TC_PRIO_MAX+1]; /* Map: logical priority -> PRIO band */ 121}; 122 123/* MULTIQ section */ 124 125struct tc_multiq_qopt { 126 __u16 bands; /* Number of bands */ 127 __u16 max_bands; /* Maximum number of queues */ 128}; 129 130/* TBF section */ 131 132struct tc_tbf_qopt { 133 struct tc_ratespec rate; 134 struct tc_ratespec peakrate; 135 __u32 limit; 136 __u32 buffer; 137 __u32 mtu; 138}; 139 140enum { 141 TCA_TBF_UNSPEC, 142 TCA_TBF_PARMS, 143 TCA_TBF_RTAB, 144 TCA_TBF_PTAB, 145 __TCA_TBF_MAX, 146}; 147 148#define TCA_TBF_MAX (__TCA_TBF_MAX - 1) 149 150 151/* TEQL section */ 152 153/* TEQL does not require any parameters */ 154 155/* SFQ section */ 156 157struct tc_sfq_qopt { 158 unsigned quantum; /* Bytes per round allocated to flow */ 159 int perturb_period; /* Period of hash perturbation */ 160 __u32 limit; /* Maximal packets in queue */ 161 unsigned divisor; /* Hash divisor */ 162 unsigned flows; /* Maximal number of flows */ 163}; 164 165struct tc_sfq_qopt_v1 { 166 struct tc_sfq_qopt v0; 167 unsigned int depth; /* max number of packets per flow */ 168 unsigned int headdrop; 169}; 170 171 172struct tc_sfq_xstats { 173 __s32 allot; 174}; 175 176/* RED section */ 177 178enum { 179 TCA_RED_UNSPEC, 180 TCA_RED_PARMS, 181 TCA_RED_STAB, 182 TCA_RED_MAX_P, 183 __TCA_RED_MAX, 184}; 185 186#define TCA_RED_MAX (__TCA_RED_MAX - 1) 187 188struct tc_red_qopt { 189 __u32 limit; /* HARD maximal queue length (bytes) */ 190 __u32 qth_min; /* Min average length threshold (bytes) */ 191 __u32 qth_max; /* Max average length threshold (bytes) */ 192 unsigned char Wlog; /* log(W) */ 193 unsigned char Plog; /* log(P_max/(qth_max-qth_min)) */ 194 unsigned char Scell_log; /* cell size for idle damping */ 195 unsigned char flags; 196#define TC_RED_ECN 1 197#define TC_RED_HARDDROP 2 198#define TC_RED_ADAPTATIVE 4 199}; 200 201struct tc_red_xstats { 202 __u32 early; /* Early drops */ 203 __u32 pdrop; /* Drops due to queue limits */ 204 __u32 other; /* Drops due to drop() calls */ 205 __u32 marked; /* Marked packets */ 206}; 207 208/* GRED section */ 209 210#define MAX_DPs 16 211 212enum { 213 TCA_GRED_UNSPEC, 214 TCA_GRED_PARMS, 215 TCA_GRED_STAB, 216 TCA_GRED_DPS, 217 TCA_GRED_MAX_P, 218 __TCA_GRED_MAX, 219}; 220 221#define TCA_GRED_MAX (__TCA_GRED_MAX - 1) 222 223struct tc_gred_qopt { 224 __u32 limit; /* HARD maximal queue length (bytes) */ 225 __u32 qth_min; /* Min average length threshold (bytes) */ 226 __u32 qth_max; /* Max average length threshold (bytes) */ 227 __u32 DP; /* up to 2^32 DPs */ 228 __u32 backlog; 229 __u32 qave; 230 __u32 forced; 231 __u32 early; 232 __u32 other; 233 __u32 pdrop; 234 __u8 Wlog; /* log(W) */ 235 __u8 Plog; /* log(P_max/(qth_max-qth_min)) */ 236 __u8 Scell_log; /* cell size for idle damping */ 237 __u8 prio; /* prio of this VQ */ 238 __u32 packets; 239 __u32 bytesin; 240}; 241 242/* gred setup */ 243struct tc_gred_sopt { 244 __u32 DPs; 245 __u32 def_DP; 246 __u8 grio; 247 __u8 flags; 248 __u16 pad1; 249}; 250 251/* CHOKe section */ 252 253enum { 254 TCA_CHOKE_UNSPEC, 255 TCA_CHOKE_PARMS, 256 TCA_CHOKE_STAB, 257 TCA_CHOKE_MAX_P, 258 __TCA_CHOKE_MAX, 259}; 260 261#define TCA_CHOKE_MAX (__TCA_CHOKE_MAX - 1) 262 263struct tc_choke_qopt { 264 __u32 limit; /* Hard queue length (packets) */ 265 __u32 qth_min; /* Min average threshold (packets) */ 266 __u32 qth_max; /* Max average threshold (packets) */ 267 unsigned char Wlog; /* log(W) */ 268 unsigned char Plog; /* log(P_max/(qth_max-qth_min)) */ 269 unsigned char Scell_log; /* cell size for idle damping */ 270 unsigned char flags; /* see RED flags */ 271}; 272 273struct tc_choke_xstats { 274 __u32 early; /* Early drops */ 275 __u32 pdrop; /* Drops due to queue limits */ 276 __u32 other; /* Drops due to drop() calls */ 277 __u32 marked; /* Marked packets */ 278 __u32 matched; /* Drops due to flow match */ 279}; 280 281/* HTB section */ 282#define TC_HTB_NUMPRIO 8 283#define TC_HTB_MAXDEPTH 8 284#define TC_HTB_PROTOVER 3 /* the same as HTB and TC's major */ 285 286struct tc_htb_opt { 287 struct tc_ratespec rate; 288 struct tc_ratespec ceil; 289 __u32 buffer; 290 __u32 cbuffer; 291 __u32 quantum; 292 __u32 level; /* out only */ 293 __u32 prio; 294}; 295struct tc_htb_glob { 296 __u32 version; /* to match HTB/TC */ 297 __u32 rate2quantum; /* bps->quantum divisor */ 298 __u32 defcls; /* default class number */ 299 __u32 debug; /* debug flags */ 300 301 /* stats */ 302 __u32 direct_pkts; /* count of non shaped packets */ 303}; 304enum { 305 TCA_HTB_UNSPEC, 306 TCA_HTB_PARMS, 307 TCA_HTB_INIT, 308 TCA_HTB_CTAB, 309 TCA_HTB_RTAB, 310 __TCA_HTB_MAX, 311}; 312 313#define TCA_HTB_MAX (__TCA_HTB_MAX - 1) 314 315struct tc_htb_xstats { 316 __u32 lends; 317 __u32 borrows; 318 __u32 giants; /* too big packets (rate will not be accurate) */ 319 __u32 tokens; 320 __u32 ctokens; 321}; 322 323/* HFSC section */ 324 325struct tc_hfsc_qopt { 326 __u16 defcls; /* default class */ 327}; 328 329struct tc_service_curve { 330 __u32 m1; /* slope of the first segment in bps */ 331 __u32 d; /* x-projection of the first segment in us */ 332 __u32 m2; /* slope of the second segment in bps */ 333}; 334 335struct tc_hfsc_stats { 336 __u64 work; /* total work done */ 337 __u64 rtwork; /* work done by real-time criteria */ 338 __u32 period; /* current period */ 339 __u32 level; /* class level in hierarchy */ 340}; 341 342enum { 343 TCA_HFSC_UNSPEC, 344 TCA_HFSC_RSC, 345 TCA_HFSC_FSC, 346 TCA_HFSC_USC, 347 __TCA_HFSC_MAX, 348}; 349 350#define TCA_HFSC_MAX (__TCA_HFSC_MAX - 1) 351 352 353/* CBQ section */ 354 355#define TC_CBQ_MAXPRIO 8 356#define TC_CBQ_MAXLEVEL 8 357#define TC_CBQ_DEF_EWMA 5 358 359struct tc_cbq_lssopt { 360 unsigned char change; 361 unsigned char flags; 362#define TCF_CBQ_LSS_BOUNDED 1 363#define TCF_CBQ_LSS_ISOLATED 2 364 unsigned char ewma_log; 365 unsigned char level; 366#define TCF_CBQ_LSS_FLAGS 1 367#define TCF_CBQ_LSS_EWMA 2 368#define TCF_CBQ_LSS_MAXIDLE 4 369#define TCF_CBQ_LSS_MINIDLE 8 370#define TCF_CBQ_LSS_OFFTIME 0x10 371#define TCF_CBQ_LSS_AVPKT 0x20 372 __u32 maxidle; 373 __u32 minidle; 374 __u32 offtime; 375 __u32 avpkt; 376}; 377 378struct tc_cbq_wrropt { 379 unsigned char flags; 380 unsigned char priority; 381 unsigned char cpriority; 382 unsigned char __reserved; 383 __u32 allot; 384 __u32 weight; 385}; 386 387struct tc_cbq_ovl { 388 unsigned char strategy; 389#define TC_CBQ_OVL_CLASSIC 0 390#define TC_CBQ_OVL_DELAY 1 391#define TC_CBQ_OVL_LOWPRIO 2 392#define TC_CBQ_OVL_DROP 3 393#define TC_CBQ_OVL_RCLASSIC 4 394 unsigned char priority2; 395 __u16 pad; 396 __u32 penalty; 397}; 398 399struct tc_cbq_police { 400 unsigned char police; 401 unsigned char __res1; 402 unsigned short __res2; 403}; 404 405struct tc_cbq_fopt { 406 __u32 split; 407 __u32 defmap; 408 __u32 defchange; 409}; 410 411struct tc_cbq_xstats { 412 __u32 borrows; 413 __u32 overactions; 414 __s32 avgidle; 415 __s32 undertime; 416}; 417 418enum { 419 TCA_CBQ_UNSPEC, 420 TCA_CBQ_LSSOPT, 421 TCA_CBQ_WRROPT, 422 TCA_CBQ_FOPT, 423 TCA_CBQ_OVL_STRATEGY, 424 TCA_CBQ_RATE, 425 TCA_CBQ_RTAB, 426 TCA_CBQ_POLICE, 427 __TCA_CBQ_MAX, 428}; 429 430#define TCA_CBQ_MAX (__TCA_CBQ_MAX - 1) 431 432/* dsmark section */ 433 434enum { 435 TCA_DSMARK_UNSPEC, 436 TCA_DSMARK_INDICES, 437 TCA_DSMARK_DEFAULT_INDEX, 438 TCA_DSMARK_SET_TC_INDEX, 439 TCA_DSMARK_MASK, 440 TCA_DSMARK_VALUE, 441 __TCA_DSMARK_MAX, 442}; 443 444#define TCA_DSMARK_MAX (__TCA_DSMARK_MAX - 1) 445 446/* ATM section */ 447 448enum { 449 TCA_ATM_UNSPEC, 450 TCA_ATM_FD, /* file/socket descriptor */ 451 TCA_ATM_PTR, /* pointer to descriptor - later */ 452 TCA_ATM_HDR, /* LL header */ 453 TCA_ATM_EXCESS, /* excess traffic class (0 for CLP) */ 454 TCA_ATM_ADDR, /* PVC address (for output only) */ 455 TCA_ATM_STATE, /* VC state (ATM_VS_*; for output only) */ 456 __TCA_ATM_MAX, 457}; 458 459#define TCA_ATM_MAX (__TCA_ATM_MAX - 1) 460 461/* Network emulator */ 462 463enum { 464 TCA_NETEM_UNSPEC, 465 TCA_NETEM_CORR, 466 TCA_NETEM_DELAY_DIST, 467 TCA_NETEM_REORDER, 468 TCA_NETEM_CORRUPT, 469 TCA_NETEM_LOSS, 470 TCA_NETEM_RATE, 471 __TCA_NETEM_MAX, 472}; 473 474#define TCA_NETEM_MAX (__TCA_NETEM_MAX - 1) 475 476struct tc_netem_qopt { 477 __u32 latency; /* added delay (us) */ 478 __u32 limit; /* fifo limit (packets) */ 479 __u32 loss; /* random packet loss (0=none ~0=100%) */ 480 __u32 gap; /* re-ordering gap (0 for none) */ 481 __u32 duplicate; /* random packet dup (0=none ~0=100%) */ 482 __u32 jitter; /* random jitter in latency (us) */ 483}; 484 485struct tc_netem_corr { 486 __u32 delay_corr; /* delay correlation */ 487 __u32 loss_corr; /* packet loss correlation */ 488 __u32 dup_corr; /* duplicate correlation */ 489}; 490 491struct tc_netem_reorder { 492 __u32 probability; 493 __u32 correlation; 494}; 495 496struct tc_netem_corrupt { 497 __u32 probability; 498 __u32 correlation; 499}; 500 501struct tc_netem_rate { 502 __u32 rate; /* byte/s */ 503 __s32 packet_overhead; 504 __u32 cell_size; 505 __s32 cell_overhead; 506}; 507 508enum { 509 NETEM_LOSS_UNSPEC, 510 NETEM_LOSS_GI, /* General Intuitive - 4 state model */ 511 NETEM_LOSS_GE, /* Gilbert Elliot models */ 512 __NETEM_LOSS_MAX 513}; 514#define NETEM_LOSS_MAX (__NETEM_LOSS_MAX - 1) 515 516/* State transition probabilities for 4 state model */ 517struct tc_netem_gimodel { 518 __u32 p13; 519 __u32 p31; 520 __u32 p32; 521 __u32 p14; 522 __u32 p23; 523}; 524 525/* Gilbert-Elliot models */ 526struct tc_netem_gemodel { 527 __u32 p; 528 __u32 r; 529 __u32 h; 530 __u32 k1; 531}; 532 533#define NETEM_DIST_SCALE 8192 534#define NETEM_DIST_MAX 16384 535 536/* DRR */ 537 538enum { 539 TCA_DRR_UNSPEC, 540 TCA_DRR_QUANTUM, 541 __TCA_DRR_MAX 542}; 543 544#define TCA_DRR_MAX (__TCA_DRR_MAX - 1) 545 546struct tc_drr_stats { 547 __u32 deficit; 548}; 549 550/* MQPRIO */ 551#define TC_QOPT_BITMASK 15 552#define TC_QOPT_MAX_QUEUE 16 553 554struct tc_mqprio_qopt { 555 __u8 num_tc; 556 __u8 prio_tc_map[TC_QOPT_BITMASK + 1]; 557 __u8 hw; 558 __u16 count[TC_QOPT_MAX_QUEUE]; 559 __u16 offset[TC_QOPT_MAX_QUEUE]; 560}; 561 562/* SFB */ 563 564enum { 565 TCA_SFB_UNSPEC, 566 TCA_SFB_PARMS, 567 __TCA_SFB_MAX, 568}; 569 570#define TCA_SFB_MAX (__TCA_SFB_MAX - 1) 571 572/* 573 * Note: increment, decrement are Q0.16 fixed-point values. 574 */ 575struct tc_sfb_qopt { 576 __u32 rehash_interval; /* delay between hash move, in ms */ 577 __u32 warmup_time; /* double buffering warmup time in ms (warmup_time < rehash_interval) */ 578 __u32 max; /* max len of qlen_min */ 579 __u32 bin_size; /* maximum queue length per bin */ 580 __u32 increment; /* probability increment, (d1 in Blue) */ 581 __u32 decrement; /* probability decrement, (d2 in Blue) */ 582 __u32 limit; /* max SFB queue length */ 583 __u32 penalty_rate; /* inelastic flows are rate limited to 'rate' pps */ 584 __u32 penalty_burst; 585}; 586 587struct tc_sfb_xstats { 588 __u32 earlydrop; 589 __u32 penaltydrop; 590 __u32 bucketdrop; 591 __u32 queuedrop; 592 __u32 childdrop; /* drops in child qdisc */ 593 __u32 marked; 594 __u32 maxqlen; 595 __u32 maxprob; 596 __u32 avgprob; 597}; 598 599#define SFB_MAX_PROB 0xFFFF 600 601/* QFQ */ 602enum { 603 TCA_QFQ_UNSPEC, 604 TCA_QFQ_WEIGHT, 605 TCA_QFQ_LMAX, 606 __TCA_QFQ_MAX 607}; 608 609#define TCA_QFQ_MAX (__TCA_QFQ_MAX - 1) 610 611struct tc_qfq_stats { 612 __u32 weight; 613 __u32 lmax; 614}; 615 616#endif 617