1 #ifndef __LINUX_PKT_SCHED_H
2 #define __LINUX_PKT_SCHED_H
4 #include <linux/types.h>
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.
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.
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
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
26 #define TC_PRIO_MAX 15
28 /* Generic queue statistics, available for all the elements.
29 Particular schedulers may have also their private records.
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 */
46 unsigned char ewma_log;
52 All the traffic control objects have 32bit identifiers, or "handles".
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.
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.
63 Macros to manipulate handles:
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))
72 #define TC_H_UNSPEC (0U)
73 #define TC_H_ROOT (0xFFFFFFFFU)
74 #define TC_H_INGRESS (0xFFFFFFF1U)
76 /* Need to corrospond to iproute2 tc/tc_core.h "enum link_layer" */
78 TC_LINKLAYER_UNAWARE, /* Indicate unaware old iproute2 util */
79 TC_LINKLAYER_ETHERNET,
82 #define TC_LINKLAYER_MASK 0x0F /* limit use to lower 4 bits */
85 unsigned char cell_log;
86 __u8 linklayer; /* lower 4 bits */
87 unsigned short overhead;
93 #define TC_RTAB_SIZE 1024
96 unsigned char cell_log;
97 unsigned char size_log;
100 unsigned int linklayer;
113 #define TCA_STAB_MAX (__TCA_STAB_MAX - 1)
117 struct tc_fifo_qopt {
118 __u32 limit; /* Queue length: bytes for bfifo, packets for pfifo */
123 #define TCQ_PRIO_BANDS 16
124 #define TCQ_MIN_PRIO_BANDS 2
126 struct tc_prio_qopt {
127 int bands; /* Number of bands */
128 __u8 priomap[TC_PRIO_MAX+1]; /* Map: logical priority -> PRIO band */
133 struct tc_multiq_qopt {
134 __u16 bands; /* Number of bands */
135 __u16 max_bands; /* Maximum number of queues */
140 #define TCQ_PLUG_BUFFER 0
141 #define TCQ_PLUG_RELEASE_ONE 1
142 #define TCQ_PLUG_RELEASE_INDEFINITE 2
143 #define TCQ_PLUG_LIMIT 3
145 struct tc_plug_qopt {
146 /* TCQ_PLUG_BUFFER: Inset a plug into the queue and
147 * buffer any incoming packets
148 * TCQ_PLUG_RELEASE_ONE: Dequeue packets from queue head
149 * to beginning of the next plug.
150 * TCQ_PLUG_RELEASE_INDEFINITE: Dequeue all packets from queue.
151 * Stop buffering packets until the next TCQ_PLUG_BUFFER
152 * command is received (just act as a pass-thru queue).
153 * TCQ_PLUG_LIMIT: Increase/decrease queue size
162 struct tc_ratespec rate;
163 struct tc_ratespec peakrate;
177 #define TCA_TBF_MAX (__TCA_TBF_MAX - 1)
182 /* TEQL does not require any parameters */
187 unsigned quantum; /* Bytes per round allocated to flow */
188 int perturb_period; /* Period of hash perturbation */
189 __u32 limit; /* Maximal packets in queue */
190 unsigned divisor; /* Hash divisor */
191 unsigned flows; /* Maximal number of flows */
194 struct tc_sfqred_stats {
195 __u32 prob_drop; /* Early drops, below max threshold */
196 __u32 forced_drop; /* Early drops, after max threshold */
197 __u32 prob_mark; /* Marked packets, below max threshold */
198 __u32 forced_mark; /* Marked packets, after max threshold */
199 __u32 prob_mark_head; /* Marked packets, below max threshold */
200 __u32 forced_mark_head;/* Marked packets, after max threshold */
203 struct tc_sfq_qopt_v1 {
204 struct tc_sfq_qopt v0;
205 unsigned int depth; /* max number of packets per flow */
206 unsigned int headdrop;
207 /* SFQRED parameters */
208 __u32 limit; /* HARD maximal flow queue length (bytes) */
209 __u32 qth_min; /* Min average length threshold (bytes) */
210 __u32 qth_max; /* Max average length threshold (bytes) */
211 unsigned char Wlog; /* log(W) */
212 unsigned char Plog; /* log(P_max/(qth_max-qth_min)) */
213 unsigned char Scell_log; /* cell size for idle damping */
215 __u32 max_P; /* probability, high resolution */
217 struct tc_sfqred_stats stats;
221 struct tc_sfq_xstats {
235 #define TCA_RED_MAX (__TCA_RED_MAX - 1)
238 __u32 limit; /* HARD maximal queue length (bytes) */
239 __u32 qth_min; /* Min average length threshold (bytes) */
240 __u32 qth_max; /* Max average length threshold (bytes) */
241 unsigned char Wlog; /* log(W) */
242 unsigned char Plog; /* log(P_max/(qth_max-qth_min)) */
243 unsigned char Scell_log; /* cell size for idle damping */
246 #define TC_RED_HARDDROP 2
247 #define TC_RED_ADAPTATIVE 4
250 struct tc_red_xstats {
251 __u32 early; /* Early drops */
252 __u32 pdrop; /* Drops due to queue limits */
253 __u32 other; /* Drops due to drop() calls */
254 __u32 marked; /* Marked packets */
270 #define TCA_GRED_MAX (__TCA_GRED_MAX - 1)
272 struct tc_gred_qopt {
273 __u32 limit; /* HARD maximal queue length (bytes) */
274 __u32 qth_min; /* Min average length threshold (bytes) */
275 __u32 qth_max; /* Max average length threshold (bytes) */
276 __u32 DP; /* up to 2^32 DPs */
283 __u8 Wlog; /* log(W) */
284 __u8 Plog; /* log(P_max/(qth_max-qth_min)) */
285 __u8 Scell_log; /* cell size for idle damping */
286 __u8 prio; /* prio of this VQ */
292 struct tc_gred_sopt {
310 #define TCA_CHOKE_MAX (__TCA_CHOKE_MAX - 1)
312 struct tc_choke_qopt {
313 __u32 limit; /* Hard queue length (packets) */
314 __u32 qth_min; /* Min average threshold (packets) */
315 __u32 qth_max; /* Max average threshold (packets) */
316 unsigned char Wlog; /* log(W) */
317 unsigned char Plog; /* log(P_max/(qth_max-qth_min)) */
318 unsigned char Scell_log; /* cell size for idle damping */
319 unsigned char flags; /* see RED flags */
322 struct tc_choke_xstats {
323 __u32 early; /* Early drops */
324 __u32 pdrop; /* Drops due to queue limits */
325 __u32 other; /* Drops due to drop() calls */
326 __u32 marked; /* Marked packets */
327 __u32 matched; /* Drops due to flow match */
331 #define TC_HTB_NUMPRIO 8
332 #define TC_HTB_MAXDEPTH 8
333 #define TC_HTB_PROTOVER 3 /* the same as HTB and TC's major */
336 struct tc_ratespec rate;
337 struct tc_ratespec ceil;
341 __u32 level; /* out only */
345 __u32 version; /* to match HTB/TC */
346 __u32 rate2quantum; /* bps->quantum divisor */
347 __u32 defcls; /* default class number */
348 __u32 debug; /* debug flags */
351 __u32 direct_pkts; /* count of non shaped packets */
365 #define TCA_HTB_MAX (__TCA_HTB_MAX - 1)
367 struct tc_htb_xstats {
370 __u32 giants; /* too big packets (rate will not be accurate) */
377 struct tc_hfsc_qopt {
378 __u16 defcls; /* default class */
381 struct tc_service_curve {
382 __u32 m1; /* slope of the first segment in bps */
383 __u32 d; /* x-projection of the first segment in us */
384 __u32 m2; /* slope of the second segment in bps */
387 struct tc_hfsc_stats {
388 __u64 work; /* total work done */
389 __u64 rtwork; /* work done by real-time criteria */
390 __u32 period; /* current period */
391 __u32 level; /* class level in hierarchy */
402 #define TCA_HFSC_MAX (__TCA_HFSC_MAX - 1)
407 #define TC_CBQ_MAXPRIO 8
408 #define TC_CBQ_MAXLEVEL 8
409 #define TC_CBQ_DEF_EWMA 5
411 struct tc_cbq_lssopt {
412 unsigned char change;
414 #define TCF_CBQ_LSS_BOUNDED 1
415 #define TCF_CBQ_LSS_ISOLATED 2
416 unsigned char ewma_log;
418 #define TCF_CBQ_LSS_FLAGS 1
419 #define TCF_CBQ_LSS_EWMA 2
420 #define TCF_CBQ_LSS_MAXIDLE 4
421 #define TCF_CBQ_LSS_MINIDLE 8
422 #define TCF_CBQ_LSS_OFFTIME 0x10
423 #define TCF_CBQ_LSS_AVPKT 0x20
430 struct tc_cbq_wrropt {
432 unsigned char priority;
433 unsigned char cpriority;
434 unsigned char __reserved;
440 unsigned char strategy;
441 #define TC_CBQ_OVL_CLASSIC 0
442 #define TC_CBQ_OVL_DELAY 1
443 #define TC_CBQ_OVL_LOWPRIO 2
444 #define TC_CBQ_OVL_DROP 3
445 #define TC_CBQ_OVL_RCLASSIC 4
446 unsigned char priority2;
451 struct tc_cbq_police {
452 unsigned char police;
453 unsigned char __res1;
454 unsigned short __res2;
463 struct tc_cbq_xstats {
475 TCA_CBQ_OVL_STRATEGY,
482 #define TCA_CBQ_MAX (__TCA_CBQ_MAX - 1)
489 TCA_DSMARK_DEFAULT_INDEX,
490 TCA_DSMARK_SET_TC_INDEX,
496 #define TCA_DSMARK_MAX (__TCA_DSMARK_MAX - 1)
502 TCA_ATM_FD, /* file/socket descriptor */
503 TCA_ATM_PTR, /* pointer to descriptor - later */
504 TCA_ATM_HDR, /* LL header */
505 TCA_ATM_EXCESS, /* excess traffic class (0 for CLP) */
506 TCA_ATM_ADDR, /* PVC address (for output only) */
507 TCA_ATM_STATE, /* VC state (ATM_VS_*; for output only) */
511 #define TCA_ATM_MAX (__TCA_ATM_MAX - 1)
513 /* Network emulator */
518 TCA_NETEM_DELAY_DIST,
527 #define TCA_NETEM_MAX (__TCA_NETEM_MAX - 1)
529 struct tc_netem_qopt {
530 __u32 latency; /* added delay (us) */
531 __u32 limit; /* fifo limit (packets) */
532 __u32 loss; /* random packet loss (0=none ~0=100%) */
533 __u32 gap; /* re-ordering gap (0 for none) */
534 __u32 duplicate; /* random packet dup (0=none ~0=100%) */
535 __u32 jitter; /* random jitter in latency (us) */
538 struct tc_netem_corr {
539 __u32 delay_corr; /* delay correlation */
540 __u32 loss_corr; /* packet loss correlation */
541 __u32 dup_corr; /* duplicate correlation */
544 struct tc_netem_reorder {
549 struct tc_netem_corrupt {
554 struct tc_netem_rate {
555 __u32 rate; /* byte/s */
556 __s32 packet_overhead;
563 NETEM_LOSS_GI, /* General Intuitive - 4 state model */
564 NETEM_LOSS_GE, /* Gilbert Elliot models */
567 #define NETEM_LOSS_MAX (__NETEM_LOSS_MAX - 1)
569 /* State transition probabilities for 4 state model */
570 struct tc_netem_gimodel {
578 /* Gilbert-Elliot models */
579 struct tc_netem_gemodel {
586 #define NETEM_DIST_SCALE 8192
587 #define NETEM_DIST_MAX 16384
597 #define TCA_DRR_MAX (__TCA_DRR_MAX - 1)
599 struct tc_drr_stats {
604 #define TC_QOPT_BITMASK 15
605 #define TC_QOPT_MAX_QUEUE 16
607 struct tc_mqprio_qopt {
609 __u8 prio_tc_map[TC_QOPT_BITMASK + 1];
611 __u16 count[TC_QOPT_MAX_QUEUE];
612 __u16 offset[TC_QOPT_MAX_QUEUE];
623 #define TCA_SFB_MAX (__TCA_SFB_MAX - 1)
626 * Note: increment, decrement are Q0.16 fixed-point values.
629 __u32 rehash_interval; /* delay between hash move, in ms */
630 __u32 warmup_time; /* double buffering warmup time in ms (warmup_time < rehash_interval) */
631 __u32 max; /* max len of qlen_min */
632 __u32 bin_size; /* maximum queue length per bin */
633 __u32 increment; /* probability increment, (d1 in Blue) */
634 __u32 decrement; /* probability decrement, (d2 in Blue) */
635 __u32 limit; /* max SFB queue length */
636 __u32 penalty_rate; /* inelastic flows are rate limited to 'rate' pps */
640 struct tc_sfb_xstats {
645 __u32 childdrop; /* drops in child qdisc */
652 #define SFB_MAX_PROB 0xFFFF
662 #define TCA_QFQ_MAX (__TCA_QFQ_MAX - 1)
664 struct tc_qfq_stats {
680 #define TCA_CODEL_MAX (__TCA_CODEL_MAX - 1)
682 struct tc_codel_xstats {
683 __u32 maxpacket; /* largest packet we've seen so far */
684 __u32 count; /* how many drops we've done since the last time we
685 * entered dropping state
687 __u32 lastcount; /* count at entry to dropping state */
688 __u32 ldelay; /* in-queue delay seen by most recently dequeued packet */
689 __s32 drop_next; /* time to drop next packet */
690 __u32 drop_overlimit; /* number of time max qdisc packet limit was hit */
691 __u32 ecn_mark; /* number of packets we ECN marked instead of dropped */
692 __u32 dropping; /* are we in dropping state ? */
701 TCA_FQ_CODEL_INTERVAL,
704 TCA_FQ_CODEL_QUANTUM,
708 #define TCA_FQ_CODEL_MAX (__TCA_FQ_CODEL_MAX - 1)
711 TCA_FQ_CODEL_XSTATS_QDISC,
712 TCA_FQ_CODEL_XSTATS_CLASS,
715 struct tc_fq_codel_qd_stats {
716 __u32 maxpacket; /* largest packet we've seen so far */
717 __u32 drop_overlimit; /* number of time max qdisc
718 * packet limit was hit
720 __u32 ecn_mark; /* number of packets we ECN marked
721 * instead of being dropped
723 __u32 new_flow_count; /* number of time packets
724 * created a 'new flow'
726 __u32 new_flows_len; /* count of flows in new list */
727 __u32 old_flows_len; /* count of flows in old list */
730 struct tc_fq_codel_cl_stats {
732 __u32 ldelay; /* in-queue delay seen by most recently
741 struct tc_fq_codel_xstats {
744 struct tc_fq_codel_qd_stats qdisc_stats;
745 struct tc_fq_codel_cl_stats class_stats;
754 TCA_FQ_PLIMIT, /* limit of total number of packets in queue */
756 TCA_FQ_FLOW_PLIMIT, /* limit of packets per flow */
758 TCA_FQ_QUANTUM, /* RR quantum */
760 TCA_FQ_INITIAL_QUANTUM, /* RR quantum for new flow */
762 TCA_FQ_RATE_ENABLE, /* enable/disable rate limiting */
764 TCA_FQ_FLOW_DEFAULT_RATE,/* for sockets with unspecified sk_rate,
765 * use the following rate
768 TCA_FQ_FLOW_MAX_RATE, /* per flow max rate */
770 TCA_FQ_BUCKETS_LOG, /* log2(number of buckets) */
774 #define TCA_FQ_MAX (__TCA_FQ_MAX - 1)
776 struct tc_fq_qd_stats {
778 __u64 highprio_packets;
783 __u64 allocation_errors;
784 __s64 time_next_delayed_flow;
786 __u32 inactive_flows;
787 __u32 throttled_flows;