1 /* SPDX-License-Identifier: GPL-2.0 WITH Linux-syscall-note */
2 #ifndef __LINUX_PKT_SCHED_H
3 #define __LINUX_PKT_SCHED_H
5 #include <linux/types.h>
7 /* Logical priority bands not depending on specific packet scheduler.
8 Every scheduler will map them to real traffic classes, if it has
9 no more precise mechanism to classify packets.
11 These numbers have no special meaning, though their coincidence
12 with obsolete IPv6 values is not occasional :-). New IPv6 drafts
13 preferred full anarchy inspired by diffserv group.
15 Note: TC_PRIO_BESTEFFORT does not mean that it is the most unhappy
16 class, actually, as rule it will be handled with more care than
20 #define TC_PRIO_BESTEFFORT 0
21 #define TC_PRIO_FILLER 1
22 #define TC_PRIO_BULK 2
23 #define TC_PRIO_INTERACTIVE_BULK 4
24 #define TC_PRIO_INTERACTIVE 6
25 #define TC_PRIO_CONTROL 7
27 #define TC_PRIO_MAX 15
29 /* Generic queue statistics, available for all the elements.
30 Particular schedulers may have also their private records.
34 __u64 bytes; /* Number of enqueued bytes */
35 __u32 packets; /* Number of enqueued packets */
36 __u32 drops; /* Packets dropped because of lack of resources */
37 __u32 overlimits; /* Number of throttle events when this
38 * flow goes out of allocated bandwidth */
39 __u32 bps; /* Current flow byte rate */
40 __u32 pps; /* Current flow packet rate */
47 unsigned char ewma_log;
53 All the traffic control objects have 32bit identifiers, or "handles".
55 They can be considered as opaque numbers from user API viewpoint,
56 but actually they always consist of two fields: major and
57 minor numbers, which are interpreted by kernel specially,
58 that may be used by applications, though not recommended.
60 F.e. qdisc handles always have minor number equal to zero,
61 classes (or flows) have major equal to parent qdisc major, and
62 minor uniquely identifying class inside qdisc.
64 Macros to manipulate handles:
67 #define TC_H_MAJ_MASK (0xFFFF0000U)
68 #define TC_H_MIN_MASK (0x0000FFFFU)
69 #define TC_H_MAJ(h) ((h)&TC_H_MAJ_MASK)
70 #define TC_H_MIN(h) ((h)&TC_H_MIN_MASK)
71 #define TC_H_MAKE(maj,min) (((maj)&TC_H_MAJ_MASK)|((min)&TC_H_MIN_MASK))
73 #define TC_H_UNSPEC (0U)
74 #define TC_H_ROOT (0xFFFFFFFFU)
75 #define TC_H_INGRESS (0xFFFFFFF1U)
76 #define TC_H_CLSACT TC_H_INGRESS
78 #define TC_H_MIN_INGRESS 0xFFF2U
79 #define TC_H_MIN_EGRESS 0xFFF3U
81 /* Need to corrospond to iproute2 tc/tc_core.h "enum link_layer" */
83 TC_LINKLAYER_UNAWARE, /* Indicate unaware old iproute2 util */
84 TC_LINKLAYER_ETHERNET,
87 #define TC_LINKLAYER_MASK 0x0F /* limit use to lower 4 bits */
90 unsigned char cell_log;
91 __u8 linklayer; /* lower 4 bits */
92 unsigned short overhead;
98 #define TC_RTAB_SIZE 1024
101 unsigned char cell_log;
102 unsigned char size_log;
105 unsigned int linklayer;
118 #define TCA_STAB_MAX (__TCA_STAB_MAX - 1)
122 struct tc_fifo_qopt {
123 __u32 limit; /* Queue length: bytes for bfifo, packets for pfifo */
128 #define TCQ_PRIO_BANDS 16
129 #define TCQ_MIN_PRIO_BANDS 2
131 struct tc_prio_qopt {
132 int bands; /* Number of bands */
133 __u8 priomap[TC_PRIO_MAX+1]; /* Map: logical priority -> PRIO band */
138 struct tc_multiq_qopt {
139 __u16 bands; /* Number of bands */
140 __u16 max_bands; /* Maximum number of queues */
145 #define TCQ_PLUG_BUFFER 0
146 #define TCQ_PLUG_RELEASE_ONE 1
147 #define TCQ_PLUG_RELEASE_INDEFINITE 2
148 #define TCQ_PLUG_LIMIT 3
150 struct tc_plug_qopt {
151 /* TCQ_PLUG_BUFFER: Inset a plug into the queue and
152 * buffer any incoming packets
153 * TCQ_PLUG_RELEASE_ONE: Dequeue packets from queue head
154 * to beginning of the next plug.
155 * TCQ_PLUG_RELEASE_INDEFINITE: Dequeue all packets from queue.
156 * Stop buffering packets until the next TCQ_PLUG_BUFFER
157 * command is received (just act as a pass-thru queue).
158 * TCQ_PLUG_LIMIT: Increase/decrease queue size
167 struct tc_ratespec rate;
168 struct tc_ratespec peakrate;
187 #define TCA_TBF_MAX (__TCA_TBF_MAX - 1)
192 /* TEQL does not require any parameters */
197 unsigned quantum; /* Bytes per round allocated to flow */
198 int perturb_period; /* Period of hash perturbation */
199 __u32 limit; /* Maximal packets in queue */
200 unsigned divisor; /* Hash divisor */
201 unsigned flows; /* Maximal number of flows */
204 struct tc_sfqred_stats {
205 __u32 prob_drop; /* Early drops, below max threshold */
206 __u32 forced_drop; /* Early drops, after max threshold */
207 __u32 prob_mark; /* Marked packets, below max threshold */
208 __u32 forced_mark; /* Marked packets, after max threshold */
209 __u32 prob_mark_head; /* Marked packets, below max threshold */
210 __u32 forced_mark_head;/* Marked packets, after max threshold */
213 struct tc_sfq_qopt_v1 {
214 struct tc_sfq_qopt v0;
215 unsigned int depth; /* max number of packets per flow */
216 unsigned int headdrop;
217 /* SFQRED parameters */
218 __u32 limit; /* HARD maximal flow queue length (bytes) */
219 __u32 qth_min; /* Min average length threshold (bytes) */
220 __u32 qth_max; /* Max average length threshold (bytes) */
221 unsigned char Wlog; /* log(W) */
222 unsigned char Plog; /* log(P_max/(qth_max-qth_min)) */
223 unsigned char Scell_log; /* cell size for idle damping */
225 __u32 max_P; /* probability, high resolution */
227 struct tc_sfqred_stats stats;
231 struct tc_sfq_xstats {
245 #define TCA_RED_MAX (__TCA_RED_MAX - 1)
248 __u32 limit; /* HARD maximal queue length (bytes) */
249 __u32 qth_min; /* Min average length threshold (bytes) */
250 __u32 qth_max; /* Max average length threshold (bytes) */
251 unsigned char Wlog; /* log(W) */
252 unsigned char Plog; /* log(P_max/(qth_max-qth_min)) */
253 unsigned char Scell_log; /* cell size for idle damping */
256 #define TC_RED_HARDDROP 2
257 #define TC_RED_ADAPTATIVE 4
260 struct tc_red_xstats {
261 __u32 early; /* Early drops */
262 __u32 pdrop; /* Drops due to queue limits */
263 __u32 other; /* Drops due to drop() calls */
264 __u32 marked; /* Marked packets */
281 #define TCA_GRED_MAX (__TCA_GRED_MAX - 1)
283 struct tc_gred_qopt {
284 __u32 limit; /* HARD maximal queue length (bytes) */
285 __u32 qth_min; /* Min average length threshold (bytes) */
286 __u32 qth_max; /* Max average length threshold (bytes) */
287 __u32 DP; /* up to 2^32 DPs */
294 __u8 Wlog; /* log(W) */
295 __u8 Plog; /* log(P_max/(qth_max-qth_min)) */
296 __u8 Scell_log; /* cell size for idle damping */
297 __u8 prio; /* prio of this VQ */
303 struct tc_gred_sopt {
321 #define TCA_CHOKE_MAX (__TCA_CHOKE_MAX - 1)
323 struct tc_choke_qopt {
324 __u32 limit; /* Hard queue length (packets) */
325 __u32 qth_min; /* Min average threshold (packets) */
326 __u32 qth_max; /* Max average threshold (packets) */
327 unsigned char Wlog; /* log(W) */
328 unsigned char Plog; /* log(P_max/(qth_max-qth_min)) */
329 unsigned char Scell_log; /* cell size for idle damping */
330 unsigned char flags; /* see RED flags */
333 struct tc_choke_xstats {
334 __u32 early; /* Early drops */
335 __u32 pdrop; /* Drops due to queue limits */
336 __u32 other; /* Drops due to drop() calls */
337 __u32 marked; /* Marked packets */
338 __u32 matched; /* Drops due to flow match */
342 #define TC_HTB_NUMPRIO 8
343 #define TC_HTB_MAXDEPTH 8
344 #define TC_HTB_PROTOVER 3 /* the same as HTB and TC's major */
347 struct tc_ratespec rate;
348 struct tc_ratespec ceil;
352 __u32 level; /* out only */
356 __u32 version; /* to match HTB/TC */
357 __u32 rate2quantum; /* bps->quantum divisor */
358 __u32 defcls; /* default class number */
359 __u32 debug; /* debug flags */
362 __u32 direct_pkts; /* count of non shaped packets */
377 #define TCA_HTB_MAX (__TCA_HTB_MAX - 1)
379 struct tc_htb_xstats {
382 __u32 giants; /* too big packets (rate will not be accurate) */
389 struct tc_hfsc_qopt {
390 __u16 defcls; /* default class */
393 struct tc_service_curve {
394 __u32 m1; /* slope of the first segment in bps */
395 __u32 d; /* x-projection of the first segment in us */
396 __u32 m2; /* slope of the second segment in bps */
399 struct tc_hfsc_stats {
400 __u64 work; /* total work done */
401 __u64 rtwork; /* work done by real-time criteria */
402 __u32 period; /* current period */
403 __u32 level; /* class level in hierarchy */
414 #define TCA_HFSC_MAX (__TCA_HFSC_MAX - 1)
419 #define TC_CBQ_MAXPRIO 8
420 #define TC_CBQ_MAXLEVEL 8
421 #define TC_CBQ_DEF_EWMA 5
423 struct tc_cbq_lssopt {
424 unsigned char change;
426 #define TCF_CBQ_LSS_BOUNDED 1
427 #define TCF_CBQ_LSS_ISOLATED 2
428 unsigned char ewma_log;
430 #define TCF_CBQ_LSS_FLAGS 1
431 #define TCF_CBQ_LSS_EWMA 2
432 #define TCF_CBQ_LSS_MAXIDLE 4
433 #define TCF_CBQ_LSS_MINIDLE 8
434 #define TCF_CBQ_LSS_OFFTIME 0x10
435 #define TCF_CBQ_LSS_AVPKT 0x20
442 struct tc_cbq_wrropt {
444 unsigned char priority;
445 unsigned char cpriority;
446 unsigned char __reserved;
452 unsigned char strategy;
453 #define TC_CBQ_OVL_CLASSIC 0
454 #define TC_CBQ_OVL_DELAY 1
455 #define TC_CBQ_OVL_LOWPRIO 2
456 #define TC_CBQ_OVL_DROP 3
457 #define TC_CBQ_OVL_RCLASSIC 4
458 unsigned char priority2;
463 struct tc_cbq_police {
464 unsigned char police;
465 unsigned char __res1;
466 unsigned short __res2;
475 struct tc_cbq_xstats {
487 TCA_CBQ_OVL_STRATEGY,
494 #define TCA_CBQ_MAX (__TCA_CBQ_MAX - 1)
501 TCA_DSMARK_DEFAULT_INDEX,
502 TCA_DSMARK_SET_TC_INDEX,
508 #define TCA_DSMARK_MAX (__TCA_DSMARK_MAX - 1)
514 TCA_ATM_FD, /* file/socket descriptor */
515 TCA_ATM_PTR, /* pointer to descriptor - later */
516 TCA_ATM_HDR, /* LL header */
517 TCA_ATM_EXCESS, /* excess traffic class (0 for CLP) */
518 TCA_ATM_ADDR, /* PVC address (for output only) */
519 TCA_ATM_STATE, /* VC state (ATM_VS_*; for output only) */
523 #define TCA_ATM_MAX (__TCA_ATM_MAX - 1)
525 /* Network emulator */
530 TCA_NETEM_DELAY_DIST,
541 #define TCA_NETEM_MAX (__TCA_NETEM_MAX - 1)
543 struct tc_netem_qopt {
544 __u32 latency; /* added delay (us) */
545 __u32 limit; /* fifo limit (packets) */
546 __u32 loss; /* random packet loss (0=none ~0=100%) */
547 __u32 gap; /* re-ordering gap (0 for none) */
548 __u32 duplicate; /* random packet dup (0=none ~0=100%) */
549 __u32 jitter; /* random jitter in latency (us) */
552 struct tc_netem_corr {
553 __u32 delay_corr; /* delay correlation */
554 __u32 loss_corr; /* packet loss correlation */
555 __u32 dup_corr; /* duplicate correlation */
558 struct tc_netem_reorder {
563 struct tc_netem_corrupt {
568 struct tc_netem_rate {
569 __u32 rate; /* byte/s */
570 __s32 packet_overhead;
577 NETEM_LOSS_GI, /* General Intuitive - 4 state model */
578 NETEM_LOSS_GE, /* Gilbert Elliot models */
581 #define NETEM_LOSS_MAX (__NETEM_LOSS_MAX - 1)
583 /* State transition probabilities for 4 state model */
584 struct tc_netem_gimodel {
592 /* Gilbert-Elliot models */
593 struct tc_netem_gemodel {
600 #define NETEM_DIST_SCALE 8192
601 #define NETEM_DIST_MAX 16384
611 #define TCA_DRR_MAX (__TCA_DRR_MAX - 1)
613 struct tc_drr_stats {
618 #define TC_QOPT_BITMASK 15
619 #define TC_QOPT_MAX_QUEUE 16
622 TC_MQPRIO_HW_OFFLOAD_NONE, /* no offload requested */
623 TC_MQPRIO_HW_OFFLOAD_TCS, /* offload TCs, no queue counts */
624 __TC_MQPRIO_HW_OFFLOAD_MAX
627 #define TC_MQPRIO_HW_OFFLOAD_MAX (__TC_MQPRIO_HW_OFFLOAD_MAX - 1)
629 struct tc_mqprio_qopt {
631 __u8 prio_tc_map[TC_QOPT_BITMASK + 1];
633 __u16 count[TC_QOPT_MAX_QUEUE];
634 __u16 offset[TC_QOPT_MAX_QUEUE];
645 #define TCA_SFB_MAX (__TCA_SFB_MAX - 1)
648 * Note: increment, decrement are Q0.16 fixed-point values.
651 __u32 rehash_interval; /* delay between hash move, in ms */
652 __u32 warmup_time; /* double buffering warmup time in ms (warmup_time < rehash_interval) */
653 __u32 max; /* max len of qlen_min */
654 __u32 bin_size; /* maximum queue length per bin */
655 __u32 increment; /* probability increment, (d1 in Blue) */
656 __u32 decrement; /* probability decrement, (d2 in Blue) */
657 __u32 limit; /* max SFB queue length */
658 __u32 penalty_rate; /* inelastic flows are rate limited to 'rate' pps */
662 struct tc_sfb_xstats {
667 __u32 childdrop; /* drops in child qdisc */
674 #define SFB_MAX_PROB 0xFFFF
684 #define TCA_QFQ_MAX (__TCA_QFQ_MAX - 1)
686 struct tc_qfq_stats {
699 TCA_CODEL_CE_THRESHOLD,
703 #define TCA_CODEL_MAX (__TCA_CODEL_MAX - 1)
705 struct tc_codel_xstats {
706 __u32 maxpacket; /* largest packet we've seen so far */
707 __u32 count; /* how many drops we've done since the last time we
708 * entered dropping state
710 __u32 lastcount; /* count at entry to dropping state */
711 __u32 ldelay; /* in-queue delay seen by most recently dequeued packet */
712 __s32 drop_next; /* time to drop next packet */
713 __u32 drop_overlimit; /* number of time max qdisc packet limit was hit */
714 __u32 ecn_mark; /* number of packets we ECN marked instead of dropped */
715 __u32 dropping; /* are we in dropping state ? */
716 __u32 ce_mark; /* number of CE marked packets because of ce_threshold */
725 TCA_FQ_CODEL_INTERVAL,
728 TCA_FQ_CODEL_QUANTUM,
729 TCA_FQ_CODEL_CE_THRESHOLD,
730 TCA_FQ_CODEL_DROP_BATCH_SIZE,
731 TCA_FQ_CODEL_MEMORY_LIMIT,
735 #define TCA_FQ_CODEL_MAX (__TCA_FQ_CODEL_MAX - 1)
738 TCA_FQ_CODEL_XSTATS_QDISC,
739 TCA_FQ_CODEL_XSTATS_CLASS,
742 struct tc_fq_codel_qd_stats {
743 __u32 maxpacket; /* largest packet we've seen so far */
744 __u32 drop_overlimit; /* number of time max qdisc
745 * packet limit was hit
747 __u32 ecn_mark; /* number of packets we ECN marked
748 * instead of being dropped
750 __u32 new_flow_count; /* number of time packets
751 * created a 'new flow'
753 __u32 new_flows_len; /* count of flows in new list */
754 __u32 old_flows_len; /* count of flows in old list */
755 __u32 ce_mark; /* packets above ce_threshold */
756 __u32 memory_usage; /* in bytes */
757 __u32 drop_overmemory;
760 struct tc_fq_codel_cl_stats {
762 __u32 ldelay; /* in-queue delay seen by most recently
771 struct tc_fq_codel_xstats {
774 struct tc_fq_codel_qd_stats qdisc_stats;
775 struct tc_fq_codel_cl_stats class_stats;
784 TCA_FQ_PLIMIT, /* limit of total number of packets in queue */
786 TCA_FQ_FLOW_PLIMIT, /* limit of packets per flow */
788 TCA_FQ_QUANTUM, /* RR quantum */
790 TCA_FQ_INITIAL_QUANTUM, /* RR quantum for new flow */
792 TCA_FQ_RATE_ENABLE, /* enable/disable rate limiting */
794 TCA_FQ_FLOW_DEFAULT_RATE,/* obsolete, do not use */
796 TCA_FQ_FLOW_MAX_RATE, /* per flow max rate */
798 TCA_FQ_BUCKETS_LOG, /* log2(number of buckets) */
800 TCA_FQ_FLOW_REFILL_DELAY, /* flow credit refill delay in usec */
802 TCA_FQ_ORPHAN_MASK, /* mask applied to orphaned skb hashes */
804 TCA_FQ_LOW_RATE_THRESHOLD, /* per packet delay under this rate */
809 #define TCA_FQ_MAX (__TCA_FQ_MAX - 1)
811 struct tc_fq_qd_stats {
813 __u64 highprio_packets;
818 __u64 allocation_errors;
819 __s64 time_next_delayed_flow;
821 __u32 inactive_flows;
822 __u32 throttled_flows;
823 __u32 unthrottle_latency_ns;
826 /* Heavy-Hitter Filter */
830 TCA_HHF_BACKLOG_LIMIT,
832 TCA_HHF_HH_FLOWS_LIMIT,
833 TCA_HHF_RESET_TIMEOUT,
835 TCA_HHF_EVICT_TIMEOUT,
836 TCA_HHF_NON_HH_WEIGHT,
840 #define TCA_HHF_MAX (__TCA_HHF_MAX - 1)
842 struct tc_hhf_xstats {
843 __u32 drop_overlimit; /* number of times max qdisc packet limit
846 __u32 hh_overlimit; /* number of times max heavy-hitters was hit */
847 __u32 hh_tot_count; /* number of captured heavy-hitters so far */
848 __u32 hh_cur_count; /* number of current heavy-hitters */
863 #define TCA_PIE_MAX (__TCA_PIE_MAX - 1)
865 struct tc_pie_xstats {
866 __u32 prob; /* current probability */
867 __u32 delay; /* current delay in ms */
868 __u32 avg_dq_rate; /* current average dq_rate in bits/pie_time */
869 __u32 packets_in; /* total number of packets enqueued */
870 __u32 dropped; /* packets dropped due to pie_action */
871 __u32 overlimit; /* dropped due to lack of space in queue */
872 __u32 maxq; /* maximum queue size */
873 __u32 ecn_mark; /* packets marked with ecn*/