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_PRIORITY 0xFFE0U
79 #define TC_H_MIN_INGRESS 0xFFF2U
80 #define TC_H_MIN_EGRESS 0xFFF3U
82 /* Need to corrospond to iproute2 tc/tc_core.h "enum link_layer" */
84 TC_LINKLAYER_UNAWARE
, /* Indicate unaware old iproute2 util */
85 TC_LINKLAYER_ETHERNET
,
88 #define TC_LINKLAYER_MASK 0x0F /* limit use to lower 4 bits */
91 unsigned char cell_log
;
92 __u8 linklayer
; /* lower 4 bits */
93 unsigned short overhead
;
99 #define TC_RTAB_SIZE 1024
102 unsigned char cell_log
;
103 unsigned char size_log
;
106 unsigned int linklayer
;
119 #define TCA_STAB_MAX (__TCA_STAB_MAX - 1)
123 struct tc_fifo_qopt
{
124 __u32 limit
; /* Queue length: bytes for bfifo, packets for pfifo */
129 #define TCQ_PRIO_BANDS 16
130 #define TCQ_MIN_PRIO_BANDS 2
132 struct tc_prio_qopt
{
133 int bands
; /* Number of bands */
134 __u8 priomap
[TC_PRIO_MAX
+1]; /* Map: logical priority -> PRIO band */
139 struct tc_multiq_qopt
{
140 __u16 bands
; /* Number of bands */
141 __u16 max_bands
; /* Maximum number of queues */
146 #define TCQ_PLUG_BUFFER 0
147 #define TCQ_PLUG_RELEASE_ONE 1
148 #define TCQ_PLUG_RELEASE_INDEFINITE 2
149 #define TCQ_PLUG_LIMIT 3
151 struct tc_plug_qopt
{
152 /* TCQ_PLUG_BUFFER: Inset a plug into the queue and
153 * buffer any incoming packets
154 * TCQ_PLUG_RELEASE_ONE: Dequeue packets from queue head
155 * to beginning of the next plug.
156 * TCQ_PLUG_RELEASE_INDEFINITE: Dequeue all packets from queue.
157 * Stop buffering packets until the next TCQ_PLUG_BUFFER
158 * command is received (just act as a pass-thru queue).
159 * TCQ_PLUG_LIMIT: Increase/decrease queue size
168 struct tc_ratespec rate
;
169 struct tc_ratespec peakrate
;
188 #define TCA_TBF_MAX (__TCA_TBF_MAX - 1)
193 /* TEQL does not require any parameters */
198 unsigned quantum
; /* Bytes per round allocated to flow */
199 int perturb_period
; /* Period of hash perturbation */
200 __u32 limit
; /* Maximal packets in queue */
201 unsigned divisor
; /* Hash divisor */
202 unsigned flows
; /* Maximal number of flows */
205 struct tc_sfqred_stats
{
206 __u32 prob_drop
; /* Early drops, below max threshold */
207 __u32 forced_drop
; /* Early drops, after max threshold */
208 __u32 prob_mark
; /* Marked packets, below max threshold */
209 __u32 forced_mark
; /* Marked packets, after max threshold */
210 __u32 prob_mark_head
; /* Marked packets, below max threshold */
211 __u32 forced_mark_head
;/* Marked packets, after max threshold */
214 struct tc_sfq_qopt_v1
{
215 struct tc_sfq_qopt v0
;
216 unsigned int depth
; /* max number of packets per flow */
217 unsigned int headdrop
;
218 /* SFQRED parameters */
219 __u32 limit
; /* HARD maximal flow queue length (bytes) */
220 __u32 qth_min
; /* Min average length threshold (bytes) */
221 __u32 qth_max
; /* Max average length threshold (bytes) */
222 unsigned char Wlog
; /* log(W) */
223 unsigned char Plog
; /* log(P_max/(qth_max-qth_min)) */
224 unsigned char Scell_log
; /* cell size for idle damping */
226 __u32 max_P
; /* probability, high resolution */
228 struct tc_sfqred_stats stats
;
232 struct tc_sfq_xstats
{
246 #define TCA_RED_MAX (__TCA_RED_MAX - 1)
249 __u32 limit
; /* HARD maximal queue length (bytes) */
250 __u32 qth_min
; /* Min average length threshold (bytes) */
251 __u32 qth_max
; /* Max average length threshold (bytes) */
252 unsigned char Wlog
; /* log(W) */
253 unsigned char Plog
; /* log(P_max/(qth_max-qth_min)) */
254 unsigned char Scell_log
; /* cell size for idle damping */
257 #define TC_RED_HARDDROP 2
258 #define TC_RED_ADAPTATIVE 4
261 struct tc_red_xstats
{
262 __u32 early
; /* Early drops */
263 __u32 pdrop
; /* Drops due to queue limits */
264 __u32 other
; /* Drops due to drop() calls */
265 __u32 marked
; /* Marked packets */
282 #define TCA_GRED_MAX (__TCA_GRED_MAX - 1)
284 struct tc_gred_qopt
{
285 __u32 limit
; /* HARD maximal queue length (bytes) */
286 __u32 qth_min
; /* Min average length threshold (bytes) */
287 __u32 qth_max
; /* Max average length threshold (bytes) */
288 __u32 DP
; /* up to 2^32 DPs */
295 __u8 Wlog
; /* log(W) */
296 __u8 Plog
; /* log(P_max/(qth_max-qth_min)) */
297 __u8 Scell_log
; /* cell size for idle damping */
298 __u8 prio
; /* prio of this VQ */
304 struct tc_gred_sopt
{
322 #define TCA_CHOKE_MAX (__TCA_CHOKE_MAX - 1)
324 struct tc_choke_qopt
{
325 __u32 limit
; /* Hard queue length (packets) */
326 __u32 qth_min
; /* Min average threshold (packets) */
327 __u32 qth_max
; /* Max average threshold (packets) */
328 unsigned char Wlog
; /* log(W) */
329 unsigned char Plog
; /* log(P_max/(qth_max-qth_min)) */
330 unsigned char Scell_log
; /* cell size for idle damping */
331 unsigned char flags
; /* see RED flags */
334 struct tc_choke_xstats
{
335 __u32 early
; /* Early drops */
336 __u32 pdrop
; /* Drops due to queue limits */
337 __u32 other
; /* Drops due to drop() calls */
338 __u32 marked
; /* Marked packets */
339 __u32 matched
; /* Drops due to flow match */
343 #define TC_HTB_NUMPRIO 8
344 #define TC_HTB_MAXDEPTH 8
345 #define TC_HTB_PROTOVER 3 /* the same as HTB and TC's major */
348 struct tc_ratespec rate
;
349 struct tc_ratespec ceil
;
353 __u32 level
; /* out only */
357 __u32 version
; /* to match HTB/TC */
358 __u32 rate2quantum
; /* bps->quantum divisor */
359 __u32 defcls
; /* default class number */
360 __u32 debug
; /* debug flags */
363 __u32 direct_pkts
; /* count of non shaped packets */
378 #define TCA_HTB_MAX (__TCA_HTB_MAX - 1)
380 struct tc_htb_xstats
{
383 __u32 giants
; /* too big packets (rate will not be accurate) */
390 struct tc_hfsc_qopt
{
391 __u16 defcls
; /* default class */
394 struct tc_service_curve
{
395 __u32 m1
; /* slope of the first segment in bps */
396 __u32 d
; /* x-projection of the first segment in us */
397 __u32 m2
; /* slope of the second segment in bps */
400 struct tc_hfsc_stats
{
401 __u64 work
; /* total work done */
402 __u64 rtwork
; /* work done by real-time criteria */
403 __u32 period
; /* current period */
404 __u32 level
; /* class level in hierarchy */
415 #define TCA_HFSC_MAX (__TCA_HFSC_MAX - 1)
420 #define TC_CBQ_MAXPRIO 8
421 #define TC_CBQ_MAXLEVEL 8
422 #define TC_CBQ_DEF_EWMA 5
424 struct tc_cbq_lssopt
{
425 unsigned char change
;
427 #define TCF_CBQ_LSS_BOUNDED 1
428 #define TCF_CBQ_LSS_ISOLATED 2
429 unsigned char ewma_log
;
431 #define TCF_CBQ_LSS_FLAGS 1
432 #define TCF_CBQ_LSS_EWMA 2
433 #define TCF_CBQ_LSS_MAXIDLE 4
434 #define TCF_CBQ_LSS_MINIDLE 8
435 #define TCF_CBQ_LSS_OFFTIME 0x10
436 #define TCF_CBQ_LSS_AVPKT 0x20
443 struct tc_cbq_wrropt
{
445 unsigned char priority
;
446 unsigned char cpriority
;
447 unsigned char __reserved
;
453 unsigned char strategy
;
454 #define TC_CBQ_OVL_CLASSIC 0
455 #define TC_CBQ_OVL_DELAY 1
456 #define TC_CBQ_OVL_LOWPRIO 2
457 #define TC_CBQ_OVL_DROP 3
458 #define TC_CBQ_OVL_RCLASSIC 4
459 unsigned char priority2
;
464 struct tc_cbq_police
{
465 unsigned char police
;
466 unsigned char __res1
;
467 unsigned short __res2
;
476 struct tc_cbq_xstats
{
488 TCA_CBQ_OVL_STRATEGY
,
495 #define TCA_CBQ_MAX (__TCA_CBQ_MAX - 1)
502 TCA_DSMARK_DEFAULT_INDEX
,
503 TCA_DSMARK_SET_TC_INDEX
,
509 #define TCA_DSMARK_MAX (__TCA_DSMARK_MAX - 1)
515 TCA_ATM_FD
, /* file/socket descriptor */
516 TCA_ATM_PTR
, /* pointer to descriptor - later */
517 TCA_ATM_HDR
, /* LL header */
518 TCA_ATM_EXCESS
, /* excess traffic class (0 for CLP) */
519 TCA_ATM_ADDR
, /* PVC address (for output only) */
520 TCA_ATM_STATE
, /* VC state (ATM_VS_*; for output only) */
524 #define TCA_ATM_MAX (__TCA_ATM_MAX - 1)
526 /* Network emulator */
531 TCA_NETEM_DELAY_DIST
,
545 #define TCA_NETEM_MAX (__TCA_NETEM_MAX - 1)
547 struct tc_netem_qopt
{
548 __u32 latency
; /* added delay (us) */
549 __u32 limit
; /* fifo limit (packets) */
550 __u32 loss
; /* random packet loss (0=none ~0=100%) */
551 __u32 gap
; /* re-ordering gap (0 for none) */
552 __u32 duplicate
; /* random packet dup (0=none ~0=100%) */
553 __u32 jitter
; /* random jitter in latency (us) */
556 struct tc_netem_corr
{
557 __u32 delay_corr
; /* delay correlation */
558 __u32 loss_corr
; /* packet loss correlation */
559 __u32 dup_corr
; /* duplicate correlation */
562 struct tc_netem_reorder
{
567 struct tc_netem_corrupt
{
572 struct tc_netem_rate
{
573 __u32 rate
; /* byte/s */
574 __s32 packet_overhead
;
579 struct tc_netem_slot
{
580 __s64 min_delay
; /* nsec */
588 NETEM_LOSS_GI
, /* General Intuitive - 4 state model */
589 NETEM_LOSS_GE
, /* Gilbert Elliot models */
592 #define NETEM_LOSS_MAX (__NETEM_LOSS_MAX - 1)
594 /* State transition probabilities for 4 state model */
595 struct tc_netem_gimodel
{
603 /* Gilbert-Elliot models */
604 struct tc_netem_gemodel
{
611 #define NETEM_DIST_SCALE 8192
612 #define NETEM_DIST_MAX 16384
622 #define TCA_DRR_MAX (__TCA_DRR_MAX - 1)
624 struct tc_drr_stats
{
629 #define TC_QOPT_BITMASK 15
630 #define TC_QOPT_MAX_QUEUE 16
633 TC_MQPRIO_HW_OFFLOAD_NONE
, /* no offload requested */
634 TC_MQPRIO_HW_OFFLOAD_TCS
, /* offload TCs, no queue counts */
635 __TC_MQPRIO_HW_OFFLOAD_MAX
638 #define TC_MQPRIO_HW_OFFLOAD_MAX (__TC_MQPRIO_HW_OFFLOAD_MAX - 1)
642 TC_MQPRIO_MODE_CHANNEL
,
646 #define __TC_MQPRIO_MODE_MAX (__TC_MQPRIO_MODE_MAX - 1)
649 TC_MQPRIO_SHAPER_DCB
,
650 TC_MQPRIO_SHAPER_BW_RATE
, /* Add new shapers below */
651 __TC_MQPRIO_SHAPER_MAX
654 #define __TC_MQPRIO_SHAPER_MAX (__TC_MQPRIO_SHAPER_MAX - 1)
656 struct tc_mqprio_qopt
{
658 __u8 prio_tc_map
[TC_QOPT_BITMASK
+ 1];
660 __u16 count
[TC_QOPT_MAX_QUEUE
];
661 __u16 offset
[TC_QOPT_MAX_QUEUE
];
664 #define TC_MQPRIO_F_MODE 0x1
665 #define TC_MQPRIO_F_SHAPER 0x2
666 #define TC_MQPRIO_F_MIN_RATE 0x4
667 #define TC_MQPRIO_F_MAX_RATE 0x8
673 TCA_MQPRIO_MIN_RATE64
,
674 TCA_MQPRIO_MAX_RATE64
,
678 #define TCA_MQPRIO_MAX (__TCA_MQPRIO_MAX - 1)
688 #define TCA_SFB_MAX (__TCA_SFB_MAX - 1)
691 * Note: increment, decrement are Q0.16 fixed-point values.
694 __u32 rehash_interval
; /* delay between hash move, in ms */
695 __u32 warmup_time
; /* double buffering warmup time in ms (warmup_time < rehash_interval) */
696 __u32 max
; /* max len of qlen_min */
697 __u32 bin_size
; /* maximum queue length per bin */
698 __u32 increment
; /* probability increment, (d1 in Blue) */
699 __u32 decrement
; /* probability decrement, (d2 in Blue) */
700 __u32 limit
; /* max SFB queue length */
701 __u32 penalty_rate
; /* inelastic flows are rate limited to 'rate' pps */
705 struct tc_sfb_xstats
{
710 __u32 childdrop
; /* drops in child qdisc */
717 #define SFB_MAX_PROB 0xFFFF
727 #define TCA_QFQ_MAX (__TCA_QFQ_MAX - 1)
729 struct tc_qfq_stats
{
742 TCA_CODEL_CE_THRESHOLD
,
746 #define TCA_CODEL_MAX (__TCA_CODEL_MAX - 1)
748 struct tc_codel_xstats
{
749 __u32 maxpacket
; /* largest packet we've seen so far */
750 __u32 count
; /* how many drops we've done since the last time we
751 * entered dropping state
753 __u32 lastcount
; /* count at entry to dropping state */
754 __u32 ldelay
; /* in-queue delay seen by most recently dequeued packet */
755 __s32 drop_next
; /* time to drop next packet */
756 __u32 drop_overlimit
; /* number of time max qdisc packet limit was hit */
757 __u32 ecn_mark
; /* number of packets we ECN marked instead of dropped */
758 __u32 dropping
; /* are we in dropping state ? */
759 __u32 ce_mark
; /* number of CE marked packets because of ce_threshold */
768 TCA_FQ_CODEL_INTERVAL
,
771 TCA_FQ_CODEL_QUANTUM
,
772 TCA_FQ_CODEL_CE_THRESHOLD
,
773 TCA_FQ_CODEL_DROP_BATCH_SIZE
,
774 TCA_FQ_CODEL_MEMORY_LIMIT
,
778 #define TCA_FQ_CODEL_MAX (__TCA_FQ_CODEL_MAX - 1)
781 TCA_FQ_CODEL_XSTATS_QDISC
,
782 TCA_FQ_CODEL_XSTATS_CLASS
,
785 struct tc_fq_codel_qd_stats
{
786 __u32 maxpacket
; /* largest packet we've seen so far */
787 __u32 drop_overlimit
; /* number of time max qdisc
788 * packet limit was hit
790 __u32 ecn_mark
; /* number of packets we ECN marked
791 * instead of being dropped
793 __u32 new_flow_count
; /* number of time packets
794 * created a 'new flow'
796 __u32 new_flows_len
; /* count of flows in new list */
797 __u32 old_flows_len
; /* count of flows in old list */
798 __u32 ce_mark
; /* packets above ce_threshold */
799 __u32 memory_usage
; /* in bytes */
800 __u32 drop_overmemory
;
803 struct tc_fq_codel_cl_stats
{
805 __u32 ldelay
; /* in-queue delay seen by most recently
814 struct tc_fq_codel_xstats
{
817 struct tc_fq_codel_qd_stats qdisc_stats
;
818 struct tc_fq_codel_cl_stats class_stats
;
827 TCA_FQ_PLIMIT
, /* limit of total number of packets in queue */
829 TCA_FQ_FLOW_PLIMIT
, /* limit of packets per flow */
831 TCA_FQ_QUANTUM
, /* RR quantum */
833 TCA_FQ_INITIAL_QUANTUM
, /* RR quantum for new flow */
835 TCA_FQ_RATE_ENABLE
, /* enable/disable rate limiting */
837 TCA_FQ_FLOW_DEFAULT_RATE
,/* obsolete, do not use */
839 TCA_FQ_FLOW_MAX_RATE
, /* per flow max rate */
841 TCA_FQ_BUCKETS_LOG
, /* log2(number of buckets) */
843 TCA_FQ_FLOW_REFILL_DELAY
, /* flow credit refill delay in usec */
845 TCA_FQ_ORPHAN_MASK
, /* mask applied to orphaned skb hashes */
847 TCA_FQ_LOW_RATE_THRESHOLD
, /* per packet delay under this rate */
852 #define TCA_FQ_MAX (__TCA_FQ_MAX - 1)
854 struct tc_fq_qd_stats
{
856 __u64 highprio_packets
;
861 __u64 allocation_errors
;
862 __s64 time_next_delayed_flow
;
864 __u32 inactive_flows
;
865 __u32 throttled_flows
;
866 __u32 unthrottle_latency_ns
;
869 /* Heavy-Hitter Filter */
873 TCA_HHF_BACKLOG_LIMIT
,
875 TCA_HHF_HH_FLOWS_LIMIT
,
876 TCA_HHF_RESET_TIMEOUT
,
878 TCA_HHF_EVICT_TIMEOUT
,
879 TCA_HHF_NON_HH_WEIGHT
,
883 #define TCA_HHF_MAX (__TCA_HHF_MAX - 1)
885 struct tc_hhf_xstats
{
886 __u32 drop_overlimit
; /* number of times max qdisc packet limit
889 __u32 hh_overlimit
; /* number of times max heavy-hitters was hit */
890 __u32 hh_tot_count
; /* number of captured heavy-hitters so far */
891 __u32 hh_cur_count
; /* number of current heavy-hitters */
906 #define TCA_PIE_MAX (__TCA_PIE_MAX - 1)
908 struct tc_pie_xstats
{
909 __u32 prob
; /* current probability */
910 __u32 delay
; /* current delay in ms */
911 __u32 avg_dq_rate
; /* current average dq_rate in bits/pie_time */
912 __u32 packets_in
; /* total number of packets enqueued */
913 __u32 dropped
; /* packets dropped due to pie_action */
914 __u32 overlimit
; /* dropped due to lack of space in queue */
915 __u32 maxq
; /* maximum queue size */
916 __u32 ecn_mark
; /* packets marked with ecn*/
935 #define TCA_CBS_MAX (__TCA_CBS_MAX - 1)