1TC(8) Linux TC(8)
2
6 tc - show / manipulate traffic control settings
7
9 tc [ OPTIONS ] qdisc [ add | change | replace | link | delete ] dev DEV
10 [ parent qdisc-id | root ] [ handle qdisc-id ] qdisc [ qdisc specific
11 parameters ]
12 tc [ OPTIONS ] class [ add | change | replace | delete ] dev DEV parent
14 qdisc-id [ classid class-id ] qdisc [ qdisc specific parameters ]
15 tc [ OPTIONS ] filter [ add | change | replace | delete | get ] dev DEV
17 [ parent qdisc-id | root ] [ handle filter-id ] protocol protocol prio
18 priority filtertype [ filtertype specific parameters ] flowid flow-id
19 tc [ OPTIONS ] [ FORMAT ] qdisc show [ dev DEV ]
21 tc [ OPTIONS ] [ FORMAT ] class show dev DEV
23 tc [ OPTIONS ] filter show dev DEV
25 OPTIONS := { [ -force ] [ -OK ] -b[atch] [ filename ] | [ -n[etns]
28 name ] | [ -nm | -nam[es] ] | [ { -cf | -c[onf] } [ filename ] ] }
29 FORMAT := { -s[tatistics] | -d[etails] | -r[aw] | -p[retty] | -i[ec] |
31 -g[raph] }
32
35 Tc is used to configure Traffic Control in the Linux kernel. Traffic
36 Control consists of the following:
37 SHAPING
40 When traffic is shaped, its rate of transmission is under con‐
41 trol. Shaping may be more than lowering the available bandwidth
42 - it is also used to smooth out bursts in traffic for better
43 network behaviour. Shaping occurs on egress.
44 SCHEDULING
47 By scheduling the transmission of packets it is possible to
48 improve interactivity for traffic that needs it while still
49 guaranteeing bandwidth to bulk transfers. Reordering is also
50 called prioritizing, and happens only on egress.
51 POLICING
54 Whereas shaping deals with transmission of traffic, policing
55 pertains to traffic arriving. Policing thus occurs on ingress.
56 DROPPING
59 Traffic exceeding a set bandwidth may also be dropped forthwith,
60 both on ingress and on egress.
61 Processing of traffic is controlled by three kinds of objects: qdiscs,
64 classes and filters.
65
68 qdisc is short for 'queueing discipline' and it is elementary to under‐
69 standing traffic control. Whenever the kernel needs to send a packet to
70 an interface, it is enqueued to the qdisc configured for that inter‐
71 face. Immediately afterwards, the kernel tries to get as many packets
72 as possible from the qdisc, for giving them to the network adaptor
73 driver.
74 A simple QDISC is the 'pfifo' one, which does no processing at all and
76 is a pure First In, First Out queue. It does however store traffic when
77 the network interface can't handle it momentarily.
78
81 Some qdiscs can contain classes, which contain further qdiscs - traffic
82 may then be enqueued in any of the inner qdiscs, which are within the
83 classes. When the kernel tries to dequeue a packet from such a class‐
84 ful qdisc it can come from any of the classes. A qdisc may for example
85 prioritize certain kinds of traffic by trying to dequeue from certain
86 classes before others.
87
90 A filter is used by a classful qdisc to determine in which class a
91 packet will be enqueued. Whenever traffic arrives at a class with sub‐
92 classes, it needs to be classified. Various methods may be employed to
93 do so, one of these are the filters. All filters attached to the class
94 are called, until one of them returns with a verdict. If no verdict was
95 made, other criteria may be available. This differs per qdisc.
96 It is important to notice that filters reside within qdiscs - they are
98 not masters of what happens.
99 The available filters are:
101 basic Filter packets based on an ematch expression. See tc-ematch(8)
103 for details.
104 bpf Filter packets using (e)BPF, see tc-bpf(8) for details.
106 cgroup Filter packets based on the control group of their process. See
108 tc-cgroup(8) for details.
109 flow, flower
111 Flow-based classifiers, filtering packets based on their flow
112 (identified by selectable keys). See tc-flow(8) and tc-flower(8)
113 for details.
114 fw Filter based on fwmark. Directly maps fwmark value to traffic
116 class. See tc-fw(8).
117 route Filter packets based on routing table. See tc-route(8) for
119 details.
120 rsvp Match Resource Reservation Protocol (RSVP) packets.
122 tcindex
124 Filter packets based on traffic control index. See tc-tcin‐
125 dex(8).
126 u32 Generic filtering on arbitrary packet data, assisted by syntax
128 to abstract common operations. See tc-u32(8) for details.
129 matchall
131 Traffic control filter that matches every packet. See tc-
132 matchall(8) for details.
133
136 The classless qdiscs are:
137 choke CHOKe (CHOose and Keep for responsive flows, CHOose and Kill for
139 unresponsive flows) is a classless qdisc designed to both iden‐
140 tify and penalize flows that monopolize the queue. CHOKe is a
141 variation of RED, and the configuration is similar to RED.
142 codel CoDel (pronounced "coddle") is an adaptive "no-knobs" active
144 queue management algorithm (AQM) scheme that was developed to
145 address the shortcomings of RED and its variants.
146 [p|b]fifo
148 Simplest usable qdisc, pure First In, First Out behaviour. Lim‐
149 ited in packets or in bytes.
150 fq Fair Queue Scheduler realises TCP pacing and scales to millions
152 of concurrent flows per qdisc.
153 fq_codel
155 Fair Queuing Controlled Delay is queuing discipline that com‐
156 bines Fair Queuing with the CoDel AQM scheme. FQ_Codel uses a
157 stochastic model to classify incoming packets into different
158 flows and is used to provide a fair share of the bandwidth to
159 all the flows using the queue. Each such flow is managed by the
160 CoDel queuing discipline. Reordering within a flow is avoided
161 since Codel internally uses a FIFO queue.
162 gred Generalized Random Early Detection combines multiple RED queues
164 in order to achieve multiple drop priorities. This is required
165 to realize Assured Forwarding (RFC 2597).
166 hhf Heavy-Hitter Filter differentiates between small flows and the
168 opposite, heavy-hitters. The goal is to catch the heavy-hitters
169 and move them to a separate queue with less priority so that
170 bulk traffic does not affect the latency of critical traffic.
171 ingress
173 This is a special qdisc as it applies to incoming traffic on an
174 interface, allowing for it to be filtered and policed.
175 mqprio The Multiqueue Priority Qdisc is a simple queuing discipline
177 that allows mapping traffic flows to hardware queue ranges using
178 priorities and a configurable priority to traffic class mapping.
179 A traffic class in this context is a set of contiguous qdisc
180 classes which map 1:1 to a set of hardware exposed queues.
181 multiq Multiqueue is a qdisc optimized for devices with multiple Tx
183 queues. It has been added for hardware that wishes to avoid
184 head-of-line blocking. It will cycle though the bands and ver‐
185 ify that the hardware queue associated with the band is not
186 stopped prior to dequeuing a packet.
187 netem Network Emulator is an enhancement of the Linux traffic control
189 facilities that allow to add delay, packet loss, duplication and
190 more other characteristics to packets outgoing from a selected
191 network interface.
192 pfifo_fast
194 Standard qdisc for 'Advanced Router' enabled kernels. Consists
195 of a three-band queue which honors Type of Service flags, as
196 well as the priority that may be assigned to a packet.
197 pie Proportional Integral controller-Enhanced (PIE) is a control
199 theoretic active queue management scheme. It is based on the
200 proportional integral controller but aims to control delay.
201 red Random Early Detection simulates physical congestion by randomly
203 dropping packets when nearing configured bandwidth allocation.
204 Well suited to very large bandwidth applications.
205 rr Round-Robin qdisc with support for multiqueue network devices.
207 Removed from Linux since kernel version 2.6.27.
208 sfb Stochastic Fair Blue is a classless qdisc to manage congestion
210 based on packet loss and link utilization history while trying
211 to prevent non-responsive flows (i.e. flows that do not react to
212 congestion marking or dropped packets) from impacting perfor‐
213 mance of responsive flows. Unlike RED, where the marking proba‐
214 bility has to be configured, BLUE tries to determine the ideal
215 marking probability automatically.
216 sfq Stochastic Fairness Queueing reorders queued traffic so each
218 'session' gets to send a packet in turn.
219 tbf The Token Bucket Filter is suited for slowing traffic down to a
221 precisely configured rate. Scales well to large bandwidths.
222
224 In the absence of classful qdiscs, classless qdiscs can only be
225 attached at the root of a device. Full syntax:
226 tc qdisc add dev DEV root QDISC QDISC-PARAMETERS
228 To remove, issue
230 tc qdisc del dev DEV root
232 The pfifo_fast qdisc is the automatic default in the absence of a con‐
234 figured qdisc.
235
238 The classful qdiscs are:
239 ATM Map flows to virtual circuits of an underlying asynchronous
241 transfer mode device.
242 CBQ Class Based Queueing implements a rich linksharing hierarchy of
244 classes. It contains shaping elements as well as prioritizing
245 capabilities. Shaping is performed using link idle time calcula‐
246 tions based on average packet size and underlying link band‐
247 width. The latter may be ill-defined for some interfaces.
248 DRR The Deficit Round Robin Scheduler is a more flexible replacement
250 for Stochastic Fairness Queuing. Unlike SFQ, there are no built-
251 in queues -- you need to add classes and then set up filters to
252 classify packets accordingly. This can be useful e.g. for using
253 RED qdiscs with different settings for particular traffic. There
254 is no default class -- if a packet cannot be classified, it is
255 dropped.
256 DSMARK Classify packets based on TOS field, change TOS field of packets
258 based on classification.
259 HFSC Hierarchical Fair Service Curve guarantees precise bandwidth and
261 delay allocation for leaf classes and allocates excess bandwidth
262 fairly. Unlike HTB, it makes use of packet dropping to achieve
263 low delays which interactive sessions benefit from.
264 HTB The Hierarchy Token Bucket implements a rich linksharing hierar‐
266 chy of classes with an emphasis on conforming to existing prac‐
267 tices. HTB facilitates guaranteeing bandwidth to classes, while
268 also allowing specification of upper limits to inter-class shar‐
269 ing. It contains shaping elements, based on TBF and can priori‐
270 tize classes.
271 PRIO The PRIO qdisc is a non-shaping container for a configurable
273 number of classes which are dequeued in order. This allows for
274 easy prioritization of traffic, where lower classes are only
275 able to send if higher ones have no packets available. To facil‐
276 itate configuration, Type Of Service bits are honored by
277 default.
278 QFQ Quick Fair Queueing is an O(1) scheduler that provides near-
280 optimal guarantees, and is the first to achieve that goal with a
281 constant cost also with respect to the number of groups and the
282 packet length. The QFQ algorithm has no loops, and uses very
283 simple instructions and data structures that lend themselves
284 very well to a hardware implementation.
285
287 Classes form a tree, where each class has a single parent. A class may
288 have multiple children. Some qdiscs allow for runtime addition of
289 classes (CBQ, HTB) while others (PRIO) are created with a static number
290 of children.
291 Qdiscs which allow dynamic addition of classes can have zero or more
293 subclasses to which traffic may be enqueued.
294 Furthermore, each class contains a leaf qdisc which by default has
296 pfifo behaviour, although another qdisc can be attached in place. This
297 qdisc may again contain classes, but each class can have only one leaf
298 qdisc.
299 When a packet enters a classful qdisc it can be classified to one of
301 the classes within. Three criteria are available, although not all
302 qdiscs will use all three:
303 tc filters
305 If tc filters are attached to a class, they are consulted first
306 for relevant instructions. Filters can match on all fields of a
307 packet header, as well as on the firewall mark applied by
308 ipchains or iptables.
309 Type of Service
311 Some qdiscs have built in rules for classifying packets based on
312 the TOS field.
313 skb->priority
315 Userspace programs can encode a class-id in the 'skb->priority'
316 field using the SO_PRIORITY option.
317 Each node within the tree can have its own filters but higher level
319 filters may also point directly to lower classes.
320 If classification did not succeed, packets are enqueued to the leaf
322 qdisc attached to that class. Check qdisc specific manpages for
323 details, however.
324
327 All qdiscs, classes and filters have IDs, which can either be specified
328 or be automatically assigned.
329 IDs consist of a major number and a minor number, separated by a colon
331 - major:minor. Both major and minor are hexadecimal numbers and are
332 limited to 16 bits. There are two special values: root is signified by
333 major and minor of all ones, and unspecified is all zeros.
334 QDISCS A qdisc, which potentially can have children, gets assigned a
337 major number, called a 'handle', leaving the minor number names‐
338 pace available for classes. The handle is expressed as '10:'.
339 It is customary to explicitly assign a handle to qdiscs expected
340 to have children.
341 CLASSES
344 Classes residing under a qdisc share their qdisc major number,
345 but each have a separate minor number called a 'classid' that
346 has no relation to their parent classes, only to their parent
347 qdisc. The same naming custom as for qdiscs applies.
348 FILTERS
351 Filters have a three part ID, which is only needed when using a
352 hashed filter hierarchy.
353
356 The following parameters are widely used in TC. For other parameters,
357 see the man pages for individual qdiscs.
358 RATES Bandwidths or rates. These parameters accept a floating point
361 number, possibly followed by a unit (both SI and IEC units sup‐
362 ported).
363 bit or a bare number
365 Bits per second
366 kbit Kilobits per second
368 mbit Megabits per second
370 gbit Gigabits per second
372 tbit Terabits per second
374 bps Bytes per second
376 kbps Kilobytes per second
378 mbps Megabytes per second
380 gbps Gigabytes per second
382 tbps Terabytes per second
384 To specify in IEC units, replace the SI prefix (k-, m-, g-, t-)
387 with IEC prefix (ki-, mi-, gi- and ti-) respectively.
388 TC store rates as a 32-bit unsigned integer in bps internally,
391 so we can specify a max rate of 4294967295 bps.
392 TIMES Length of time. Can be specified as a floating point number fol‐
395 lowed by an optional unit:
396 s, sec or secs
398 Whole seconds
399 ms, msec or msecs
401 Milliseconds
402 us, usec, usecs or a bare number
404 Microseconds.
405 TC defined its own time unit (equal to microsecond) and stores
408 time values as 32-bit unsigned integer, thus we can specify a
409 max time value of 4294967295 usecs.
410 SIZES Amounts of data. Can be specified as a floating point number
413 followed by an optional unit:
414 b or a bare number
416 Bytes.
417 kbit Kilobits
419 kb or k
421 Kilobytes
422 mbit Megabits
424 mb or m
426 Megabytes
427 gbit Gigabits
429 gb or g
431 Gigabytes
432 TC stores sizes internally as 32-bit unsigned integer in byte,
435 so we can specify a max size of 4294967295 bytes.
436 VALUES Other values without a unit. These parameters are interpreted
439 as decimal by default, but you can indicate TC to interpret them
440 as octal and hexadecimal by adding a '0' or '0x' prefix respec‐
441 tively.
442
445 The following commands are available for qdiscs, classes and filter:
446 add Add a qdisc, class or filter to a node. For all entities, a par‐
448 ent must be passed, either by passing its ID or by attaching
449 directly to the root of a device. When creating a qdisc or a
450 filter, it can be named with the handle parameter. A class is
451 named with the classid parameter.
452 delete A qdisc can be deleted by specifying its handle, which may also
455 be 'root'. All subclasses and their leaf qdiscs are automati‐
456 cally deleted, as well as any filters attached to them.
457 change Some entities can be modified 'in place'. Shares the syntax of
460 'add', with the exception that the handle cannot be changed and
461 neither can the parent. In other words, change cannot move a
462 node.
463 replace
466 Performs a nearly atomic remove/add on an existing node id. If
467 the node does not exist yet it is created.
468 get Displays a single filter given the interface DEV, qdisc-id, pri‐
471 ority, protocol and filter-id.
472 show Displays all filters attached to the given interface. A valid
475 parent ID must be passed.
476 link Only available for qdiscs and performs a replace where the node
479 must exist already.
480
483 -b, -b filename, -batch, -batch filename
484 read commands from provided file or standard input and invoke
485 them. First failure will cause termination of tc.
486 -force don't terminate tc on errors in batch mode. If there were any
489 errors during execution of the commands, the application return
490 code will be non zero.
491 -OK in batch mode, print OK and a new line on standard output after
494 each successfully interpreted command.
495 -n, -net, -netns <NETNS>
498 switches tc to the specified network namespace NETNS. Actually
499 it just simplifies executing of:
500 ip netns exec NETNS tc [ OPTIONS ] OBJECT { COMMAND | help }
502 to
504 tc -n[etns] NETNS [ OPTIONS ] OBJECT { COMMAND | help }
506 -cf, -conf <FILENAME>
509 specifies path to the config file. This option is used in con‐
510 junction with other options (e.g. -nm).
511
514 The show command has additional formatting options:
515 -s, -stats, -statistics
518 output more statistics about packet usage.
519 -d, -details
522 output more detailed information about rates and cell sizes.
523 -r, -raw
526 output raw hex values for handles.
527 -p, -pretty
530 decode filter offset and mask values to equivalent filter com‐
531 mands based on TCP/IP.
532 -iec print rates in IEC units (ie. 1K = 1024).
535 -g, -graph
538 shows classes as ASCII graph. Prints generic stats info under
539 each class if -s option was specified. Classes can be filtered
540 only by dev option.
541 -nm, -name
544 resolve class name from /etc/iproute2/tc_cls file or from file
545 specified by -cf option. This file is just a mapping of classid
546 to class name:
547 # Here is comment
549 1:40 voip # Here is another comment
550 1:50 web
551 1:60 ftp
552 1:2 home
553 tc will not fail if -nm was specified without -cf option but
555 /etc/iproute2/tc_cls file does not exist, which makes it possi‐
556 ble to pass -nm option for creating tc alias.
557
560 tc -g class show dev eth0
561 Shows classes as ASCII graph on eth0 interface.
562 tc -g -s class show dev eth0
564 Shows classes as ASCII graph with stats info under each class.
565
568 tc was written by Alexey N. Kuznetsov and added in Linux 2.2.
569
571 tc-basic(8), tc-bfifo(8), tc-bpf(8), tc-cbq(8), tc-cgroup(8), tc-
572 choke(8), tc-codel(8), tc-drr(8), tc-ematch(8), tc-flow(8), tc-
573 flower(8), tc-fq(8), tc-fq_codel(8), tc-fw(8), tc-hfsc(7), tc-hfsc(8),
574 tc-htb(8), tc-mqprio(8), tc-pfifo(8), tc-pfifo_fast(8), tc-red(8), tc-
575 route(8), tc-sfb(8), tc-sfq(8), tc-stab(8), tc-tbf(8), tc-tcindex(8),
576 tc-u32(8),
577 User documentation at http://lartc.org/, but please direct bugreports
578 and patches to: <netdev@vger.kernel.org>
579
582 Manpage maintained by bert hubert (ahu@ds9a.nl)
583iproute2 16 December 2001 TC(8)