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 ] [ ingress_block
11 BLOCK_INDEX ] [ egress_block BLOCK_INDEX ] qdisc [ qdisc specific
12 parameters ]
13 tc [ OPTIONS ] class [ add | change | replace | delete ] dev DEV parent
15 qdisc-id [ classid class-id ] qdisc [ qdisc specific parameters ]
16 tc [ OPTIONS ] filter [ add | change | replace | delete | get ] dev DEV
18 [ parent qdisc-id | root ] [ handle filter-id ] protocol protocol prio
19 priority filtertype [ filtertype specific parameters ] flowid flow-id
20 tc [ OPTIONS ] filter [ add | change | replace | delete | get ] block
22 BLOCK_INDEX [ handle filter-id ] protocol protocol prio priority fil‐
23 tertype [ filtertype specific parameters ] flowid flow-id
24 tc [ OPTIONS ] [ FORMAT ] qdisc show [ dev DEV ]
26 tc [ OPTIONS ] [ FORMAT ] class show dev DEV
28 tc [ OPTIONS ] filter show dev DEV
30 tc [ OPTIONS ] filter show block BLOCK_INDEX
32 tc [ OPTIONS ] monitor [ file FILENAME ]
35 OPTIONS := { [ -force ] -b[atch] [ filename ] | [ -n[etns] name ] | [
38 -nm | -nam[es] ] | [ { -cf | -c[onf] } [ filename ] ] [ -t[imestamp] ]
39 | [ -t[short] | [ -o[neline] ] }
40 FORMAT := { -s[tatistics] | -d[etails] | -r[aw] | -i[ec] | -g[raph] |
42 -j[json] | -p[retty] | -col[or] }
43
46 Tc is used to configure Traffic Control in the Linux kernel. Traffic
47 Control consists of the following:
48 SHAPING
51 When traffic is shaped, its rate of transmission is under con‐
52 trol. Shaping may be more than lowering the available bandwidth
53 - it is also used to smooth out bursts in traffic for better
54 network behaviour. Shaping occurs on egress.
55 SCHEDULING
58 By scheduling the transmission of packets it is possible to
59 improve interactivity for traffic that needs it while still
60 guaranteeing bandwidth to bulk transfers. Reordering is also
61 called prioritizing, and happens only on egress.
62 POLICING
65 Whereas shaping deals with transmission of traffic, policing
66 pertains to traffic arriving. Policing thus occurs on ingress.
67 DROPPING
70 Traffic exceeding a set bandwidth may also be dropped forthwith,
71 both on ingress and on egress.
72 Processing of traffic is controlled by three kinds of objects: qdiscs,
75 classes and filters.
76
79 qdisc is short for 'queueing discipline' and it is elementary to under‐
80 standing traffic control. Whenever the kernel needs to send a packet to
81 an interface, it is enqueued to the qdisc configured for that inter‐
82 face. Immediately afterwards, the kernel tries to get as many packets
83 as possible from the qdisc, for giving them to the network adaptor
84 driver.
85 A simple QDISC is the 'pfifo' one, which does no processing at all and
87 is a pure First In, First Out queue. It does however store traffic when
88 the network interface can't handle it momentarily.
89
92 Some qdiscs can contain classes, which contain further qdiscs - traffic
93 may then be enqueued in any of the inner qdiscs, which are within the
94 classes. When the kernel tries to dequeue a packet from such a class‐
95 ful qdisc it can come from any of the classes. A qdisc may for example
96 prioritize certain kinds of traffic by trying to dequeue from certain
97 classes before others.
98
101 A filter is used by a classful qdisc to determine in which class a
102 packet will be enqueued. Whenever traffic arrives at a class with sub‐
103 classes, it needs to be classified. Various methods may be employed to
104 do so, one of these are the filters. All filters attached to the class
105 are called, until one of them returns with a verdict. If no verdict was
106 made, other criteria may be available. This differs per qdisc.
107 It is important to notice that filters reside within qdiscs - they are
109 not masters of what happens.
110 The available filters are:
112 basic Filter packets based on an ematch expression. See tc-ematch(8)
114 for details.
115 bpf Filter packets using (e)BPF, see tc-bpf(8) for details.
117 cgroup Filter packets based on the control group of their process. See
119 tc-cgroup(8) for details.
120 flow, flower
122 Flow-based classifiers, filtering packets based on their flow
123 (identified by selectable keys). See tc-flow(8) and tc-flower(8)
124 for details.
125 fw Filter based on fwmark. Directly maps fwmark value to traffic
127 class. See tc-fw(8).
128 route Filter packets based on routing table. See tc-route(8) for
130 details.
131 rsvp Match Resource Reservation Protocol (RSVP) packets.
133 tcindex
135 Filter packets based on traffic control index. See tc-tcin‐
136 dex(8).
137 u32 Generic filtering on arbitrary packet data, assisted by syntax
139 to abstract common operations. See tc-u32(8) for details.
140 matchall
142 Traffic control filter that matches every packet. See tc-
143 matchall(8) for details.
144
147 The classless qdiscs are:
148 choke CHOKe (CHOose and Keep for responsive flows, CHOose and Kill for
150 unresponsive flows) is a classless qdisc designed to both iden‐
151 tify and penalize flows that monopolize the queue. CHOKe is a
152 variation of RED, and the configuration is similar to RED.
153 codel CoDel (pronounced "coddle") is an adaptive "no-knobs" active
155 queue management algorithm (AQM) scheme that was developed to
156 address the shortcomings of RED and its variants.
157 [p|b]fifo
159 Simplest usable qdisc, pure First In, First Out behaviour. Lim‐
160 ited in packets or in bytes.
161 fq Fair Queue Scheduler realises TCP pacing and scales to millions
163 of concurrent flows per qdisc.
164 fq_codel
166 Fair Queuing Controlled Delay is queuing discipline that com‐
167 bines Fair Queuing with the CoDel AQM scheme. FQ_Codel uses a
168 stochastic model to classify incoming packets into different
169 flows and is used to provide a fair share of the bandwidth to
170 all the flows using the queue. Each such flow is managed by the
171 CoDel queuing discipline. Reordering within a flow is avoided
172 since Codel internally uses a FIFO queue.
173 gred Generalized Random Early Detection combines multiple RED queues
175 in order to achieve multiple drop priorities. This is required
176 to realize Assured Forwarding (RFC 2597).
177 hhf Heavy-Hitter Filter differentiates between small flows and the
179 opposite, heavy-hitters. The goal is to catch the heavy-hitters
180 and move them to a separate queue with less priority so that
181 bulk traffic does not affect the latency of critical traffic.
182 ingress
184 This is a special qdisc as it applies to incoming traffic on an
185 interface, allowing for it to be filtered and policed.
186 mqprio The Multiqueue Priority Qdisc is a simple queuing discipline
188 that allows mapping traffic flows to hardware queue ranges using
189 priorities and a configurable priority to traffic class mapping.
190 A traffic class in this context is a set of contiguous qdisc
191 classes which map 1:1 to a set of hardware exposed queues.
192 multiq Multiqueue is a qdisc optimized for devices with multiple Tx
194 queues. It has been added for hardware that wishes to avoid
195 head-of-line blocking. It will cycle though the bands and ver‐
196 ify that the hardware queue associated with the band is not
197 stopped prior to dequeuing a packet.
198 netem Network Emulator is an enhancement of the Linux traffic control
200 facilities that allow to add delay, packet loss, duplication and
201 more other characteristics to packets outgoing from a selected
202 network interface.
203 pfifo_fast
205 Standard qdisc for 'Advanced Router' enabled kernels. Consists
206 of a three-band queue which honors Type of Service flags, as
207 well as the priority that may be assigned to a packet.
208 pie Proportional Integral controller-Enhanced (PIE) is a control
210 theoretic active queue management scheme. It is based on the
211 proportional integral controller but aims to control delay.
212 red Random Early Detection simulates physical congestion by randomly
214 dropping packets when nearing configured bandwidth allocation.
215 Well suited to very large bandwidth applications.
216 rr Round-Robin qdisc with support for multiqueue network devices.
218 Removed from Linux since kernel version 2.6.27.
219 sfb Stochastic Fair Blue is a classless qdisc to manage congestion
221 based on packet loss and link utilization history while trying
222 to prevent non-responsive flows (i.e. flows that do not react to
223 congestion marking or dropped packets) from impacting perfor‐
224 mance of responsive flows. Unlike RED, where the marking proba‐
225 bility has to be configured, BLUE tries to determine the ideal
226 marking probability automatically.
227 sfq Stochastic Fairness Queueing reorders queued traffic so each
229 'session' gets to send a packet in turn.
230 tbf The Token Bucket Filter is suited for slowing traffic down to a
232 precisely configured rate. Scales well to large bandwidths.
233
235 In the absence of classful qdiscs, classless qdiscs can only be
236 attached at the root of a device. Full syntax:
237 tc qdisc add dev DEV root QDISC QDISC-PARAMETERS
239 To remove, issue
241 tc qdisc del dev DEV root
243 The pfifo_fast qdisc is the automatic default in the absence of a con‐
245 figured qdisc.
246
249 The classful qdiscs are:
250 ATM Map flows to virtual circuits of an underlying asynchronous
252 transfer mode device.
253 CBQ Class Based Queueing implements a rich linksharing hierarchy of
255 classes. It contains shaping elements as well as prioritizing
256 capabilities. Shaping is performed using link idle time calcula‐
257 tions based on average packet size and underlying link band‐
258 width. The latter may be ill-defined for some interfaces.
259 DRR The Deficit Round Robin Scheduler is a more flexible replacement
261 for Stochastic Fairness Queuing. Unlike SFQ, there are no built-
262 in queues -- you need to add classes and then set up filters to
263 classify packets accordingly. This can be useful e.g. for using
264 RED qdiscs with different settings for particular traffic. There
265 is no default class -- if a packet cannot be classified, it is
266 dropped.
267 DSMARK Classify packets based on TOS field, change TOS field of packets
269 based on classification.
270 HFSC Hierarchical Fair Service Curve guarantees precise bandwidth and
272 delay allocation for leaf classes and allocates excess bandwidth
273 fairly. Unlike HTB, it makes use of packet dropping to achieve
274 low delays which interactive sessions benefit from.
275 HTB The Hierarchy Token Bucket implements a rich linksharing hierar‐
277 chy of classes with an emphasis on conforming to existing prac‐
278 tices. HTB facilitates guaranteeing bandwidth to classes, while
279 also allowing specification of upper limits to inter-class shar‐
280 ing. It contains shaping elements, based on TBF and can priori‐
281 tize classes.
282 PRIO The PRIO qdisc is a non-shaping container for a configurable
284 number of classes which are dequeued in order. This allows for
285 easy prioritization of traffic, where lower classes are only
286 able to send if higher ones have no packets available. To facil‐
287 itate configuration, Type Of Service bits are honored by
288 default.
289 QFQ Quick Fair Queueing is an O(1) scheduler that provides near-
291 optimal guarantees, and is the first to achieve that goal with a
292 constant cost also with respect to the number of groups and the
293 packet length. The QFQ algorithm has no loops, and uses very
294 simple instructions and data structures that lend themselves
295 very well to a hardware implementation.
296
298 Classes form a tree, where each class has a single parent. A class may
299 have multiple children. Some qdiscs allow for runtime addition of
300 classes (CBQ, HTB) while others (PRIO) are created with a static number
301 of children.
302 Qdiscs which allow dynamic addition of classes can have zero or more
304 subclasses to which traffic may be enqueued.
305 Furthermore, each class contains a leaf qdisc which by default has
307 pfifo behaviour, although another qdisc can be attached in place. This
308 qdisc may again contain classes, but each class can have only one leaf
309 qdisc.
310 When a packet enters a classful qdisc it can be classified to one of
312 the classes within. Three criteria are available, although not all
313 qdiscs will use all three:
314 tc filters
316 If tc filters are attached to a class, they are consulted first
317 for relevant instructions. Filters can match on all fields of a
318 packet header, as well as on the firewall mark applied by
319 ipchains or iptables.
320 Type of Service
322 Some qdiscs have built in rules for classifying packets based on
323 the TOS field.
324 skb->priority
326 Userspace programs can encode a class-id in the 'skb->priority'
327 field using the SO_PRIORITY option.
328 Each node within the tree can have its own filters but higher level
330 filters may also point directly to lower classes.
331 If classification did not succeed, packets are enqueued to the leaf
333 qdisc attached to that class. Check qdisc specific manpages for
334 details, however.
335
338 All qdiscs, classes and filters have IDs, which can either be specified
339 or be automatically assigned.
340 IDs consist of a major number and a minor number, separated by a colon
342 - major:minor. Both major and minor are hexadecimal numbers and are
343 limited to 16 bits. There are two special values: root is signified by
344 major and minor of all ones, and unspecified is all zeros.
345 QDISCS A qdisc, which potentially can have children, gets assigned a
348 major number, called a 'handle', leaving the minor number names‐
349 pace available for classes. The handle is expressed as '10:'.
350 It is customary to explicitly assign a handle to qdiscs expected
351 to have children.
352 CLASSES
355 Classes residing under a qdisc share their qdisc major number,
356 but each have a separate minor number called a 'classid' that
357 has no relation to their parent classes, only to their parent
358 qdisc. The same naming custom as for qdiscs applies.
359 FILTERS
362 Filters have a three part ID, which is only needed when using a
363 hashed filter hierarchy.
364
367 The following parameters are widely used in TC. For other parameters,
368 see the man pages for individual qdiscs.
369 RATES Bandwidths or rates. These parameters accept a floating point
372 number, possibly followed by either a unit (both SI and IEC
373 units supported), or a float followed by a '%' character to
374 specify the rate as a percentage of the device's speed (e.g. 5%,
375 99.5%). Warning: specifying the rate as a percentage means a
376 fraction of the current speed; if the speed changes, the value
377 will not be recalculated.
378 bit or a bare number
380 Bits per second
381 kbit Kilobits per second
383 mbit Megabits per second
385 gbit Gigabits per second
387 tbit Terabits per second
389 bps Bytes per second
391 kbps Kilobytes per second
393 mbps Megabytes per second
395 gbps Gigabytes per second
397 tbps Terabytes per second
399 To specify in IEC units, replace the SI prefix (k-, m-, g-, t-)
402 with IEC prefix (ki-, mi-, gi- and ti-) respectively.
403 TC store rates as a 32-bit unsigned integer in bps internally,
406 so we can specify a max rate of 4294967295 bps.
407 TIMES Length of time. Can be specified as a floating point number fol‐
410 lowed by an optional unit:
411 s, sec or secs
413 Whole seconds
414 ms, msec or msecs
416 Milliseconds
417 us, usec, usecs or a bare number
419 Microseconds.
420 TC defined its own time unit (equal to microsecond) and stores
423 time values as 32-bit unsigned integer, thus we can specify a
424 max time value of 4294967295 usecs.
425 SIZES Amounts of data. Can be specified as a floating point number
428 followed by an optional unit:
429 b or a bare number
431 Bytes.
432 kbit Kilobits
434 kb or k
436 Kilobytes
437 mbit Megabits
439 mb or m
441 Megabytes
442 gbit Gigabits
444 gb or g
446 Gigabytes
447 TC stores sizes internally as 32-bit unsigned integer in byte,
450 so we can specify a max size of 4294967295 bytes.
451 VALUES Other values without a unit. These parameters are interpreted
454 as decimal by default, but you can indicate TC to interpret them
455 as octal and hexadecimal by adding a '0' or '0x' prefix respec‐
456 tively.
457
460 The following commands are available for qdiscs, classes and filter:
461 add Add a qdisc, class or filter to a node. For all entities, a par‐
463 ent must be passed, either by passing its ID or by attaching
464 directly to the root of a device. When creating a qdisc or a
465 filter, it can be named with the handle parameter. A class is
466 named with the classid parameter.
467 delete A qdisc can be deleted by specifying its handle, which may also
470 be 'root'. All subclasses and their leaf qdiscs are automati‐
471 cally deleted, as well as any filters attached to them.
472 change Some entities can be modified 'in place'. Shares the syntax of
475 'add', with the exception that the handle cannot be changed and
476 neither can the parent. In other words, change cannot move a
477 node.
478 replace
481 Performs a nearly atomic remove/add on an existing node id. If
482 the node does not exist yet it is created.
483 get Displays a single filter given the interface DEV, qdisc-id, pri‐
486 ority, protocol and filter-id.
487 show Displays all filters attached to the given interface. A valid
490 parent ID must be passed.
491 link Only available for qdiscs and performs a replace where the node
494 must exist already.
495
498 The tc utility can monitor events generated by the kernel such as
499 adding/deleting qdiscs, filters or actions, or modifying existing ones.
500 The following command is available for monitor :
502 file If the file option is given, the tc does not listen to kernel
504 events, but opens the given file and dumps its contents. The
505 file has to be in binary format and contain netlink messages.
506
509 -b, -b filename, -batch, -batch filename
510 read commands from provided file or standard input and invoke
511 them. First failure will cause termination of tc.
512 -force don't terminate tc on errors in batch mode. If there were any
515 errors during execution of the commands, the application return
516 code will be non zero.
517 -o, -oneline
520 output each record on a single line, replacing line feeds with
521 the '\' character. This is convenient when you want to count
522 records with wc(1) or to grep(1) the output.
523 -n, -net, -netns <NETNS>
526 switches tc to the specified network namespace NETNS. Actually
527 it just simplifies executing of:
528 ip netns exec NETNS tc [ OPTIONS ] OBJECT { COMMAND | help }
530 to
532 tc -n[etns] NETNS [ OPTIONS ] OBJECT { COMMAND | help }
534 -cf, -conf <FILENAME>
537 specifies path to the config file. This option is used in con‐
538 junction with other options (e.g. -nm).
539 -t, -timestamp
542 When tc monitor runs, print timestamp before the event message
543 in format:
544 Timestamp: <Day> <Month> <DD> <hh:mm:ss> <YYYY> <usecs> usec
545 -ts, -tshort
548 When tc monitor runs, prints short timestamp before the event
549 message in format:
550 [<YYYY>-<MM>-<DD>T<hh:mm:ss>.<ms>]
551
554 The show command has additional formatting options:
555 -s, -stats, -statistics
558 output more statistics about packet usage.
559 -d, -details
562 output more detailed information about rates and cell sizes.
563 -r, -raw
566 output raw hex values for handles.
567 -p, -pretty
570 for u32 filter, decode offset and mask values to equivalent fil‐
571 ter commands based on TCP/IP. In JSON output, add whitespace to
572 improve readability.
573 -iec print rates in IEC units (ie. 1K = 1024).
576 -g, -graph
579 shows classes as ASCII graph. Prints generic stats info under
580 each class if -s option was specified. Classes can be filtered
581 only by dev option.
582 -color
585 Use color output.
586 -j, -json
589 Display results in JSON format.
590 -nm, -name
593 resolve class name from /etc/iproute2/tc_cls file or from file
594 specified by -cf option. This file is just a mapping of classid
595 to class name:
596 # Here is comment
598 1:40 voip # Here is another comment
599 1:50 web
600 1:60 ftp
601 1:2 home
602 tc will not fail if -nm was specified without -cf option but
604 /etc/iproute2/tc_cls file does not exist, which makes it possi‐
605 ble to pass -nm option for creating tc alias.
606
609 tc -g class show dev eth0
610 Shows classes as ASCII graph on eth0 interface.
611 tc -g -s class show dev eth0
613 Shows classes as ASCII graph with stats info under each class.
614
617 tc was written by Alexey N. Kuznetsov and added in Linux 2.2.
618
620 tc-basic(8), tc-bfifo(8), tc-bpf(8), tc-cbq(8), tc-cgroup(8), tc-
621 choke(8), tc-codel(8), tc-drr(8), tc-ematch(8), tc-flow(8), tc-
622 flower(8), tc-fq(8), tc-fq_codel(8), tc-fw(8), tc-hfsc(7), tc-hfsc(8),
623 tc-htb(8), tc-mqprio(8), tc-pfifo(8), tc-pfifo_fast(8), tc-red(8), tc-
624 route(8), tc-sfb(8), tc-sfq(8), tc-stab(8), tc-tbf(8), tc-tcindex(8),
625 tc-u32(8),
626 User documentation at http://lartc.org/, but please direct bugreports
627 and patches to: <netdev@vger.kernel.org>
628
631 Manpage maintained by bert hubert (ahu@ds9a.nl)
632iproute2 16 December 2001 TC(8)