1TC(8) Linux TC(8)
2
6 tc - show / manipulate traffic control settings
7
9 tc qdisc [ add | change | replace | link ] dev DEV [ parent qdisc-id |
10 root ] [ handle qdisc-id ] qdisc [ qdisc specific parameters ]
11 tc class [ add | change | replace ] dev DEV parent qdisc-id [ classid
13 class-id ] qdisc [ qdisc specific parameters ]
14 tc filter [ add | change | replace ] dev DEV [ parent qdisc-id | root ]
16 protocol protocol prio priority filtertype [ filtertype specific param‐
17 eters ] flowid flow-id
18 tc [ FORMAT ] qdisc show [ dev DEV ]
20 tc [ FORMAT ] class show dev DEV
22 tc filter show dev DEV
24 tc [ -force ] [ -OK ] -b[atch] [ filename ]
27
29
32 Tc is used to configure Traffic Control in the Linux kernel. Traffic
33 Control consists of the following:
34 SHAPING
37 When traffic is shaped, its rate of transmission is under con‐
38 trol. Shaping may be more than lowering the available bandwidth
39 - it is also used to smooth out bursts in traffic for better
40 network behaviour. Shaping occurs on egress.
41 SCHEDULING
44 By scheduling the transmission of packets it is possible to
45 improve interactivity for traffic that needs it while still
46 guaranteeing bandwidth to bulk transfers. Reordering is also
47 called prioritizing, and happens only on egress.
48 POLICING
51 Whereas shaping deals with transmission of traffic, policing
52 pertains to traffic arriving. Policing thus occurs on ingress.
53 DROPPING
56 Traffic exceeding a set bandwidth may also be dropped forthwith,
57 both on ingress and on egress.
58 Processing of traffic is controlled by three kinds of objects: qdiscs,
61 classes and filters.
62
65 qdisc is short for 'queueing discipline' and it is elementary to under‐
66 standing traffic control. Whenever the kernel needs to send a packet to
67 an interface, it is enqueued to the qdisc configured for that inter‐
68 face. Immediately afterwards, the kernel tries to get as many packets
69 as possible from the qdisc, for giving them to the network adaptor
70 driver.
71 A simple QDISC is the 'pfifo' one, which does no processing at all and
73 is a pure First In, First Out queue. It does however store traffic when
74 the network interface can't handle it momentarily.
75
78 Some qdiscs can contain classes, which contain further qdiscs - traffic
79 may then be enqueued in any of the inner qdiscs, which are within the
80 classes. When the kernel tries to dequeue a packet from such a class‐
81 ful qdisc it can come from any of the classes. A qdisc may for example
82 prioritize certain kinds of traffic by trying to dequeue from certain
83 classes before others.
84
87 A filter is used by a classful qdisc to determine in which class a
88 packet will be enqueued. Whenever traffic arrives at a class with sub‐
89 classes, it needs to be classified. Various methods may be employed to
90 do so, one of these are the filters. All filters attached to the class
91 are called, until one of them returns with a verdict. If no verdict was
92 made, other criteria may be available. This differs per qdisc.
93 It is important to notice that filters reside within qdiscs - they are
95 not masters of what happens.
96 The available filters are:
98 basic Filter packets based on an ematch expression. See tc-ematch(8)
100 for details.
101 cgroup Filter packets based on the control group of their process. See
103 tc-cgroup(8) for details.
104 flow Flow-based classifier, filtering packets based on their flow
106 (identified by selectable keys). See tc-flow(8) for details.
107 fw Filter based on fwmark. Directly maps fwmark value to traffic
109 class. See tc-fw(8).
110 route Filter packets based on routing table. See tc-route(8) for
112 details.
113 rsvp Match Resource Reservation Protocol (RSVP) packets.
115 tcindex
117 Filter packets based on traffic control index. See tc-index(8).
118 u32 Generic filtering on arbitrary packet data, assisted by syntax
120 to abstract common operations. See tc-u32(8) for details.
121
124 The classless qdiscs are:
125 [p|b]fifo
127 Simplest usable qdisc, pure First In, First Out behaviour. Lim‐
128 ited in packets or in bytes.
129 pfifo_fast
131 Standard qdisc for 'Advanced Router' enabled kernels. Consists
132 of a three-band queue which honors Type of Service flags, as
133 well as the priority that may be assigned to a packet.
134 red Random Early Detection simulates physical congestion by randomly
136 dropping packets when nearing configured bandwidth allocation.
137 Well suited to very large bandwidth applications.
138 sfq Stochastic Fairness Queueing reorders queued traffic so each
140 'session' gets to send a packet in turn.
141 tbf The Token Bucket Filter is suited for slowing traffic down to a
143 precisely configured rate. Scales well to large bandwidths.
144
146 In the absence of classful qdiscs, classless qdiscs can only be
147 attached at the root of a device. Full syntax:
148 tc qdisc add dev DEV root QDISC QDISC-PARAMETERS
150 To remove, issue
152 tc qdisc del dev DEV root
154 The pfifo_fast qdisc is the automatic default in the absence of a con‐
156 figured qdisc.
157
160 The classful qdiscs are:
161 CBQ Class Based Queueing implements a rich linksharing hierarchy of
163 classes. It contains shaping elements as well as prioritizing
164 capabilities. Shaping is performed using link idle time calcula‐
165 tions based on average packet size and underlying link band‐
166 width. The latter may be ill-defined for some interfaces.
167 HTB The Hierarchy Token Bucket implements a rich linksharing hierar‐
169 chy of classes with an emphasis on conforming to existing prac‐
170 tices. HTB facilitates guaranteeing bandwidth to classes, while
171 also allowing specification of upper limits to inter-class shar‐
172 ing. It contains shaping elements, based on TBF and can priori‐
173 tize classes.
174 PRIO The PRIO qdisc is a non-shaping container for a configurable
176 number of classes which are dequeued in order. This allows for
177 easy prioritization of traffic, where lower classes are only
178 able to send if higher ones have no packets available. To facil‐
179 itate configuration, Type Of Service bits are honored by
180 default.
181
183 Classes form a tree, where each class has a single parent. A class may
184 have multiple children. Some qdiscs allow for runtime addition of
185 classes (CBQ, HTB) while others (PRIO) are created with a static number
186 of children.
187 Qdiscs which allow dynamic addition of classes can have zero or more
189 subclasses to which traffic may be enqueued.
190 Furthermore, each class contains a leaf qdisc which by default has
192 pfifo behaviour, although another qdisc can be attached in place. This
193 qdisc may again contain classes, but each class can have only one leaf
194 qdisc.
195 When a packet enters a classful qdisc it can be classified to one of
197 the classes within. Three criteria are available, although not all
198 qdiscs will use all three:
199 tc filters
201 If tc filters are attached to a class, they are consulted first
202 for relevant instructions. Filters can match on all fields of a
203 packet header, as well as on the firewall mark applied by
204 ipchains or iptables.
205 Type of Service
207 Some qdiscs have built in rules for classifying packets based on
208 the TOS field.
209 skb->priority
211 Userspace programs can encode a class-id in the 'skb->priority'
212 field using the SO_PRIORITY option.
213 Each node within the tree can have its own filters but higher level
215 filters may also point directly to lower classes.
216 If classification did not succeed, packets are enqueued to the leaf
218 qdisc attached to that class. Check qdisc specific manpages for
219 details, however.
220
223 All qdiscs, classes and filters have IDs, which can either be specified
224 or be automatically assigned.
225 IDs consist of a major number and a minor number, separated by a colon.
227 Both major and minor number are limited to 16 bits. There are two spe‐
228 cial values: root is signified by major and minor of all ones, and
229 unspecified is all zeros.
230 QDISCS A qdisc, which potentially can have children, gets assigned a
233 major number, called a 'handle', leaving the minor number names‐
234 pace available for classes. The handle is expressed as '10:'.
235 It is customary to explicitly assign a handle to qdiscs expected
236 to have children.
237 CLASSES
240 Classes residing under a qdisc share their qdisc major number,
241 but each have a separate minor number called a 'classid' that
242 has no relation to their parent classes, only to their parent
243 qdisc. The same naming custom as for qdiscs applies.
244 FILTERS
247 Filters have a three part ID, which is only needed when using a
248 hashed filter hierarchy.
249
251 All parameters accept a floating point number, possibly followed by a
252 unit.
253 Bandwidths or rates can be specified in:
255 kbps Kilobytes per second
257 mbps Megabytes per second
259 kbit Kilobits per second
261 mbit Megabits per second
263 bps or a bare number
265 Bytes per second
266 Amounts of data can be specified in:
268 kb or k
270 Kilobytes
271 mb or m
273 Megabytes
274 mbit Megabits
276 kbit Kilobits
278 b or a bare number
280 Bytes.
281 Lengths of time can be specified in:
283 s, sec or secs
285 Whole seconds
286 ms, msec or msecs
288 Milliseconds
289 us, usec, usecs or a bare number
291 Microseconds.
292
295 The following commands are available for qdiscs, classes and filter:
296 add Add a qdisc, class or filter to a node. For all entities, a par‐
298 ent must be passed, either by passing its ID or by attaching
299 directly to the root of a device. When creating a qdisc or a
300 filter, it can be named with the handle parameter. A class is
301 named with the classid parameter.
302 remove A qdisc can be removed by specifying its handle, which may also
305 be 'root'. All subclasses and their leaf qdiscs are automati‐
306 cally deleted, as well as any filters attached to them.
307 change Some entities can be modified 'in place'. Shares the syntax of
310 'add', with the exception that the handle cannot be changed and
311 neither can the parent. In other words, change cannot move a
312 node.
313 replace
316 Performs a nearly atomic remove/add on an existing node id. If
317 the node does not exist yet it is created.
318 link Only available for qdiscs and performs a replace where the node
321 must exist already.
322
325 The show command has additional formatting options:
326 -s, -stats, -statistics
329 output more statistics about packet usage.
330 -d, -details
333 output more detailed information about rates and cell sizes.
334 -r, -raw
337 output raw hex values for handles.
338 -p, -pretty
341 decode filter offset and mask values to equivalent filter com‐
342 mands based on TCP/IP.
343 -iec print rates in IEC units (ie. 1K = 1024).
346 -b, -b filename, -batch, -batch filename
349 read commands from provided file or standard input and invoke
350 them. First failure will cause termination of tc.
351 -force don't terminate tc on errors in batch mode. If there were any
354 errors during execution of the commands, the application return
355 code will be non zero.
356 -OK in batch mode, print OK and a new line on standard output after
359 each successfully interpreted command.
360
363 tc was written by Alexey N. Kuznetsov and added in Linux 2.2.
364
366 tc-basic(8), tc-bfifo(8), tc-cbq(8), tc-cgroup(8), tc-drr(8), tc-
367 ematch(8), tc-flow(8), tc-fw(8), tc-htb(8), tc-pfifo(8), tc-
368 pfifo_fast(8), tc-red(8), tc-route(8), tc-sfq(8), tc-tbf(8), tc-tcin‐
369 dex(8), tc-u32(8),
370 User documentation at http://lartc.org/, but please direct bugreports
371 and patches to: <netdev@vger.kernel.org>
372
375 Manpage maintained by bert hubert (ahu@ds9a.nl)
376iproute2 16 December 2001 TC(8)