1Flow filter in tc(8) Linux Flow filter in tc(8)
2
3
4
6 flow - flow based traffic control filter
7
9 Mapping mode:
10
11 tc filter ... flow map key KEY [ OPS ] [ OPTIONS ]
12
13 Hashing mode:
14
15 tc filter ... flow hash keys KEY_LIST [ perturb secs ] [ OPTIONS
16 ]
17
18 OPS := [ OPS ] OP
19
20 OPTIONS := [ divisor NUM ] [ baseclass ID ] [ match EMATCH_TREE ] [
21 action ACTION_SPEC ]
22
23 KEY_LIST := [ KEY_LIST ] KEY
24
25 OP := { or | and | xor | rshift | addend } NUM
26
27 ID := X:Y
28
29 KEY := { src | dst | proto | proto-src | proto-dst | iif | priority |
30 mark | nfct | nfct-src | nfct-dst | nfct-proto-src | nfct-
31 proto-dst | rt-classid | sk-uid | sk-gid | vlan-tag | rxhash }
32
34 The flow classifier is meant to extend the SFQ hashing capabilities
35 without hard-coding new hash functions. It also allows deterministic
36 mappings of keys to classes.
37
39 action ACTION_SPEC
40 Apply an action from the generic actions framework on matching
41 packets.
42
43 baseclass ID
44 An offset for the resulting class ID. ID may be root, none or a
45 hexadecimal class ID in the form [X:]Y. X must match
46 qdisc's/class's major handle (if omitted, the correct value is
47 chosen automatically). If the whole baseclass is omitted, Y
48 defaults to 1.
49
50 divisor NUM
51 Number of buckets to use for sorting into. Keys are calculated
52 modulo NUM.
53
54 hash keys KEY-LIST
55 Perform a jhash2 operation over the keys in KEY-LIST, the result
56 (modulo the divisor if given) is taken as class ID, optionally
57 offset by the value of baseclass. It is possible to specify an
58 interval (in seconds) after which jhash2's entropy source is
59 recreated using the perturb parameter.
60
61 map key KEY
62 Packet data identified by KEY is translated into class IDs to
63 push the packet into. The value may be mangled by OPS before
64 using it for the mapping. They are applied in the order listed
65 here:
66
67 and NUM
68 Perform bitwise AND operation with numeric value NUM.
69
70 or NUM
71 Perform bitwise OR operation with numeric value NUM.
72
73 xor NUM
74 Perform bitwise XOR operation with numeric value NUM.
75
76 rshift NUM
77 Shift the value of KEY to the right by NUM bits.
78
79 addend NUM
80 Add NUM to the value of KEY.
81
82 For the or, and, xor and rshift operations, NUM is assumed to be
83 an unsigned, 32bit integer value. For the addend operation, NUM
84 may be much more complex: It may be prefixed by a minus ('-')
85 sign to cause subtraction instead of addition and for keys of
86 src, dst, nfct-src and nfct-dst it may be given in IP address
87 notation. See below for an illustrating example.
88
89 match EMATCH_TREE
90 Match packets using the extended match infrastructure. See tc-
91 ematch(8) for a detailed description of the allowed syntax in
92 EMATCH_TREE.
93
95 In mapping mode, a single key is used (after optional permutation) to
96 build a class ID. The resulting ID is deducible in most cases. In hash‐
97 ing more, a number of keys may be specified which are then hashed and
98 the output used as class ID. This ID is not deducible in beforehand,
99 and may even change over time for a given flow if a perturb interval
100 has been given.
101
102 The range of class IDs can be limited by the divisor option, which is
103 used for a modulus.
104
105 src, dst
106 Use source or destination address as key. In case of IPv4 and
107 TIPC, this is the actual address value. For IPv6, the 128bit
108 address is folded into a 32bit value by XOR'ing the four 32bit
109 words. In all other cases, the kernel-internal socket address is
110 used (after folding into 32bits on 64bit systems).
111
112 proto Use the layer four protocol number as key.
113
114 proto-src
115 Use the layer four source port as key. If not available, the
116 kernel-internal socket address is used instead.
117
118 proto-dst
119 Use the layer four destination port as key. If not available,
120 the associated kernel-internal dst_entry address is used after
121 XOR'ing with the packet's layer three protocol number.
122
123 iif Use the incoming interface index as key.
124
125 priority
126 Use the packet's priority as key. Usually this is the IP
127 header's DSCP/ECN value.
128
129 mark Use the netfilter fwmark as key.
130
131 nfct Use the associated conntrack entry address as key.
132
133 nfct-src, nfct-dst, nfct-proto-src, nfct-proto-dst
134 These are conntrack-aware variants of src, dst, proto-src and
135 proto-dst. In case of NAT, these are basically the packet
136 header's values before NAT was applied.
137
138 rt-classid
139 Use the packet's destination routing table entry's realm as key.
140
141 sk-uid
142 sk-gid For locally generated packets, use the user or group ID the
143 originating socket belongs to as key.
144
145 vlan-tag
146 Use the packet's vlan ID as key.
147
148 rxhash Use the flow hash as key.
149
150
152 Classic SFQ hash:
153
154 tc filter add ... flow hash \
155 keys src,dst,proto,proto-src,proto-dst divisor 1024
156
157 Classic SFQ hash, but using information from conntrack to work properly
158 in combination with NAT:
159
160 tc filter add ... flow hash \
161 keys nfct-src,nfct-dst,proto,nfct-proto-src,nfct-proto-dst \
162 divisor 1024
163
164 Map destination IPs of 192.168.0.0/24 to classids 1-256:
165
166 tc filter add ... flow map \
167 key dst addend -192.168.0.0 divisor 256
168
169 Alternative to the above:
170
171 tc filter add ... flow map \
172 key dst and 0xff
173
174 The same, but in reverse order:
175
176 tc filter add ... flow map \
177 key dst and 0xff xor 0xff
178
180 tc(8), tc-ematch(8), tc-sfq(8)
181
182
183
184iproute2 20 Oct 2015 Flow filter in tc(8)