2 This documented is slightly dated but should give you idea of how things
8 An extension to the filtering/classification architecture of Linux Traffic
10 Up to 2.6.8 the only action that could be "attached" to a filter was policing.
11 i.e you could say something like:
14 tc filter add dev lo parent ffff: protocol ip prio 10 u32 match ip src \
15 127.0.0.1/32 flowid 1:1 police mtu 4000 rate 1500kbit burst 90k
18 which implies "if a packet is seen on the ingress of the lo device with
19 a source IP address of 127.0.0.1/32 we give it a classification id of 1:1 and
20 we execute a policing action which rate limits its bandwidth utilization
23 The new extensions allow for more than just policing actions to be added.
24 They are also fully backward compatible. If you have a kernel that doesnt
25 understand them, then the effect is null i.e if you have a newer tc
26 but older kernel, the actions are not installed. Likewise if you
27 have a newer kernel but older tc, obviously the tc will use current
28 syntax which will work fine. Of course to get the required effect you need
29 both newer tc and kernel. If you are reading this you have the
32 A side effect is that we can now get stateless firewalling to work with tc.
33 Essentially this is now an alternative to iptables.
34 I wont go into details of my dislike for iptables at times, but
35 scalability is one of the main issues; however, if you need stateful
36 classification - use netfilter (for now).
38 This stuff works on both ingress and egress qdiscs.
43 1) new additional syntax and actions enabled. Note old syntax is still valid.
45 Essentially this is still the same syntax as tc with a new construct
46 "action". The syntax is of the form:
47 tc filter add <DEVICE> parent 1:0 protocol ip prio 10 <Filter description>
48 flowid 1:1 action <ACTION description>*
50 You can have as many actions as you want (within sensible reasoning).
52 In the past the only real action was the policer; i.e you could do something
54 tc filter add dev lo parent ffff: protocol ip prio 10 u32 \
55 match ip src 127.0.0.1/32 flowid 1:1 \
56 police mtu 4000 rate 1500kbit burst 90k
58 Although you can still use the same syntax, now you can say:
60 tc filter add dev lo parent 1:0 protocol ip prio 10 u32 \
61 match ip src 127.0.0.1/32 flowid 1:1 \
62 action police mtu 4000 rate 1500kbit burst 90k
64 " generic Actions" (gact) at the moment are:
65 { drop, pass, reclassify, continue}
66 (If you have others, no listed here give me a reason and we will add them)
67 +drop says to drop the packet
68 +pass and ok (are equivalent) says to accept it
69 +reclassify requests for reclassification of the packet
70 +continue requests for next lookup to match
72 2)In order to take advantage of some of the targets written by the
73 iptables people, a classifier can have a packet being massaged by an
74 iptable target. I have only tested with mangler targets up to now.
75 (infact anything that is not in the mangling table is disabled right now)
78 *ingress is mapped to pre-routing hook
79 *egress is mapped to post-routing hook
80 I dont see much value in the other hooks, if you see it and email me good
81 reasons, the addition is trivial.
83 Example syntax for iptables targets usage becomes:
84 tc filter add ..... u32 <u32 syntax> action ipt -j <iptables target syntax>
87 tc filter add dev lo parent ffff: protocol ip prio 8 u32 \
88 match ip dst 127.0.0.8/32 flowid 1:12 \
89 action ipt -j mark --set-mark 2
91 NOTE: flowid 1:12 is parsed flowid 0x1:0x12. Make sure if you want flowid
92 decimal 12, then use flowid 1:c.
94 3) A feature i call pipe
95 The motivation is derived from Unix pipe mechanism but applied to packets.
96 Essentially take a matching packet and pass it through
97 action1 | action2 | action3 etc.
98 You could do something similar to this with the tc policer and the "continue"
99 operator but this rather restricts it to just the policer and requires
100 multiple rules (and lookups, hence quiet inefficient);
102 as an example -- and please note that this is just an example _not_ The
103 Word Youve Been Waiting For (yes i have had problems giving examples
104 which ended becoming dogma in documents and people modifying them a little
107 i selected the metering rates to be small so that i can show better how
110 The script below does the following:
111 - an incoming packet from 10.0.0.21 is first given a firewall mark of 1.
113 - It is then metered to make sure it does not exceed its allocated rate of
114 1Kbps. If it doesnt exceed rate, this is where we terminate action execution.
116 - If it does exceed its rate, its "color" changes to a mark of 2 and it is
117 then passed through a second meter.
119 -The second meter is shared across all flows on that device [i am suprised
120 that this seems to be not a well know feature of the policer; Bert was telling
121 me that someone was writing a qdisc just to do sharing across multiple devices;
122 it must be the summer heat again; weve had someone doing that every year around
123 summer -- the key to sharing is to use a operator "index" in your policer
124 rules (example "index 20"). All your rules have to use the same index to
127 -If the second meter is exceeded the color of the flow changes further to 3.
129 -We then pass the packet to another meter which is shared across all devices
130 in the system. If this meter is exceeded we drop the packet.
132 Note the mark can be used further up the system to do things like policy
133 or more interesting things on the egress.
135 ------------------ cut here -------------------------------
137 # Add an ingress qdisc on eth0
138 tc qdisc add dev eth0 ingress
140 #if you see an incoming packet from 10.0.0.21
141 tc filter add dev eth0 parent ffff: protocol ip prio 1 \
142 u32 match ip src 10.0.0.21/32 flowid 1:15 \
144 # first give it a mark of 1
145 action ipt -j mark --set-mark 1 index 2 \
147 # then pass it through a policer which allows 1kbps; if the flow
148 # doesnt exceed that rate, this is where we stop, if it exceeds we
149 # pipe the packet to the next action
150 action police rate 1kbit burst 9k pipe \
152 # which marks the packet fwmark as 2 and pipes
153 action ipt -j mark --set-mark 2 \
155 # next attempt to borrow b/width from a meter
156 # used across all flows incoming on eth0("index 30")
157 # and if that is exceeded we pipe to the next action
158 action police index 30 mtu 5000 rate 1kbit burst 10k pipe \
159 # mark it as fwmark 3 if exceeded
160 action ipt -j mark --set-mark 3 \
161 # and then attempt to borrow from a meter used by all devices in the
162 # system. Should this be exceeded, drop the packet on the floor.
163 action police index 20 mtu 5000 rate 1kbit burst 90k drop
164 ---------------------------------
166 Now lets see the actions installed with
167 "tc filter show parent ffff: dev eth0"
169 -------- output -----------
170 jroot# tc filter show parent ffff: dev eth0
171 filter protocol ip pref 1 u32
172 filter protocol ip pref 1 u32 fh 800: ht divisor 1
173 filter protocol ip pref 1 u32 fh 800::800 order 2048 key ht 800 bkt 0 flowid 1:15
175 action order 1: tablename: mangle hook: NF_IP_PRE_ROUTING
176 target MARK set 0x1 index 2
178 action order 2: police 1 action pipe rate 1Kbit burst 9Kb mtu 2Kb
180 action order 3: tablename: mangle hook: NF_IP_PRE_ROUTING
181 target MARK set 0x2 index 1
183 action order 4: police 30 action pipe rate 1Kbit burst 10Kb mtu 5000b
185 action order 5: tablename: mangle hook: NF_IP_PRE_ROUTING
186 target MARK set 0x3 index 3
188 action order 6: police 20 action drop rate 1Kbit burst 90Kb mtu 5000b
190 match 0a000015/ffffffff at 12
191 -------------------------------
193 Note the ordering of the actions is based on the order in which we entered
194 them. In the future i will add explicit priorities.
196 Now lets run a ping -f from 10.0.0.21 to this host; stop the ping after
197 you see a few lines of dots
200 [root@jzny hadi]# ping -f 10.0.0.22
201 PING 10.0.0.22 (10.0.0.22): 56 data bytes
202 ....................................................................................................................................................................................................................................................................................................................................................................................................................................................
203 --- 10.0.0.22 ping statistics ---
204 2248 packets transmitted, 1811 packets received, 19% packet loss
205 round-trip min/avg/max = 0.7/9.3/20.1 ms
206 -----------------------------
208 Now lets take a look at the stats with "tc -s filter show parent ffff: dev eth0"
211 jroot# tc -s filter show parent ffff: dev eth0
212 filter protocol ip pref 1 u32
213 filter protocol ip pref 1 u32 fh 800: ht divisor 1
214 filter protocol ip pref 1 u32 fh 800::800 order 2048 key ht 800 bkt 0 flowid 1:1
217 action order 1: tablename: mangle hook: NF_IP_PRE_ROUTING
218 target MARK set 0x1 index 2
219 Sent 188832 bytes 2248 pkts (dropped 0, overlimits 0)
221 action order 2: police 1 action pipe rate 1Kbit burst 9Kb mtu 2Kb
222 Sent 188832 bytes 2248 pkts (dropped 0, overlimits 2122)
224 action order 3: tablename: mangle hook: NF_IP_PRE_ROUTING
225 target MARK set 0x2 index 1
226 Sent 178248 bytes 2122 pkts (dropped 0, overlimits 0)
228 action order 4: police 30 action pipe rate 1Kbit burst 10Kb mtu 5000b
229 Sent 178248 bytes 2122 pkts (dropped 0, overlimits 1945)
231 action order 5: tablename: mangle hook: NF_IP_PRE_ROUTING
232 target MARK set 0x3 index 3
233 Sent 163380 bytes 1945 pkts (dropped 0, overlimits 0)
235 action order 6: police 20 action drop rate 1Kbit burst 90Kb mtu 5000b
236 Sent 163380 bytes 1945 pkts (dropped 0, overlimits 437)
238 match 0a000015/ffffffff at 12
239 -------------------------------
244 Wanna write an action module?
245 ------------------------------
246 Its easy. Either look at the code or send me email. I will document at
247 some point; will also accept documentation.
252 Lotsa goodies/features coming. Requests also being accepted.
253 At the moment the focus has been on getting the architecture in place.
254 Expect new things in the spurious time i have to work on this
255 (particularly around end of year when i have typically get time off