3 HOWTO for the linux packet generator
4 ------------------------------------
8 Enable CONFIG_NET_PKTGEN to compile and build pktgen.o either in kernel
9 or as module. Module is preferred. insmod pktgen if needed. Once running
10 pktgen creates a thread on each CPU where each thread has affinity to its CPU.
11 Monitoring and controlling is done via /proc. Easiest to select a suitable
12 a sample script and configure.
17 root 129 0.3 0.0 0 0 ? SW 2003 523:20 [pktgen/0]
18 root 130 0.3 0.0 0 0 ? SW 2003 509:50 [pktgen/1]
21 For monitoring and control pktgen creates:
22 /proc/net/pktgen/pgctrl
23 /proc/net/pktgen/kpktgend_X
27 Tuning NIC for max performance
28 ==============================
30 The default NIC setting are (likely) not tuned for pktgen's artificial
31 overload type of benchmarking, as this could hurt the normal use-case.
33 Specifically increasing the TX ring buffer in the NIC:
34 # ethtool -G ethX tx 1024
36 A larger TX ring can improve pktgen's performance, while it can hurt
37 in the general case, 1) because the TX ring buffer might get larger
38 than the CPUs L1/L2 cache, 2) because it allow more queueing in the
39 NIC HW layer (which is bad for bufferbloat).
41 One should be careful to conclude, that packets/descriptors in the HW
42 TX ring cause delay. Drivers usually delay cleaning up the
43 ring-buffers (for various performance reasons), thus packets stalling
44 the TX ring, might just be waiting for cleanup.
46 This cleanup issues is specifically the case, for the driver ixgbe
47 (Intel 82599 chip). This driver (ixgbe) combine TX+RX ring cleanups,
48 and the cleanup interval is affected by the ethtool --coalesce setting
49 of parameter "rx-usecs".
51 For ixgbe use e.g "30" resulting in approx 33K interrupts/sec (1/30*10^6):
52 # ethtool -C ethX rx-usecs 30
57 /proc/net/pktgen/kpktgend_0
58 Name: kpktgend_0 max_before_softirq: 10000
61 Result: OK: max_before_softirq=10000
63 Most important the devices assigned to thread. Note! A device can only belong
70 Parm section holds configured info. Current hold running stats.
71 Result is printed after run or after interruption. Example:
75 Params: count 10000000 min_pkt_size: 60 max_pkt_size: 60
76 frags: 0 delay: 0 clone_skb: 1000000 ifname: eth1
78 dst_min: 10.10.11.2 dst_max:
80 src_mac: 00:00:00:00:00:00 dst_mac: 00:04:23:AC:FD:82
81 udp_src_min: 9 udp_src_max: 9 udp_dst_min: 9 udp_dst_max: 9
82 src_mac_count: 0 dst_mac_count: 0
85 pkts-sofar: 10000000 errors: 39664
86 started: 1103053986245187us stopped: 1103053999346329us idle: 880401us
87 seq_num: 10000011 cur_dst_mac_offset: 0 cur_src_mac_offset: 0
88 cur_saddr: 0x10a0a0a cur_daddr: 0x20b0a0a
89 cur_udp_dst: 9 cur_udp_src: 9
91 Result: OK: 13101142(c12220741+d880401) usec, 10000000 (60byte,0frags)
92 763292pps 390Mb/sec (390805504bps) errors: 39664
94 Configuring threads and devices
95 ================================
96 This is done via the /proc interface easiest done via pgset in the scripts
100 pgset "clone_skb 1" sets the number of copies of the same packet
101 pgset "clone_skb 0" use single SKB for all transmits
102 pgset "pkt_size 9014" sets packet size to 9014
103 pgset "frags 5" packet will consist of 5 fragments
104 pgset "count 200000" sets number of packets to send, set to zero
105 for continuous sends until explicitly stopped.
107 pgset "delay 5000" adds delay to hard_start_xmit(). nanoseconds
109 pgset "dst 10.0.0.1" sets IP destination address
110 (BEWARE! This generator is very aggressive!)
112 pgset "dst_min 10.0.0.1" Same as dst
113 pgset "dst_max 10.0.0.254" Set the maximum destination IP.
114 pgset "src_min 10.0.0.1" Set the minimum (or only) source IP.
115 pgset "src_max 10.0.0.254" Set the maximum source IP.
116 pgset "dst6 fec0::1" IPV6 destination address
117 pgset "src6 fec0::2" IPV6 source address
118 pgset "dstmac 00:00:00:00:00:00" sets MAC destination address
119 pgset "srcmac 00:00:00:00:00:00" sets MAC source address
121 pgset "queue_map_min 0" Sets the min value of tx queue interval
122 pgset "queue_map_max 7" Sets the max value of tx queue interval, for multiqueue devices
123 To select queue 1 of a given device,
124 use queue_map_min=1 and queue_map_max=1
126 pgset "src_mac_count 1" Sets the number of MACs we'll range through.
127 The 'minimum' MAC is what you set with srcmac.
129 pgset "dst_mac_count 1" Sets the number of MACs we'll range through.
130 The 'minimum' MAC is what you set with dstmac.
132 pgset "flag [name]" Set a flag to determine behaviour. Current flags
133 are: IPSRC_RND # IP source is random (between min/max)
134 IPDST_RND # IP destination is random
135 UDPSRC_RND, UDPDST_RND,
136 MACSRC_RND, MACDST_RND
138 MPLS_RND, VID_RND, SVID_RND
140 QUEUE_MAP_RND # queue map random
141 QUEUE_MAP_CPU # queue map mirrors smp_processor_id()
143 IPSEC # IPsec encapsulation (needs CONFIG_XFRM)
144 NODE_ALLOC # node specific memory allocation
146 pgset spi SPI_VALUE Set specific SA used to transform packet.
148 pgset "udp_src_min 9" set UDP source port min, If < udp_src_max, then
149 cycle through the port range.
151 pgset "udp_src_max 9" set UDP source port max.
152 pgset "udp_dst_min 9" set UDP destination port min, If < udp_dst_max, then
153 cycle through the port range.
154 pgset "udp_dst_max 9" set UDP destination port max.
156 pgset "mpls 0001000a,0002000a,0000000a" set MPLS labels (in this example
157 outer label=16,middle label=32,
158 inner label=0 (IPv4 NULL)) Note that
159 there must be no spaces between the
160 arguments. Leading zeros are required.
161 Do not set the bottom of stack bit,
162 that's done automatically. If you do
163 set the bottom of stack bit, that
164 indicates that you want to randomly
165 generate that address and the flag
166 MPLS_RND will be turned on. You
167 can have any mix of random and fixed
168 labels in the label stack.
170 pgset "mpls 0" turn off mpls (or any invalid argument works too!)
172 pgset "vlan_id 77" set VLAN ID 0-4095
173 pgset "vlan_p 3" set priority bit 0-7 (default 0)
174 pgset "vlan_cfi 0" set canonical format identifier 0-1 (default 0)
176 pgset "svlan_id 22" set SVLAN ID 0-4095
177 pgset "svlan_p 3" set priority bit 0-7 (default 0)
178 pgset "svlan_cfi 0" set canonical format identifier 0-1 (default 0)
180 pgset "vlan_id 9999" > 4095 remove vlan and svlan tags
181 pgset "svlan 9999" > 4095 remove svlan tag
184 pgset "tos XX" set former IPv4 TOS field (e.g. "tos 28" for AF11 no ECN, default 00)
185 pgset "traffic_class XX" set former IPv6 TRAFFIC CLASS (e.g. "traffic_class B8" for EF no ECN, default 00)
187 pgset stop aborts injection. Also, ^C aborts generator.
189 pgset "rate 300M" set rate to 300 Mb/s
190 pgset "ratep 1000000" set rate to 1Mpps
195 A collection of small tutorial scripts for pktgen is in examples dir.
197 pktgen.conf-1-1 # 1 CPU 1 dev
198 pktgen.conf-1-2 # 1 CPU 2 dev
199 pktgen.conf-2-1 # 2 CPU's 1 dev
200 pktgen.conf-2-2 # 2 CPU's 2 dev
201 pktgen.conf-1-1-rdos # 1 CPU 1 dev w. route DoS
202 pktgen.conf-1-1-ip6 # 1 CPU 1 dev ipv6
203 pktgen.conf-1-1-ip6-rdos # 1 CPU 1 dev ipv6 w. route DoS
204 pktgen.conf-1-1-flows # 1 CPU 1 dev multiple flows.
206 Run in shell: ./pktgen.conf-X-Y It does all the setup including sending.
211 Note when adding devices to a specific CPU there good idea to also assign
212 /proc/irq/XX/smp_affinity so the TX-interrupts gets bound to the same CPU.
213 as this reduces cache bouncing when freeing skb's.
217 Default IPsec transformation with ESP encapsulation plus Transport mode
218 could be enabled by simply setting:
223 To avoid breaking existing testbed scripts for using AH type and tunnel mode,
224 user could use "pgset spi SPI_VALUE" to specify which formal of transformation
228 Current commands and configuration options
229 ==========================================
231 ** Pgcontrol commands:
307 ftp://robur.slu.se/pub/Linux/net-development/pktgen-testing/
308 ftp://robur.slu.se/pub/Linux/net-development/pktgen-testing/examples/
310 Paper from Linux-Kongress in Erlangen 2004.
311 ftp://robur.slu.se/pub/Linux/net-development/pktgen-testing/pktgen_paper.pdf
314 Grant Grundler for testing on IA-64 and parisc, Harald Welte, Lennert Buytenhek
315 Stephen Hemminger, Andi Kleen, Dave Miller and many others.
318 Good luck with the linux net-development.