2 Copyright(c) 2010-2014 Intel Corporation. All rights reserved.
5 Redistribution and use in source and binary forms, with or without
6 modification, are permitted provided that the following conditions
9 * Redistributions of source code must retain the above copyright
10 notice, this list of conditions and the following disclaimer.
11 * Redistributions in binary form must reproduce the above copyright
12 notice, this list of conditions and the following disclaimer in
13 the documentation and/or other materials provided with the
15 * Neither the name of Intel Corporation nor the names of its
16 contributors may be used to endorse or promote products derived
17 from this software without specific prior written permission.
19 THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
20 "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
21 LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
22 A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
23 OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
24 SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
25 LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
26 DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
27 THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
28 (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
29 OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
31 Exception Path Sample Application
32 =================================
34 The Exception Path sample application is a simple example that demonstrates the use of the DPDK
35 to set up an exception path for packets to go through the Linux* kernel.
36 This is done by using virtual TAP network interfaces.
37 These can be read from and written to by the DPDK application and
38 appear to the kernel as a standard network interface.
43 The application creates two threads for each NIC port being used.
44 One thread reads from the port and writes the data unmodified to a thread-specific TAP interface.
45 The second thread reads from a TAP interface and writes the data unmodified to the NIC port.
47 The packet flow through the exception path application is as shown in the following figure.
49 .. _figure_exception_path_example:
51 .. figure:: img/exception_path_example.*
56 To make throughput measurements, kernel bridges must be setup to forward data between the bridges appropriately.
58 Compiling the Application
59 -------------------------
61 #. Go to example directory:
63 .. code-block:: console
65 export RTE_SDK=/path/to/rte_sdk
66 cd ${RTE_SDK}/examples/exception_path
68 #. Set the target (a default target will be used if not specified).
71 .. code-block:: console
73 export RTE_TARGET=x86_64-native-linuxapp-gcc
75 This application is intended as a linuxapp only.
76 See the *DPDK Getting Started Guide* for possible RTE_TARGET values.
78 #. Build the application:
80 .. code-block:: console
84 Running the Application
85 -----------------------
87 The application requires a number of command line options:
89 .. code-block:: console
91 .build/exception_path [EAL options] -- -p PORTMASK -i IN_CORES -o OUT_CORES
95 * -p PORTMASK: A hex bitmask of ports to use
97 * -i IN_CORES: A hex bitmask of cores which read from NIC
99 * -o OUT_CORES: A hex bitmask of cores which write to NIC
101 Refer to the *DPDK Getting Started Guide* for general information on running applications
102 and the Environment Abstraction Layer (EAL) options.
104 The number of bits set in each bitmask must be the same.
105 The coremask -c parameter of the EAL options should include IN_CORES and OUT_CORES.
106 The same bit must not be set in IN_CORES and OUT_CORES.
107 The affinities between ports and cores are set beginning with the least significant bit of each mask, that is,
108 the port represented by the lowest bit in PORTMASK is read from by the core represented by the lowest bit in IN_CORES,
109 and written to by the core represented by the lowest bit in OUT_CORES.
111 For example to run the application with two ports and four cores:
113 .. code-block:: console
115 ./build/exception_path -c f -n 4 -- -p 3 -i 3 -o c
120 While the application is running, statistics on packets sent and
121 received can be displayed by sending the SIGUSR1 signal to the application from another terminal:
123 .. code-block:: console
125 killall -USR1 exception_path
127 The statistics can be reset by sending a SIGUSR2 signal in a similar way.
132 The following sections provide some explanation of the code.
137 Setup of the mbuf pool, driver and queues is similar to the setup done in the L2 Forwarding sample application
138 (see Chapter 9 "L2 forwarding Sample Application (in Real and Virtualized Environments" for details).
139 In addition, the TAP interfaces must also be created.
140 A TAP interface is created for each lcore that is being used.
141 The code for creating the TAP interface is as follows:
146 * Create a tap network interface, or use existing one with same name.
147 * If name[0]='\0' then a name is automatically assigned and returned in name.
150 static int tap_create(char *name)
155 fd = open("/dev/net/tun", O_RDWR);
159 memset(&ifr, 0, sizeof(ifr));
161 /* TAP device without packet information */
163 ifr.ifr_flags = IFF_TAP | IFF_NO_PI;
165 rte_snprinf(ifr.ifr_name, IFNAMSIZ, name);
167 ret = ioctl(fd, TUNSETIFF, (void *) &ifr);
176 rte_snprintf(name, IFNAMSIZ, ifr.ifr_name);
181 The other step in the initialization process that is unique to this sample application
182 is the association of each port with two cores:
184 * One core to read from the port and write to a TAP interface
186 * A second core to read from a TAP interface and write to the port
188 This is done using an array called port_ids[], which is indexed by the lcore IDs.
189 The population of this array is shown below:
196 RTE_LCORE_FOREACH(i) {
197 if (input_cores_mask & (1ULL << i)) {
198 /* Skip ports that are not enabled */
199 while ((ports_mask & (1 << rx_port)) == 0) {
201 if (rx_port > (sizeof(ports_mask) * 8))
202 goto fail; /* not enough ports */
204 port_ids[i] = rx_port++;
205 } else if (output_cores_mask & (1ULL << i)) {
206 /* Skip ports that are not enabled */
207 while ((ports_mask & (1 << tx_port)) == 0) {
209 if (tx_port > (sizeof(ports_mask) * 8))
210 goto fail; /* not enough ports */
212 port_ids[i] = tx_port++;
219 After the initialization steps are complete, the main_loop() function is run on each lcore.
220 This function first checks the lcore_id against the user provided input_cores_mask and output_cores_mask to see
221 if this core is reading from or writing to a TAP interface.
223 For the case that reads from a NIC port, the packet reception is the same as in the L2 Forwarding sample application
224 (see Section 9.4.6, "Receive, Process and Transmit Packets").
225 The packet transmission is done by calling write() with the file descriptor of the appropriate TAP interface
226 and then explicitly freeing the mbuf back to the pool.
230 /* Loop forever reading from NIC and writing to tap */
233 struct rte_mbuf *pkts_burst[PKT_BURST_SZ];
236 const unsigned nb_rx = rte_eth_rx_burst(port_ids[lcore_id], 0, pkts_burst, PKT_BURST_SZ);
238 lcore_stats[lcore_id].rx += nb_rx;
240 for (i = 0; likely(i < nb_rx); i++) {
241 struct rte_mbuf *m = pkts_burst[i];
242 int ret = write(tap_fd, rte_pktmbuf_mtod(m, void*),
244 rte_pktmbuf_data_len(m));
247 lcore_stats[lcore_id].dropped++;
249 lcore_stats[lcore_id].tx++;
253 For the other case that reads from a TAP interface and writes to a NIC port,
254 packets are retrieved by doing a read() from the file descriptor of the appropriate TAP interface.
255 This fills in the data into the mbuf, then other fields are set manually.
256 The packet can then be transmitted as normal.
260 /* Loop forever reading from tap and writing to NIC */
264 struct rte_mbuf *m = rte_pktmbuf_alloc(pktmbuf_pool);
269 ret = read(tap_fd, m->pkt.data, MAX_PACKET_SZ); lcore_stats[lcore_id].rx++;
270 if (unlikely(ret < 0)) {
271 FATAL_ERROR("Reading from %s interface failed", tap_name);
276 m->pkt.data_len = (uint16_t)ret;
278 ret = rte_eth_tx_burst(port_ids[lcore_id], 0, &m, 1);
279 if (unlikely(ret < 1)) {
281 lcore_stats[lcore_id].dropped++;
284 lcore_stats[lcore_id].tx++;
288 To set up loops for measuring throughput, TAP interfaces can be connected using bridging.
289 The steps to do this are described in the section that follows.
291 Managing TAP Interfaces and Bridges
292 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
294 The Exception Path sample application creates TAP interfaces with names of the format tap_dpdk_nn,
295 where nn is the lcore ID. These TAP interfaces need to be configured for use:
297 .. code-block:: console
299 ifconfig tap_dpdk_00 up
301 To set up a bridge between two interfaces so that packets sent to one interface can be read from another,
304 .. code-block:: console
307 brctl addif br0 tap_dpdk_00
308 brctl addif br0 tap_dpdk_03
311 The TAP interfaces created by this application exist only when the application is running,
312 so the steps above need to be repeated each time the application is run.
313 To avoid this, persistent TAP interfaces can be created using openvpn:
315 .. code-block:: console
317 openvpn --mktun --dev tap_dpdk_00
319 If this method is used, then the steps above have to be done only once and
320 the same TAP interfaces can be reused each time the application is run.
321 To remove bridges and persistent TAP interfaces, the following commands are used:
323 .. code-block:: console
327 openvpn --rmtun --dev tap_dpdk_00