80. IP fragmentation Tests

The IP fragmentation results are produced using ‘’ip_fragmentation’’ application. The test application should run with both IPv4 and IPv6 fragmentation.

The suit support NIC: Intel® Ethernet 700 Series, Intel® Ethernet 800 Series, 82599 and igc driver NIC.

80.1. Prerequisites

  1. Hardware requirements:

    • For each CPU socket, each memory channel should be populated with at least 1x DIMM

    • Board is populated with at least 2x 1GbE or 10GbE ports. Special PCIe restrictions may be required for performance. For example, the following requirements should be met for Intel 82599 NICs:

      • NICs are plugged into PCIe Gen2 or Gen3 slots
      • For PCIe Gen2 slots, the number of lanes should be 8x or higher
      • A single port from each NIC should be used, so for 2x ports, 2x NICs should be used

    • NIC ports connected to traffic generator. It is assumed that the NIC ports P0, P1, P2, P3 (as identified by the DPDK application) are connected to the traffic generator ports TG0, TG1, TG2, TG3. The application-side port mask of NIC ports P0, P1, P2, P3 is noted as PORTMASK in this section. Traffic generator should support sending jumbo frames with size up to 9K.

  2. BIOS requirements:

    • Intel Hyper-Threading Technology is ENABLED
    • Hardware Prefetcher is DISABLED
    • Adjacent Cache Line Prefetch is DISABLED
    • Direct Cache Access is DISABLED

  3. Linux kernel requirements:

    • Linux kernel has the following features enabled: huge page support, UIO, HPET
    • Appropriate number of huge pages are reserved at kernel boot time
    • The IDs of the hardware threads (logical cores) per each CPU socket can be determined by parsing the file /proc/cpuinfo. The naming convention for the logical cores is: C{x.y.z} = hyper-thread z of physical core y of CPU socket x, with typical values of x = 0 .. 3, y = 0 .. 7, z = 0 .. 1. Logical cores C{0.0.1} and C{0.0.1} should be avoided while executing the test, as they are used by the Linux kernel for running regular processes.

  4. Software application requirements

  5. If using vfio the kernel must be >= 3.6+ and VT-d must be enabled in bios.When using vfio, use the following commands to load the vfio driver and bind it to the device under test:

    modprobe vfio
modprobe vfio-pci
usertools/dpdk-devbind.py --bind=vfio-pci device_bus_id
    • The test can be run with IPv4 package. The LPM table used for IPv4 packet routing is:
    Entry # LPM prefix (IP/length) Output port
    0 100.10.0.0/16 P2
    1 100.20.0.0/16 P2
    2 100.30.0.0/16 P0
    3 100.40.0.0/16 P0
    • The test can be run with IPv6 package, which follows rules below.
    • There is no support for Hop-by-Hop or Routing extension headers in the packet to be fragmented. All other optional headers, which are not part of the unfragmentable part of the IPv6 packet are supported.
    • When a fragment is generated, its identification field in the IPv6 fragmentation extension header is set to 0. This is not RFC compliant, but proper identification number generation is out of the scope of the application and routers in an IPv6 path are not allowed to fragment in the first place... Generating that identification number is the job of a proper IP stack.
    • The LPM table used for IPv6 packet routing is:
    Entry # LPM prefix (IP/length) Output port
    0 101:101:101:101:101:101:101:101/48 P2
    1 201:101:101:101:101:101:101:101/48 P2
    2 301:101:101:101:101:101:101:101/48 P0
    3 401:101:101:101:101:101:101:101/48 P0

    The following items are configured through the command line interface of the application:

    • The set of one or several RX queues to be enabled for each NIC port
    • The set of logical cores to execute the packet forwarding task
    • Mapping of the NIC RX queues to logical cores handling them.

  6. Compile examples/ip_fragmentation:

    meson configure -Dexamples=ip_fragmentation x86_64-native-linuxapp-gcc
ninja -C x86_64-native-linuxapp-gcc

80.2. Test Case 1: IP Fragmentation normal ip packet forward

With 1 input and 1 output port make sure that IP header and contents of the header are forwarded correctly for the frame sizes: 64, 128, 256, 512,1024, 1518 bytes.

80.3. Test Case 2: IP Fragmentation Don’t fragment

In TG set IP flag “Don’t fragment” and make sure that frames with size 1519 bytes are discarded by ip_frag.

80.4. Test Case 3: IP Fragmentation May fragment

In TG set IP flag “May fragment” and send frames with the following sizes: 1519 bytes, 2K, 3K, 4K, 5K, 6K, 7K, 8K, 9K. For each of them check that:

  1. Check number of output packets.
  2. Check header of each output packet: length, ID, fragment offset, flags.
  3. Check payload: size and contents as expected, not corrupted.

80.5. Test Case 4: Throughput test

The test report should provide the throughput rate measurements (in mpps and % of the line rate for 2x NIC ports) for the following input frame sizes: 64 bytes, 1518 bytes, 1519 bytes, 2K, 9k.

The following configurations should be tested:


# of ports Socket/Core/HyperThread Total # of sw threads
2 1S/1C/1T 1
2 1S/1C/2T 2
2 1S/2C/1T 2
2 2S/1C/1T 2

Command line:

./x86_64-native-linuxapp-gcc/examples/dpdk-ip_fragmentation -c <LCOREMASK> -n 4 -- [-P] -p PORTMASK
   -q <NUM_OF_PORTS_PER_THREAD>