DPDK Test Plans
stable
  • 1. Port Blacklist Tests
  • 2. RX/TX Checksum Offload Tests
  • 3. Cloud filter Support through Ethtool Tests
  • 4. Coremask Tests
  • 5. Cryptodev Performance Application Tests
  • 6. Fortville DDP (Dynamic Device Personalization) Tests
  • 7. Dual VLAN Offload Tests
  • 8. Dynamic Driver Configuration Tests
  • 9. External Tag (E-tag) Tests
  • 10. External Mempool Handler Tests
  • 11. Niantic Flow Director Tests
  • 12. VEB Switch and floating VEB Tests
  • 13. Fortville Granularity Configuration of RSS and 32-bit GRE key Tests
  • 14. FM10k FTAG Forwarding Tests
  • 15. Generic Filter Tests
  • 16. DPDK Hotplug API Tests
  • 17. IEEE1588 Precise Time Protocol Tests
  • 18. One-shot Rx Interrupt Tests
  • 19. IP fragmentation Tests
  • 20. Generic Routing Encapsulation (GRE) Tests
  • 21. IP Pipeline Application Tests
  • 22. IP Reassembly Tests
  • 23. Jumbo Frame Tests
  • 24. Kernel NIC Interface (KNI) Tests
  • 25. CryptoDev API Tests
  • 26. L2 Forwarding Tests
  • 27. L3 Forwarding Exact Match Tests
  • 28. L3 Forwarding Tests
  • 29. Ethernet Link Flow Control Tests
    • 29.1. Prerequisites
    • 29.2. Test Case: test_perf_flowctrl_on_pause_fwd_on
    • 29.3. Test Case: test_perf_flowctrl_on_pause_fwd_off
    • 29.4. Test Case: test_perf_flowctrl_rx_on
    • 29.5. Test Case: test_perf_flowctrl_off_pause_fwd_off
    • 29.6. Test Case: test_perf_flowctrl_off_pause_fwd_on
    • 29.7. Test Case: test_perf_flowctrl_tx_on
  • 30. Link Status Detection Tests
  • 31. Whitelisting Tests
  • 32. Niantic Media Access Control Security (MACsec) Tests
  • 33. External Mempool Handler Tests
  • 34. NIC Statistics Tests
  • 35. Fortville NVGRE Tests
  • 36. Bonding Tests
  • 37. TestPMD PCAP Tests
  • 38. Fortville RSS - Configuring Hash Function Tests
  • 39. Niantic Reta (Redirection table) Tests
  • 40. Niantic PMD Tests
  • 41. PTYPE Mapping Tests
  • 42. Shutdown API Queue Tests
  • 43. Scattered Packets Tests
  • 44. Short-lived Application Tests
  • 45. Shutdown API Feature Tests
  • 46. SRIOV and InterVM Communication Tests
  • 47. Stability Tests
  • 48. Transmit Segmentation Offload (TSO) Tests
  • 49. Tx Preparation Forwarding Tests
  • 50. Unified Packet Type Tests
  • 51. Userspace Ethtool Tests
  • 52. VEB Switch and floating VEB Tests
  • 53. VFD as SRIOV Policy Manager Tests
  • 54. VF Jumboframe Tests
  • 55. VF One-shot Rx Interrupt Tests
  • 56. VF MAC Filter Tests
  • 57. VF Offload
  • 58. VF Packet RxTX Tests
  • 59. VF PF Reset Tests
  • 60. VF Port Start Stop Tests
  • 61. VF RSS - Configuring Hash Function Tests
  • 62. VF to VF Bridge Tests
  • 63. VF VLAN Tests
  • 64. Vhost PMD Xstats Tests
  • 65. Vhost TSO Tests
  • 66. Vhost User Live Migration Tests
  • 67. Virtio-1.0 Support Tests
  • 68. VLAN Ethertype Config Tests
  • 69. VLAN Offload Tests
  • 70. VMDQ Tests
  • 71. VM Power Management Tests
  • 72. Fortville Vxlan Tests
  • 73. Niantic ixgbe_get_vf_queue Include Extra Information Tests
  • 74. Fortville Configure RSS Queue Regions Tests
  • 75. Niantic Inline IPsec Tests
  • 76. Eventdev Pipeline SW PMD Tests
  • 77. Fortville Dynamic Mapping of Flow Types to PCTYPEs Tests
  • 78. VFD as SRIOV Policy Manager Tests
  • 79. Multiple Pthread Test
  • 80. Fortville Cloud filters for QinQ steering Tests
  • 81. Fortville DDP GTP-C/GTP-U Tests
  • 82. Generic filter/flow api
  • 83. DDP GTP Qregion
  • 84. Dynamically Configure VF Queue Number
  • 85. Vhost/Virtio multiple queue qemu test plan
  • 86. Vhost MTU Test Plan
  • 87. Unit Tests: Cmdline
  • 88. Unit Tests: CRC
  • 89. Unit Tests: Cryptodev
  • 90. Unit Tests: Dump Log History
  • 91. Unit Tests: Dump Ring
  • 92. Unit Tests: Dump Mempool
  • 93. Unit Tests: Dump Physical Memory
  • 94. Unit Tests: Dump Memzone
  • 95. Unit Tests: Dump Struct Size
  • 96. Unit Tests: EAL
  • 97. Unit Tests: KNI
  • 98. Unit Tests: LPM
  • 99. Unit Tests: LPM_ipv6
  • 100. Unit Tests: Mbuf
  • 101. Unit Tests: Mempool
  • 102. Unit Tests: PMD Performance
  • 103. Unit Tests: Power Library
  • 104. Unit Tests: Random Early Detection (RED)
  • 105. Unit Tests: Metering
  • 106. Unit tests: Scheduler
  • 107. Unit Tests: Ring Pmd
  • 108. Unit Tests: Ring
  • 109. Unit Tests: Ring Performance
  • 110. Unit tests: Timer
  • 111. Sample Application Tests: Cmdline Example
  • 112. Sample Application Tests: Hello World Example
  • 113. Sample Application Tests: Keep Alive Example
  • 114. Sample Application Tests: Multi-Process
  • 115. Sample Application Tests: Netmap Compatibility
  • 116. Sample Application Tests: Quota and Water-mark
  • 117. Sample Application Tests: RX/TX Callbacks
  • 118. Sample Application Tests: Basic Forwarding/Skeleton Application
  • 119. Sample Application Tests: Timer Example
  • 120. Sample Application Tests: Vxlan Example
  • 121. Sample Application Tests: IEEE1588
  • 122. Sample Application Tests: Packet distributor
  • 123. Sample Application Tests: Elastic Flow Distributor
DPDK Test Plans
  • Docs »
  • 29. Ethernet Link Flow Control Tests
  • View page source

29. Ethernet Link Flow Control Tests¶

The support of Ethernet link flow control features by Poll Mode Drivers consists in:

  • At the receive side, if packet buffer is not enough, NIC will send out the pause frame to peer and ask the peer to slow down the Ethernet frame # transmission.
  • At the transmit side, if pause frame is received, NIC will slow down the Ethernet frame transmission according to the pause frame.

MAC Control Frame Forwarding consists in:

  • Control frames (PAUSE Frames) are taken by the NIC and do not pass to the host.
  • When Flow Control and MAC Control Frame Forwarding are enabled the PAUSE frames will be passed to the host and can be handled by testpmd.

Note: Priority flow control is not included in this test plan.

Note: the high_water, low_water, pause_time, send_xon are configured into the NIC register. It is not necessary to validate the accuracy of these parameters. And what change it can cause. The port_id is used to indicate the NIC to be configured. In certain case, a system can contain multiple NIC. However the NIC need not be configured multiple times.

29.1. Prerequisites¶

Assuming that ports 0 and 2 are connected to a traffic generator, launch the testpmd with the following arguments:

./build/app/testpmd -cffffff -n 3 -- -i --burst=1 --txpt=32 \
--txht=8 --txwt=0 --txfreet=0 --rxfreet=64 --mbcache=250 --portmask=0x5

The -n command is used to select the number of memory channels. It should match the number of memory channels on that setup.

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 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

29.2. Test Case: test_perf_flowctrl_on_pause_fwd_on¶

testpmd> set flowctrl rx on tx on high_water low_water pause_time
send_xon mac_ctrl_frame_fwd on autoneg on port_id

Setup the csum forwarding mode:

testpmd> set fwd csum
Set csum packet forwarding mode

Start the packet forwarding:

testpmd> start
  csum packet forwarding - CRC stripping disabled - packets/burst=32
  nb forwarding cores=1 - nb forwarding ports=10
  RX queues=1 - RX desc=128 - RX free threshold=64
  RX threshold registers: pthresh=8 hthresh=8 wthresh=4
  TX queues=1 - TX desc=512 - TX free threshold=0
  TX threshold registers: pthresh=32 hthresh=8 wthresh=8

Validate the NIC can generate the pause frame? Configure the traffic generator to send IPv4/UDP packet at the length of 66Byte at the line speed (10G). Because the 66Byte packet cannot reach line rate when running with testpmd, so it is expected that the pause frame will be sent to the peer (traffic generator). Ideally this mechanism can avoid the packet loss. And this depends on high_water/low_water and other parameters are configured properly. It is strongly recommended that the user look into the data sheet before doing any flow control configuration. By default, the flow control on 10G is disabled. the flow control for 1G is enabled.

Validate the NIC can deal with the pause frame. Configure the traffic generator to send out large amount of pause frames, this will cause the NIC to disable / slow down the packet transmission according to the pause time. Once the traffic generator stop sending the pause frame, the NIC will restore the packet transmission to the expected rate.

29.3. Test Case: test_perf_flowctrl_on_pause_fwd_off¶

testpmd> set flowctrl rx on tx on high_water low_water pause_time
send_xon mac_ctrl_frame_fwd off autoneg on port_id

Validate same behavior as test_perf_flowctrl_on_pause_fwd_on

29.4. Test Case: test_perf_flowctrl_rx_on¶

testpmd> set flowctrl rx on tx on high_water low_water pause_time
send_xon mac_ctrl_frame_fwd off autoneg on port_id

Validate same behavior as test_perf_flowctrl_on_pause_fwd_on

29.5. Test Case: test_perf_flowctrl_off_pause_fwd_off¶

This is the default mode for 10G PMD, by default, testpmd is running on this mode. no need to execute any command:

testpmd> set flowctrl rx off tx off high_water low_water pause_time
send_xon mac_ctrl_frame_fwd off autoneg on port_id

Validate the NIC won’t generate the pause frame when the packet buffer is not enough. Packet loss can be observed. Validate the NIC will not slow down the packet transmission after receiving the pause frame.

29.6. Test Case: test_perf_flowctrl_off_pause_fwd_on¶

testpmd> set flowctrl rx off tx off high_water low_water pause_time
send_xon mac_ctrl_frame_fwd on autoneg on port_id

Validate same behavior as test_perf_flowctrl_off_pause_fwd_off

29.7. Test Case: test_perf_flowctrl_tx_on¶

testpmd> set flowctrl rx off tx on high_water low_water pause_time
send_xon mac_ctrl_frame_fwd off autoneg on port_id

Validate same behavior as test_perf_flowctrl_on_pause_fwd_off

Next Previous

© Copyright 2017, dpdk.org. Revision a7c8df3d.

Built with Sphinx using a theme provided by Read the Docs.