diff options
Diffstat (limited to 'docs')
| -rw-r--r-- | docs/testing/user/userguide/testlist.rst | 281 |
1 files changed, 281 insertions, 0 deletions
diff --git a/docs/testing/user/userguide/testlist.rst b/docs/testing/user/userguide/testlist.rst index 4b535494..2b0e9d7f 100644 --- a/docs/testing/user/userguide/testlist.rst +++ b/docs/testing/user/userguide/testlist.rst @@ -107,3 +107,284 @@ List of integration testcases above can be obtained by execution of: .. code-block:: bash $ ./vsperf --integration --list + +OVS/DPDK Regression TestCases +^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ + +These regression tests verify several DPDK features used internally by Open vSwitch. Tests +can be used for verification of performance and correct functionality of upcoming DPDK +and OVS releases and release candidates. + +These tests are part of integration testcases and they must be executed with +``--integration`` CLI parameter. + +Example of execution of all OVS/DPDK regression tests: + +.. code-block:: bash + + $ ./vsperf --integration --tests ovsdpdk_ + +Testcases are defined in the file ``conf/integration/01b_dpdk_regression_tests.conf``. This file +contains a set of configuration options with prefix ``OVSDPDK_``. These parameters can be used +for customization of regression tests and they will override some of standard VSPERF configuration +options. It is recommended to check OVSDPDK configuration parameters and modify them in accordance +with VSPERF configuration. + +At least following parameters should be examined. Their values shall ensure, that DPDK and +QEMU threads are pinned to cpu cores of the same NUMA slot, where tested NICs are connected. + +.. code-block:: python + + _OVSDPDK_1st_PMD_CORE + _OVSDPDK_2nd_PMD_CORE + _OVSDPDK_GUEST_5_CORES + +DPDK NIC Support +++++++++++++++++ + +A set of performance tests to verify support of DPDK accelerated network interface cards. +Testcases use standard physical to physical network scenario with several vSwitch and +traffic configurations, which includes one and two PMD threads, uni and bidirectional traffic +and RFC2544 Continuous or RFC2544 Throughput with 0% packet loss traffic types. + +======================================== ====================================================================================== +Testcase Name Description +======================================== ====================================================================================== +ovsdpdk_nic_p2p_single_pmd_unidir_cont P2P with single PMD in OVS and unidirectional traffic. +ovsdpdk_nic_p2p_single_pmd_bidir_cont P2P with single PMD in OVS and bidirectional traffic. +ovsdpdk_nic_p2p_two_pmd_bidir_cont P2P with two PMDs in OVS and bidirectional traffic. +ovsdpdk_nic_p2p_single_pmd_unidir_tput P2P with single PMD in OVS and unidirectional traffic. +ovsdpdk_nic_p2p_single_pmd_bidir_tput P2P with single PMD in OVS and bidirectional traffic. +ovsdpdk_nic_p2p_two_pmd_bidir_tput P2P with two PMDs in OVS and bidirectional traffic. +======================================== ====================================================================================== + +DPDK Hotplug Support +++++++++++++++++++++ + +A set of functional tests to verify DPDK hotplug support. Tests verify, that it is possible +to use port, which was not bound to DPDK driver during vSwitch startup. There is also +a test which verifies a possibility to detach port from DPDK driver. However support +for manual detachment of a port from DPDK has been removed from recent OVS versions and +thus this testcase is expected to fail. + +======================================== ====================================================================================== +Testcase Name Description +======================================== ====================================================================================== +ovsdpdk_hotplug_attach Ensure successful port-add after binding a device to igb_uio after + ovs-vswitchd is launched. +ovsdpdk_hotplug_detach Same as ovsdpdk_hotplug_attach, but delete and detach the device + after the hotplug. Note Support of netdev-dpdk/detach has been + removed from OVS, so testcase will fail with recent OVS/DPDK + versions. +======================================== ====================================================================================== + +RX Checksum Support ++++++++++++++++++++ + +A set of functional tests for verification of RX checksum calculation for tunneled traffic. +Open vSwitch enables RX checksum offloading by default if NIC supports it. It is to note, +that it is not possible to disable or enable RX checksum offloading. In order to verify +correct RX checksum calculation in software, user has to execute these testcases +at NIC without HW offloading capabilities. + +Testcases utilize existing overlay physical to physical (op2p) network deployment +implemented in vsperf. This deployment expects, that traffic generator sends unidirectional +tunneled traffic (e.g. vxlan) and Open vSwitch performs data decapsulation and sends them +back to the traffic generator via second port. + +======================================== ====================================================================================== +Testcase Name Description +======================================== ====================================================================================== +ovsdpdk_checksum_l3 Test verifies RX IP header checksum (offloading) validation for + tunneling protocols. +ovsdpdk_checksum_l4 Test verifies RX UDP header checksum (offloading) validation for + tunneling protocols. +======================================== ====================================================================================== + +Flow Control Support +++++++++++++++++++++ + +A set of functional testcases for the validation of flow control support in Open vSwitch +with DPDK support. If flow control is enabled in both OVS and Traffic Generator, +the network endpoint (OVS or TGEN) is not able to process incoming data and +thus it detects a RX buffer overflow. It then sends an ethernet pause frame (as defined at 802.3x) +to the TX side. This mechanism will ensure, that the TX side will slow down traffic transmission +and thus no data is lost at RX side. + +Introduced testcases use physical to physical scenario to forward data between +traffic generator ports. It is expected that the processing of small frames in OVS is slower +than line rate. It means that with flow control disabled, traffic generator will +report a frame loss. On the other hand with flow control enabled, there should be 0% +frame loss reported by traffic generator. + +======================================== ====================================================================================== +Testcase Name Description +======================================== ====================================================================================== +ovsdpdk_flow_ctrl_rx Test the rx flow control functionality of DPDK PHY ports. +ovsdpdk_flow_ctrl_rx_dynamic Change the rx flow control support at run time and ensure the system + honored the changes. +======================================== ====================================================================================== + +Multiqueue Support +++++++++++++++++++ + +A set of functional testcases for validation of multiqueue support for both physical +and vHost User DPDK ports. Testcases utilize P2P and PVP network deployments and +native support of multiqueue configuration available in VSPERF. + +======================================== ====================================================================================== +Testcase Name Description +======================================== ====================================================================================== +ovsdpdk_mq_p2p_rxqs Setup rxqs on NIC port. +ovsdpdk_mq_p2p_rxqs_same_core_affinity Affinitize rxqs to the same core. +ovsdpdk_mq_p2p_rxqs_multi_core_affinity Affinitize rxqs to separate cores. +ovsdpdk_mq_pvp_rxqs Setup rxqs on vhost user port. +ovsdpdk_mq_pvp_rxqs_linux_bridge Confirm traffic received over vhost RXQs with Linux virtio device in + guest. +ovsdpdk_mq_pvp_rxqs_testpmd Confirm traffic received over vhost RXQs with DPDK device in guest. +======================================== ====================================================================================== + +Vhost User +++++++++++ + +A set of functional testcases for validation of vHost User Client and vHost User +Server modes in OVS. + +**NOTE:** Vhost User Server mode is deprecated and it will be removed from OVS +in the future. + +======================================== ====================================================================================== +Testcase Name Description +======================================== ====================================================================================== +ovsdpdk_vhostuser_client Test vhost-user client mode +ovsdpdk_vhostuser_client_reconnect Test vhost-user client mode reconnect feature +ovsdpdk_vhostuser_server Test vhost-user server mode +ovsdpdk_vhostuser_sock_dir Verify functionality of vhost-sock-dir flag +======================================== ====================================================================================== + +Virtual Devices Support ++++++++++++++++++++++++ + +A set of functional testcases for verification of correct functionality of virtual +device PMD drivers. + +======================================== ====================================================================================== +Testcase Name Description +======================================== ====================================================================================== +ovsdpdk_vdev_add_null_pmd Test addition of port using the null DPDK PMD driver. +ovsdpdk_vdev_del_null_pmd Test deletion of port using the null DPDK PMD driver. +ovsdpdk_vdev_add_af_packet_pmd Test addition of port using the af_packet DPDK PMD driver. +ovsdpdk_vdev_del_af_packet_pmd Test deletion of port using the af_packet DPDK PMD driver. +======================================== ====================================================================================== + +NUMA Support +++++++++++++ + +A functional testcase for validation of NUMA awareness feature in OVS. + +======================================== ====================================================================================== +Testcase Name Description +======================================== ====================================================================================== +ovsdpdk_numa Test vhost-user NUMA support. Vhostuser PMD threads should migrate to + the same numa slot, where QEMU is executed. +======================================== ====================================================================================== + +Jumbo Frame Support ++++++++++++++++++++ + +A set of functional testcases for verification of jumbo frame support in OVS. +Testcases utilize P2P and PVP network deployments and native support of jumbo +frames available in VSPERF. + +============================================ ================================================================================== +Testcase Name Description +============================================ ================================================================================== +ovsdpdk_jumbo_increase_mtu_phy_port_ovsdb Ensure that the increased MTU for a DPDK physical port is updated in + OVSDB. +ovsdpdk_jumbo_increase_mtu_vport_ovsdb Ensure that the increased MTU for a DPDK vhost-user port is updated in + OVSDB. +ovsdpdk_jumbo_reduce_mtu_phy_port_ovsdb Ensure that the reduced MTU for a DPDK physical port is updated in + OVSDB. +ovsdpdk_jumbo_reduce_mtu_vport_ovsdb Ensure that the reduced MTU for a DPDK vhost-user port is updated in + OVSDB. +ovsdpdk_jumbo_increase_mtu_phy_port_datapath Ensure that the MTU for a DPDK physical port is updated in the + datapath itself when increased to a valid value. +ovsdpdk_jumbo_increase_mtu_vport_datapath Ensure that the MTU for a DPDK vhost-user port is updated in the + datapath itself when increased to a valid value. +ovsdpdk_jumbo_reduce_mtu_phy_port_datapath + Ensure that the MTU for a DPDK physical port is updated in the + datapath itself when decreased to a valid value. +ovsdpdk_jumbo_reduce_mtu_vport_datapath Ensure that the MTU for a DPDK vhost-user port is updated in the + datapath itself when decreased to a valid value. +ovsdpdk_jumbo_mtu_upper_bound_phy_port Verify that the upper bound limit is enforced for OvS DPDK Phy ports. +ovsdpdk_jumbo_mtu_upper_bound_vport Verify that the upper bound limit is enforced for OvS DPDK vhost-user + ports. +ovsdpdk_jumbo_mtu_lower_bound_phy_port Verify that the lower bound limit is enforced for OvS DPDK Phy ports. +ovsdpdk_jumbo_mtu_lower_bound_vport Verify that the lower bound limit is enforced for OvS DPDK vhost-user + ports. +ovsdpdk_jumbo_p2p Ensure that jumbo frames are received, processed and forwarded + correctly by DPDK physical ports. +ovsdpdk_jumbo_pvp Ensure that jumbo frames are received, processed and forwarded + correctly by DPDK vhost-user ports. +ovsdpdk_jumbo_p2p_upper_bound Ensure that jumbo frames above the configured Rx port's MTU are not + accepted +============================================ ================================================================================== + +Rate Limiting ++++++++++++++ + +A set of functional testcases for validation of rate limiting support. This feature +allows to configure an ingress policing for both physical and vHost User DPDK +ports. + +**NOTE:** Desired maximum rate is specified in kilo bits per second and it defines +the rate of payload only. + +======================================== ====================================================================================== +Testcase Name Description +======================================== ====================================================================================== +ovsdpdk_rate_create_phy_port Ensure a rate limiting interface can be created on a physical DPDK + port. +ovsdpdk_rate_delete_phy_port Ensure a rate limiting interface can be destroyed on a physical DPDK + port. +ovsdpdk_rate_create_vport Ensure a rate limiting interface can be created on a vhost-user port. +ovsdpdk_rate_delete_vport Ensure a rate limiting interface can be destroyed on a vhost-user + port. +ovsdpdk_rate_no_policing Ensure when a user attempts to create a rate limiting interface but + is missing policing rate argument, no rate limitiner is created. +ovsdpdk_rate_no_burst Ensure when a user attempts to create a rate limiting interface but + is missing policing burst argument, rate limitiner is created. +ovsdpdk_rate_p2p Ensure when a user creates a rate limiting physical interface that + the traffic is limited to the specified policer rate in a p2p setup. +ovsdpdk_rate_pvp Ensure when a user creates a rate limiting vHost User interface that + the traffic is limited to the specified policer rate in a pvp setup. +ovsdpdk_rate_p2p_multi_pkt_sizes Ensure that rate limiting works for various frame sizes. +======================================== ====================================================================================== + +Quality of Service +++++++++++++++++++ + +A set of functional testcases for validation of QoS support. This feature +allows to configure an egress policing for both physical and vHost User DPDK +ports. + +**NOTE:** Desired maximum rate is specified in bytes per second and it defines +the rate of payload only. + +======================================== ====================================================================================== +Testcase Name Description +======================================== ====================================================================================== +ovsdpdk_qos_create_phy_port Ensure a QoS policy can be created on a physical DPDK port +ovsdpdk_qos_delete_phy_port Ensure an existing QoS policy can be destroyed on a physical DPDK + port. +ovsdpdk_qos_create_vport Ensure a QoS policy can be created on a virtual vhost user port. +ovsdpdk_qos_delete_vport Ensure an existing QoS policy can be destroyed on a vhost user port. +ovsdpdk_qos_create_no_cir Ensure that a QoS policy cannot be created if the egress policer cir + argument is missing. +ovsdpdk_qos_create_no_cbs Ensure that a QoS policy cannot be created if the egress policer cbs + argument is missing. +ovsdpdk_qos_p2p In a p2p setup, ensure when a QoS egress policer is created that the + traffic is limited to the specified rate. +ovsdpdk_qos_pvp In a pvp setup, ensure when a QoS egress policer is created that the + traffic is limited to the specified rate. +======================================== ====================================================================================== |