1.. SPDX-License-Identifier: BSD-3-Clause 2 Copyright(c) 2016 QLogic Corporation 3 Copyright(c) 2017 Cavium, Inc 4 5QEDE Poll Mode Driver 6====================== 7 8The QEDE poll mode driver library (**librte_net_qede**) implements support 9for **QLogic FastLinQ QL4xxxx 10G/25G/40G/50G/100G Intelligent Ethernet Adapters (IEA) and Converged Network Adapters (CNA)** family of adapters as well as SR-IOV virtual functions (VF). It is supported on 10several standard Linux distros like RHEL, SLES, Ubuntu etc. 11It is compile-tested under FreeBSD OS. 12 13More information can be found at `QLogic Corporation's Website 14<http://www.qlogic.com>`_. 15 16Supported Features 17------------------ 18 19- Unicast/Multicast filtering 20- Promiscuous mode 21- Allmulti mode 22- Port hardware statistics 23- Jumbo frames 24- Multiple MAC address 25- MTU change 26- Default pause flow control 27- Multiprocess aware 28- Scatter-Gather 29- Multiple Rx/Tx queues 30- RSS (with RETA/hash table/key) 31- TSS 32- Stateless checksum offloads (IPv4/IPv6/TCP/UDP) 33- LRO/TSO 34- VLAN offload - Filtering and stripping 35- N-tuple filter and flow director (limited support) 36- NPAR (NIC Partitioning) 37- SR-IOV PF and VF 38- GRE Tunneling offload 39- GENEVE Tunneling offload 40- VXLAN Tunneling offload 41- MPLSoUDP Tx Tunneling offload 42- Generic flow API 43 44Co-existence considerations 45--------------------------- 46 47- QLogic FastLinQ QL4xxxx CNAs support Ethernet, RDMA, iSCSI and FCoE 48 functionalities. These functionalities are supported using 49 QLogic Linux kernel drivers qed, qede, qedr, qedi and qedf. DPDK is 50 supported on these adapters using qede PMD. 51 52- When SR-IOV is not enabled on the adapter, 53 QLogic Linux kernel drivers (qed, qede, qedr, qedi and qedf) and qede 54 PMD can’t be attached to different PFs on a given QLogic FastLinQ 55 QL4xxx adapter. 56 A given adapter needs to be completely used by DPDK or Linux drivers 57 Before binding DPDK driver to one or more PFs on the adapter, 58 please make sure to unbind Linux drivers from all PFs of the adapter. 59 If there are multiple adapters on the system, one or more adapters 60 can be used by DPDK driver completely and other adapters can be used 61 by Linux drivers completely. 62 63- When SR-IOV is enabled on the adapter, 64 Linux kernel drivers (qed, qede, qedr, qedi and qedf) can be bound 65 to the PFs of a given adapter and either qede PMD or Linux drivers 66 (qed and qede) can be bound to the VFs of the adapter. 67 68- For sharing an adapter between DPDK and Linux drivers, SRIOV needs 69 to be enabled. Bind all the PFs to Linux Drivers(qed/qede). Create 70 a VF on PFs where DPDK is desired and bind these VFs to qede_pmd. 71 Binding of PFs simultaneously to DPDK and Linux drivers on a given 72 adapter is not supported. 73 74Supported QLogic Adapters 75------------------------- 76 77- QLogic FastLinQ QL4xxxx 10G/25G/40G/50G/100G Intelligent Ethernet Adapters (IEA) and Converged Network Adapters (CNA) 78 79Prerequisites 80------------- 81 82- Requires storm firmware version **8.40.33.0**. Firmware may be available 83 inbox in certain newer Linux distros under the standard directory 84 ``E.g. /lib/firmware/qed/qed_init_values-8.40.33.0.bin``. 85 If the required firmware files are not available then download it from 86 `linux-firmware git repository <http://git.kernel.org/pub/scm/linux/kernel/git/firmware/linux-firmware.git/tree/qed>`_. 87 88- Requires the NIC be updated minimally with **8.30.x.x** Management firmware(MFW) version supported for that NIC. 89 It is highly recommended that the NIC be updated with the latest available management firmware version to get latest feature set. 90 Management Firmware and Firmware Upgrade Utility for Cavium FastLinQ(r) branded adapters can be downloaded from 91 `Driver Download Center <http://driverdownloads.qlogic.com/QLogicDriverDownloads_UI/DefaultNewSearch.aspx>`_. 92 For downloading Firmware Upgrade Utility, select NIC category, model and Linux distro. 93 To update the management firmware, refer to the instructions in the Firmware Upgrade Utility Readme document. 94 For OEM branded adapters please follow the instruction provided by the OEM to update the Management Firmware on the NIC. 95 96- SR-IOV requires Linux PF driver version **8.20.x.x** or higher. 97 If the required PF driver is not available then download it from 98 `QLogic Driver Download Center <http://driverdownloads.qlogic.com/QLogicDriverDownloads_UI/DefaultNewSearch.aspx>`_. 99 For downloading PF driver, select adapter category, model and Linux distro. 100 101Performance note 102~~~~~~~~~~~~~~~~ 103 104- For better performance, it is recommended to use 4K or higher RX/TX rings. 105 106Config Options 107~~~~~~~~~~~~~~ 108 109The following option can be modified in the ``config/rte_config.h`` file. 110 111- ``RTE_LIBRTE_QEDE_FW`` (default **""**) 112 113 Gives absolute path of firmware file. 114 ``Eg: "/lib/firmware/qed/qed_init_values-8.40.33.0.bin"`` 115 Empty string indicates driver will pick up the firmware file 116 from the default location /lib/firmware/qed. 117 CAUTION this option is more for custom firmware, it is not 118 recommended for use under normal condition. 119 120The following options can be enabled with Meson flags. 121 122- ``RTE_LIBRTE_QEDE_DEBUG_TX`` (default **disabled**) 123 124 Toggle display of transmit fast path run-time messages. 125 126- ``RTE_LIBRTE_QEDE_DEBUG_RX`` (default **disabled**) 127 128 Toggle display of receive fast path run-time messages. 129 130Config notes 131~~~~~~~~~~~~ 132 133When there are multiple adapters and/or large number of Rx/Tx queues 134configured on the adapters, the default (2560) number of memzone 135descriptors may not be enough. Please increase the number of memzone 136descriptors to a higher number as needed. When sufficient number of 137memzone descriptors are not configured, user can potentially run into 138following error. 139 140 .. code-block:: console 141 142 EAL: memzone_reserve_aligned_thread_unsafe(): No more room in config 143 144Driver compilation and testing 145------------------------------ 146 147Refer to the document :ref:`compiling and testing a PMD for a NIC <pmd_build_and_test>` 148for details. 149 150RTE Flow Support 151---------------- 152 153QLogic FastLinQ QL4xxxx NICs has support for the following patterns and 154actions. 155 156Patterns: 157 158.. _table_qede_supported_flow_item_types: 159 160.. table:: Item types 161 162 +----+--------------------------------+ 163 | # | Pattern Type | 164 +====+================================+ 165 | 1 | RTE_FLOW_ITEM_TYPE_IPV4 | 166 +----+--------------------------------+ 167 | 2 | RTE_FLOW_ITEM_TYPE_IPV6 | 168 +----+--------------------------------+ 169 | 3 | RTE_FLOW_ITEM_TYPE_UDP | 170 +----+--------------------------------+ 171 | 4 | RTE_FLOW_ITEM_TYPE_TCP | 172 +----+--------------------------------+ 173 174Actions: 175 176.. _table_qede_supported_ingress_action_types: 177 178.. table:: Ingress action types 179 180 +----+--------------------------------+ 181 | # | Action Type | 182 +====+================================+ 183 | 1 | RTE_FLOW_ACTION_TYPE_QUEUE | 184 +----+--------------------------------+ 185 | 2 | RTE_FLOW_ACTION_TYPE_DROP | 186 +----+--------------------------------+ 187 188SR-IOV: Prerequisites and Sample Application Notes 189-------------------------------------------------- 190 191This section provides instructions to configure SR-IOV with Linux OS. 192 193**Note**: librte_net_qede will be used to bind to SR-IOV VF device and Linux native kernel driver (qede) will function as SR-IOV PF driver. Requires PF driver to be 8.20.x.x or higher. 194 195#. Verify SR-IOV and ARI capability is enabled on the adapter using ``lspci``: 196 197 .. code-block:: console 198 199 lspci -s <slot> -vvv 200 201 Example output: 202 203 .. code-block:: console 204 205 [...] 206 Capabilities: [1b8 v1] Alternative Routing-ID Interpretation (ARI) 207 [...] 208 Capabilities: [1c0 v1] Single Root I/O Virtualization (SR-IOV) 209 [...] 210 Kernel driver in use: igb_uio 211 212#. Load the kernel module: 213 214 .. code-block:: console 215 216 modprobe qede 217 218 Example output: 219 220 .. code-block:: console 221 222 systemd-udevd[4848]: renamed network interface eth0 to ens5f0 223 systemd-udevd[4848]: renamed network interface eth1 to ens5f1 224 225#. Bring up the PF ports: 226 227 .. code-block:: console 228 229 ifconfig ens5f0 up 230 ifconfig ens5f1 up 231 232#. Create VF device(s): 233 234 Echo the number of VFs to be created into ``"sriov_numvfs"`` sysfs entry 235 of the parent PF. 236 237 Example output: 238 239 .. code-block:: console 240 241 echo 2 > /sys/devices/pci0000:00/0000:00:03.0/0000:81:00.0/sriov_numvfs 242 243 244#. Assign VF MAC address: 245 246 Assign MAC address to the VF using iproute2 utility. The syntax is:: 247 248 ip link set <PF iface> vf <VF id> mac <macaddr> 249 250 Example output: 251 252 .. code-block:: console 253 254 ip link set ens5f0 vf 0 mac 52:54:00:2f:9d:e8 255 256 257#. PCI Passthrough: 258 259 The VF devices may be passed through to the guest VM using ``virt-manager`` or 260 ``virsh``. QEDE PMD should be used to bind the VF devices in the guest VM 261 using the instructions from Driver compilation and testing section above. 262 263 264#. Running testpmd 265 (Supply ``--log-level="pmd.net.qede.driver:info`` to view informational messages): 266 267 Refer to the document 268 :ref:`compiling and testing a PMD for a NIC <pmd_build_and_test>` to run 269 ``testpmd`` application. 270 271 Example output: 272 273 .. code-block:: console 274 275 dpdk-testpmd -l 0,4-11 -n 4 -- -i --nb-cores=8 --portmask=0xf --rxd=4096 \ 276 --txd=4096 --txfreet=4068 --enable-rx-cksum --rxq=4 --txq=4 \ 277 --rss-ip --rss-udp 278 279 [...] 280 281 EAL: PCI device 0000:84:00.0 on NUMA socket 1 282 EAL: probe driver: 1077:1634 rte_qede_pmd 283 EAL: Not managed by a supported kernel driver, skipped 284 EAL: PCI device 0000:84:00.1 on NUMA socket 1 285 EAL: probe driver: 1077:1634 rte_qede_pmd 286 EAL: Not managed by a supported kernel driver, skipped 287 EAL: PCI device 0000:88:00.0 on NUMA socket 1 288 EAL: probe driver: 1077:1656 rte_qede_pmd 289 EAL: PCI memory mapped at 0x7f738b200000 290 EAL: PCI memory mapped at 0x7f738b280000 291 EAL: PCI memory mapped at 0x7f738b300000 292 PMD: Chip details : BB1 293 PMD: Driver version : QEDE PMD 8.7.9.0_1.0.0 294 PMD: Firmware version : 8.7.7.0 295 PMD: Management firmware version : 8.7.8.0 296 PMD: Firmware file : /lib/firmware/qed/qed_init_values_zipped-8.7.7.0.bin 297 [QEDE PMD: (84:00.0:dpdk-port-0)]qede_common_dev_init:macaddr \ 298 00:0e:1e:d2:09:9c 299 [...] 300 [QEDE PMD: (84:00.0:dpdk-port-0)]qede_tx_queue_setup:txq 0 num_desc 4096 \ 301 tx_free_thresh 4068 socket 0 302 [QEDE PMD: (84:00.0:dpdk-port-0)]qede_tx_queue_setup:txq 1 num_desc 4096 \ 303 tx_free_thresh 4068 socket 0 304 [QEDE PMD: (84:00.0:dpdk-port-0)]qede_tx_queue_setup:txq 2 num_desc 4096 \ 305 tx_free_thresh 4068 socket 0 306 [QEDE PMD: (84:00.0:dpdk-port-0)]qede_tx_queue_setup:txq 3 num_desc 4096 \ 307 tx_free_thresh 4068 socket 0 308 [QEDE PMD: (84:00.0:dpdk-port-0)]qede_rx_queue_setup:rxq 0 num_desc 4096 \ 309 rx_buf_size=2148 socket 0 310 [QEDE PMD: (84:00.0:dpdk-port-0)]qede_rx_queue_setup:rxq 1 num_desc 4096 \ 311 rx_buf_size=2148 socket 0 312 [QEDE PMD: (84:00.0:dpdk-port-0)]qede_rx_queue_setup:rxq 2 num_desc 4096 \ 313 rx_buf_size=2148 socket 0 314 [QEDE PMD: (84:00.0:dpdk-port-0)]qede_rx_queue_setup:rxq 3 num_desc 4096 \ 315 rx_buf_size=2148 socket 0 316 [QEDE PMD: (84:00.0:dpdk-port-0)]qede_dev_start:port 0 317 [QEDE PMD: (84:00.0:dpdk-port-0)]qede_dev_start:link status: down 318 [...] 319 Checking link statuses... 320 Port 0 Link Up - speed 25000 Mbps - full-duplex 321 Port 1 Link Up - speed 25000 Mbps - full-duplex 322 Port 2 Link Up - speed 25000 Mbps - full-duplex 323 Port 3 Link Up - speed 25000 Mbps - full-duplex 324 Done 325 testpmd> 326