1.. SPDX-License-Identifier: BSD-3-Clause 2 Copyright (c) 2016-2019 Solarflare Communications Inc. 3 All rights reserved. 4 5 This software was jointly developed between OKTET Labs (under contract 6 for Solarflare) and Solarflare Communications, Inc. 7 8Solarflare libefx-based Poll Mode Driver 9======================================== 10 11The SFC EFX PMD (**librte_pmd_sfc_efx**) provides poll mode driver support 12for **Solarflare SFN7xxx and SFN8xxx** family of 10/40 Gbps adapters and 13**Solarflare XtremeScale X2xxx** family of 10/25/40/50/100 Gbps adapters. 14SFC EFX PMD has support for the latest Linux and FreeBSD operating systems. 15 16More information can be found at `Solarflare Communications website 17<http://solarflare.com>`_. 18 19 20Features 21-------- 22 23SFC EFX PMD has support for: 24 25- Multiple transmit and receive queues 26 27- Link state information including link status change interrupt 28 29- IPv4/IPv6 TCP/UDP transmit checksum offload 30 31- Inner IPv4/IPv6 TCP/UDP transmit checksum offload 32 33- Port hardware statistics 34 35- Extended statistics (see Solarflare Server Adapter User's Guide for 36 the statistics description) 37 38- Basic flow control 39 40- MTU update 41 42- Jumbo frames up to 9K 43 44- Promiscuous mode 45 46- Allmulticast mode 47 48- TCP segmentation offload (TSO) including VXLAN and GENEVE encapsulated 49 50- Multicast MAC filter 51 52- IPv4/IPv6 TCP/UDP receive checksum offload 53 54- Inner IPv4/IPv6 TCP/UDP receive checksum offload 55 56- Received packet type information 57 58- Receive side scaling (RSS) 59 60- RSS hash 61 62- Scattered Rx DMA for packet that are larger that a single Rx descriptor 63 64- Receive queue interrupts 65 66- Deferred receive and transmit queue start 67 68- Transmit VLAN insertion (if running firmware variant supports it) 69 70- Flow API 71 72- Loopback 73 74 75Non-supported Features 76---------------------- 77 78The features not yet supported include: 79 80- Priority-based flow control 81 82- Configurable RX CRC stripping (always stripped) 83 84- Header split on receive 85 86- VLAN filtering 87 88- VLAN stripping 89 90- LRO 91 92 93Limitations 94----------- 95 96Due to requirements on receive buffer alignment and usage of the receive 97buffer for the auxiliary packet information provided by the NIC up to 98extra 269 (14 bytes prefix plus up to 255 bytes for end padding) bytes may be 99required in the receive buffer. 100It should be taken into account when mbuf pool for receive is created. 101 102 103Equal stride super-buffer mode 104~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ 105 106When the receive queue uses equal stride super-buffer DMA mode, one HW Rx 107descriptor carries many Rx buffers which contiguously follow each other 108with some stride (equal to total size of rte_mbuf as mempool object). 109Each Rx buffer is an independent rte_mbuf. 110However dedicated mempool manager must be used when mempool for the Rx 111queue is created. The manager must support dequeue of the contiguous 112block of objects and provide mempool info API to get the block size. 113 114Another limitation of a equal stride super-buffer mode, imposed by the 115firmware, is that it allows for a single RSS context. 116 117 118Tunnels support 119--------------- 120 121NVGRE, VXLAN and GENEVE tunnels are supported on SFN8xxx and X2xxx family 122adapters with full-feature firmware variant running. 123**sfboot** should be used to configure NIC to run full-feature firmware variant. 124See Solarflare Server Adapter User's Guide for details. 125 126SFN8xxx and X2xxx family adapters provide either inner or outer packet classes. 127If adapter firmware advertises support for tunnels then the PMD 128configures the hardware to report inner classes, and outer classes are 129not reported in received packets. 130However, for VXLAN and GENEVE tunnels the PMD does report UDP as the 131outer layer 4 packet type. 132 133SFN8xxx and X2xxx family adapters report GENEVE packets as VXLAN. 134If UDP ports are configured for only one tunnel type then it is safe to 135treat VXLAN packet type indication as the corresponding UDP tunnel type. 136 137 138Flow API support 139---------------- 140 141Supported attributes: 142 143- Ingress 144 145Supported pattern items: 146 147- VOID 148 149- ETH (exact match of source/destination addresses, individual/group match 150 of destination address, EtherType in the outer frame and exact match of 151 destination addresses, individual/group match of destination address in 152 the inner frame) 153 154- VLAN (exact match of VID, double-tagging is supported) 155 156- IPV4 (exact match of source/destination addresses, 157 IP transport protocol) 158 159- IPV6 (exact match of source/destination addresses, 160 IP transport protocol) 161 162- TCP (exact match of source/destination ports) 163 164- UDP (exact match of source/destination ports) 165 166- VXLAN (exact match of VXLAN network identifier) 167 168- GENEVE (exact match of virtual network identifier, only Ethernet (0x6558) 169 protocol type is supported) 170 171- NVGRE (exact match of virtual subnet ID) 172 173Supported actions: 174 175- VOID 176 177- QUEUE 178 179- RSS 180 181- DROP 182 183- FLAG (supported only with ef10_essb Rx datapath) 184 185- MARK (supported only with ef10_essb Rx datapath) 186 187Validating flow rules depends on the firmware variant. 188 189The :ref:`flow_isolated_mode` is supported. 190 191Ethernet destination individual/group match 192~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ 193 194Ethernet item supports I/G matching, if only the corresponding bit is set 195in the mask of destination address. If destination address in the spec is 196multicast, it matches all multicast (and broadcast) packets, otherwise it 197matches unicast packets that are not filtered by other flow rules. 198 199Exceptions to flow rules 200~~~~~~~~~~~~~~~~~~~~~~~~ 201 202There is a list of exceptional flow rule patterns which will not be 203accepted by the PMD. A pattern will be rejected if at least one of the 204conditions is met: 205 206- Filtering by IPv4 or IPv6 EtherType without pattern items of internet 207 layer and above. 208 209- The last item is IPV4 or IPV6, and it's empty. 210 211- Filtering by TCP or UDP IP transport protocol without pattern items of 212 transport layer and above. 213 214- The last item is TCP or UDP, and it's empty. 215 216 217Supported NICs 218-------------- 219 220- Solarflare XtremeScale Adapters: 221 222 - Solarflare X2522 Dual Port SFP28 10/25GbE Adapter 223 224 - Solarflare X2541 Single Port QSFP28 10/25G/100G Adapter 225 226 - Solarflare X2542 Dual Port QSFP28 10/25G/100G Adapter 227 228- Solarflare Flareon [Ultra] Server Adapters: 229 230 - Solarflare SFN8522 Dual Port SFP+ Server Adapter 231 232 - Solarflare SFN8522M Dual Port SFP+ Server Adapter 233 234 - Solarflare SFN8042 Dual Port QSFP+ Server Adapter 235 236 - Solarflare SFN8542 Dual Port QSFP+ Server Adapter 237 238 - Solarflare SFN8722 Dual Port SFP+ OCP Server Adapter 239 240 - Solarflare SFN7002F Dual Port SFP+ Server Adapter 241 242 - Solarflare SFN7004F Quad Port SFP+ Server Adapter 243 244 - Solarflare SFN7042Q Dual Port QSFP+ Server Adapter 245 246 - Solarflare SFN7122F Dual Port SFP+ Server Adapter 247 248 - Solarflare SFN7124F Quad Port SFP+ Server Adapter 249 250 - Solarflare SFN7142Q Dual Port QSFP+ Server Adapter 251 252 - Solarflare SFN7322F Precision Time Synchronization Server Adapter 253 254 255Prerequisites 256------------- 257 258- Requires firmware version: 259 260 - SFN7xxx: **4.7.1.1001** or higher 261 262 - SFN8xxx: **6.0.2.1004** or higher 263 264Visit `Solarflare Support Downloads <https://support.solarflare.com>`_ to get 265Solarflare Utilities (either Linux or FreeBSD) with the latest firmware. 266Follow instructions from Solarflare Server Adapter User's Guide to 267update firmware and configure the adapter. 268 269 270Pre-Installation Configuration 271------------------------------ 272 273 274Config File Options 275~~~~~~~~~~~~~~~~~~~ 276 277The following options can be modified in the ``.config`` file. 278Please note that enabling debugging options may affect system performance. 279 280- ``CONFIG_RTE_LIBRTE_SFC_EFX_PMD`` (default **y**) 281 282 Enable compilation of Solarflare libefx-based poll-mode driver. 283 284- ``CONFIG_RTE_LIBRTE_SFC_EFX_DEBUG`` (default **n**) 285 286 Enable compilation of the extra run-time consistency checks. 287 288 289Per-Device Parameters 290~~~~~~~~~~~~~~~~~~~~~ 291 292The following per-device parameters can be passed via EAL PCI device 293whitelist option like "-w 02:00.0,arg1=value1,...". 294 295Case-insensitive 1/y/yes/on or 0/n/no/off may be used to specify 296boolean parameters value. 297 298- ``rx_datapath`` [auto|efx|ef10|ef10_esps] (default **auto**) 299 300 Choose receive datapath implementation. 301 **auto** allows the driver itself to make a choice based on firmware 302 features available and required by the datapath implementation. 303 **efx** chooses libefx-based datapath which supports Rx scatter. 304 **ef10** chooses EF10 (SFN7xxx, SFN8xxx, X2xxx) native datapath which is 305 more efficient than libefx-based and provides richer packet type 306 classification. 307 **ef10_esps** chooses SFNX2xxx equal stride packed stream datapath 308 which may be used on DPDK firmware variant only 309 (see notes about its limitations above). 310 311- ``tx_datapath`` [auto|efx|ef10|ef10_simple] (default **auto**) 312 313 Choose transmit datapath implementation. 314 **auto** allows the driver itself to make a choice based on firmware 315 features available and required by the datapath implementation. 316 **efx** chooses libefx-based datapath which supports VLAN insertion 317 (full-feature firmware variant only), TSO and multi-segment mbufs. 318 Mbuf segments may come from different mempools, and mbuf reference 319 counters are treated responsibly. 320 **ef10** chooses EF10 (SFN7xxx, SFN8xxx, X2xxx) native datapath which is 321 more efficient than libefx-based but has no VLAN insertion support yet. 322 Mbuf segments may come from different mempools, and mbuf reference 323 counters are treated responsibly. 324 **ef10_simple** chooses EF10 (SFN7xxx, SFN8xxx, X2xxx) native datapath which 325 is even more faster then **ef10** but does not support multi-segment 326 mbufs, disallows multiple mempools and neglects mbuf reference counters. 327 328- ``perf_profile`` [auto|throughput|low-latency] (default **throughput**) 329 330 Choose hardware tuning to be optimized for either throughput or 331 low-latency. 332 **auto** allows NIC firmware to make a choice based on 333 installed licenses and firmware variant configured using **sfboot**. 334 335- ``stats_update_period_ms`` [long] (default **1000**) 336 337 Adjust period in milliseconds to update port hardware statistics. 338 The accepted range is 0 to 65535. The value of **0** may be used 339 to disable periodic statistics update. One should note that it's 340 only possible to set an arbitrary value on SFN8xxx and X2xxx provided that 341 firmware version is 6.2.1.1033 or higher, otherwise any positive 342 value will select a fixed update period of **1000** milliseconds 343 344- ``fw_variant`` [dont-care|full-feature|ultra-low-latency| 345 capture-packed-stream|dpdk] (default **dont-care**) 346 347 Choose the preferred firmware variant to use. In order for the selected 348 option to have an effect, the **sfboot** utility must be configured with the 349 **auto** firmware-variant option. The preferred firmware variant applies to 350 all ports on the NIC. 351 **dont-care** ensures that the driver can attach to an unprivileged function. 352 The datapath firmware type to use is controlled by the **sfboot** 353 utility. 354 **full-feature** chooses full featured firmware. 355 **ultra-low-latency** chooses firmware with fewer features but lower latency. 356 **capture-packed-stream** chooses firmware for SolarCapture packed stream 357 mode. 358 **dpdk** chooses DPDK firmware with equal stride super-buffer Rx mode 359 for higher Rx packet rate and packet marks support and firmware subvariant 360 without checksumming on transmit for higher Tx packet rate if 361 checksumming is not required. 362 363- ``rxd_wait_timeout_ns`` [long] (default **200 us**) 364 365 Adjust timeout in nanoseconds to head-of-line block to wait for 366 Rx descriptors. 367 The accepted range is 0 to 400 ms. 368 Flow control should be enabled to make it work. 369 The value of **0** disables it and packets are dropped immediately. 370 When a packet is dropped because of no Rx descriptors, 371 ``rx_nodesc_drop_cnt`` counter grows. 372 The feature is supported only by the DPDK firmware variant when equal 373 stride super-buffer Rx mode is used. 374 375 376Dynamic Logging Parameters 377~~~~~~~~~~~~~~~~~~~~~~~~~~ 378 379One may leverage EAL option "--log-level" to change default levels 380for the log types supported by the driver. The option is used with 381an argument typically consisting of two parts separated by a colon. 382 383Level value is the last part which takes a symbolic name (or integer). 384Log type is the former part which may shell match syntax. 385Depending on the choice of the expression, the given log level may 386be used either for some specific log type or for a subset of types. 387 388SFC EFX PMD provides the following log types available for control: 389 390- ``pmd.net.sfc.driver`` (default level is **notice**) 391 392 Affects driver-wide messages unrelated to any particular devices. 393 394- ``pmd.net.sfc.main`` (default level is **notice**) 395 396 Matches a subset of per-port log types registered during runtime. 397 A full name for a particular type may be obtained by appending a 398 dot and a PCI device identifier (``XXXX:XX:XX.X``) to the prefix. 399 400- ``pmd.net.sfc.mcdi`` (default level is **notice**) 401 402 Extra logging of the communication with the NIC's management CPU. 403 The format of the log is consumed by the Solarflare netlogdecode 404 cross-platform tool. May be managed per-port, as explained above. 405