1 .. SPDX-License-Identifier: BSD-3-Clause
2 Copyright(c) 2019-2020 Xilinx, Inc.
3 Copyright(c) 2016-2019 Solarflare Communications Inc.
5 This software was jointly developed between OKTET Labs (under contract
6 for Solarflare) and Solarflare Communications, Inc.
8 Solarflare libefx-based Poll Mode Driver
9 ========================================
11 The SFC EFX PMD (**librte_net_sfc_efx**) provides poll mode driver support
12 for **Solarflare SFN7xxx and SFN8xxx** family of 10/40 Gbps adapters,
13 **Solarflare XtremeScale X2xxx** family of 10/25/40/50/100 Gbps adapters and
14 **Alveo SN1000 SmartNICs** family of 10/25/40/50/100 Gbps adapters.
15 SFC EFX PMD has support for the latest Linux and FreeBSD operating systems.
17 More information can be found at `Solarflare Communications website
18 <http://solarflare.com>`_.
24 SFC EFX PMD has support for:
26 - Multiple transmit and receive queues
28 - Link state information including link status change interrupt
30 - IPv4/IPv6 TCP/UDP transmit checksum offload
32 - Inner IPv4/IPv6 TCP/UDP transmit checksum offload
34 - Port hardware statistics
36 - Extended statistics (see Solarflare Server Adapter User's Guide for
37 the statistics description)
43 - Jumbo frames up to 9K
49 - TCP segmentation offload (TSO) including VXLAN and GENEVE encapsulated
51 - Multicast MAC filter
53 - IPv4/IPv6 TCP/UDP receive checksum offload
55 - Inner IPv4/IPv6 TCP/UDP receive checksum offload
57 - Received packet type information
59 - Receive side scaling (RSS)
63 - Scattered Rx DMA for packet that are larger that a single Rx descriptor
65 - Receive queue interrupts
67 - Deferred receive and transmit queue start
69 - Transmit VLAN insertion (if running firmware variant supports it)
78 Non-supported Features
79 ----------------------
81 The features not yet supported include:
83 - Priority-based flow control
85 - Configurable RX CRC stripping (always stripped)
87 - Header split on receive
99 Due to requirements on receive buffer alignment and usage of the receive
100 buffer for the auxiliary packet information provided by the NIC up to
101 extra 269 (14 bytes prefix plus up to 255 bytes for end padding) bytes may be
102 required in the receive buffer.
103 It should be taken into account when mbuf pool for receive is created.
106 Equal stride super-buffer mode
107 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
109 When the receive queue uses equal stride super-buffer DMA mode, one HW Rx
110 descriptor carries many Rx buffers which contiguously follow each other
111 with some stride (equal to total size of rte_mbuf as mempool object).
112 Each Rx buffer is an independent rte_mbuf.
113 However dedicated mempool manager must be used when mempool for the Rx
114 queue is created. The manager must support dequeue of the contiguous
115 block of objects and provide mempool info API to get the block size.
117 Another limitation of a equal stride super-buffer mode, imposed by the
118 firmware, is that it allows for a single RSS context.
124 NVGRE, VXLAN and GENEVE tunnels are supported on SFN8xxx and X2xxx family
125 adapters with full-feature firmware variant running.
126 **sfboot** should be used to configure NIC to run full-feature firmware variant.
127 See Solarflare Server Adapter User's Guide for details.
129 SFN8xxx and X2xxx family adapters provide either inner or outer packet classes.
130 If adapter firmware advertises support for tunnels then the PMD
131 configures the hardware to report inner classes, and outer classes are
132 not reported in received packets.
133 However, for VXLAN and GENEVE tunnels the PMD does report UDP as the
134 outer layer 4 packet type.
136 SFN8xxx and X2xxx family adapters report GENEVE packets as VXLAN.
137 If UDP ports are configured for only one tunnel type then it is safe to
138 treat VXLAN packet type indication as the corresponding UDP tunnel type.
144 Supported attributes:
149 Supported pattern items (***non-transfer*** rules):
153 - ETH (exact match of source/destination addresses, individual/group match
154 of destination address, EtherType in the outer frame and exact match of
155 destination addresses, individual/group match of destination address in
158 - VLAN (exact match of VID, double-tagging is supported)
160 - IPV4 (exact match of source/destination addresses,
161 IP transport protocol)
163 - IPV6 (exact match of source/destination addresses,
164 IP transport protocol)
166 - TCP (exact match of source/destination ports)
168 - UDP (exact match of source/destination ports)
170 - VXLAN (exact match of VXLAN network identifier)
172 - GENEVE (exact match of virtual network identifier, only Ethernet (0x6558)
173 protocol type is supported)
175 - NVGRE (exact match of virtual subnet ID)
177 Supported actions (***non-transfer*** rules):
187 - FLAG (supported only with ef10_essb Rx datapath)
189 - MARK (supported only with ef10_essb Rx datapath)
191 Supported pattern items (***transfer*** rules):
193 - PHY_PORT (cannot repeat; conflicts with other traffic source items)
195 - PF (cannot repeat; conflicts with other traffic source items)
197 - VF (cannot repeat; conflicts with other traffic source items)
201 Supported actions (***transfer*** rules):
223 Validating flow rules depends on the firmware variant.
225 The :ref:`flow_isolated_mode` is supported.
227 Ethernet destination individual/group match
228 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
230 Ethernet item supports I/G matching, if only the corresponding bit is set
231 in the mask of destination address. If destination address in the spec is
232 multicast, it matches all multicast (and broadcast) packets, otherwise it
233 matches unicast packets that are not filtered by other flow rules.
235 Exceptions to flow rules
236 ~~~~~~~~~~~~~~~~~~~~~~~~
238 There is a list of exceptional flow rule patterns which will not be
239 accepted by the PMD. A pattern will be rejected if at least one of the
242 - Filtering by IPv4 or IPv6 EtherType without pattern items of internet
245 - The last item is IPV4 or IPV6, and it's empty.
247 - Filtering by TCP or UDP IP transport protocol without pattern items of
248 transport layer and above.
250 - The last item is TCP or UDP, and it's empty.
258 - Alveo SN1022 SmartNIC
260 - Solarflare XtremeScale Adapters:
262 - Solarflare X2522 Dual Port SFP28 10/25GbE Adapter
264 - Solarflare X2541 Single Port QSFP28 10/25G/100G Adapter
266 - Solarflare X2542 Dual Port QSFP28 10/25G/100G Adapter
268 - Solarflare Flareon [Ultra] Server Adapters:
270 - Solarflare SFN8522 Dual Port SFP+ Server Adapter
272 - Solarflare SFN8522M Dual Port SFP+ Server Adapter
274 - Solarflare SFN8042 Dual Port QSFP+ Server Adapter
276 - Solarflare SFN8542 Dual Port QSFP+ Server Adapter
278 - Solarflare SFN8722 Dual Port SFP+ OCP Server Adapter
280 - Solarflare SFN7002F Dual Port SFP+ Server Adapter
282 - Solarflare SFN7004F Quad Port SFP+ Server Adapter
284 - Solarflare SFN7042Q Dual Port QSFP+ Server Adapter
286 - Solarflare SFN7122F Dual Port SFP+ Server Adapter
288 - Solarflare SFN7124F Quad Port SFP+ Server Adapter
290 - Solarflare SFN7142Q Dual Port QSFP+ Server Adapter
292 - Solarflare SFN7322F Precision Time Synchronization Server Adapter
298 - Requires firmware version:
300 - SFN7xxx: **4.7.1.1001** or higher
302 - SFN8xxx: **6.0.2.1004** or higher
304 Visit `Solarflare Support Downloads <https://support.solarflare.com>`_ to get
305 Solarflare Utilities (either Linux or FreeBSD) with the latest firmware.
306 Follow instructions from Solarflare Server Adapter User's Guide to
307 update firmware and configure the adapter.
310 Pre-Installation Configuration
311 ------------------------------
317 The following build-time options may be enabled on build time using
318 ``-Dc_args=`` meson argument (e.g. ``-Dc_args=-DRTE_LIBRTE_SFC_EFX_DEBUG``).
320 Please note that enabling debugging options may affect system performance.
322 - ``RTE_LIBRTE_SFC_EFX_DEBUG`` (undefined by default)
324 Enable compilation of the extra run-time consistency checks.
327 Per-Device Parameters
328 ~~~~~~~~~~~~~~~~~~~~~
330 The following per-device parameters can be passed via EAL PCI device
331 whitelist option like "-w 02:00.0,arg1=value1,...".
333 Case-insensitive 1/y/yes/on or 0/n/no/off may be used to specify
334 boolean parameters value.
336 - ``rx_datapath`` [auto|efx|ef10|ef10_essb] (default **auto**)
338 Choose receive datapath implementation.
339 **auto** allows the driver itself to make a choice based on firmware
340 features available and required by the datapath implementation.
341 **efx** chooses libefx-based datapath which supports Rx scatter.
342 Supported for SFN7xxx, SFN8xxx and X2xxx family adapters only.
343 **ef10** chooses EF10 (SFN7xxx, SFN8xxx, X2xxx) native datapath which is
344 more efficient than libefx-based and provides richer packet type
346 **ef10_essb** chooses SFNX2xxx equal stride super-buffer datapath
347 which may be used on DPDK firmware variant only
348 (see notes about its limitations above).
349 **ef100** chooses EF100 native datapath which is the only supported
350 Rx datapath for EF100 architecture based NICs.
352 - ``tx_datapath`` [auto|efx|ef10|ef10_simple|ef100] (default **auto**)
354 Choose transmit datapath implementation.
355 **auto** allows the driver itself to make a choice based on firmware
356 features available and required by the datapath implementation.
357 **efx** chooses libefx-based datapath which supports VLAN insertion
358 (full-feature firmware variant only), TSO and multi-segment mbufs.
359 Mbuf segments may come from different mempools, and mbuf reference
360 counters are treated responsibly.
361 Supported for SFN7xxx, SFN8xxx and X2xxx family adapters only.
362 **ef10** chooses EF10 (SFN7xxx, SFN8xxx, X2xxx) native datapath which is
363 more efficient than libefx-based but has no VLAN insertion support yet.
364 Mbuf segments may come from different mempools, and mbuf reference
365 counters are treated responsibly.
366 **ef10_simple** chooses EF10 (SFN7xxx, SFN8xxx, X2xxx) native datapath which
367 is even more faster then **ef10** but does not support multi-segment
368 mbufs, disallows multiple mempools and neglects mbuf reference counters.
369 **ef100** chooses EF100 native datapath which supports multi-segment
370 mbufs, VLAN insertion, inner/outer IPv4 and TCP/UDP checksum and TCP
371 segmentation offloads including VXLAN and GENEVE IPv4/IPv6 tunnels.
373 - ``perf_profile`` [auto|throughput|low-latency] (default **throughput**)
375 Choose hardware tuning to be optimized for either throughput or
377 **auto** allows NIC firmware to make a choice based on
378 installed licenses and firmware variant configured using **sfboot**.
380 - ``stats_update_period_ms`` [long] (default **1000**)
382 Adjust period in milliseconds to update port hardware statistics.
383 The accepted range is 0 to 65535. The value of **0** may be used
384 to disable periodic statistics update. One should note that it's
385 only possible to set an arbitrary value on SFN8xxx and X2xxx provided that
386 firmware version is 6.2.1.1033 or higher, otherwise any positive
387 value will select a fixed update period of **1000** milliseconds
389 - ``fw_variant`` [dont-care|full-feature|ultra-low-latency|
390 capture-packed-stream|dpdk] (default **dont-care**)
392 Choose the preferred firmware variant to use.
393 The parameter is supported for SFN7xxX, SFN8xxx and X2xxx families only.
394 In order for the selected option to have an effect, the **sfboot** utility
395 must be configured with the **auto** firmware-variant option.
396 The preferred firmware variant applies to all ports on the NIC.
397 **dont-care** ensures that the driver can attach to an unprivileged function.
398 The datapath firmware type to use is controlled by the **sfboot**
400 **full-feature** chooses full featured firmware.
401 **ultra-low-latency** chooses firmware with fewer features but lower latency.
402 **capture-packed-stream** chooses firmware for SolarCapture packed stream
404 **dpdk** chooses DPDK firmware with equal stride super-buffer Rx mode
405 for higher Rx packet rate and packet marks support and firmware subvariant
406 without checksumming on transmit for higher Tx packet rate if
407 checksumming is not required.
409 - ``rxd_wait_timeout_ns`` [long] (default **200 us**)
411 Adjust timeout in nanoseconds to head-of-line block to wait for
413 The accepted range is 0 to 400 ms.
414 Flow control should be enabled to make it work.
415 The value of **0** disables it and packets are dropped immediately.
416 When a packet is dropped because of no Rx descriptors,
417 ``rx_nodesc_drop_cnt`` counter grows.
418 The feature is supported only by the DPDK firmware variant when equal
419 stride super-buffer Rx mode is used.
422 Dynamic Logging Parameters
423 ~~~~~~~~~~~~~~~~~~~~~~~~~~
425 One may leverage EAL option "--log-level" to change default levels
426 for the log types supported by the driver. The option is used with
427 an argument typically consisting of two parts separated by a colon.
429 Level value is the last part which takes a symbolic name (or integer).
430 Log type is the former part which may shell match syntax.
431 Depending on the choice of the expression, the given log level may
432 be used either for some specific log type or for a subset of types.
434 SFC EFX PMD provides the following log types available for control:
436 - ``pmd.net.sfc.driver`` (default level is **notice**)
438 Affects driver-wide messages unrelated to any particular devices.
440 - ``pmd.net.sfc.main`` (default level is **notice**)
442 Matches a subset of per-port log types registered during runtime.
443 A full name for a particular type may be obtained by appending a
444 dot and a PCI device identifier (``XXXX:XX:XX.X``) to the prefix.
446 - ``pmd.net.sfc.mcdi`` (default level is **notice**)
448 Extra logging of the communication with the NIC's management CPU.
449 The format of the log is consumed by the Solarflare netlogdecode
450 cross-platform tool. May be managed per-port, as explained above.