1 /* SPDX-License-Identifier: BSD-3-Clause
2 * Copyright(c) 2010-2017 Intel Corporation
11 * RTE Ethernet Device API
13 * The Ethernet Device API is composed of two parts:
15 * - The application-oriented Ethernet API that includes functions to setup
16 * an Ethernet device (configure it, setup its RX and TX queues and start it),
17 * to get its MAC address, the speed and the status of its physical link,
18 * to receive and to transmit packets, and so on.
20 * - The driver-oriented Ethernet API that exports functions allowing
21 * an Ethernet Poll Mode Driver (PMD) to allocate an Ethernet device instance,
22 * create memzone for HW rings and process registered callbacks, and so on.
23 * PMDs should include ethdev_driver.h instead of this header.
25 * By default, all the functions of the Ethernet Device API exported by a PMD
26 * are lock-free functions which assume to not be invoked in parallel on
27 * different logical cores to work on the same target object. For instance,
28 * the receive function of a PMD cannot be invoked in parallel on two logical
29 * cores to poll the same RX queue [of the same port]. Of course, this function
30 * can be invoked in parallel by different logical cores on different RX queues.
31 * It is the responsibility of the upper level application to enforce this rule.
33 * If needed, parallel accesses by multiple logical cores to shared queues
34 * shall be explicitly protected by dedicated inline lock-aware functions
35 * built on top of their corresponding lock-free functions of the PMD API.
37 * In all functions of the Ethernet API, the Ethernet device is
38 * designated by an integer >= 0 named the device port identifier.
40 * At the Ethernet driver level, Ethernet devices are represented by a generic
41 * data structure of type *rte_eth_dev*.
43 * Ethernet devices are dynamically registered during the PCI probing phase
44 * performed at EAL initialization time.
45 * When an Ethernet device is being probed, an *rte_eth_dev* structure and
46 * a new port identifier are allocated for that device. Then, the eth_dev_init()
47 * function supplied by the Ethernet driver matching the probed PCI
48 * device is invoked to properly initialize the device.
50 * The role of the device init function consists of resetting the hardware,
51 * checking access to Non-volatile Memory (NVM), reading the MAC address
54 * If the device init operation is successful, the correspondence between
55 * the port identifier assigned to the new device and its associated
56 * *rte_eth_dev* structure is effectively registered.
57 * Otherwise, both the *rte_eth_dev* structure and the port identifier are
60 * The functions exported by the application Ethernet API to setup a device
61 * designated by its port identifier must be invoked in the following order:
62 * - rte_eth_dev_configure()
63 * - rte_eth_tx_queue_setup()
64 * - rte_eth_rx_queue_setup()
65 * - rte_eth_dev_start()
67 * Then, the network application can invoke, in any order, the functions
68 * exported by the Ethernet API to get the MAC address of a given device, to
69 * get the speed and the status of a device physical link, to receive/transmit
70 * [burst of] packets, and so on.
72 * If the application wants to change the configuration (i.e. call
73 * rte_eth_dev_configure(), rte_eth_tx_queue_setup(), or
74 * rte_eth_rx_queue_setup()), it must call rte_eth_dev_stop() first to stop the
75 * device and then do the reconfiguration before calling rte_eth_dev_start()
76 * again. The transmit and receive functions should not be invoked when the
79 * Please note that some configuration is not stored between calls to
80 * rte_eth_dev_stop()/rte_eth_dev_start(). The following configuration will
84 * - flow control settings
85 * - receive mode configuration (promiscuous mode, all-multicast mode,
86 * hardware checksum mode, RSS/VMDQ settings etc.)
87 * - VLAN filtering configuration
88 * - default MAC address
89 * - MAC addresses supplied to MAC address array
90 * - flow director filtering mode (but not filtering rules)
91 * - NIC queue statistics mappings
93 * Any other configuration will not be stored and will need to be re-entered
94 * before a call to rte_eth_dev_start().
96 * Finally, a network application can close an Ethernet device by invoking the
97 * rte_eth_dev_close() function.
99 * Each function of the application Ethernet API invokes a specific function
100 * of the PMD that controls the target device designated by its port
102 * For this purpose, all device-specific functions of an Ethernet driver are
103 * supplied through a set of pointers contained in a generic structure of type
105 * The address of the *eth_dev_ops* structure is stored in the *rte_eth_dev*
106 * structure by the device init function of the Ethernet driver, which is
107 * invoked during the PCI probing phase, as explained earlier.
109 * In other words, each function of the Ethernet API simply retrieves the
110 * *rte_eth_dev* structure associated with the device port identifier and
111 * performs an indirect invocation of the corresponding driver function
112 * supplied in the *eth_dev_ops* structure of the *rte_eth_dev* structure.
114 * For performance reasons, the address of the burst-oriented RX and TX
115 * functions of the Ethernet driver are not contained in the *eth_dev_ops*
116 * structure. Instead, they are directly stored at the beginning of the
117 * *rte_eth_dev* structure to avoid an extra indirect memory access during
120 * RTE ethernet device drivers do not use interrupts for transmitting or
121 * receiving. Instead, Ethernet drivers export Poll-Mode receive and transmit
122 * functions to applications.
123 * Both receive and transmit functions are packet-burst oriented to minimize
124 * their cost per packet through the following optimizations:
126 * - Sharing among multiple packets the incompressible cost of the
127 * invocation of receive/transmit functions.
129 * - Enabling receive/transmit functions to take advantage of burst-oriented
130 * hardware features (L1 cache, prefetch instructions, NIC head/tail
131 * registers) to minimize the number of CPU cycles per packet, for instance,
132 * by avoiding useless read memory accesses to ring descriptors, or by
133 * systematically using arrays of pointers that exactly fit L1 cache line
134 * boundaries and sizes.
136 * The burst-oriented receive function does not provide any error notification,
137 * to avoid the corresponding overhead. As a hint, the upper-level application
138 * might check the status of the device link once being systematically returned
139 * a 0 value by the receive function of the driver for a given number of tries.
148 /* Use this macro to check if LRO API is supported */
149 #define RTE_ETHDEV_HAS_LRO_SUPPORT
151 /* Alias RTE_LIBRTE_ETHDEV_DEBUG for backward compatibility. */
152 #ifdef RTE_LIBRTE_ETHDEV_DEBUG
153 #define RTE_ETHDEV_DEBUG_RX
154 #define RTE_ETHDEV_DEBUG_TX
157 #include <rte_compat.h>
159 #include <rte_interrupts.h>
161 #include <rte_devargs.h>
162 #include <rte_errno.h>
163 #include <rte_common.h>
164 #include <rte_config.h>
165 #include <rte_ether.h>
166 #include <rte_power_intrinsics.h>
168 #include "rte_ethdev_trace_fp.h"
169 #include "rte_dev_info.h"
171 extern int rte_eth_dev_logtype;
173 #define RTE_ETHDEV_LOG(level, ...) \
174 rte_log(RTE_LOG_ ## level, rte_eth_dev_logtype, "" __VA_ARGS__)
179 * Initializes a device iterator.
181 * This iterator allows accessing a list of devices matching some devargs.
184 * Device iterator handle initialized by the function.
185 * The fields bus_str and cls_str might be dynamically allocated,
186 * and could be freed by calling rte_eth_iterator_cleanup().
189 * Device description string.
192 * 0 on successful initialization, negative otherwise.
194 int rte_eth_iterator_init(struct rte_dev_iterator *iter, const char *devargs);
197 * Iterates on devices with devargs filter.
198 * The ownership is not checked.
200 * The next port id is returned, and the iterator is updated.
203 * Device iterator handle initialized by rte_eth_iterator_init().
204 * Some fields bus_str and cls_str might be freed when no more port is found,
205 * by calling rte_eth_iterator_cleanup().
208 * A port id if found, RTE_MAX_ETHPORTS otherwise.
210 uint16_t rte_eth_iterator_next(struct rte_dev_iterator *iter);
213 * Free some allocated fields of the iterator.
215 * This function is automatically called by rte_eth_iterator_next()
216 * on the last iteration (i.e. when no more matching port is found).
218 * It is safe to call this function twice; it will do nothing more.
221 * Device iterator handle initialized by rte_eth_iterator_init().
222 * The fields bus_str and cls_str are freed if needed.
224 void rte_eth_iterator_cleanup(struct rte_dev_iterator *iter);
227 * Macro to iterate over all ethdev ports matching some devargs.
229 * If a break is done before the end of the loop,
230 * the function rte_eth_iterator_cleanup() must be called.
233 * Iterated port id of type uint16_t.
235 * Device parameters input as string of type char*.
237 * Iterator handle of type struct rte_dev_iterator, used internally.
239 #define RTE_ETH_FOREACH_MATCHING_DEV(id, devargs, iter) \
240 for (rte_eth_iterator_init(iter, devargs), \
241 id = rte_eth_iterator_next(iter); \
242 id != RTE_MAX_ETHPORTS; \
243 id = rte_eth_iterator_next(iter))
246 * A structure used to retrieve statistics for an Ethernet port.
247 * Not all statistics fields in struct rte_eth_stats are supported
248 * by any type of network interface card (NIC). If any statistics
249 * field is not supported, its value is 0.
250 * All byte-related statistics do not include Ethernet FCS regardless
251 * of whether these bytes have been delivered to the application
252 * (see DEV_RX_OFFLOAD_KEEP_CRC).
254 struct rte_eth_stats {
255 uint64_t ipackets; /**< Total number of successfully received packets. */
256 uint64_t opackets; /**< Total number of successfully transmitted packets.*/
257 uint64_t ibytes; /**< Total number of successfully received bytes. */
258 uint64_t obytes; /**< Total number of successfully transmitted bytes. */
260 /**< Total of RX packets dropped by the HW,
261 * because there are no available buffer (i.e. RX queues are full).
263 uint64_t ierrors; /**< Total number of erroneous received packets. */
264 uint64_t oerrors; /**< Total number of failed transmitted packets. */
265 uint64_t rx_nombuf; /**< Total number of RX mbuf allocation failures. */
266 /* Queue stats are limited to max 256 queues */
267 uint64_t q_ipackets[RTE_ETHDEV_QUEUE_STAT_CNTRS];
268 /**< Total number of queue RX packets. */
269 uint64_t q_opackets[RTE_ETHDEV_QUEUE_STAT_CNTRS];
270 /**< Total number of queue TX packets. */
271 uint64_t q_ibytes[RTE_ETHDEV_QUEUE_STAT_CNTRS];
272 /**< Total number of successfully received queue bytes. */
273 uint64_t q_obytes[RTE_ETHDEV_QUEUE_STAT_CNTRS];
274 /**< Total number of successfully transmitted queue bytes. */
275 uint64_t q_errors[RTE_ETHDEV_QUEUE_STAT_CNTRS];
276 /**< Total number of queue packets received that are dropped. */
279 /**@{@name Link speed capabilities
280 * Device supported speeds bitmap flags
282 #define ETH_LINK_SPEED_AUTONEG (0 << 0) /**< Autonegotiate (all speeds) */
283 #define ETH_LINK_SPEED_FIXED (1 << 0) /**< Disable autoneg (fixed speed) */
284 #define ETH_LINK_SPEED_10M_HD (1 << 1) /**< 10 Mbps half-duplex */
285 #define ETH_LINK_SPEED_10M (1 << 2) /**< 10 Mbps full-duplex */
286 #define ETH_LINK_SPEED_100M_HD (1 << 3) /**< 100 Mbps half-duplex */
287 #define ETH_LINK_SPEED_100M (1 << 4) /**< 100 Mbps full-duplex */
288 #define ETH_LINK_SPEED_1G (1 << 5) /**< 1 Gbps */
289 #define ETH_LINK_SPEED_2_5G (1 << 6) /**< 2.5 Gbps */
290 #define ETH_LINK_SPEED_5G (1 << 7) /**< 5 Gbps */
291 #define ETH_LINK_SPEED_10G (1 << 8) /**< 10 Gbps */
292 #define ETH_LINK_SPEED_20G (1 << 9) /**< 20 Gbps */
293 #define ETH_LINK_SPEED_25G (1 << 10) /**< 25 Gbps */
294 #define ETH_LINK_SPEED_40G (1 << 11) /**< 40 Gbps */
295 #define ETH_LINK_SPEED_50G (1 << 12) /**< 50 Gbps */
296 #define ETH_LINK_SPEED_56G (1 << 13) /**< 56 Gbps */
297 #define ETH_LINK_SPEED_100G (1 << 14) /**< 100 Gbps */
298 #define ETH_LINK_SPEED_200G (1 << 15) /**< 200 Gbps */
301 /**@{@name Link speed
302 * Ethernet numeric link speeds in Mbps
304 #define ETH_SPEED_NUM_NONE 0 /**< Not defined */
305 #define ETH_SPEED_NUM_10M 10 /**< 10 Mbps */
306 #define ETH_SPEED_NUM_100M 100 /**< 100 Mbps */
307 #define ETH_SPEED_NUM_1G 1000 /**< 1 Gbps */
308 #define ETH_SPEED_NUM_2_5G 2500 /**< 2.5 Gbps */
309 #define ETH_SPEED_NUM_5G 5000 /**< 5 Gbps */
310 #define ETH_SPEED_NUM_10G 10000 /**< 10 Gbps */
311 #define ETH_SPEED_NUM_20G 20000 /**< 20 Gbps */
312 #define ETH_SPEED_NUM_25G 25000 /**< 25 Gbps */
313 #define ETH_SPEED_NUM_40G 40000 /**< 40 Gbps */
314 #define ETH_SPEED_NUM_50G 50000 /**< 50 Gbps */
315 #define ETH_SPEED_NUM_56G 56000 /**< 56 Gbps */
316 #define ETH_SPEED_NUM_100G 100000 /**< 100 Gbps */
317 #define ETH_SPEED_NUM_200G 200000 /**< 200 Gbps */
318 #define ETH_SPEED_NUM_UNKNOWN UINT32_MAX /**< Unknown */
322 * A structure used to retrieve link-level information of an Ethernet port.
325 struct rte_eth_link {
326 uint32_t link_speed; /**< ETH_SPEED_NUM_ */
327 uint16_t link_duplex : 1; /**< ETH_LINK_[HALF/FULL]_DUPLEX */
328 uint16_t link_autoneg : 1; /**< ETH_LINK_[AUTONEG/FIXED] */
329 uint16_t link_status : 1; /**< ETH_LINK_[DOWN/UP] */
330 } __rte_aligned(8); /**< aligned for atomic64 read/write */
332 /**@{@name Link negotiation
333 * Constants used in link management.
335 #define ETH_LINK_HALF_DUPLEX 0 /**< Half-duplex connection (see link_duplex). */
336 #define ETH_LINK_FULL_DUPLEX 1 /**< Full-duplex connection (see link_duplex). */
337 #define ETH_LINK_DOWN 0 /**< Link is down (see link_status). */
338 #define ETH_LINK_UP 1 /**< Link is up (see link_status). */
339 #define ETH_LINK_FIXED 0 /**< No autonegotiation (see link_autoneg). */
340 #define ETH_LINK_AUTONEG 1 /**< Autonegotiated (see link_autoneg). */
341 #define RTE_ETH_LINK_MAX_STR_LEN 40 /**< Max length of default link string. */
345 * A structure used to configure the ring threshold registers of an RX/TX
346 * queue for an Ethernet port.
348 struct rte_eth_thresh {
349 uint8_t pthresh; /**< Ring prefetch threshold. */
350 uint8_t hthresh; /**< Ring host threshold. */
351 uint8_t wthresh; /**< Ring writeback threshold. */
354 /**@{@name Multi-queue mode
355 * @see rte_eth_conf.rxmode.mq_mode.
357 #define ETH_MQ_RX_RSS_FLAG 0x1 /**< Enable RSS. @see rte_eth_rss_conf */
358 #define ETH_MQ_RX_DCB_FLAG 0x2 /**< Enable DCB. */
359 #define ETH_MQ_RX_VMDQ_FLAG 0x4 /**< Enable VMDq. */
363 * A set of values to identify what method is to be used to route
364 * packets to multiple queues.
366 enum rte_eth_rx_mq_mode {
367 /** None of DCB,RSS or VMDQ mode */
370 /** For RX side, only RSS is on */
371 ETH_MQ_RX_RSS = ETH_MQ_RX_RSS_FLAG,
372 /** For RX side,only DCB is on. */
373 ETH_MQ_RX_DCB = ETH_MQ_RX_DCB_FLAG,
374 /** Both DCB and RSS enable */
375 ETH_MQ_RX_DCB_RSS = ETH_MQ_RX_RSS_FLAG | ETH_MQ_RX_DCB_FLAG,
377 /** Only VMDQ, no RSS nor DCB */
378 ETH_MQ_RX_VMDQ_ONLY = ETH_MQ_RX_VMDQ_FLAG,
379 /** RSS mode with VMDQ */
380 ETH_MQ_RX_VMDQ_RSS = ETH_MQ_RX_RSS_FLAG | ETH_MQ_RX_VMDQ_FLAG,
381 /** Use VMDQ+DCB to route traffic to queues */
382 ETH_MQ_RX_VMDQ_DCB = ETH_MQ_RX_VMDQ_FLAG | ETH_MQ_RX_DCB_FLAG,
383 /** Enable both VMDQ and DCB in VMDq */
384 ETH_MQ_RX_VMDQ_DCB_RSS = ETH_MQ_RX_RSS_FLAG | ETH_MQ_RX_DCB_FLAG |
389 * for rx mq mode backward compatible
391 #define ETH_RSS ETH_MQ_RX_RSS
392 #define VMDQ_DCB ETH_MQ_RX_VMDQ_DCB
393 #define ETH_DCB_RX ETH_MQ_RX_DCB
396 * A set of values to identify what method is to be used to transmit
397 * packets using multi-TCs.
399 enum rte_eth_tx_mq_mode {
400 ETH_MQ_TX_NONE = 0, /**< It is in neither DCB nor VT mode. */
401 ETH_MQ_TX_DCB, /**< For TX side,only DCB is on. */
402 ETH_MQ_TX_VMDQ_DCB, /**< For TX side,both DCB and VT is on. */
403 ETH_MQ_TX_VMDQ_ONLY, /**< Only VT on, no DCB */
407 * for tx mq mode backward compatible
409 #define ETH_DCB_NONE ETH_MQ_TX_NONE
410 #define ETH_VMDQ_DCB_TX ETH_MQ_TX_VMDQ_DCB
411 #define ETH_DCB_TX ETH_MQ_TX_DCB
414 * A structure used to configure the RX features of an Ethernet port.
416 struct rte_eth_rxmode {
417 /** The multi-queue packet distribution mode to be used, e.g. RSS. */
418 enum rte_eth_rx_mq_mode mq_mode;
419 uint32_t mtu; /**< Requested MTU. */
420 /** Maximum allowed size of LRO aggregated packet. */
421 uint32_t max_lro_pkt_size;
422 uint16_t split_hdr_size; /**< hdr buf size (header_split enabled).*/
424 * Per-port Rx offloads to be set using DEV_RX_OFFLOAD_* flags.
425 * Only offloads set on rx_offload_capa field on rte_eth_dev_info
426 * structure are allowed to be set.
430 uint64_t reserved_64s[2]; /**< Reserved for future fields */
431 void *reserved_ptrs[2]; /**< Reserved for future fields */
435 * VLAN types to indicate if it is for single VLAN, inner VLAN or outer VLAN.
436 * Note that single VLAN is treated the same as inner VLAN.
439 ETH_VLAN_TYPE_UNKNOWN = 0,
440 ETH_VLAN_TYPE_INNER, /**< Inner VLAN. */
441 ETH_VLAN_TYPE_OUTER, /**< Single VLAN, or outer VLAN. */
446 * A structure used to describe a vlan filter.
447 * If the bit corresponding to a VID is set, such VID is on.
449 struct rte_vlan_filter_conf {
454 * A structure used to configure the Receive Side Scaling (RSS) feature
455 * of an Ethernet port.
456 * If not NULL, the *rss_key* pointer of the *rss_conf* structure points
457 * to an array holding the RSS key to use for hashing specific header
458 * fields of received packets. The length of this array should be indicated
459 * by *rss_key_len* below. Otherwise, a default random hash key is used by
462 * The *rss_key_len* field of the *rss_conf* structure indicates the length
463 * in bytes of the array pointed by *rss_key*. To be compatible, this length
464 * will be checked in i40e only. Others assume 40 bytes to be used as before.
466 * The *rss_hf* field of the *rss_conf* structure indicates the different
467 * types of IPv4/IPv6 packets to which the RSS hashing must be applied.
468 * Supplying an *rss_hf* equal to zero disables the RSS feature.
470 struct rte_eth_rss_conf {
471 uint8_t *rss_key; /**< If not NULL, 40-byte hash key. */
472 uint8_t rss_key_len; /**< hash key length in bytes. */
473 uint64_t rss_hf; /**< Hash functions to apply - see below. */
477 * A packet can be identified by hardware as different flow types. Different
478 * NIC hardware may support different flow types.
479 * Basically, the NIC hardware identifies the flow type as deep protocol as
480 * possible, and exclusively. For example, if a packet is identified as
481 * 'RTE_ETH_FLOW_NONFRAG_IPV4_TCP', it will not be any of other flow types,
482 * though it is an actual IPV4 packet.
484 #define RTE_ETH_FLOW_UNKNOWN 0
485 #define RTE_ETH_FLOW_RAW 1
486 #define RTE_ETH_FLOW_IPV4 2
487 #define RTE_ETH_FLOW_FRAG_IPV4 3
488 #define RTE_ETH_FLOW_NONFRAG_IPV4_TCP 4
489 #define RTE_ETH_FLOW_NONFRAG_IPV4_UDP 5
490 #define RTE_ETH_FLOW_NONFRAG_IPV4_SCTP 6
491 #define RTE_ETH_FLOW_NONFRAG_IPV4_OTHER 7
492 #define RTE_ETH_FLOW_IPV6 8
493 #define RTE_ETH_FLOW_FRAG_IPV6 9
494 #define RTE_ETH_FLOW_NONFRAG_IPV6_TCP 10
495 #define RTE_ETH_FLOW_NONFRAG_IPV6_UDP 11
496 #define RTE_ETH_FLOW_NONFRAG_IPV6_SCTP 12
497 #define RTE_ETH_FLOW_NONFRAG_IPV6_OTHER 13
498 #define RTE_ETH_FLOW_L2_PAYLOAD 14
499 #define RTE_ETH_FLOW_IPV6_EX 15
500 #define RTE_ETH_FLOW_IPV6_TCP_EX 16
501 #define RTE_ETH_FLOW_IPV6_UDP_EX 17
502 #define RTE_ETH_FLOW_PORT 18
503 /**< Consider device port number as a flow differentiator */
504 #define RTE_ETH_FLOW_VXLAN 19 /**< VXLAN protocol based flow */
505 #define RTE_ETH_FLOW_GENEVE 20 /**< GENEVE protocol based flow */
506 #define RTE_ETH_FLOW_NVGRE 21 /**< NVGRE protocol based flow */
507 #define RTE_ETH_FLOW_VXLAN_GPE 22 /**< VXLAN-GPE protocol based flow */
508 #define RTE_ETH_FLOW_GTPU 23 /**< GTPU protocol based flow */
509 #define RTE_ETH_FLOW_MAX 24
512 * Below macros are defined for RSS offload types, they can be used to
513 * fill rte_eth_rss_conf.rss_hf or rte_flow_action_rss.types.
515 #define ETH_RSS_IPV4 (1ULL << 2)
516 #define ETH_RSS_FRAG_IPV4 (1ULL << 3)
517 #define ETH_RSS_NONFRAG_IPV4_TCP (1ULL << 4)
518 #define ETH_RSS_NONFRAG_IPV4_UDP (1ULL << 5)
519 #define ETH_RSS_NONFRAG_IPV4_SCTP (1ULL << 6)
520 #define ETH_RSS_NONFRAG_IPV4_OTHER (1ULL << 7)
521 #define ETH_RSS_IPV6 (1ULL << 8)
522 #define ETH_RSS_FRAG_IPV6 (1ULL << 9)
523 #define ETH_RSS_NONFRAG_IPV6_TCP (1ULL << 10)
524 #define ETH_RSS_NONFRAG_IPV6_UDP (1ULL << 11)
525 #define ETH_RSS_NONFRAG_IPV6_SCTP (1ULL << 12)
526 #define ETH_RSS_NONFRAG_IPV6_OTHER (1ULL << 13)
527 #define ETH_RSS_L2_PAYLOAD (1ULL << 14)
528 #define ETH_RSS_IPV6_EX (1ULL << 15)
529 #define ETH_RSS_IPV6_TCP_EX (1ULL << 16)
530 #define ETH_RSS_IPV6_UDP_EX (1ULL << 17)
531 #define ETH_RSS_PORT (1ULL << 18)
532 #define ETH_RSS_VXLAN (1ULL << 19)
533 #define ETH_RSS_GENEVE (1ULL << 20)
534 #define ETH_RSS_NVGRE (1ULL << 21)
535 #define ETH_RSS_GTPU (1ULL << 23)
536 #define ETH_RSS_ETH (1ULL << 24)
537 #define ETH_RSS_S_VLAN (1ULL << 25)
538 #define ETH_RSS_C_VLAN (1ULL << 26)
539 #define ETH_RSS_ESP (1ULL << 27)
540 #define ETH_RSS_AH (1ULL << 28)
541 #define ETH_RSS_L2TPV3 (1ULL << 29)
542 #define ETH_RSS_PFCP (1ULL << 30)
543 #define ETH_RSS_PPPOE (1ULL << 31)
544 #define ETH_RSS_ECPRI (1ULL << 32)
545 #define ETH_RSS_MPLS (1ULL << 33)
546 #define ETH_RSS_IPV4_CHKSUM (1ULL << 34)
549 * The ETH_RSS_L4_CHKSUM works on checksum field of any L4 header.
550 * It is similar to ETH_RSS_PORT that they don't specify the specific type of
551 * L4 header. This macro is defined to replace some specific L4 (TCP/UDP/SCTP)
552 * checksum type for constructing the use of RSS offload bits.
554 * Due to above reason, some old APIs (and configuration) don't support
555 * ETH_RSS_L4_CHKSUM. The rte_flow RSS API supports it.
557 * For the case that checksum is not used in an UDP header,
558 * it takes the reserved value 0 as input for the hash function.
560 #define ETH_RSS_L4_CHKSUM (1ULL << 35)
563 * We use the following macros to combine with above ETH_RSS_* for
564 * more specific input set selection. These bits are defined starting
565 * from the high end of the 64 bits.
566 * Note: If we use above ETH_RSS_* without SRC/DST_ONLY, it represents
567 * both SRC and DST are taken into account. If SRC_ONLY and DST_ONLY of
568 * the same level are used simultaneously, it is the same case as none of
571 #define ETH_RSS_L3_SRC_ONLY (1ULL << 63)
572 #define ETH_RSS_L3_DST_ONLY (1ULL << 62)
573 #define ETH_RSS_L4_SRC_ONLY (1ULL << 61)
574 #define ETH_RSS_L4_DST_ONLY (1ULL << 60)
575 #define ETH_RSS_L2_SRC_ONLY (1ULL << 59)
576 #define ETH_RSS_L2_DST_ONLY (1ULL << 58)
579 * Only select IPV6 address prefix as RSS input set according to
580 * https://tools.ietf.org/html/rfc6052
581 * Must be combined with ETH_RSS_IPV6, ETH_RSS_NONFRAG_IPV6_UDP,
582 * ETH_RSS_NONFRAG_IPV6_TCP, ETH_RSS_NONFRAG_IPV6_SCTP.
584 #define RTE_ETH_RSS_L3_PRE32 (1ULL << 57)
585 #define RTE_ETH_RSS_L3_PRE40 (1ULL << 56)
586 #define RTE_ETH_RSS_L3_PRE48 (1ULL << 55)
587 #define RTE_ETH_RSS_L3_PRE56 (1ULL << 54)
588 #define RTE_ETH_RSS_L3_PRE64 (1ULL << 53)
589 #define RTE_ETH_RSS_L3_PRE96 (1ULL << 52)
592 * Use the following macros to combine with the above layers
593 * to choose inner and outer layers or both for RSS computation.
594 * Bits 50 and 51 are reserved for this.
598 * level 0, requests the default behavior.
599 * Depending on the packet type, it can mean outermost, innermost,
600 * anything in between or even no RSS.
601 * It basically stands for the innermost encapsulation level RSS
602 * can be performed on according to PMD and device capabilities.
604 #define ETH_RSS_LEVEL_PMD_DEFAULT (0ULL << 50)
607 * level 1, requests RSS to be performed on the outermost packet
608 * encapsulation level.
610 #define ETH_RSS_LEVEL_OUTERMOST (1ULL << 50)
613 * level 2, requests RSS to be performed on the specified inner packet
614 * encapsulation level, from outermost to innermost (lower to higher values).
616 #define ETH_RSS_LEVEL_INNERMOST (2ULL << 50)
617 #define ETH_RSS_LEVEL_MASK (3ULL << 50)
619 #define ETH_RSS_LEVEL(rss_hf) ((rss_hf & ETH_RSS_LEVEL_MASK) >> 50)
622 * For input set change of hash filter, if SRC_ONLY and DST_ONLY of
623 * the same level are used simultaneously, it is the same case as
624 * none of them are added.
627 * RSS types with SRC/DST_ONLY.
631 static inline uint64_t
632 rte_eth_rss_hf_refine(uint64_t rss_hf)
634 if ((rss_hf & ETH_RSS_L3_SRC_ONLY) && (rss_hf & ETH_RSS_L3_DST_ONLY))
635 rss_hf &= ~(ETH_RSS_L3_SRC_ONLY | ETH_RSS_L3_DST_ONLY);
637 if ((rss_hf & ETH_RSS_L4_SRC_ONLY) && (rss_hf & ETH_RSS_L4_DST_ONLY))
638 rss_hf &= ~(ETH_RSS_L4_SRC_ONLY | ETH_RSS_L4_DST_ONLY);
643 #define ETH_RSS_IPV6_PRE32 ( \
645 RTE_ETH_RSS_L3_PRE32)
647 #define ETH_RSS_IPV6_PRE40 ( \
649 RTE_ETH_RSS_L3_PRE40)
651 #define ETH_RSS_IPV6_PRE48 ( \
653 RTE_ETH_RSS_L3_PRE48)
655 #define ETH_RSS_IPV6_PRE56 ( \
657 RTE_ETH_RSS_L3_PRE56)
659 #define ETH_RSS_IPV6_PRE64 ( \
661 RTE_ETH_RSS_L3_PRE64)
663 #define ETH_RSS_IPV6_PRE96 ( \
665 RTE_ETH_RSS_L3_PRE96)
667 #define ETH_RSS_IPV6_PRE32_UDP ( \
668 ETH_RSS_NONFRAG_IPV6_UDP | \
669 RTE_ETH_RSS_L3_PRE32)
671 #define ETH_RSS_IPV6_PRE40_UDP ( \
672 ETH_RSS_NONFRAG_IPV6_UDP | \
673 RTE_ETH_RSS_L3_PRE40)
675 #define ETH_RSS_IPV6_PRE48_UDP ( \
676 ETH_RSS_NONFRAG_IPV6_UDP | \
677 RTE_ETH_RSS_L3_PRE48)
679 #define ETH_RSS_IPV6_PRE56_UDP ( \
680 ETH_RSS_NONFRAG_IPV6_UDP | \
681 RTE_ETH_RSS_L3_PRE56)
683 #define ETH_RSS_IPV6_PRE64_UDP ( \
684 ETH_RSS_NONFRAG_IPV6_UDP | \
685 RTE_ETH_RSS_L3_PRE64)
687 #define ETH_RSS_IPV6_PRE96_UDP ( \
688 ETH_RSS_NONFRAG_IPV6_UDP | \
689 RTE_ETH_RSS_L3_PRE96)
691 #define ETH_RSS_IPV6_PRE32_TCP ( \
692 ETH_RSS_NONFRAG_IPV6_TCP | \
693 RTE_ETH_RSS_L3_PRE32)
695 #define ETH_RSS_IPV6_PRE40_TCP ( \
696 ETH_RSS_NONFRAG_IPV6_TCP | \
697 RTE_ETH_RSS_L3_PRE40)
699 #define ETH_RSS_IPV6_PRE48_TCP ( \
700 ETH_RSS_NONFRAG_IPV6_TCP | \
701 RTE_ETH_RSS_L3_PRE48)
703 #define ETH_RSS_IPV6_PRE56_TCP ( \
704 ETH_RSS_NONFRAG_IPV6_TCP | \
705 RTE_ETH_RSS_L3_PRE56)
707 #define ETH_RSS_IPV6_PRE64_TCP ( \
708 ETH_RSS_NONFRAG_IPV6_TCP | \
709 RTE_ETH_RSS_L3_PRE64)
711 #define ETH_RSS_IPV6_PRE96_TCP ( \
712 ETH_RSS_NONFRAG_IPV6_TCP | \
713 RTE_ETH_RSS_L3_PRE96)
715 #define ETH_RSS_IPV6_PRE32_SCTP ( \
716 ETH_RSS_NONFRAG_IPV6_SCTP | \
717 RTE_ETH_RSS_L3_PRE32)
719 #define ETH_RSS_IPV6_PRE40_SCTP ( \
720 ETH_RSS_NONFRAG_IPV6_SCTP | \
721 RTE_ETH_RSS_L3_PRE40)
723 #define ETH_RSS_IPV6_PRE48_SCTP ( \
724 ETH_RSS_NONFRAG_IPV6_SCTP | \
725 RTE_ETH_RSS_L3_PRE48)
727 #define ETH_RSS_IPV6_PRE56_SCTP ( \
728 ETH_RSS_NONFRAG_IPV6_SCTP | \
729 RTE_ETH_RSS_L3_PRE56)
731 #define ETH_RSS_IPV6_PRE64_SCTP ( \
732 ETH_RSS_NONFRAG_IPV6_SCTP | \
733 RTE_ETH_RSS_L3_PRE64)
735 #define ETH_RSS_IPV6_PRE96_SCTP ( \
736 ETH_RSS_NONFRAG_IPV6_SCTP | \
737 RTE_ETH_RSS_L3_PRE96)
739 #define ETH_RSS_IP ( \
741 ETH_RSS_FRAG_IPV4 | \
742 ETH_RSS_NONFRAG_IPV4_OTHER | \
744 ETH_RSS_FRAG_IPV6 | \
745 ETH_RSS_NONFRAG_IPV6_OTHER | \
748 #define ETH_RSS_UDP ( \
749 ETH_RSS_NONFRAG_IPV4_UDP | \
750 ETH_RSS_NONFRAG_IPV6_UDP | \
753 #define ETH_RSS_TCP ( \
754 ETH_RSS_NONFRAG_IPV4_TCP | \
755 ETH_RSS_NONFRAG_IPV6_TCP | \
758 #define ETH_RSS_SCTP ( \
759 ETH_RSS_NONFRAG_IPV4_SCTP | \
760 ETH_RSS_NONFRAG_IPV6_SCTP)
762 #define ETH_RSS_TUNNEL ( \
767 #define ETH_RSS_VLAN ( \
771 /**< Mask of valid RSS hash protocols */
772 #define ETH_RSS_PROTO_MASK ( \
774 ETH_RSS_FRAG_IPV4 | \
775 ETH_RSS_NONFRAG_IPV4_TCP | \
776 ETH_RSS_NONFRAG_IPV4_UDP | \
777 ETH_RSS_NONFRAG_IPV4_SCTP | \
778 ETH_RSS_NONFRAG_IPV4_OTHER | \
780 ETH_RSS_FRAG_IPV6 | \
781 ETH_RSS_NONFRAG_IPV6_TCP | \
782 ETH_RSS_NONFRAG_IPV6_UDP | \
783 ETH_RSS_NONFRAG_IPV6_SCTP | \
784 ETH_RSS_NONFRAG_IPV6_OTHER | \
785 ETH_RSS_L2_PAYLOAD | \
787 ETH_RSS_IPV6_TCP_EX | \
788 ETH_RSS_IPV6_UDP_EX | \
796 * Definitions used for redirection table entry size.
797 * Some RSS RETA sizes may not be supported by some drivers, check the
798 * documentation or the description of relevant functions for more details.
800 #define ETH_RSS_RETA_SIZE_64 64
801 #define ETH_RSS_RETA_SIZE_128 128
802 #define ETH_RSS_RETA_SIZE_256 256
803 #define ETH_RSS_RETA_SIZE_512 512
804 #define RTE_RETA_GROUP_SIZE 64
806 /**@{@name VMDq and DCB maximums */
807 #define ETH_VMDQ_MAX_VLAN_FILTERS 64 /**< Maximum nb. of VMDQ vlan filters. */
808 #define ETH_DCB_NUM_USER_PRIORITIES 8 /**< Maximum nb. of DCB priorities. */
809 #define ETH_VMDQ_DCB_NUM_QUEUES 128 /**< Maximum nb. of VMDQ DCB queues. */
810 #define ETH_DCB_NUM_QUEUES 128 /**< Maximum nb. of DCB queues. */
813 /**@{@name DCB capabilities */
814 #define ETH_DCB_PG_SUPPORT 0x00000001 /**< Priority Group(ETS) support. */
815 #define ETH_DCB_PFC_SUPPORT 0x00000002 /**< Priority Flow Control support. */
818 /**@{@name VLAN offload bits */
819 #define ETH_VLAN_STRIP_OFFLOAD 0x0001 /**< VLAN Strip On/Off */
820 #define ETH_VLAN_FILTER_OFFLOAD 0x0002 /**< VLAN Filter On/Off */
821 #define ETH_VLAN_EXTEND_OFFLOAD 0x0004 /**< VLAN Extend On/Off */
822 #define ETH_QINQ_STRIP_OFFLOAD 0x0008 /**< QINQ Strip On/Off */
824 #define ETH_VLAN_STRIP_MASK 0x0001 /**< VLAN Strip setting mask */
825 #define ETH_VLAN_FILTER_MASK 0x0002 /**< VLAN Filter setting mask*/
826 #define ETH_VLAN_EXTEND_MASK 0x0004 /**< VLAN Extend setting mask*/
827 #define ETH_QINQ_STRIP_MASK 0x0008 /**< QINQ Strip setting mask */
828 #define ETH_VLAN_ID_MAX 0x0FFF /**< VLAN ID is in lower 12 bits*/
831 /* Definitions used for receive MAC address */
832 #define ETH_NUM_RECEIVE_MAC_ADDR 128 /**< Maximum nb. of receive mac addr. */
834 /* Definitions used for unicast hash */
835 #define ETH_VMDQ_NUM_UC_HASH_ARRAY 128 /**< Maximum nb. of UC hash array. */
837 /**@{@name VMDq Rx mode
838 * @see rte_eth_vmdq_rx_conf.rx_mode
840 #define ETH_VMDQ_ACCEPT_UNTAG 0x0001 /**< accept untagged packets. */
841 #define ETH_VMDQ_ACCEPT_HASH_MC 0x0002 /**< accept packets in multicast table . */
842 #define ETH_VMDQ_ACCEPT_HASH_UC 0x0004 /**< accept packets in unicast table. */
843 #define ETH_VMDQ_ACCEPT_BROADCAST 0x0008 /**< accept broadcast packets. */
844 #define ETH_VMDQ_ACCEPT_MULTICAST 0x0010 /**< multicast promiscuous. */
848 * A structure used to configure 64 entries of Redirection Table of the
849 * Receive Side Scaling (RSS) feature of an Ethernet port. To configure
850 * more than 64 entries supported by hardware, an array of this structure
853 struct rte_eth_rss_reta_entry64 {
855 /**< Mask bits indicate which entries need to be updated/queried. */
856 uint16_t reta[RTE_RETA_GROUP_SIZE];
857 /**< Group of 64 redirection table entries. */
861 * This enum indicates the possible number of traffic classes
862 * in DCB configurations
864 enum rte_eth_nb_tcs {
865 ETH_4_TCS = 4, /**< 4 TCs with DCB. */
866 ETH_8_TCS = 8 /**< 8 TCs with DCB. */
870 * This enum indicates the possible number of queue pools
871 * in VMDQ configurations.
873 enum rte_eth_nb_pools {
874 ETH_8_POOLS = 8, /**< 8 VMDq pools. */
875 ETH_16_POOLS = 16, /**< 16 VMDq pools. */
876 ETH_32_POOLS = 32, /**< 32 VMDq pools. */
877 ETH_64_POOLS = 64 /**< 64 VMDq pools. */
880 /* This structure may be extended in future. */
881 struct rte_eth_dcb_rx_conf {
882 enum rte_eth_nb_tcs nb_tcs; /**< Possible DCB TCs, 4 or 8 TCs */
883 /** Traffic class each UP mapped to. */
884 uint8_t dcb_tc[ETH_DCB_NUM_USER_PRIORITIES];
887 struct rte_eth_vmdq_dcb_tx_conf {
888 enum rte_eth_nb_pools nb_queue_pools; /**< With DCB, 16 or 32 pools. */
889 /** Traffic class each UP mapped to. */
890 uint8_t dcb_tc[ETH_DCB_NUM_USER_PRIORITIES];
893 struct rte_eth_dcb_tx_conf {
894 enum rte_eth_nb_tcs nb_tcs; /**< Possible DCB TCs, 4 or 8 TCs. */
895 /** Traffic class each UP mapped to. */
896 uint8_t dcb_tc[ETH_DCB_NUM_USER_PRIORITIES];
899 struct rte_eth_vmdq_tx_conf {
900 enum rte_eth_nb_pools nb_queue_pools; /**< VMDq mode, 64 pools. */
904 * A structure used to configure the VMDQ+DCB feature
905 * of an Ethernet port.
907 * Using this feature, packets are routed to a pool of queues, based
908 * on the vlan id in the vlan tag, and then to a specific queue within
909 * that pool, using the user priority vlan tag field.
911 * A default pool may be used, if desired, to route all traffic which
912 * does not match the vlan filter rules.
914 struct rte_eth_vmdq_dcb_conf {
915 enum rte_eth_nb_pools nb_queue_pools; /**< With DCB, 16 or 32 pools */
916 uint8_t enable_default_pool; /**< If non-zero, use a default pool */
917 uint8_t default_pool; /**< The default pool, if applicable */
918 uint8_t nb_pool_maps; /**< We can have up to 64 filters/mappings */
920 uint16_t vlan_id; /**< The vlan id of the received frame */
921 uint64_t pools; /**< Bitmask of pools for packet rx */
922 } pool_map[ETH_VMDQ_MAX_VLAN_FILTERS]; /**< VMDq vlan pool maps. */
923 uint8_t dcb_tc[ETH_DCB_NUM_USER_PRIORITIES];
924 /**< Selects a queue in a pool */
928 * A structure used to configure the VMDQ feature of an Ethernet port when
929 * not combined with the DCB feature.
931 * Using this feature, packets are routed to a pool of queues. By default,
932 * the pool selection is based on the MAC address, the vlan id in the
933 * vlan tag as specified in the pool_map array.
934 * Passing the ETH_VMDQ_ACCEPT_UNTAG in the rx_mode field allows pool
935 * selection using only the MAC address. MAC address to pool mapping is done
936 * using the rte_eth_dev_mac_addr_add function, with the pool parameter
937 * corresponding to the pool id.
939 * Queue selection within the selected pool will be done using RSS when
940 * it is enabled or revert to the first queue of the pool if not.
942 * A default pool may be used, if desired, to route all traffic which
943 * does not match the vlan filter rules or any pool MAC address.
945 struct rte_eth_vmdq_rx_conf {
946 enum rte_eth_nb_pools nb_queue_pools; /**< VMDq only mode, 8 or 64 pools */
947 uint8_t enable_default_pool; /**< If non-zero, use a default pool */
948 uint8_t default_pool; /**< The default pool, if applicable */
949 uint8_t enable_loop_back; /**< Enable VT loop back */
950 uint8_t nb_pool_maps; /**< We can have up to 64 filters/mappings */
951 uint32_t rx_mode; /**< Flags from ETH_VMDQ_ACCEPT_* */
953 uint16_t vlan_id; /**< The vlan id of the received frame */
954 uint64_t pools; /**< Bitmask of pools for packet rx */
955 } pool_map[ETH_VMDQ_MAX_VLAN_FILTERS]; /**< VMDq vlan pool maps. */
959 * A structure used to configure the TX features of an Ethernet port.
961 struct rte_eth_txmode {
962 enum rte_eth_tx_mq_mode mq_mode; /**< TX multi-queues mode. */
964 * Per-port Tx offloads to be set using DEV_TX_OFFLOAD_* flags.
965 * Only offloads set on tx_offload_capa field on rte_eth_dev_info
966 * structure are allowed to be set.
972 uint8_t hw_vlan_reject_tagged : 1,
973 /**< If set, reject sending out tagged pkts */
974 hw_vlan_reject_untagged : 1,
975 /**< If set, reject sending out untagged pkts */
976 hw_vlan_insert_pvid : 1;
977 /**< If set, enable port based VLAN insertion */
979 uint64_t reserved_64s[2]; /**< Reserved for future fields */
980 void *reserved_ptrs[2]; /**< Reserved for future fields */
985 * @b EXPERIMENTAL: this structure may change without prior notice.
987 * A structure used to configure an Rx packet segment to split.
989 * If RTE_ETH_RX_OFFLOAD_BUFFER_SPLIT flag is set in offloads field,
990 * the PMD will split the received packets into multiple segments
991 * according to the specification in the description array:
993 * - The first network buffer will be allocated from the memory pool,
994 * specified in the first array element, the second buffer, from the
995 * pool in the second element, and so on.
997 * - The offsets from the segment description elements specify
998 * the data offset from the buffer beginning except the first mbuf.
999 * The first segment offset is added with RTE_PKTMBUF_HEADROOM.
1001 * - The lengths in the elements define the maximal data amount
1002 * being received to each segment. The receiving starts with filling
1003 * up the first mbuf data buffer up to specified length. If the
1004 * there are data remaining (packet is longer than buffer in the first
1005 * mbuf) the following data will be pushed to the next segment
1006 * up to its own length, and so on.
1008 * - If the length in the segment description element is zero
1009 * the actual buffer size will be deduced from the appropriate
1010 * memory pool properties.
1012 * - If there is not enough elements to describe the buffer for entire
1013 * packet of maximal length the following parameters will be used
1014 * for the all remaining segments:
1015 * - pool from the last valid element
1016 * - the buffer size from this pool
1019 struct rte_eth_rxseg_split {
1020 struct rte_mempool *mp; /**< Memory pool to allocate segment from. */
1021 uint16_t length; /**< Segment data length, configures split point. */
1022 uint16_t offset; /**< Data offset from beginning of mbuf data buffer. */
1023 uint32_t reserved; /**< Reserved field. */
1028 * @b EXPERIMENTAL: this structure may change without prior notice.
1030 * A common structure used to describe Rx packet segment properties.
1032 union rte_eth_rxseg {
1033 /* The settings for buffer split offload. */
1034 struct rte_eth_rxseg_split split;
1035 /* The other features settings should be added here. */
1039 * A structure used to configure an RX ring of an Ethernet port.
1041 struct rte_eth_rxconf {
1042 struct rte_eth_thresh rx_thresh; /**< RX ring threshold registers. */
1043 uint16_t rx_free_thresh; /**< Drives the freeing of RX descriptors. */
1044 uint8_t rx_drop_en; /**< Drop packets if no descriptors are available. */
1045 uint8_t rx_deferred_start; /**< Do not start queue with rte_eth_dev_start(). */
1046 uint16_t rx_nseg; /**< Number of descriptions in rx_seg array. */
1048 * Per-queue Rx offloads to be set using DEV_RX_OFFLOAD_* flags.
1049 * Only offloads set on rx_queue_offload_capa or rx_offload_capa
1050 * fields on rte_eth_dev_info structure are allowed to be set.
1054 * Points to the array of segment descriptions for an entire packet.
1055 * Array elements are properties for consecutive Rx segments.
1057 * The supported capabilities of receiving segmentation is reported
1058 * in rte_eth_dev_info.rx_seg_capa field.
1060 union rte_eth_rxseg *rx_seg;
1062 uint64_t reserved_64s[2]; /**< Reserved for future fields */
1063 void *reserved_ptrs[2]; /**< Reserved for future fields */
1067 * A structure used to configure a TX ring of an Ethernet port.
1069 struct rte_eth_txconf {
1070 struct rte_eth_thresh tx_thresh; /**< TX ring threshold registers. */
1071 uint16_t tx_rs_thresh; /**< Drives the setting of RS bit on TXDs. */
1072 uint16_t tx_free_thresh; /**< Start freeing TX buffers if there are
1073 less free descriptors than this value. */
1075 uint8_t tx_deferred_start; /**< Do not start queue with rte_eth_dev_start(). */
1077 * Per-queue Tx offloads to be set using DEV_TX_OFFLOAD_* flags.
1078 * Only offloads set on tx_queue_offload_capa or tx_offload_capa
1079 * fields on rte_eth_dev_info structure are allowed to be set.
1083 uint64_t reserved_64s[2]; /**< Reserved for future fields */
1084 void *reserved_ptrs[2]; /**< Reserved for future fields */
1089 * @b EXPERIMENTAL: this API may change, or be removed, without prior notice
1091 * A structure used to return the hairpin capabilities that are supported.
1093 struct rte_eth_hairpin_cap {
1094 /** The max number of hairpin queues (different bindings). */
1095 uint16_t max_nb_queues;
1096 /** Max number of Rx queues to be connected to one Tx queue. */
1097 uint16_t max_rx_2_tx;
1098 /** Max number of Tx queues to be connected to one Rx queue. */
1099 uint16_t max_tx_2_rx;
1100 uint16_t max_nb_desc; /**< The max num of descriptors. */
1103 #define RTE_ETH_MAX_HAIRPIN_PEERS 32
1107 * @b EXPERIMENTAL: this API may change, or be removed, without prior notice
1109 * A structure used to hold hairpin peer data.
1111 struct rte_eth_hairpin_peer {
1112 uint16_t port; /**< Peer port. */
1113 uint16_t queue; /**< Peer queue. */
1118 * @b EXPERIMENTAL: this API may change, or be removed, without prior notice
1120 * A structure used to configure hairpin binding.
1122 struct rte_eth_hairpin_conf {
1123 uint32_t peer_count:16; /**< The number of peers. */
1126 * Explicit Tx flow rule mode.
1127 * One hairpin pair of queues should have the same attribute.
1129 * - When set, the user should be responsible for inserting the hairpin
1130 * Tx part flows and removing them.
1131 * - When clear, the PMD will try to handle the Tx part of the flows,
1132 * e.g., by splitting one flow into two parts.
1134 uint32_t tx_explicit:1;
1137 * Manually bind hairpin queues.
1138 * One hairpin pair of queues should have the same attribute.
1140 * - When set, to enable hairpin, the user should call the hairpin bind
1141 * function after all the queues are set up properly and the ports are
1142 * started. Also, the hairpin unbind function should be called
1143 * accordingly before stopping a port that with hairpin configured.
1144 * - When clear, the PMD will try to enable the hairpin with the queues
1145 * configured automatically during port start.
1147 uint32_t manual_bind:1;
1148 uint32_t reserved:14; /**< Reserved bits. */
1149 struct rte_eth_hairpin_peer peers[RTE_ETH_MAX_HAIRPIN_PEERS];
1153 * A structure contains information about HW descriptor ring limitations.
1155 struct rte_eth_desc_lim {
1156 uint16_t nb_max; /**< Max allowed number of descriptors. */
1157 uint16_t nb_min; /**< Min allowed number of descriptors. */
1158 uint16_t nb_align; /**< Number of descriptors should be aligned to. */
1161 * Max allowed number of segments per whole packet.
1163 * - For TSO packet this is the total number of data descriptors allowed
1166 * @see nb_mtu_seg_max
1168 uint16_t nb_seg_max;
1171 * Max number of segments per one MTU.
1173 * - For non-TSO packet, this is the maximum allowed number of segments
1174 * in a single transmit packet.
1176 * - For TSO packet each segment within the TSO may span up to this
1181 uint16_t nb_mtu_seg_max;
1185 * This enum indicates the flow control mode
1187 enum rte_eth_fc_mode {
1188 RTE_FC_NONE = 0, /**< Disable flow control. */
1189 RTE_FC_RX_PAUSE, /**< RX pause frame, enable flowctrl on TX side. */
1190 RTE_FC_TX_PAUSE, /**< TX pause frame, enable flowctrl on RX side. */
1191 RTE_FC_FULL /**< Enable flow control on both side. */
1195 * A structure used to configure Ethernet flow control parameter.
1196 * These parameters will be configured into the register of the NIC.
1197 * Please refer to the corresponding data sheet for proper value.
1199 struct rte_eth_fc_conf {
1200 uint32_t high_water; /**< High threshold value to trigger XOFF */
1201 uint32_t low_water; /**< Low threshold value to trigger XON */
1202 uint16_t pause_time; /**< Pause quota in the Pause frame */
1203 uint16_t send_xon; /**< Is XON frame need be sent */
1204 enum rte_eth_fc_mode mode; /**< Link flow control mode */
1205 uint8_t mac_ctrl_frame_fwd; /**< Forward MAC control frames */
1206 uint8_t autoneg; /**< Use Pause autoneg */
1210 * A structure used to configure Ethernet priority flow control parameter.
1211 * These parameters will be configured into the register of the NIC.
1212 * Please refer to the corresponding data sheet for proper value.
1214 struct rte_eth_pfc_conf {
1215 struct rte_eth_fc_conf fc; /**< General flow control parameter. */
1216 uint8_t priority; /**< VLAN User Priority. */
1220 * Tunnel type for device-specific classifier configuration.
1221 * @see rte_eth_udp_tunnel
1223 enum rte_eth_tunnel_type {
1224 RTE_TUNNEL_TYPE_NONE = 0,
1225 RTE_TUNNEL_TYPE_VXLAN,
1226 RTE_TUNNEL_TYPE_GENEVE,
1227 RTE_TUNNEL_TYPE_TEREDO,
1228 RTE_TUNNEL_TYPE_NVGRE,
1229 RTE_TUNNEL_TYPE_IP_IN_GRE,
1230 RTE_L2_TUNNEL_TYPE_E_TAG,
1231 RTE_TUNNEL_TYPE_VXLAN_GPE,
1232 RTE_TUNNEL_TYPE_ECPRI,
1233 RTE_TUNNEL_TYPE_MAX,
1236 /* Deprecated API file for rte_eth_dev_filter_* functions */
1237 #include "rte_eth_ctrl.h"
1240 * Memory space that can be configured to store Flow Director filters
1241 * in the board memory.
1243 enum rte_fdir_pballoc_type {
1244 RTE_FDIR_PBALLOC_64K = 0, /**< 64k. */
1245 RTE_FDIR_PBALLOC_128K, /**< 128k. */
1246 RTE_FDIR_PBALLOC_256K, /**< 256k. */
1250 * Select report mode of FDIR hash information in RX descriptors.
1252 enum rte_fdir_status_mode {
1253 RTE_FDIR_NO_REPORT_STATUS = 0, /**< Never report FDIR hash. */
1254 RTE_FDIR_REPORT_STATUS, /**< Only report FDIR hash for matching pkts. */
1255 RTE_FDIR_REPORT_STATUS_ALWAYS, /**< Always report FDIR hash. */
1259 * A structure used to configure the Flow Director (FDIR) feature
1260 * of an Ethernet port.
1262 * If mode is RTE_FDIR_MODE_NONE, the pballoc value is ignored.
1264 struct rte_fdir_conf {
1265 enum rte_fdir_mode mode; /**< Flow Director mode. */
1266 enum rte_fdir_pballoc_type pballoc; /**< Space for FDIR filters. */
1267 enum rte_fdir_status_mode status; /**< How to report FDIR hash. */
1268 /** RX queue of packets matching a "drop" filter in perfect mode. */
1270 struct rte_eth_fdir_masks mask;
1271 struct rte_eth_fdir_flex_conf flex_conf;
1272 /**< Flex payload configuration. */
1276 * UDP tunneling configuration.
1278 * Used to configure the classifier of a device,
1279 * associating an UDP port with a type of tunnel.
1281 * Some NICs may need such configuration to properly parse a tunnel
1282 * with any standard or custom UDP port.
1284 struct rte_eth_udp_tunnel {
1285 uint16_t udp_port; /**< UDP port used for the tunnel. */
1286 uint8_t prot_type; /**< Tunnel type. @see rte_eth_tunnel_type */
1290 * A structure used to enable/disable specific device interrupts.
1292 struct rte_intr_conf {
1293 /** enable/disable lsc interrupt. 0 (default) - disable, 1 enable */
1295 /** enable/disable rxq interrupt. 0 (default) - disable, 1 enable */
1297 /** enable/disable rmv interrupt. 0 (default) - disable, 1 enable */
1302 * A structure used to configure an Ethernet port.
1303 * Depending upon the RX multi-queue mode, extra advanced
1304 * configuration settings may be needed.
1306 struct rte_eth_conf {
1307 uint32_t link_speeds; /**< bitmap of ETH_LINK_SPEED_XXX of speeds to be
1308 used. ETH_LINK_SPEED_FIXED disables link
1309 autonegotiation, and a unique speed shall be
1310 set. Otherwise, the bitmap defines the set of
1311 speeds to be advertised. If the special value
1312 ETH_LINK_SPEED_AUTONEG (0) is used, all speeds
1313 supported are advertised. */
1314 struct rte_eth_rxmode rxmode; /**< Port RX configuration. */
1315 struct rte_eth_txmode txmode; /**< Port TX configuration. */
1316 uint32_t lpbk_mode; /**< Loopback operation mode. By default the value
1317 is 0, meaning the loopback mode is disabled.
1318 Read the datasheet of given ethernet controller
1319 for details. The possible values of this field
1320 are defined in implementation of each driver. */
1322 struct rte_eth_rss_conf rss_conf; /**< Port RSS configuration */
1323 struct rte_eth_vmdq_dcb_conf vmdq_dcb_conf;
1324 /**< Port vmdq+dcb configuration. */
1325 struct rte_eth_dcb_rx_conf dcb_rx_conf;
1326 /**< Port dcb RX configuration. */
1327 struct rte_eth_vmdq_rx_conf vmdq_rx_conf;
1328 /**< Port vmdq RX configuration. */
1329 } rx_adv_conf; /**< Port RX filtering configuration. */
1331 struct rte_eth_vmdq_dcb_tx_conf vmdq_dcb_tx_conf;
1332 /**< Port vmdq+dcb TX configuration. */
1333 struct rte_eth_dcb_tx_conf dcb_tx_conf;
1334 /**< Port dcb TX configuration. */
1335 struct rte_eth_vmdq_tx_conf vmdq_tx_conf;
1336 /**< Port vmdq TX configuration. */
1337 } tx_adv_conf; /**< Port TX DCB configuration (union). */
1338 /** Currently,Priority Flow Control(PFC) are supported,if DCB with PFC
1339 is needed,and the variable must be set ETH_DCB_PFC_SUPPORT. */
1340 uint32_t dcb_capability_en;
1341 struct rte_fdir_conf fdir_conf; /**< FDIR configuration. DEPRECATED */
1342 struct rte_intr_conf intr_conf; /**< Interrupt mode configuration. */
1346 * RX offload capabilities of a device.
1348 #define DEV_RX_OFFLOAD_VLAN_STRIP 0x00000001
1349 #define DEV_RX_OFFLOAD_IPV4_CKSUM 0x00000002
1350 #define DEV_RX_OFFLOAD_UDP_CKSUM 0x00000004
1351 #define DEV_RX_OFFLOAD_TCP_CKSUM 0x00000008
1352 #define DEV_RX_OFFLOAD_TCP_LRO 0x00000010
1353 #define DEV_RX_OFFLOAD_QINQ_STRIP 0x00000020
1354 #define DEV_RX_OFFLOAD_OUTER_IPV4_CKSUM 0x00000040
1355 #define DEV_RX_OFFLOAD_MACSEC_STRIP 0x00000080
1356 #define DEV_RX_OFFLOAD_HEADER_SPLIT 0x00000100
1357 #define DEV_RX_OFFLOAD_VLAN_FILTER 0x00000200
1358 #define DEV_RX_OFFLOAD_VLAN_EXTEND 0x00000400
1359 #define DEV_RX_OFFLOAD_SCATTER 0x00002000
1361 * Timestamp is set by the driver in RTE_MBUF_DYNFIELD_TIMESTAMP_NAME
1362 * and RTE_MBUF_DYNFLAG_RX_TIMESTAMP_NAME is set in ol_flags.
1363 * The mbuf field and flag are registered when the offload is configured.
1365 #define DEV_RX_OFFLOAD_TIMESTAMP 0x00004000
1366 #define DEV_RX_OFFLOAD_SECURITY 0x00008000
1367 #define DEV_RX_OFFLOAD_KEEP_CRC 0x00010000
1368 #define DEV_RX_OFFLOAD_SCTP_CKSUM 0x00020000
1369 #define DEV_RX_OFFLOAD_OUTER_UDP_CKSUM 0x00040000
1370 #define DEV_RX_OFFLOAD_RSS_HASH 0x00080000
1371 #define RTE_ETH_RX_OFFLOAD_BUFFER_SPLIT 0x00100000
1373 #define DEV_RX_OFFLOAD_CHECKSUM (DEV_RX_OFFLOAD_IPV4_CKSUM | \
1374 DEV_RX_OFFLOAD_UDP_CKSUM | \
1375 DEV_RX_OFFLOAD_TCP_CKSUM)
1376 #define DEV_RX_OFFLOAD_VLAN (DEV_RX_OFFLOAD_VLAN_STRIP | \
1377 DEV_RX_OFFLOAD_VLAN_FILTER | \
1378 DEV_RX_OFFLOAD_VLAN_EXTEND | \
1379 DEV_RX_OFFLOAD_QINQ_STRIP)
1382 * If new Rx offload capabilities are defined, they also must be
1383 * mentioned in rte_rx_offload_names in rte_ethdev.c file.
1387 * TX offload capabilities of a device.
1389 #define DEV_TX_OFFLOAD_VLAN_INSERT 0x00000001
1390 #define DEV_TX_OFFLOAD_IPV4_CKSUM 0x00000002
1391 #define DEV_TX_OFFLOAD_UDP_CKSUM 0x00000004
1392 #define DEV_TX_OFFLOAD_TCP_CKSUM 0x00000008
1393 #define DEV_TX_OFFLOAD_SCTP_CKSUM 0x00000010
1394 #define DEV_TX_OFFLOAD_TCP_TSO 0x00000020
1395 #define DEV_TX_OFFLOAD_UDP_TSO 0x00000040
1396 #define DEV_TX_OFFLOAD_OUTER_IPV4_CKSUM 0x00000080 /**< Used for tunneling packet. */
1397 #define DEV_TX_OFFLOAD_QINQ_INSERT 0x00000100
1398 #define DEV_TX_OFFLOAD_VXLAN_TNL_TSO 0x00000200 /**< Used for tunneling packet. */
1399 #define DEV_TX_OFFLOAD_GRE_TNL_TSO 0x00000400 /**< Used for tunneling packet. */
1400 #define DEV_TX_OFFLOAD_IPIP_TNL_TSO 0x00000800 /**< Used for tunneling packet. */
1401 #define DEV_TX_OFFLOAD_GENEVE_TNL_TSO 0x00001000 /**< Used for tunneling packet. */
1402 #define DEV_TX_OFFLOAD_MACSEC_INSERT 0x00002000
1403 #define DEV_TX_OFFLOAD_MT_LOCKFREE 0x00004000
1404 /**< Multiple threads can invoke rte_eth_tx_burst() concurrently on the same
1405 * tx queue without SW lock.
1407 #define DEV_TX_OFFLOAD_MULTI_SEGS 0x00008000
1408 /**< Device supports multi segment send. */
1409 #define DEV_TX_OFFLOAD_MBUF_FAST_FREE 0x00010000
1410 /**< Device supports optimization for fast release of mbufs.
1411 * When set application must guarantee that per-queue all mbufs comes from
1412 * the same mempool and has refcnt = 1.
1414 #define DEV_TX_OFFLOAD_SECURITY 0x00020000
1416 * Device supports generic UDP tunneled packet TSO.
1417 * Application must set PKT_TX_TUNNEL_UDP and other mbuf fields required
1420 #define DEV_TX_OFFLOAD_UDP_TNL_TSO 0x00040000
1422 * Device supports generic IP tunneled packet TSO.
1423 * Application must set PKT_TX_TUNNEL_IP and other mbuf fields required
1426 #define DEV_TX_OFFLOAD_IP_TNL_TSO 0x00080000
1427 /** Device supports outer UDP checksum */
1428 #define DEV_TX_OFFLOAD_OUTER_UDP_CKSUM 0x00100000
1430 * Device sends on time read from RTE_MBUF_DYNFIELD_TIMESTAMP_NAME
1431 * if RTE_MBUF_DYNFLAG_TX_TIMESTAMP_NAME is set in ol_flags.
1432 * The mbuf field and flag are registered when the offload is configured.
1434 #define DEV_TX_OFFLOAD_SEND_ON_TIMESTAMP 0x00200000
1436 * If new Tx offload capabilities are defined, they also must be
1437 * mentioned in rte_tx_offload_names in rte_ethdev.c file.
1440 /**@{@name Device capabilities
1441 * Non-offload capabilities reported in rte_eth_dev_info.dev_capa.
1443 /** Device supports Rx queue setup after device started. */
1444 #define RTE_ETH_DEV_CAPA_RUNTIME_RX_QUEUE_SETUP 0x00000001
1445 /** Device supports Tx queue setup after device started. */
1446 #define RTE_ETH_DEV_CAPA_RUNTIME_TX_QUEUE_SETUP 0x00000002
1450 * Fallback default preferred Rx/Tx port parameters.
1451 * These are used if an application requests default parameters
1452 * but the PMD does not provide preferred values.
1454 #define RTE_ETH_DEV_FALLBACK_RX_RINGSIZE 512
1455 #define RTE_ETH_DEV_FALLBACK_TX_RINGSIZE 512
1456 #define RTE_ETH_DEV_FALLBACK_RX_NBQUEUES 1
1457 #define RTE_ETH_DEV_FALLBACK_TX_NBQUEUES 1
1460 * Preferred Rx/Tx port parameters.
1461 * There are separate instances of this structure for transmission
1462 * and reception respectively.
1464 struct rte_eth_dev_portconf {
1465 uint16_t burst_size; /**< Device-preferred burst size */
1466 uint16_t ring_size; /**< Device-preferred size of queue rings */
1467 uint16_t nb_queues; /**< Device-preferred number of queues */
1471 * Default values for switch domain id when ethdev does not support switch
1472 * domain definitions.
1474 #define RTE_ETH_DEV_SWITCH_DOMAIN_ID_INVALID (UINT16_MAX)
1477 * Ethernet device associated switch information
1479 struct rte_eth_switch_info {
1480 const char *name; /**< switch name */
1481 uint16_t domain_id; /**< switch domain id */
1484 * mapping to the devices physical switch port as enumerated from the
1485 * perspective of the embedded interconnect/switch. For SR-IOV enabled
1486 * device this may correspond to the VF_ID of each virtual function,
1487 * but each driver should explicitly define the mapping of switch
1488 * port identifier to that physical interconnect/switch
1494 * @b EXPERIMENTAL: this structure may change without prior notice.
1496 * Ethernet device Rx buffer segmentation capabilities.
1498 struct rte_eth_rxseg_capa {
1500 uint32_t multi_pools:1; /**< Supports receiving to multiple pools.*/
1501 uint32_t offset_allowed:1; /**< Supports buffer offsets. */
1502 uint32_t offset_align_log2:4; /**< Required offset alignment. */
1503 uint16_t max_nseg; /**< Maximum amount of segments to split. */
1504 uint16_t reserved; /**< Reserved field. */
1508 * Ethernet device information
1512 * Ethernet device representor port type.
1514 enum rte_eth_representor_type {
1515 RTE_ETH_REPRESENTOR_NONE, /**< not a representor. */
1516 RTE_ETH_REPRESENTOR_VF, /**< representor of Virtual Function. */
1517 RTE_ETH_REPRESENTOR_SF, /**< representor of Sub Function. */
1518 RTE_ETH_REPRESENTOR_PF, /**< representor of Physical Function. */
1522 * A structure used to retrieve the contextual information of
1523 * an Ethernet device, such as the controlling driver of the
1526 struct rte_eth_dev_info {
1527 struct rte_device *device; /** Generic device information */
1528 const char *driver_name; /**< Device Driver name. */
1529 unsigned int if_index; /**< Index to bound host interface, or 0 if none.
1530 Use if_indextoname() to translate into an interface name. */
1531 uint16_t min_mtu; /**< Minimum MTU allowed */
1532 uint16_t max_mtu; /**< Maximum MTU allowed */
1533 const uint32_t *dev_flags; /**< Device flags */
1534 uint32_t min_rx_bufsize; /**< Minimum size of RX buffer. */
1535 uint32_t max_rx_pktlen; /**< Maximum configurable length of RX pkt. */
1536 /** Maximum configurable size of LRO aggregated packet. */
1537 uint32_t max_lro_pkt_size;
1538 uint16_t max_rx_queues; /**< Maximum number of RX queues. */
1539 uint16_t max_tx_queues; /**< Maximum number of TX queues. */
1540 uint32_t max_mac_addrs; /**< Maximum number of MAC addresses. */
1541 uint32_t max_hash_mac_addrs;
1542 /** Maximum number of hash MAC addresses for MTA and UTA. */
1543 uint16_t max_vfs; /**< Maximum number of VFs. */
1544 uint16_t max_vmdq_pools; /**< Maximum number of VMDq pools. */
1545 struct rte_eth_rxseg_capa rx_seg_capa; /**< Segmentation capability.*/
1546 uint64_t rx_offload_capa;
1547 /**< All RX offload capabilities including all per-queue ones */
1548 uint64_t tx_offload_capa;
1549 /**< All TX offload capabilities including all per-queue ones */
1550 uint64_t rx_queue_offload_capa;
1551 /**< Device per-queue RX offload capabilities. */
1552 uint64_t tx_queue_offload_capa;
1553 /**< Device per-queue TX offload capabilities. */
1555 /**< Device redirection table size, the total number of entries. */
1556 uint8_t hash_key_size; /**< Hash key size in bytes */
1557 /** Bit mask of RSS offloads, the bit offset also means flow type */
1558 uint64_t flow_type_rss_offloads;
1559 struct rte_eth_rxconf default_rxconf; /**< Default RX configuration */
1560 struct rte_eth_txconf default_txconf; /**< Default TX configuration */
1561 uint16_t vmdq_queue_base; /**< First queue ID for VMDQ pools. */
1562 uint16_t vmdq_queue_num; /**< Queue number for VMDQ pools. */
1563 uint16_t vmdq_pool_base; /**< First ID of VMDQ pools. */
1564 struct rte_eth_desc_lim rx_desc_lim; /**< RX descriptors limits */
1565 struct rte_eth_desc_lim tx_desc_lim; /**< TX descriptors limits */
1566 uint32_t speed_capa; /**< Supported speeds bitmap (ETH_LINK_SPEED_). */
1567 /** Configured number of rx/tx queues */
1568 uint16_t nb_rx_queues; /**< Number of RX queues. */
1569 uint16_t nb_tx_queues; /**< Number of TX queues. */
1570 /** Rx parameter recommendations */
1571 struct rte_eth_dev_portconf default_rxportconf;
1572 /** Tx parameter recommendations */
1573 struct rte_eth_dev_portconf default_txportconf;
1574 /** Generic device capabilities (RTE_ETH_DEV_CAPA_). */
1577 * Switching information for ports on a device with a
1578 * embedded managed interconnect/switch.
1580 struct rte_eth_switch_info switch_info;
1582 uint64_t reserved_64s[2]; /**< Reserved for future fields */
1583 void *reserved_ptrs[2]; /**< Reserved for future fields */
1586 /**@{@name Rx/Tx queue states */
1587 #define RTE_ETH_QUEUE_STATE_STOPPED 0 /**< Queue stopped. */
1588 #define RTE_ETH_QUEUE_STATE_STARTED 1 /**< Queue started. */
1589 #define RTE_ETH_QUEUE_STATE_HAIRPIN 2 /**< Queue used for hairpin. */
1593 * Ethernet device RX queue information structure.
1594 * Used to retrieve information about configured queue.
1596 struct rte_eth_rxq_info {
1597 struct rte_mempool *mp; /**< mempool used by that queue. */
1598 struct rte_eth_rxconf conf; /**< queue config parameters. */
1599 uint8_t scattered_rx; /**< scattered packets RX supported. */
1600 uint8_t queue_state; /**< one of RTE_ETH_QUEUE_STATE_*. */
1601 uint16_t nb_desc; /**< configured number of RXDs. */
1602 uint16_t rx_buf_size; /**< hardware receive buffer size. */
1603 } __rte_cache_min_aligned;
1606 * Ethernet device TX queue information structure.
1607 * Used to retrieve information about configured queue.
1609 struct rte_eth_txq_info {
1610 struct rte_eth_txconf conf; /**< queue config parameters. */
1611 uint16_t nb_desc; /**< configured number of TXDs. */
1612 uint8_t queue_state; /**< one of RTE_ETH_QUEUE_STATE_*. */
1613 } __rte_cache_min_aligned;
1615 /* Generic Burst mode flag definition, values can be ORed. */
1618 * If the queues have different burst mode description, this bit will be set
1619 * by PMD, then the application can iterate to retrieve burst description for
1622 #define RTE_ETH_BURST_FLAG_PER_QUEUE (1ULL << 0)
1625 * Ethernet device RX/TX queue packet burst mode information structure.
1626 * Used to retrieve information about packet burst mode setting.
1628 struct rte_eth_burst_mode {
1629 uint64_t flags; /**< The ORed values of RTE_ETH_BURST_FLAG_xxx */
1631 #define RTE_ETH_BURST_MODE_INFO_SIZE 1024 /**< Maximum size for information */
1632 char info[RTE_ETH_BURST_MODE_INFO_SIZE]; /**< burst mode information */
1635 /** Maximum name length for extended statistics counters */
1636 #define RTE_ETH_XSTATS_NAME_SIZE 64
1639 * An Ethernet device extended statistic structure
1641 * This structure is used by rte_eth_xstats_get() to provide
1642 * statistics that are not provided in the generic *rte_eth_stats*
1644 * It maps a name id, corresponding to an index in the array returned
1645 * by rte_eth_xstats_get_names(), to a statistic value.
1647 struct rte_eth_xstat {
1648 uint64_t id; /**< The index in xstats name array. */
1649 uint64_t value; /**< The statistic counter value. */
1653 * A name element for extended statistics.
1655 * An array of this structure is returned by rte_eth_xstats_get_names().
1656 * It lists the names of extended statistics for a PMD. The *rte_eth_xstat*
1657 * structure references these names by their array index.
1659 * The xstats should follow a common naming scheme.
1660 * Some names are standardized in rte_stats_strings.
1662 * - rx_missed_errors
1664 * - tx_size_128_to_255_packets
1666 struct rte_eth_xstat_name {
1667 char name[RTE_ETH_XSTATS_NAME_SIZE]; /**< The statistic name. */
1670 #define ETH_DCB_NUM_TCS 8
1671 #define ETH_MAX_VMDQ_POOL 64
1674 * A structure used to get the information of queue and
1675 * TC mapping on both TX and RX paths.
1677 struct rte_eth_dcb_tc_queue_mapping {
1678 /** rx queues assigned to tc per Pool */
1682 } tc_rxq[ETH_MAX_VMDQ_POOL][ETH_DCB_NUM_TCS];
1683 /** rx queues assigned to tc per Pool */
1687 } tc_txq[ETH_MAX_VMDQ_POOL][ETH_DCB_NUM_TCS];
1691 * A structure used to get the information of DCB.
1692 * It includes TC UP mapping and queue TC mapping.
1694 struct rte_eth_dcb_info {
1695 uint8_t nb_tcs; /**< number of TCs */
1696 uint8_t prio_tc[ETH_DCB_NUM_USER_PRIORITIES]; /**< Priority to tc */
1697 uint8_t tc_bws[ETH_DCB_NUM_TCS]; /**< TX BW percentage for each TC */
1698 /** rx queues assigned to tc */
1699 struct rte_eth_dcb_tc_queue_mapping tc_queue;
1703 * This enum indicates the possible Forward Error Correction (FEC) modes
1704 * of an ethdev port.
1706 enum rte_eth_fec_mode {
1707 RTE_ETH_FEC_NOFEC = 0, /**< FEC is off */
1708 RTE_ETH_FEC_AUTO, /**< FEC autonegotiation modes */
1709 RTE_ETH_FEC_BASER, /**< FEC using common algorithm */
1710 RTE_ETH_FEC_RS, /**< FEC using RS algorithm */
1713 /* Translate from FEC mode to FEC capa */
1714 #define RTE_ETH_FEC_MODE_TO_CAPA(x) (1U << (x))
1716 /* This macro indicates FEC capa mask */
1717 #define RTE_ETH_FEC_MODE_CAPA_MASK(x) (1U << (RTE_ETH_FEC_ ## x))
1719 /* A structure used to get capabilities per link speed */
1720 struct rte_eth_fec_capa {
1721 uint32_t speed; /**< Link speed (see ETH_SPEED_NUM_*) */
1722 uint32_t capa; /**< FEC capabilities bitmask */
1725 #define RTE_ETH_ALL RTE_MAX_ETHPORTS
1727 /* Macros to check for valid port */
1728 #define RTE_ETH_VALID_PORTID_OR_ERR_RET(port_id, retval) do { \
1729 if (!rte_eth_dev_is_valid_port(port_id)) { \
1730 RTE_ETHDEV_LOG(ERR, "Invalid port_id=%u\n", port_id); \
1735 #define RTE_ETH_VALID_PORTID_OR_RET(port_id) do { \
1736 if (!rte_eth_dev_is_valid_port(port_id)) { \
1737 RTE_ETHDEV_LOG(ERR, "Invalid port_id=%u\n", port_id); \
1742 /**@{@name L2 tunnel configuration */
1743 /**< l2 tunnel enable mask */
1744 #define ETH_L2_TUNNEL_ENABLE_MASK 0x00000001
1745 /**< l2 tunnel insertion mask */
1746 #define ETH_L2_TUNNEL_INSERTION_MASK 0x00000002
1747 /**< l2 tunnel stripping mask */
1748 #define ETH_L2_TUNNEL_STRIPPING_MASK 0x00000004
1749 /**< l2 tunnel forwarding mask */
1750 #define ETH_L2_TUNNEL_FORWARDING_MASK 0x00000008
1754 * Function type used for RX packet processing packet callbacks.
1756 * The callback function is called on RX with a burst of packets that have
1757 * been received on the given port and queue.
1760 * The Ethernet port on which RX is being performed.
1762 * The queue on the Ethernet port which is being used to receive the packets.
1764 * The burst of packets that have just been received.
1766 * The number of packets in the burst pointed to by "pkts".
1768 * The max number of packets that can be stored in the "pkts" array.
1770 * The arbitrary user parameter passed in by the application when the callback
1771 * was originally configured.
1773 * The number of packets returned to the user.
1775 typedef uint16_t (*rte_rx_callback_fn)(uint16_t port_id, uint16_t queue,
1776 struct rte_mbuf *pkts[], uint16_t nb_pkts, uint16_t max_pkts,
1780 * Function type used for TX packet processing packet callbacks.
1782 * The callback function is called on TX with a burst of packets immediately
1783 * before the packets are put onto the hardware queue for transmission.
1786 * The Ethernet port on which TX is being performed.
1788 * The queue on the Ethernet port which is being used to transmit the packets.
1790 * The burst of packets that are about to be transmitted.
1792 * The number of packets in the burst pointed to by "pkts".
1794 * The arbitrary user parameter passed in by the application when the callback
1795 * was originally configured.
1797 * The number of packets to be written to the NIC.
1799 typedef uint16_t (*rte_tx_callback_fn)(uint16_t port_id, uint16_t queue,
1800 struct rte_mbuf *pkts[], uint16_t nb_pkts, void *user_param);
1803 * Possible states of an ethdev port.
1805 enum rte_eth_dev_state {
1806 /** Device is unused before being probed. */
1807 RTE_ETH_DEV_UNUSED = 0,
1808 /** Device is attached when allocated in probing. */
1809 RTE_ETH_DEV_ATTACHED,
1810 /** Device is in removed state when plug-out is detected. */
1811 RTE_ETH_DEV_REMOVED,
1814 struct rte_eth_dev_sriov {
1815 uint8_t active; /**< SRIOV is active with 16, 32 or 64 pools */
1816 uint8_t nb_q_per_pool; /**< rx queue number per pool */
1817 uint16_t def_vmdq_idx; /**< Default pool num used for PF */
1818 uint16_t def_pool_q_idx; /**< Default pool queue start reg index */
1820 #define RTE_ETH_DEV_SRIOV(dev) ((dev)->data->sriov)
1822 #define RTE_ETH_NAME_MAX_LEN RTE_DEV_NAME_MAX_LEN
1824 #define RTE_ETH_DEV_NO_OWNER 0
1826 #define RTE_ETH_MAX_OWNER_NAME_LEN 64
1828 struct rte_eth_dev_owner {
1829 uint64_t id; /**< The owner unique identifier. */
1830 char name[RTE_ETH_MAX_OWNER_NAME_LEN]; /**< The owner name. */
1833 /**@{@name Device flags
1834 * Flags internally saved in rte_eth_dev_data.dev_flags
1835 * and reported in rte_eth_dev_info.dev_flags.
1837 /** PMD supports thread-safe flow operations */
1838 #define RTE_ETH_DEV_FLOW_OPS_THREAD_SAFE 0x0001
1839 /** Device supports link state interrupt */
1840 #define RTE_ETH_DEV_INTR_LSC 0x0002
1841 /** Device is a bonded slave */
1842 #define RTE_ETH_DEV_BONDED_SLAVE 0x0004
1843 /** Device supports device removal interrupt */
1844 #define RTE_ETH_DEV_INTR_RMV 0x0008
1845 /** Device is port representor */
1846 #define RTE_ETH_DEV_REPRESENTOR 0x0010
1847 /** Device does not support MAC change after started */
1848 #define RTE_ETH_DEV_NOLIVE_MAC_ADDR 0x0020
1850 * Queue xstats filled automatically by ethdev layer.
1851 * PMDs filling the queue xstats themselves should not set this flag
1853 #define RTE_ETH_DEV_AUTOFILL_QUEUE_XSTATS 0x0040
1857 * Iterates over valid ethdev ports owned by a specific owner.
1860 * The id of the next possible valid owned port.
1862 * The owner identifier.
1863 * RTE_ETH_DEV_NO_OWNER means iterate over all valid ownerless ports.
1865 * Next valid port id owned by owner_id, RTE_MAX_ETHPORTS if there is none.
1867 uint64_t rte_eth_find_next_owned_by(uint16_t port_id,
1868 const uint64_t owner_id);
1871 * Macro to iterate over all enabled ethdev ports owned by a specific owner.
1873 #define RTE_ETH_FOREACH_DEV_OWNED_BY(p, o) \
1874 for (p = rte_eth_find_next_owned_by(0, o); \
1875 (unsigned int)p < (unsigned int)RTE_MAX_ETHPORTS; \
1876 p = rte_eth_find_next_owned_by(p + 1, o))
1879 * Iterates over valid ethdev ports.
1882 * The id of the next possible valid port.
1884 * Next valid port id, RTE_MAX_ETHPORTS if there is none.
1886 uint16_t rte_eth_find_next(uint16_t port_id);
1889 * Macro to iterate over all enabled and ownerless ethdev ports.
1891 #define RTE_ETH_FOREACH_DEV(p) \
1892 RTE_ETH_FOREACH_DEV_OWNED_BY(p, RTE_ETH_DEV_NO_OWNER)
1895 * Iterates over ethdev ports of a specified device.
1897 * @param port_id_start
1898 * The id of the next possible valid port.
1900 * The generic device behind the ports to iterate.
1902 * Next port id of the device, possibly port_id_start,
1903 * RTE_MAX_ETHPORTS if there is none.
1906 rte_eth_find_next_of(uint16_t port_id_start,
1907 const struct rte_device *parent);
1910 * Macro to iterate over all ethdev ports of a specified device.
1913 * The id of the matching port being iterated.
1915 * The rte_device pointer matching the iterated ports.
1917 #define RTE_ETH_FOREACH_DEV_OF(port_id, parent) \
1918 for (port_id = rte_eth_find_next_of(0, parent); \
1919 port_id < RTE_MAX_ETHPORTS; \
1920 port_id = rte_eth_find_next_of(port_id + 1, parent))
1923 * Iterates over sibling ethdev ports (i.e. sharing the same rte_device).
1925 * @param port_id_start
1926 * The id of the next possible valid sibling port.
1927 * @param ref_port_id
1928 * The id of a reference port to compare rte_device with.
1930 * Next sibling port id, possibly port_id_start or ref_port_id itself,
1931 * RTE_MAX_ETHPORTS if there is none.
1934 rte_eth_find_next_sibling(uint16_t port_id_start, uint16_t ref_port_id);
1937 * Macro to iterate over all ethdev ports sharing the same rte_device
1938 * as the specified port.
1939 * Note: the specified reference port is part of the loop iterations.
1942 * The id of the matching port being iterated.
1943 * @param ref_port_id
1944 * The id of the port being compared.
1946 #define RTE_ETH_FOREACH_DEV_SIBLING(port_id, ref_port_id) \
1947 for (port_id = rte_eth_find_next_sibling(0, ref_port_id); \
1948 port_id < RTE_MAX_ETHPORTS; \
1949 port_id = rte_eth_find_next_sibling(port_id + 1, ref_port_id))
1953 * @b EXPERIMENTAL: this API may change without prior notice.
1955 * Get a new unique owner identifier.
1956 * An owner identifier is used to owns Ethernet devices by only one DPDK entity
1957 * to avoid multiple management of device by different entities.
1960 * Owner identifier pointer.
1962 * Negative errno value on error, 0 on success.
1965 int rte_eth_dev_owner_new(uint64_t *owner_id);
1969 * @b EXPERIMENTAL: this API may change without prior notice.
1971 * Set an Ethernet device owner.
1974 * The identifier of the port to own.
1976 * The owner pointer.
1978 * Negative errno value on error, 0 on success.
1981 int rte_eth_dev_owner_set(const uint16_t port_id,
1982 const struct rte_eth_dev_owner *owner);
1986 * @b EXPERIMENTAL: this API may change without prior notice.
1988 * Unset Ethernet device owner to make the device ownerless.
1991 * The identifier of port to make ownerless.
1993 * The owner identifier.
1995 * 0 on success, negative errno value on error.
1998 int rte_eth_dev_owner_unset(const uint16_t port_id,
1999 const uint64_t owner_id);
2003 * @b EXPERIMENTAL: this API may change without prior notice.
2005 * Remove owner from all Ethernet devices owned by a specific owner.
2008 * The owner identifier.
2010 * 0 on success, negative errno value on error.
2013 int rte_eth_dev_owner_delete(const uint64_t owner_id);
2017 * @b EXPERIMENTAL: this API may change without prior notice.
2019 * Get the owner of an Ethernet device.
2022 * The port identifier.
2024 * The owner structure pointer to fill.
2026 * 0 on success, negative errno value on error..
2029 int rte_eth_dev_owner_get(const uint16_t port_id,
2030 struct rte_eth_dev_owner *owner);
2033 * Get the number of ports which are usable for the application.
2035 * These devices must be iterated by using the macro
2036 * ``RTE_ETH_FOREACH_DEV`` or ``RTE_ETH_FOREACH_DEV_OWNED_BY``
2037 * to deal with non-contiguous ranges of devices.
2040 * The count of available Ethernet devices.
2042 uint16_t rte_eth_dev_count_avail(void);
2045 * Get the total number of ports which are allocated.
2047 * Some devices may not be available for the application.
2050 * The total count of Ethernet devices.
2052 uint16_t rte_eth_dev_count_total(void);
2055 * Convert a numerical speed in Mbps to a bitmap flag that can be used in
2056 * the bitmap link_speeds of the struct rte_eth_conf
2059 * Numerical speed value in Mbps
2061 * ETH_LINK_[HALF/FULL]_DUPLEX (only for 10/100M speeds)
2063 * 0 if the speed cannot be mapped
2065 uint32_t rte_eth_speed_bitflag(uint32_t speed, int duplex);
2068 * Get DEV_RX_OFFLOAD_* flag name.
2073 * Offload name or 'UNKNOWN' if the flag cannot be recognised.
2075 const char *rte_eth_dev_rx_offload_name(uint64_t offload);
2078 * Get DEV_TX_OFFLOAD_* flag name.
2083 * Offload name or 'UNKNOWN' if the flag cannot be recognised.
2085 const char *rte_eth_dev_tx_offload_name(uint64_t offload);
2088 * Configure an Ethernet device.
2089 * This function must be invoked first before any other function in the
2090 * Ethernet API. This function can also be re-invoked when a device is in the
2094 * The port identifier of the Ethernet device to configure.
2095 * @param nb_rx_queue
2096 * The number of receive queues to set up for the Ethernet device.
2097 * @param nb_tx_queue
2098 * The number of transmit queues to set up for the Ethernet device.
2100 * The pointer to the configuration data to be used for the Ethernet device.
2101 * The *rte_eth_conf* structure includes:
2102 * - the hardware offload features to activate, with dedicated fields for
2103 * each statically configurable offload hardware feature provided by
2104 * Ethernet devices, such as IP checksum or VLAN tag stripping for
2106 * The Rx offload bitfield API is obsolete and will be deprecated.
2107 * Applications should set the ignore_bitfield_offloads bit on *rxmode*
2108 * structure and use offloads field to set per-port offloads instead.
2109 * - Any offloading set in eth_conf->[rt]xmode.offloads must be within
2110 * the [rt]x_offload_capa returned from rte_eth_dev_info_get().
2111 * Any type of device supported offloading set in the input argument
2112 * eth_conf->[rt]xmode.offloads to rte_eth_dev_configure() is enabled
2113 * on all queues and it can't be disabled in rte_eth_[rt]x_queue_setup()
2114 * - the Receive Side Scaling (RSS) configuration when using multiple RX
2115 * queues per port. Any RSS hash function set in eth_conf->rss_conf.rss_hf
2116 * must be within the flow_type_rss_offloads provided by drivers via
2117 * rte_eth_dev_info_get() API.
2119 * Embedding all configuration information in a single data structure
2120 * is the more flexible method that allows the addition of new features
2121 * without changing the syntax of the API.
2123 * - 0: Success, device configured.
2124 * - <0: Error code returned by the driver configuration function.
2126 int rte_eth_dev_configure(uint16_t port_id, uint16_t nb_rx_queue,
2127 uint16_t nb_tx_queue, const struct rte_eth_conf *eth_conf);
2131 * @b EXPERIMENTAL: this API may change without prior notice.
2133 * Check if an Ethernet device was physically removed.
2136 * The port identifier of the Ethernet device.
2138 * 1 when the Ethernet device is removed, otherwise 0.
2142 rte_eth_dev_is_removed(uint16_t port_id);
2145 * Allocate and set up a receive queue for an Ethernet device.
2147 * The function allocates a contiguous block of memory for *nb_rx_desc*
2148 * receive descriptors from a memory zone associated with *socket_id*
2149 * and initializes each receive descriptor with a network buffer allocated
2150 * from the memory pool *mb_pool*.
2153 * The port identifier of the Ethernet device.
2154 * @param rx_queue_id
2155 * The index of the receive queue to set up.
2156 * The value must be in the range [0, nb_rx_queue - 1] previously supplied
2157 * to rte_eth_dev_configure().
2159 * The number of receive descriptors to allocate for the receive ring.
2161 * The *socket_id* argument is the socket identifier in case of NUMA.
2162 * The value can be *SOCKET_ID_ANY* if there is no NUMA constraint for
2163 * the DMA memory allocated for the receive descriptors of the ring.
2165 * The pointer to the configuration data to be used for the receive queue.
2166 * NULL value is allowed, in which case default RX configuration
2168 * The *rx_conf* structure contains an *rx_thresh* structure with the values
2169 * of the Prefetch, Host, and Write-Back threshold registers of the receive
2171 * In addition it contains the hardware offloads features to activate using
2172 * the DEV_RX_OFFLOAD_* flags.
2173 * If an offloading set in rx_conf->offloads
2174 * hasn't been set in the input argument eth_conf->rxmode.offloads
2175 * to rte_eth_dev_configure(), it is a new added offloading, it must be
2176 * per-queue type and it is enabled for the queue.
2177 * No need to repeat any bit in rx_conf->offloads which has already been
2178 * enabled in rte_eth_dev_configure() at port level. An offloading enabled
2179 * at port level can't be disabled at queue level.
2180 * The configuration structure also contains the pointer to the array
2181 * of the receiving buffer segment descriptions, see rx_seg and rx_nseg
2182 * fields, this extended configuration might be used by split offloads like
2183 * RTE_ETH_RX_OFFLOAD_BUFFER_SPLIT. If mb_pool is not NULL,
2184 * the extended configuration fields must be set to NULL and zero.
2186 * The pointer to the memory pool from which to allocate *rte_mbuf* network
2187 * memory buffers to populate each descriptor of the receive ring. There are
2188 * two options to provide Rx buffer configuration:
2190 * mb_pool is not NULL, rx_conf.rx_nseg is 0.
2191 * - multiple segments description:
2192 * mb_pool is NULL, rx_conf.rx_seg is not NULL, rx_conf.rx_nseg is not 0.
2193 * Taken only if flag RTE_ETH_RX_OFFLOAD_BUFFER_SPLIT is set in offloads.
2196 * - 0: Success, receive queue correctly set up.
2197 * - -EIO: if device is removed.
2198 * - -ENODEV: if *port_id* is invalid.
2199 * - -EINVAL: The memory pool pointer is null or the size of network buffers
2200 * which can be allocated from this memory pool does not fit the various
2201 * buffer sizes allowed by the device controller.
2202 * - -ENOMEM: Unable to allocate the receive ring descriptors or to
2203 * allocate network memory buffers from the memory pool when
2204 * initializing receive descriptors.
2206 int rte_eth_rx_queue_setup(uint16_t port_id, uint16_t rx_queue_id,
2207 uint16_t nb_rx_desc, unsigned int socket_id,
2208 const struct rte_eth_rxconf *rx_conf,
2209 struct rte_mempool *mb_pool);
2213 * @b EXPERIMENTAL: this API may change, or be removed, without prior notice
2215 * Allocate and set up a hairpin receive queue for an Ethernet device.
2217 * The function set up the selected queue to be used in hairpin.
2220 * The port identifier of the Ethernet device.
2221 * @param rx_queue_id
2222 * The index of the receive queue to set up.
2223 * The value must be in the range [0, nb_rx_queue - 1] previously supplied
2224 * to rte_eth_dev_configure().
2226 * The number of receive descriptors to allocate for the receive ring.
2227 * 0 means the PMD will use default value.
2229 * The pointer to the hairpin configuration.
2232 * - (0) if successful.
2233 * - (-ENODEV) if *port_id* is invalid.
2234 * - (-ENOTSUP) if hardware doesn't support.
2235 * - (-EINVAL) if bad parameter.
2236 * - (-ENOMEM) if unable to allocate the resources.
2239 int rte_eth_rx_hairpin_queue_setup
2240 (uint16_t port_id, uint16_t rx_queue_id, uint16_t nb_rx_desc,
2241 const struct rte_eth_hairpin_conf *conf);
2244 * Allocate and set up a transmit queue for an Ethernet device.
2247 * The port identifier of the Ethernet device.
2248 * @param tx_queue_id
2249 * The index of the transmit queue to set up.
2250 * The value must be in the range [0, nb_tx_queue - 1] previously supplied
2251 * to rte_eth_dev_configure().
2253 * The number of transmit descriptors to allocate for the transmit ring.
2255 * The *socket_id* argument is the socket identifier in case of NUMA.
2256 * Its value can be *SOCKET_ID_ANY* if there is no NUMA constraint for
2257 * the DMA memory allocated for the transmit descriptors of the ring.
2259 * The pointer to the configuration data to be used for the transmit queue.
2260 * NULL value is allowed, in which case default TX configuration
2262 * The *tx_conf* structure contains the following data:
2263 * - The *tx_thresh* structure with the values of the Prefetch, Host, and
2264 * Write-Back threshold registers of the transmit ring.
2265 * When setting Write-Back threshold to the value greater then zero,
2266 * *tx_rs_thresh* value should be explicitly set to one.
2267 * - The *tx_free_thresh* value indicates the [minimum] number of network
2268 * buffers that must be pending in the transmit ring to trigger their
2269 * [implicit] freeing by the driver transmit function.
2270 * - The *tx_rs_thresh* value indicates the [minimum] number of transmit
2271 * descriptors that must be pending in the transmit ring before setting the
2272 * RS bit on a descriptor by the driver transmit function.
2273 * The *tx_rs_thresh* value should be less or equal then
2274 * *tx_free_thresh* value, and both of them should be less then
2276 * - The *offloads* member contains Tx offloads to be enabled.
2277 * If an offloading set in tx_conf->offloads
2278 * hasn't been set in the input argument eth_conf->txmode.offloads
2279 * to rte_eth_dev_configure(), it is a new added offloading, it must be
2280 * per-queue type and it is enabled for the queue.
2281 * No need to repeat any bit in tx_conf->offloads which has already been
2282 * enabled in rte_eth_dev_configure() at port level. An offloading enabled
2283 * at port level can't be disabled at queue level.
2285 * Note that setting *tx_free_thresh* or *tx_rs_thresh* value to 0 forces
2286 * the transmit function to use default values.
2288 * - 0: Success, the transmit queue is correctly set up.
2289 * - -ENOMEM: Unable to allocate the transmit ring descriptors.
2291 int rte_eth_tx_queue_setup(uint16_t port_id, uint16_t tx_queue_id,
2292 uint16_t nb_tx_desc, unsigned int socket_id,
2293 const struct rte_eth_txconf *tx_conf);
2297 * @b EXPERIMENTAL: this API may change, or be removed, without prior notice
2299 * Allocate and set up a transmit hairpin queue for an Ethernet device.
2302 * The port identifier of the Ethernet device.
2303 * @param tx_queue_id
2304 * The index of the transmit queue to set up.
2305 * The value must be in the range [0, nb_tx_queue - 1] previously supplied
2306 * to rte_eth_dev_configure().
2308 * The number of transmit descriptors to allocate for the transmit ring.
2309 * 0 to set default PMD value.
2311 * The hairpin configuration.
2314 * - (0) if successful.
2315 * - (-ENODEV) if *port_id* is invalid.
2316 * - (-ENOTSUP) if hardware doesn't support.
2317 * - (-EINVAL) if bad parameter.
2318 * - (-ENOMEM) if unable to allocate the resources.
2321 int rte_eth_tx_hairpin_queue_setup
2322 (uint16_t port_id, uint16_t tx_queue_id, uint16_t nb_tx_desc,
2323 const struct rte_eth_hairpin_conf *conf);
2327 * @b EXPERIMENTAL: this API may change, or be removed, without prior notice
2329 * Get all the hairpin peer Rx / Tx ports of the current port.
2330 * The caller should ensure that the array is large enough to save the ports
2334 * The port identifier of the Ethernet device.
2336 * Pointer to the array to store the peer ports list.
2338 * Length of the array to store the port identifiers.
2340 * Current port to peer port direction
2341 * positive - current used as Tx to get all peer Rx ports.
2342 * zero - current used as Rx to get all peer Tx ports.
2345 * - (0 or positive) actual peer ports number.
2346 * - (-EINVAL) if bad parameter.
2347 * - (-ENODEV) if *port_id* invalid
2348 * - (-ENOTSUP) if hardware doesn't support.
2349 * - Others detailed errors from PMD drivers.
2352 int rte_eth_hairpin_get_peer_ports(uint16_t port_id, uint16_t *peer_ports,
2353 size_t len, uint32_t direction);
2357 * @b EXPERIMENTAL: this API may change, or be removed, without prior notice
2359 * Bind all hairpin Tx queues of one port to the Rx queues of the peer port.
2360 * It is only allowed to call this function after all hairpin queues are
2361 * configured properly and the devices are in started state.
2364 * The identifier of the Tx port.
2366 * The identifier of peer Rx port.
2367 * RTE_MAX_ETHPORTS is allowed for the traversal of all devices.
2368 * Rx port ID could have the same value as Tx port ID.
2371 * - (0) if successful.
2372 * - (-ENODEV) if Tx port ID is invalid.
2373 * - (-EBUSY) if device is not in started state.
2374 * - (-ENOTSUP) if hardware doesn't support.
2375 * - Others detailed errors from PMD drivers.
2378 int rte_eth_hairpin_bind(uint16_t tx_port, uint16_t rx_port);
2382 * @b EXPERIMENTAL: this API may change, or be removed, without prior notice
2384 * Unbind all hairpin Tx queues of one port from the Rx queues of the peer port.
2385 * This should be called before closing the Tx or Rx devices, if the bind
2386 * function is called before.
2387 * After unbinding the hairpin ports pair, it is allowed to bind them again.
2388 * Changing queues configuration should be after stopping the device(s).
2391 * The identifier of the Tx port.
2393 * The identifier of peer Rx port.
2394 * RTE_MAX_ETHPORTS is allowed for traversal of all devices.
2395 * Rx port ID could have the same value as Tx port ID.
2398 * - (0) if successful.
2399 * - (-ENODEV) if Tx port ID is invalid.
2400 * - (-EBUSY) if device is in stopped state.
2401 * - (-ENOTSUP) if hardware doesn't support.
2402 * - Others detailed errors from PMD drivers.
2405 int rte_eth_hairpin_unbind(uint16_t tx_port, uint16_t rx_port);
2408 * Return the NUMA socket to which an Ethernet device is connected
2411 * The port identifier of the Ethernet device
2413 * The NUMA socket id to which the Ethernet device is connected or
2414 * a default of zero if the socket could not be determined.
2415 * -1 is returned is the port_id value is out of range.
2417 int rte_eth_dev_socket_id(uint16_t port_id);
2420 * Check if port_id of device is attached
2423 * The port identifier of the Ethernet device
2425 * - 0 if port is out of range or not attached
2426 * - 1 if device is attached
2428 int rte_eth_dev_is_valid_port(uint16_t port_id);
2431 * Start specified RX queue of a port. It is used when rx_deferred_start
2432 * flag of the specified queue is true.
2435 * The port identifier of the Ethernet device
2436 * @param rx_queue_id
2437 * The index of the rx queue to update the ring.
2438 * The value must be in the range [0, nb_rx_queue - 1] previously supplied
2439 * to rte_eth_dev_configure().
2441 * - 0: Success, the receive queue is started.
2442 * - -ENODEV: if *port_id* is invalid.
2443 * - -EINVAL: The queue_id out of range or belong to hairpin.
2444 * - -EIO: if device is removed.
2445 * - -ENOTSUP: The function not supported in PMD driver.
2447 int rte_eth_dev_rx_queue_start(uint16_t port_id, uint16_t rx_queue_id);
2450 * Stop specified RX queue of a port
2453 * The port identifier of the Ethernet device
2454 * @param rx_queue_id
2455 * The index of the rx queue to update the ring.
2456 * The value must be in the range [0, nb_rx_queue - 1] previously supplied
2457 * to rte_eth_dev_configure().
2459 * - 0: Success, the receive queue is stopped.
2460 * - -ENODEV: if *port_id* is invalid.
2461 * - -EINVAL: The queue_id out of range or belong to hairpin.
2462 * - -EIO: if device is removed.
2463 * - -ENOTSUP: The function not supported in PMD driver.
2465 int rte_eth_dev_rx_queue_stop(uint16_t port_id, uint16_t rx_queue_id);
2468 * Start TX for specified queue of a port. It is used when tx_deferred_start
2469 * flag of the specified queue is true.
2472 * The port identifier of the Ethernet device
2473 * @param tx_queue_id
2474 * The index of the tx queue to update the ring.
2475 * The value must be in the range [0, nb_tx_queue - 1] previously supplied
2476 * to rte_eth_dev_configure().
2478 * - 0: Success, the transmit queue is started.
2479 * - -ENODEV: if *port_id* is invalid.
2480 * - -EINVAL: The queue_id out of range or belong to hairpin.
2481 * - -EIO: if device is removed.
2482 * - -ENOTSUP: The function not supported in PMD driver.
2484 int rte_eth_dev_tx_queue_start(uint16_t port_id, uint16_t tx_queue_id);
2487 * Stop specified TX queue of a port
2490 * The port identifier of the Ethernet device
2491 * @param tx_queue_id
2492 * The index of the tx queue to update the ring.
2493 * The value must be in the range [0, nb_tx_queue - 1] previously supplied
2494 * to rte_eth_dev_configure().
2496 * - 0: Success, the transmit queue is stopped.
2497 * - -ENODEV: if *port_id* is invalid.
2498 * - -EINVAL: The queue_id out of range or belong to hairpin.
2499 * - -EIO: if device is removed.
2500 * - -ENOTSUP: The function not supported in PMD driver.
2502 int rte_eth_dev_tx_queue_stop(uint16_t port_id, uint16_t tx_queue_id);
2505 * Start an Ethernet device.
2507 * The device start step is the last one and consists of setting the configured
2508 * offload features and in starting the transmit and the receive units of the
2511 * Device RTE_ETH_DEV_NOLIVE_MAC_ADDR flag causes MAC address to be set before
2512 * PMD port start callback function is invoked.
2514 * On success, all basic functions exported by the Ethernet API (link status,
2515 * receive/transmit, and so on) can be invoked.
2518 * The port identifier of the Ethernet device.
2520 * - 0: Success, Ethernet device started.
2521 * - <0: Error code of the driver device start function.
2523 int rte_eth_dev_start(uint16_t port_id);
2526 * Stop an Ethernet device. The device can be restarted with a call to
2527 * rte_eth_dev_start()
2530 * The port identifier of the Ethernet device.
2532 * - 0: Success, Ethernet device stopped.
2533 * - <0: Error code of the driver device stop function.
2535 int rte_eth_dev_stop(uint16_t port_id);
2538 * Link up an Ethernet device.
2540 * Set device link up will re-enable the device rx/tx
2541 * functionality after it is previously set device linked down.
2544 * The port identifier of the Ethernet device.
2546 * - 0: Success, Ethernet device linked up.
2547 * - <0: Error code of the driver device link up function.
2549 int rte_eth_dev_set_link_up(uint16_t port_id);
2552 * Link down an Ethernet device.
2553 * The device rx/tx functionality will be disabled if success,
2554 * and it can be re-enabled with a call to
2555 * rte_eth_dev_set_link_up()
2558 * The port identifier of the Ethernet device.
2560 int rte_eth_dev_set_link_down(uint16_t port_id);
2563 * Close a stopped Ethernet device. The device cannot be restarted!
2564 * The function frees all port resources.
2567 * The port identifier of the Ethernet device.
2569 * - Zero if the port is closed successfully.
2570 * - Negative if something went wrong.
2572 int rte_eth_dev_close(uint16_t port_id);
2575 * Reset a Ethernet device and keep its port id.
2577 * When a port has to be reset passively, the DPDK application can invoke
2578 * this function. For example when a PF is reset, all its VFs should also
2579 * be reset. Normally a DPDK application can invoke this function when
2580 * RTE_ETH_EVENT_INTR_RESET event is detected, but can also use it to start
2581 * a port reset in other circumstances.
2583 * When this function is called, it first stops the port and then calls the
2584 * PMD specific dev_uninit( ) and dev_init( ) to return the port to initial
2585 * state, in which no Tx and Rx queues are setup, as if the port has been
2586 * reset and not started. The port keeps the port id it had before the
2589 * After calling rte_eth_dev_reset( ), the application should use
2590 * rte_eth_dev_configure( ), rte_eth_rx_queue_setup( ),
2591 * rte_eth_tx_queue_setup( ), and rte_eth_dev_start( )
2592 * to reconfigure the device as appropriate.
2594 * Note: To avoid unexpected behavior, the application should stop calling
2595 * Tx and Rx functions before calling rte_eth_dev_reset( ). For thread
2596 * safety, all these controlling functions should be called from the same
2600 * The port identifier of the Ethernet device.
2603 * - (0) if successful.
2604 * - (-ENODEV) if *port_id* is invalid.
2605 * - (-ENOTSUP) if hardware doesn't support this function.
2606 * - (-EPERM) if not ran from the primary process.
2607 * - (-EIO) if re-initialisation failed or device is removed.
2608 * - (-ENOMEM) if the reset failed due to OOM.
2609 * - (-EAGAIN) if the reset temporarily failed and should be retried later.
2611 int rte_eth_dev_reset(uint16_t port_id);
2614 * Enable receipt in promiscuous mode for an Ethernet device.
2617 * The port identifier of the Ethernet device.
2619 * - (0) if successful.
2620 * - (-ENOTSUP) if support for promiscuous_enable() does not exist
2622 * - (-ENODEV) if *port_id* invalid.
2624 int rte_eth_promiscuous_enable(uint16_t port_id);
2627 * Disable receipt in promiscuous mode for an Ethernet device.
2630 * The port identifier of the Ethernet device.
2632 * - (0) if successful.
2633 * - (-ENOTSUP) if support for promiscuous_disable() does not exist
2635 * - (-ENODEV) if *port_id* invalid.
2637 int rte_eth_promiscuous_disable(uint16_t port_id);
2640 * Return the value of promiscuous mode for an Ethernet device.
2643 * The port identifier of the Ethernet device.
2645 * - (1) if promiscuous is enabled
2646 * - (0) if promiscuous is disabled.
2649 int rte_eth_promiscuous_get(uint16_t port_id);
2652 * Enable the receipt of any multicast frame by an Ethernet device.
2655 * The port identifier of the Ethernet device.
2657 * - (0) if successful.
2658 * - (-ENOTSUP) if support for allmulticast_enable() does not exist
2660 * - (-ENODEV) if *port_id* invalid.
2662 int rte_eth_allmulticast_enable(uint16_t port_id);
2665 * Disable the receipt of all multicast frames by an Ethernet device.
2668 * The port identifier of the Ethernet device.
2670 * - (0) if successful.
2671 * - (-ENOTSUP) if support for allmulticast_disable() does not exist
2673 * - (-ENODEV) if *port_id* invalid.
2675 int rte_eth_allmulticast_disable(uint16_t port_id);
2678 * Return the value of allmulticast mode for an Ethernet device.
2681 * The port identifier of the Ethernet device.
2683 * - (1) if allmulticast is enabled
2684 * - (0) if allmulticast is disabled.
2687 int rte_eth_allmulticast_get(uint16_t port_id);
2690 * Retrieve the link status (up/down), the duplex mode (half/full),
2691 * the negotiation (auto/fixed), and if available, the speed (Mbps).
2693 * It might need to wait up to 9 seconds.
2694 * @see rte_eth_link_get_nowait.
2697 * The port identifier of the Ethernet device.
2699 * Link information written back.
2701 * - (0) if successful.
2702 * - (-ENOTSUP) if the function is not supported in PMD driver.
2703 * - (-ENODEV) if *port_id* invalid.
2704 * - (-EINVAL) if bad parameter.
2706 int rte_eth_link_get(uint16_t port_id, struct rte_eth_link *link);
2709 * Retrieve the link status (up/down), the duplex mode (half/full),
2710 * the negotiation (auto/fixed), and if available, the speed (Mbps).
2713 * The port identifier of the Ethernet device.
2715 * Link information written back.
2717 * - (0) if successful.
2718 * - (-ENOTSUP) if the function is not supported in PMD driver.
2719 * - (-ENODEV) if *port_id* invalid.
2720 * - (-EINVAL) if bad parameter.
2722 int rte_eth_link_get_nowait(uint16_t port_id, struct rte_eth_link *link);
2726 * @b EXPERIMENTAL: this API may change without prior notice.
2728 * The function converts a link_speed to a string. It handles all special
2729 * values like unknown or none speed.
2732 * link_speed of rte_eth_link struct
2734 * Link speed in textual format. It's pointer to immutable memory.
2735 * No free is required.
2738 const char *rte_eth_link_speed_to_str(uint32_t link_speed);
2742 * @b EXPERIMENTAL: this API may change without prior notice.
2744 * The function converts a rte_eth_link struct representing a link status to
2748 * A pointer to a string to be filled with textual representation of
2749 * device status. At least ETH_LINK_MAX_STR_LEN bytes should be allocated to
2750 * store default link status text.
2752 * Length of available memory at 'str' string.
2754 * Link status returned by rte_eth_link_get function
2756 * Number of bytes written to str array or -EINVAL if bad parameter.
2759 int rte_eth_link_to_str(char *str, size_t len,
2760 const struct rte_eth_link *eth_link);
2763 * Retrieve the general I/O statistics of an Ethernet device.
2766 * The port identifier of the Ethernet device.
2768 * A pointer to a structure of type *rte_eth_stats* to be filled with
2769 * the values of device counters for the following set of statistics:
2770 * - *ipackets* with the total of successfully received packets.
2771 * - *opackets* with the total of successfully transmitted packets.
2772 * - *ibytes* with the total of successfully received bytes.
2773 * - *obytes* with the total of successfully transmitted bytes.
2774 * - *ierrors* with the total of erroneous received packets.
2775 * - *oerrors* with the total of failed transmitted packets.
2777 * Zero if successful. Non-zero otherwise.
2779 int rte_eth_stats_get(uint16_t port_id, struct rte_eth_stats *stats);
2782 * Reset the general I/O statistics of an Ethernet device.
2785 * The port identifier of the Ethernet device.
2787 * - (0) if device notified to reset stats.
2788 * - (-ENOTSUP) if hardware doesn't support.
2789 * - (-ENODEV) if *port_id* invalid.
2790 * - (<0): Error code of the driver stats reset function.
2792 int rte_eth_stats_reset(uint16_t port_id);
2795 * Retrieve names of extended statistics of an Ethernet device.
2797 * There is an assumption that 'xstat_names' and 'xstats' arrays are matched
2799 * xstats_names[i].name => xstats[i].value
2801 * And the array index is same with id field of 'struct rte_eth_xstat':
2804 * This assumption makes key-value pair matching less flexible but simpler.
2807 * The port identifier of the Ethernet device.
2808 * @param xstats_names
2809 * An rte_eth_xstat_name array of at least *size* elements to
2810 * be filled. If set to NULL, the function returns the required number
2813 * The size of the xstats_names array (number of elements).
2815 * - A positive value lower or equal to size: success. The return value
2816 * is the number of entries filled in the stats table.
2817 * - A positive value higher than size: error, the given statistics table
2818 * is too small. The return value corresponds to the size that should
2819 * be given to succeed. The entries in the table are not valid and
2820 * shall not be used by the caller.
2821 * - A negative value on error (invalid port id).
2823 int rte_eth_xstats_get_names(uint16_t port_id,
2824 struct rte_eth_xstat_name *xstats_names,
2828 * Retrieve extended statistics of an Ethernet device.
2830 * There is an assumption that 'xstat_names' and 'xstats' arrays are matched
2832 * xstats_names[i].name => xstats[i].value
2834 * And the array index is same with id field of 'struct rte_eth_xstat':
2837 * This assumption makes key-value pair matching less flexible but simpler.
2840 * The port identifier of the Ethernet device.
2842 * A pointer to a table of structure of type *rte_eth_xstat*
2843 * to be filled with device statistics ids and values.
2844 * This parameter can be set to NULL if n is 0.
2846 * The size of the xstats array (number of elements).
2848 * - A positive value lower or equal to n: success. The return value
2849 * is the number of entries filled in the stats table.
2850 * - A positive value higher than n: error, the given statistics table
2851 * is too small. The return value corresponds to the size that should
2852 * be given to succeed. The entries in the table are not valid and
2853 * shall not be used by the caller.
2854 * - A negative value on error (invalid port id).
2856 int rte_eth_xstats_get(uint16_t port_id, struct rte_eth_xstat *xstats,
2860 * Retrieve names of extended statistics of an Ethernet device.
2863 * The port identifier of the Ethernet device.
2864 * @param xstats_names
2865 * Array to be filled in with names of requested device statistics.
2866 * Must not be NULL if @p ids are specified (not NULL).
2868 * Number of elements in @p xstats_names array (if not NULL) and in
2869 * @p ids array (if not NULL). Must be 0 if both array pointers are NULL.
2871 * IDs array given by app to retrieve specific statistics. May be NULL to
2872 * retrieve names of all available statistics or, if @p xstats_names is
2873 * NULL as well, just the number of available statistics.
2875 * - A positive value lower or equal to size: success. The return value
2876 * is the number of entries filled in the stats table.
2877 * - A positive value higher than size: success. The given statistics table
2878 * is too small. The return value corresponds to the size that should
2879 * be given to succeed. The entries in the table are not valid and
2880 * shall not be used by the caller.
2881 * - A negative value on error.
2884 rte_eth_xstats_get_names_by_id(uint16_t port_id,
2885 struct rte_eth_xstat_name *xstats_names, unsigned int size,
2889 * Retrieve extended statistics of an Ethernet device.
2892 * The port identifier of the Ethernet device.
2894 * IDs array given by app to retrieve specific statistics. May be NULL to
2895 * retrieve all available statistics or, if @p values is NULL as well,
2896 * just the number of available statistics.
2898 * Array to be filled in with requested device statistics.
2899 * Must not be NULL if ids are specified (not NULL).
2901 * Number of elements in @p values array (if not NULL) and in @p ids
2902 * array (if not NULL). Must be 0 if both array pointers are NULL.
2904 * - A positive value lower or equal to size: success. The return value
2905 * is the number of entries filled in the stats table.
2906 * - A positive value higher than size: success: The given statistics table
2907 * is too small. The return value corresponds to the size that should
2908 * be given to succeed. The entries in the table are not valid and
2909 * shall not be used by the caller.
2910 * - A negative value on error.
2912 int rte_eth_xstats_get_by_id(uint16_t port_id, const uint64_t *ids,
2913 uint64_t *values, unsigned int size);
2916 * Gets the ID of a statistic from its name.
2918 * This function searches for the statistics using string compares, and
2919 * as such should not be used on the fast-path. For fast-path retrieval of
2920 * specific statistics, store the ID as provided in *id* from this function,
2921 * and pass the ID to rte_eth_xstats_get()
2923 * @param port_id The port to look up statistics from
2924 * @param xstat_name The name of the statistic to return
2925 * @param[out] id A pointer to an app-supplied uint64_t which should be
2926 * set to the ID of the stat if the stat exists.
2929 * -ENODEV for invalid port_id,
2930 * -EIO if device is removed,
2931 * -EINVAL if the xstat_name doesn't exist in port_id
2932 * -ENOMEM if bad parameter.
2934 int rte_eth_xstats_get_id_by_name(uint16_t port_id, const char *xstat_name,
2938 * Reset extended statistics of an Ethernet device.
2941 * The port identifier of the Ethernet device.
2943 * - (0) if device notified to reset extended stats.
2944 * - (-ENOTSUP) if pmd doesn't support both
2945 * extended stats and basic stats reset.
2946 * - (-ENODEV) if *port_id* invalid.
2947 * - (<0): Error code of the driver xstats reset function.
2949 int rte_eth_xstats_reset(uint16_t port_id);
2952 * Set a mapping for the specified transmit queue to the specified per-queue
2953 * statistics counter.
2956 * The port identifier of the Ethernet device.
2957 * @param tx_queue_id
2958 * The index of the transmit queue for which a queue stats mapping is required.
2959 * The value must be in the range [0, nb_tx_queue - 1] previously supplied
2960 * to rte_eth_dev_configure().
2962 * The per-queue packet statistics functionality number that the transmit
2963 * queue is to be assigned.
2964 * The value must be in the range [0, RTE_ETHDEV_QUEUE_STAT_CNTRS - 1].
2965 * Max RTE_ETHDEV_QUEUE_STAT_CNTRS being 256.
2967 * Zero if successful. Non-zero otherwise.
2969 int rte_eth_dev_set_tx_queue_stats_mapping(uint16_t port_id,
2970 uint16_t tx_queue_id, uint8_t stat_idx);
2973 * Set a mapping for the specified receive queue to the specified per-queue
2974 * statistics counter.
2977 * The port identifier of the Ethernet device.
2978 * @param rx_queue_id
2979 * The index of the receive queue for which a queue stats mapping is required.
2980 * The value must be in the range [0, nb_rx_queue - 1] previously supplied
2981 * to rte_eth_dev_configure().
2983 * The per-queue packet statistics functionality number that the receive
2984 * queue is to be assigned.
2985 * The value must be in the range [0, RTE_ETHDEV_QUEUE_STAT_CNTRS - 1].
2986 * Max RTE_ETHDEV_QUEUE_STAT_CNTRS being 256.
2988 * Zero if successful. Non-zero otherwise.
2990 int rte_eth_dev_set_rx_queue_stats_mapping(uint16_t port_id,
2991 uint16_t rx_queue_id,
2995 * Retrieve the Ethernet address of an Ethernet device.
2998 * The port identifier of the Ethernet device.
3000 * A pointer to a structure of type *ether_addr* to be filled with
3001 * the Ethernet address of the Ethernet device.
3003 * - (0) if successful
3004 * - (-ENODEV) if *port_id* invalid.
3005 * - (-EINVAL) if bad parameter.
3007 int rte_eth_macaddr_get(uint16_t port_id, struct rte_ether_addr *mac_addr);
3011 * @b EXPERIMENTAL: this API may change without prior notice
3013 * Retrieve the Ethernet addresses of an Ethernet device.
3016 * The port identifier of the Ethernet device.
3018 * A pointer to an array of structures of type *ether_addr* to be filled with
3019 * the Ethernet addresses of the Ethernet device.
3021 * Number of elements in the @p ma array.
3022 * Note that rte_eth_dev_info::max_mac_addrs can be used to retrieve
3023 * max number of Ethernet addresses for given port.
3025 * - number of retrieved addresses if successful
3026 * - (-ENODEV) if *port_id* invalid.
3027 * - (-EINVAL) if bad parameter.
3030 int rte_eth_macaddrs_get(uint16_t port_id, struct rte_ether_addr *ma,
3034 * Retrieve the contextual information of an Ethernet device.
3036 * As part of this function, a number of of fields in dev_info will be
3037 * initialized as follows:
3042 * Where lim is defined within the rte_eth_dev_info_get as
3044 * const struct rte_eth_desc_lim lim = {
3045 * .nb_max = UINT16_MAX,
3048 * .nb_seg_max = UINT16_MAX,
3049 * .nb_mtu_seg_max = UINT16_MAX,
3052 * device = dev->device
3053 * min_mtu = RTE_ETHER_MIN_LEN - RTE_ETHER_HDR_LEN - RTE_ETHER_CRC_LEN
3054 * max_mtu = UINT16_MAX
3056 * The following fields will be populated if support for dev_infos_get()
3057 * exists for the device and the rte_eth_dev 'dev' has been populated
3058 * successfully with a call to it:
3060 * driver_name = dev->device->driver->name
3061 * nb_rx_queues = dev->data->nb_rx_queues
3062 * nb_tx_queues = dev->data->nb_tx_queues
3063 * dev_flags = &dev->data->dev_flags
3066 * The port identifier of the Ethernet device.
3068 * A pointer to a structure of type *rte_eth_dev_info* to be filled with
3069 * the contextual information of the Ethernet device.
3071 * - (0) if successful.
3072 * - (-ENOTSUP) if support for dev_infos_get() does not exist for the device.
3073 * - (-ENODEV) if *port_id* invalid.
3074 * - (-EINVAL) if bad parameter.
3076 int rte_eth_dev_info_get(uint16_t port_id, struct rte_eth_dev_info *dev_info);
3080 * @b EXPERIMENTAL: this API may change without prior notice.
3082 * Retrieve the configuration of an Ethernet device.
3085 * The port identifier of the Ethernet device.
3087 * Location for Ethernet device configuration to be filled in.
3089 * - (0) if successful.
3090 * - (-ENODEV) if *port_id* invalid.
3091 * - (-EINVAL) if bad parameter.
3094 int rte_eth_dev_conf_get(uint16_t port_id, struct rte_eth_conf *dev_conf);
3097 * Retrieve the firmware version of a device.
3100 * The port identifier of the device.
3102 * A pointer to a string array storing the firmware version of a device,
3103 * the string includes terminating null. This pointer is allocated by caller.
3105 * The size of the string array pointed by fw_version, which should be
3106 * large enough to store firmware version of the device.
3108 * - (0) if successful.
3109 * - (-ENOTSUP) if operation is not supported.
3110 * - (-ENODEV) if *port_id* invalid.
3111 * - (-EIO) if device is removed.
3112 * - (-EINVAL) if bad parameter.
3113 * - (>0) if *fw_size* is not enough to store firmware version, return
3114 * the size of the non truncated string.
3116 int rte_eth_dev_fw_version_get(uint16_t port_id,
3117 char *fw_version, size_t fw_size);
3120 * Retrieve the supported packet types of an Ethernet device.
3122 * When a packet type is announced as supported, it *must* be recognized by
3123 * the PMD. For instance, if RTE_PTYPE_L2_ETHER, RTE_PTYPE_L2_ETHER_VLAN
3124 * and RTE_PTYPE_L3_IPV4 are announced, the PMD must return the following
3125 * packet types for these packets:
3126 * - Ether/IPv4 -> RTE_PTYPE_L2_ETHER | RTE_PTYPE_L3_IPV4
3127 * - Ether/Vlan/IPv4 -> RTE_PTYPE_L2_ETHER_VLAN | RTE_PTYPE_L3_IPV4
3128 * - Ether/[anything else] -> RTE_PTYPE_L2_ETHER
3129 * - Ether/Vlan/[anything else] -> RTE_PTYPE_L2_ETHER_VLAN
3131 * When a packet is received by a PMD, the most precise type must be
3132 * returned among the ones supported. However a PMD is allowed to set
3133 * packet type that is not in the supported list, at the condition that it
3134 * is more precise. Therefore, a PMD announcing no supported packet types
3135 * can still set a matching packet type in a received packet.
3138 * Better to invoke this API after the device is already started or rx burst
3139 * function is decided, to obtain correct supported ptypes.
3141 * if a given PMD does not report what ptypes it supports, then the supported
3142 * ptype count is reported as 0.
3144 * The port identifier of the Ethernet device.
3146 * A hint of what kind of packet type which the caller is interested in.
3148 * An array pointer to store adequate packet types, allocated by caller.
3150 * Size of the array pointed by param ptypes.
3152 * - (>=0) Number of supported ptypes. If the number of types exceeds num,
3153 * only num entries will be filled into the ptypes array, but the full
3154 * count of supported ptypes will be returned.
3155 * - (-ENODEV) if *port_id* invalid.
3156 * - (-EINVAL) if bad parameter.
3158 int rte_eth_dev_get_supported_ptypes(uint16_t port_id, uint32_t ptype_mask,
3159 uint32_t *ptypes, int num);
3161 * Inform Ethernet device about reduced range of packet types to handle.
3163 * Application can use this function to set only specific ptypes that it's
3164 * interested. This information can be used by the PMD to optimize Rx path.
3166 * The function accepts an array `set_ptypes` allocated by the caller to
3167 * store the packet types set by the driver, the last element of the array
3168 * is set to RTE_PTYPE_UNKNOWN. The size of the `set_ptype` array should be
3169 * `rte_eth_dev_get_supported_ptypes() + 1` else it might only be filled
3173 * The port identifier of the Ethernet device.
3175 * The ptype family that application is interested in should be bitwise OR of
3176 * RTE_PTYPE_*_MASK or 0.
3178 * An array pointer to store set packet types, allocated by caller. The
3179 * function marks the end of array with RTE_PTYPE_UNKNOWN.
3181 * Size of the array pointed by param ptypes.
3182 * Should be rte_eth_dev_get_supported_ptypes() + 1 to accommodate the
3186 * - (-ENODEV) if *port_id* invalid.
3187 * - (-EINVAL) if *ptype_mask* is invalid (or) set_ptypes is NULL and
3190 int rte_eth_dev_set_ptypes(uint16_t port_id, uint32_t ptype_mask,
3191 uint32_t *set_ptypes, unsigned int num);
3194 * Retrieve the MTU of an Ethernet device.
3197 * The port identifier of the Ethernet device.
3199 * A pointer to a uint16_t where the retrieved MTU is to be stored.
3201 * - (0) if successful.
3202 * - (-ENODEV) if *port_id* invalid.
3203 * - (-EINVAL) if bad parameter.
3205 int rte_eth_dev_get_mtu(uint16_t port_id, uint16_t *mtu);
3208 * Change the MTU of an Ethernet device.
3211 * The port identifier of the Ethernet device.
3213 * A uint16_t for the MTU to be applied.
3215 * - (0) if successful.
3216 * - (-ENOTSUP) if operation is not supported.
3217 * - (-ENODEV) if *port_id* invalid.
3218 * - (-EIO) if device is removed.
3219 * - (-EINVAL) if *mtu* invalid, validation of mtu can occur within
3220 * rte_eth_dev_set_mtu if dev_infos_get is supported by the device or
3221 * when the mtu is set using dev->dev_ops->mtu_set.
3222 * - (-EBUSY) if operation is not allowed when the port is running
3224 int rte_eth_dev_set_mtu(uint16_t port_id, uint16_t mtu);
3227 * Enable/Disable hardware filtering by an Ethernet device of received
3228 * VLAN packets tagged with a given VLAN Tag Identifier.
3231 * The port identifier of the Ethernet device.
3233 * The VLAN Tag Identifier whose filtering must be enabled or disabled.
3235 * If > 0, enable VLAN filtering of VLAN packets tagged with *vlan_id*.
3236 * Otherwise, disable VLAN filtering of VLAN packets tagged with *vlan_id*.
3238 * - (0) if successful.
3239 * - (-ENOTSUP) if hardware-assisted VLAN filtering not configured.
3240 * - (-ENODEV) if *port_id* invalid.
3241 * - (-EIO) if device is removed.
3242 * - (-ENOSYS) if VLAN filtering on *port_id* disabled.
3243 * - (-EINVAL) if *vlan_id* > 4095.
3245 int rte_eth_dev_vlan_filter(uint16_t port_id, uint16_t vlan_id, int on);
3248 * Enable/Disable hardware VLAN Strip by a rx queue of an Ethernet device.
3251 * The port identifier of the Ethernet device.
3252 * @param rx_queue_id
3253 * The index of the receive queue for which a queue stats mapping is required.
3254 * The value must be in the range [0, nb_rx_queue - 1] previously supplied
3255 * to rte_eth_dev_configure().
3257 * If 1, Enable VLAN Stripping of the receive queue of the Ethernet port.
3258 * If 0, Disable VLAN Stripping of the receive queue of the Ethernet port.
3260 * - (0) if successful.
3261 * - (-ENOTSUP) if hardware-assisted VLAN stripping not configured.
3262 * - (-ENODEV) if *port_id* invalid.
3263 * - (-EINVAL) if *rx_queue_id* invalid.
3265 int rte_eth_dev_set_vlan_strip_on_queue(uint16_t port_id, uint16_t rx_queue_id,
3269 * Set the Outer VLAN Ether Type by an Ethernet device, it can be inserted to
3273 * The port identifier of the Ethernet device.
3277 * The Tag Protocol ID
3279 * - (0) if successful.
3280 * - (-ENOTSUP) if hardware-assisted VLAN TPID setup is not supported.
3281 * - (-ENODEV) if *port_id* invalid.
3282 * - (-EIO) if device is removed.
3284 int rte_eth_dev_set_vlan_ether_type(uint16_t port_id,
3285 enum rte_vlan_type vlan_type,
3289 * Set VLAN offload configuration on an Ethernet device.
3292 * The port identifier of the Ethernet device.
3293 * @param offload_mask
3294 * The VLAN Offload bit mask can be mixed use with "OR"
3295 * ETH_VLAN_STRIP_OFFLOAD
3296 * ETH_VLAN_FILTER_OFFLOAD
3297 * ETH_VLAN_EXTEND_OFFLOAD
3298 * ETH_QINQ_STRIP_OFFLOAD
3300 * - (0) if successful.
3301 * - (-ENOTSUP) if hardware-assisted VLAN filtering not configured.
3302 * - (-ENODEV) if *port_id* invalid.
3303 * - (-EIO) if device is removed.
3305 int rte_eth_dev_set_vlan_offload(uint16_t port_id, int offload_mask);
3308 * Read VLAN Offload configuration from an Ethernet device
3311 * The port identifier of the Ethernet device.
3313 * - (>0) if successful. Bit mask to indicate
3314 * ETH_VLAN_STRIP_OFFLOAD
3315 * ETH_VLAN_FILTER_OFFLOAD
3316 * ETH_VLAN_EXTEND_OFFLOAD
3317 * ETH_QINQ_STRIP_OFFLOAD
3318 * - (-ENODEV) if *port_id* invalid.
3320 int rte_eth_dev_get_vlan_offload(uint16_t port_id);
3323 * Set port based TX VLAN insertion on or off.
3326 * The port identifier of the Ethernet device.
3328 * Port based TX VLAN identifier together with user priority.
3330 * Turn on or off the port based TX VLAN insertion.
3333 * - (0) if successful.
3334 * - negative if failed.
3336 int rte_eth_dev_set_vlan_pvid(uint16_t port_id, uint16_t pvid, int on);
3338 typedef void (*buffer_tx_error_fn)(struct rte_mbuf **unsent, uint16_t count,
3342 * Structure used to buffer packets for future TX
3343 * Used by APIs rte_eth_tx_buffer and rte_eth_tx_buffer_flush
3345 struct rte_eth_dev_tx_buffer {
3346 buffer_tx_error_fn error_callback;
3347 void *error_userdata;
3348 uint16_t size; /**< Size of buffer for buffered tx */
3349 uint16_t length; /**< Number of packets in the array */
3350 struct rte_mbuf *pkts[];
3351 /**< Pending packets to be sent on explicit flush or when full */
3355 * Calculate the size of the tx buffer.
3358 * Number of stored packets.
3360 #define RTE_ETH_TX_BUFFER_SIZE(sz) \
3361 (sizeof(struct rte_eth_dev_tx_buffer) + (sz) * sizeof(struct rte_mbuf *))
3364 * Initialize default values for buffered transmitting
3367 * Tx buffer to be initialized.
3374 rte_eth_tx_buffer_init(struct rte_eth_dev_tx_buffer *buffer, uint16_t size);
3377 * Configure a callback for buffered packets which cannot be sent
3379 * Register a specific callback to be called when an attempt is made to send
3380 * all packets buffered on an ethernet port, but not all packets can
3381 * successfully be sent. The callback registered here will be called only
3382 * from calls to rte_eth_tx_buffer() and rte_eth_tx_buffer_flush() APIs.
3383 * The default callback configured for each queue by default just frees the
3384 * packets back to the calling mempool. If additional behaviour is required,
3385 * for example, to count dropped packets, or to retry transmission of packets
3386 * which cannot be sent, this function should be used to register a suitable
3387 * callback function to implement the desired behaviour.
3388 * The example callback "rte_eth_count_unsent_packet_callback()" is also
3389 * provided as reference.
3392 * The port identifier of the Ethernet device.
3394 * The function to be used as the callback.
3396 * Arbitrary parameter to be passed to the callback function
3398 * 0 on success, or -EINVAL if bad parameter
3401 rte_eth_tx_buffer_set_err_callback(struct rte_eth_dev_tx_buffer *buffer,
3402 buffer_tx_error_fn callback, void *userdata);
3405 * Callback function for silently dropping unsent buffered packets.
3407 * This function can be passed to rte_eth_tx_buffer_set_err_callback() to
3408 * adjust the default behavior when buffered packets cannot be sent. This
3409 * function drops any unsent packets silently and is used by tx buffered
3410 * operations as default behavior.
3412 * NOTE: this function should not be called directly, instead it should be used
3413 * as a callback for packet buffering.
3415 * NOTE: when configuring this function as a callback with
3416 * rte_eth_tx_buffer_set_err_callback(), the final, userdata parameter
3417 * should point to an uint64_t value.
3420 * The previously buffered packets which could not be sent
3422 * The number of unsent packets in the pkts array
3427 rte_eth_tx_buffer_drop_callback(struct rte_mbuf **pkts, uint16_t unsent,
3431 * Callback function for tracking unsent buffered packets.
3433 * This function can be passed to rte_eth_tx_buffer_set_err_callback() to
3434 * adjust the default behavior when buffered packets cannot be sent. This
3435 * function drops any unsent packets, but also updates a user-supplied counter
3436 * to track the overall number of packets dropped. The counter should be an
3437 * uint64_t variable.
3439 * NOTE: this function should not be called directly, instead it should be used
3440 * as a callback for packet buffering.
3442 * NOTE: when configuring this function as a callback with
3443 * rte_eth_tx_buffer_set_err_callback(), the final, userdata parameter
3444 * should point to an uint64_t value.
3447 * The previously buffered packets which could not be sent
3449 * The number of unsent packets in the pkts array
3451 * Pointer to an uint64_t value, which will be incremented by unsent
3454 rte_eth_tx_buffer_count_callback(struct rte_mbuf **pkts, uint16_t unsent,
3458 * Request the driver to free mbufs currently cached by the driver. The
3459 * driver will only free the mbuf if it is no longer in use. It is the
3460 * application's responsibility to ensure rte_eth_tx_buffer_flush(..) is
3464 * The port identifier of the Ethernet device.
3466 * The index of the transmit queue through which output packets must be
3468 * The value must be in the range [0, nb_tx_queue - 1] previously supplied
3469 * to rte_eth_dev_configure().
3471 * Maximum number of packets to free. Use 0 to indicate all possible packets
3472 * should be freed. Note that a packet may be using multiple mbufs.
3475 * -ENODEV: Invalid interface
3476 * -EIO: device is removed
3477 * -ENOTSUP: Driver does not support function
3479 * 0-n: Number of packets freed. More packets may still remain in ring that
3483 rte_eth_tx_done_cleanup(uint16_t port_id, uint16_t queue_id, uint32_t free_cnt);
3486 * Subtypes for IPsec offload event(@ref RTE_ETH_EVENT_IPSEC) raised by
3489 enum rte_eth_event_ipsec_subtype {
3490 RTE_ETH_EVENT_IPSEC_UNKNOWN = 0,
3491 /**< Unknown event type */
3492 RTE_ETH_EVENT_IPSEC_ESN_OVERFLOW,
3493 /**< Sequence number overflow */
3494 RTE_ETH_EVENT_IPSEC_SA_TIME_EXPIRY,
3495 /**< Soft time expiry of SA */
3496 RTE_ETH_EVENT_IPSEC_SA_BYTE_EXPIRY,
3497 /**< Soft byte expiry of SA */
3498 RTE_ETH_EVENT_IPSEC_MAX
3499 /**< Max value of this enum */
3503 * Descriptor for @ref RTE_ETH_EVENT_IPSEC event. Used by eth dev to send extra
3504 * information of the IPsec offload event.
3506 struct rte_eth_event_ipsec_desc {
3507 enum rte_eth_event_ipsec_subtype subtype;
3508 /**< Type of RTE_ETH_EVENT_IPSEC_* event */
3510 /**< Event specific metadata
3512 * For the following events, *userdata* registered
3513 * with the *rte_security_session* would be returned
3516 * - @ref RTE_ETH_EVENT_IPSEC_ESN_OVERFLOW
3517 * - @ref RTE_ETH_EVENT_IPSEC_SA_TIME_EXPIRY
3518 * - @ref RTE_ETH_EVENT_IPSEC_SA_BYTE_EXPIRY
3520 * @see struct rte_security_session_conf
3526 * The eth device event type for interrupt, and maybe others in the future.
3528 enum rte_eth_event_type {
3529 RTE_ETH_EVENT_UNKNOWN, /**< unknown event type */
3530 RTE_ETH_EVENT_INTR_LSC, /**< lsc interrupt event */
3531 RTE_ETH_EVENT_QUEUE_STATE,
3532 /**< queue state event (enabled/disabled) */
3533 RTE_ETH_EVENT_INTR_RESET,
3534 /**< reset interrupt event, sent to VF on PF reset */
3535 RTE_ETH_EVENT_VF_MBOX, /**< message from the VF received by PF */
3536 RTE_ETH_EVENT_MACSEC, /**< MACsec offload related event */
3537 RTE_ETH_EVENT_INTR_RMV, /**< device removal event */
3538 RTE_ETH_EVENT_NEW, /**< port is probed */
3539 RTE_ETH_EVENT_DESTROY, /**< port is released */
3540 RTE_ETH_EVENT_IPSEC, /**< IPsec offload related event */
3541 RTE_ETH_EVENT_FLOW_AGED,/**< New aged-out flows is detected */
3542 RTE_ETH_EVENT_MAX /**< max value of this enum */
3545 typedef int (*rte_eth_dev_cb_fn)(uint16_t port_id,
3546 enum rte_eth_event_type event, void *cb_arg, void *ret_param);
3547 /**< user application callback to be registered for interrupts */
3550 * Register a callback function for port event.
3554 * RTE_ETH_ALL means register the event for all port ids.
3558 * User supplied callback function to be called.
3560 * Pointer to the parameters for the registered callback.
3563 * - On success, zero.
3564 * - On failure, a negative value.
3566 int rte_eth_dev_callback_register(uint16_t port_id,
3567 enum rte_eth_event_type event,
3568 rte_eth_dev_cb_fn cb_fn, void *cb_arg);
3571 * Unregister a callback function for port event.
3575 * RTE_ETH_ALL means unregister the event for all port ids.
3579 * User supplied callback function to be called.
3581 * Pointer to the parameters for the registered callback. -1 means to
3582 * remove all for the same callback address and same event.
3585 * - On success, zero.
3586 * - On failure, a negative value.
3588 int rte_eth_dev_callback_unregister(uint16_t port_id,
3589 enum rte_eth_event_type event,
3590 rte_eth_dev_cb_fn cb_fn, void *cb_arg);
3593 * When there is no rx packet coming in Rx Queue for a long time, we can
3594 * sleep lcore related to RX Queue for power saving, and enable rx interrupt
3595 * to be triggered when Rx packet arrives.
3597 * The rte_eth_dev_rx_intr_enable() function enables rx queue
3598 * interrupt on specific rx queue of a port.
3601 * The port identifier of the Ethernet device.
3603 * The index of the receive queue from which to retrieve input packets.
3604 * The value must be in the range [0, nb_rx_queue - 1] previously supplied
3605 * to rte_eth_dev_configure().
3607 * - (0) if successful.
3608 * - (-ENOTSUP) if underlying hardware OR driver doesn't support
3610 * - (-ENODEV) if *port_id* invalid.
3611 * - (-EIO) if device is removed.
3613 int rte_eth_dev_rx_intr_enable(uint16_t port_id, uint16_t queue_id);
3616 * When lcore wakes up from rx interrupt indicating packet coming, disable rx
3617 * interrupt and returns to polling mode.
3619 * The rte_eth_dev_rx_intr_disable() function disables rx queue
3620 * interrupt on specific rx queue of a port.
3623 * The port identifier of the Ethernet device.
3625 * The index of the receive queue from which to retrieve input packets.
3626 * The value must be in the range [0, nb_rx_queue - 1] previously supplied
3627 * to rte_eth_dev_configure().
3629 * - (0) if successful.
3630 * - (-ENOTSUP) if underlying hardware OR driver doesn't support
3632 * - (-ENODEV) if *port_id* invalid.
3633 * - (-EIO) if device is removed.
3635 int rte_eth_dev_rx_intr_disable(uint16_t port_id, uint16_t queue_id);
3638 * RX Interrupt control per port.
3641 * The port identifier of the Ethernet device.
3643 * Epoll instance fd which the intr vector associated to.
3644 * Using RTE_EPOLL_PER_THREAD allows to use per thread epoll instance.
3646 * The operation be performed for the vector.
3647 * Operation type of {RTE_INTR_EVENT_ADD, RTE_INTR_EVENT_DEL}.
3651 * - On success, zero.
3652 * - On failure, a negative value.
3654 int rte_eth_dev_rx_intr_ctl(uint16_t port_id, int epfd, int op, void *data);
3657 * RX Interrupt control per queue.
3660 * The port identifier of the Ethernet device.
3662 * The index of the receive queue from which to retrieve input packets.
3663 * The value must be in the range [0, nb_rx_queue - 1] previously supplied
3664 * to rte_eth_dev_configure().
3666 * Epoll instance fd which the intr vector associated to.
3667 * Using RTE_EPOLL_PER_THREAD allows to use per thread epoll instance.
3669 * The operation be performed for the vector.
3670 * Operation type of {RTE_INTR_EVENT_ADD, RTE_INTR_EVENT_DEL}.
3674 * - On success, zero.
3675 * - On failure, a negative value.
3677 int rte_eth_dev_rx_intr_ctl_q(uint16_t port_id, uint16_t queue_id,
3678 int epfd, int op, void *data);
3681 * Get interrupt fd per Rx queue.
3684 * The port identifier of the Ethernet device.
3686 * The index of the receive queue from which to retrieve input packets.
3687 * The value must be in the range [0, nb_rx_queue - 1] previously supplied
3688 * to rte_eth_dev_configure().
3690 * - (>=0) the interrupt fd associated to the requested Rx queue if
3695 rte_eth_dev_rx_intr_ctl_q_get_fd(uint16_t port_id, uint16_t queue_id);
3698 * Turn on the LED on the Ethernet device.
3699 * This function turns on the LED on the Ethernet device.
3702 * The port identifier of the Ethernet device.
3704 * - (0) if successful.
3705 * - (-ENOTSUP) if underlying hardware OR driver doesn't support
3707 * - (-ENODEV) if *port_id* invalid.
3708 * - (-EIO) if device is removed.
3710 int rte_eth_led_on(uint16_t port_id);
3713 * Turn off the LED on the Ethernet device.
3714 * This function turns off the LED on the Ethernet device.
3717 * The port identifier of the Ethernet device.
3719 * - (0) if successful.
3720 * - (-ENOTSUP) if underlying hardware OR driver doesn't support
3722 * - (-ENODEV) if *port_id* invalid.
3723 * - (-EIO) if device is removed.
3725 int rte_eth_led_off(uint16_t port_id);
3729 * @b EXPERIMENTAL: this API may change, or be removed, without prior notice
3731 * Get Forward Error Correction(FEC) capability.
3734 * The port identifier of the Ethernet device.
3735 * @param speed_fec_capa
3736 * speed_fec_capa is out only with per-speed capabilities.
3737 * If set to NULL, the function returns the required number
3738 * of required array entries.
3740 * a number of elements in an speed_fec_capa array.
3743 * - A non-negative value lower or equal to num: success. The return value
3744 * is the number of entries filled in the fec capa array.
3745 * - A non-negative value higher than num: error, the given fec capa array
3746 * is too small. The return value corresponds to the num that should
3747 * be given to succeed. The entries in fec capa array are not valid and
3748 * shall not be used by the caller.
3749 * - (-ENOTSUP) if underlying hardware OR driver doesn't support.
3751 * - (-EIO) if device is removed.
3752 * - (-ENODEV) if *port_id* invalid.
3753 * - (-EINVAL) if *num* or *speed_fec_capa* invalid
3756 int rte_eth_fec_get_capability(uint16_t port_id,
3757 struct rte_eth_fec_capa *speed_fec_capa,
3762 * @b EXPERIMENTAL: this API may change, or be removed, without prior notice
3764 * Get current Forward Error Correction(FEC) mode.
3765 * If link is down and AUTO is enabled, AUTO is returned, otherwise,
3766 * configured FEC mode is returned.
3767 * If link is up, current FEC mode is returned.
3770 * The port identifier of the Ethernet device.
3772 * A bitmask of enabled FEC modes. If AUTO bit is set, other
3773 * bits specify FEC modes which may be negotiated. If AUTO
3774 * bit is clear, specify FEC modes to be used (only one valid
3775 * mode per speed may be set).
3777 * - (0) if successful.
3778 * - (-ENOTSUP) if underlying hardware OR driver doesn't support.
3780 * - (-EIO) if device is removed.
3781 * - (-ENODEV) if *port_id* invalid.
3784 int rte_eth_fec_get(uint16_t port_id, uint32_t *fec_capa);
3788 * @b EXPERIMENTAL: this API may change, or be removed, without prior notice
3790 * Set Forward Error Correction(FEC) mode.
3793 * The port identifier of the Ethernet device.
3795 * A bitmask of allowed FEC modes. If AUTO bit is set, other
3796 * bits specify FEC modes which may be negotiated. If AUTO
3797 * bit is clear, specify FEC modes to be used (only one valid
3798 * mode per speed may be set).
3800 * - (0) if successful.
3801 * - (-EINVAL) if the FEC mode is not valid.
3802 * - (-ENOTSUP) if underlying hardware OR driver doesn't support.
3803 * - (-EIO) if device is removed.
3804 * - (-ENODEV) if *port_id* invalid.
3807 int rte_eth_fec_set(uint16_t port_id, uint32_t fec_capa);
3810 * Get current status of the Ethernet link flow control for Ethernet device
3813 * The port identifier of the Ethernet device.
3815 * The pointer to the structure where to store the flow control parameters.
3817 * - (0) if successful.
3818 * - (-ENOTSUP) if hardware doesn't support flow control.
3819 * - (-ENODEV) if *port_id* invalid.
3820 * - (-EIO) if device is removed.
3821 * - (-EINVAL) if bad parameter.
3823 int rte_eth_dev_flow_ctrl_get(uint16_t port_id,
3824 struct rte_eth_fc_conf *fc_conf);
3827 * Configure the Ethernet link flow control for Ethernet device
3830 * The port identifier of the Ethernet device.
3832 * The pointer to the structure of the flow control parameters.
3834 * - (0) if successful.
3835 * - (-ENOTSUP) if hardware doesn't support flow control mode.
3836 * - (-ENODEV) if *port_id* invalid.
3837 * - (-EINVAL) if bad parameter
3838 * - (-EIO) if flow control setup failure or device is removed.
3840 int rte_eth_dev_flow_ctrl_set(uint16_t port_id,
3841 struct rte_eth_fc_conf *fc_conf);
3844 * Configure the Ethernet priority flow control under DCB environment
3845 * for Ethernet device.
3848 * The port identifier of the Ethernet device.
3850 * The pointer to the structure of the priority flow control parameters.
3852 * - (0) if successful.
3853 * - (-ENOTSUP) if hardware doesn't support priority flow control mode.
3854 * - (-ENODEV) if *port_id* invalid.
3855 * - (-EINVAL) if bad parameter
3856 * - (-EIO) if flow control setup failure or device is removed.
3858 int rte_eth_dev_priority_flow_ctrl_set(uint16_t port_id,
3859 struct rte_eth_pfc_conf *pfc_conf);
3862 * Add a MAC address to the set used for filtering incoming packets.
3865 * The port identifier of the Ethernet device.
3867 * The MAC address to add.
3869 * VMDq pool index to associate address with (if VMDq is enabled). If VMDq is
3870 * not enabled, this should be set to 0.
3872 * - (0) if successfully added or *mac_addr* was already added.
3873 * - (-ENOTSUP) if hardware doesn't support this feature.
3874 * - (-ENODEV) if *port* is invalid.
3875 * - (-EIO) if device is removed.
3876 * - (-ENOSPC) if no more MAC addresses can be added.
3877 * - (-EINVAL) if MAC address is invalid.
3879 int rte_eth_dev_mac_addr_add(uint16_t port_id, struct rte_ether_addr *mac_addr,
3883 * Remove a MAC address from the internal array of addresses.
3886 * The port identifier of the Ethernet device.
3888 * MAC address to remove.
3890 * - (0) if successful, or *mac_addr* didn't exist.
3891 * - (-ENOTSUP) if hardware doesn't support.
3892 * - (-ENODEV) if *port* invalid.
3893 * - (-EADDRINUSE) if attempting to remove the default MAC address.
3894 * - (-EINVAL) if MAC address is invalid.
3896 int rte_eth_dev_mac_addr_remove(uint16_t port_id,
3897 struct rte_ether_addr *mac_addr);
3900 * Set the default MAC address.
3903 * The port identifier of the Ethernet device.
3905 * New default MAC address.
3907 * - (0) if successful, or *mac_addr* didn't exist.
3908 * - (-ENOTSUP) if hardware doesn't support.
3909 * - (-ENODEV) if *port* invalid.
3910 * - (-EINVAL) if MAC address is invalid.
3912 int rte_eth_dev_default_mac_addr_set(uint16_t port_id,
3913 struct rte_ether_addr *mac_addr);
3916 * Update Redirection Table(RETA) of Receive Side Scaling of Ethernet device.
3919 * The port identifier of the Ethernet device.
3923 * Redirection table size. The table size can be queried by
3924 * rte_eth_dev_info_get().
3926 * - (0) if successful.
3927 * - (-ENODEV) if *port_id* is invalid.
3928 * - (-ENOTSUP) if hardware doesn't support.
3929 * - (-EINVAL) if bad parameter.
3930 * - (-EIO) if device is removed.
3932 int rte_eth_dev_rss_reta_update(uint16_t port_id,
3933 struct rte_eth_rss_reta_entry64 *reta_conf,
3934 uint16_t reta_size);
3937 * Query Redirection Table(RETA) of Receive Side Scaling of Ethernet device.
3940 * The port identifier of the Ethernet device.
3942 * RETA to query. For each requested reta entry, corresponding bit
3943 * in mask must be set.
3945 * Redirection table size. The table size can be queried by
3946 * rte_eth_dev_info_get().
3948 * - (0) if successful.
3949 * - (-ENODEV) if *port_id* is invalid.
3950 * - (-ENOTSUP) if hardware doesn't support.
3951 * - (-EINVAL) if bad parameter.
3952 * - (-EIO) if device is removed.
3954 int rte_eth_dev_rss_reta_query(uint16_t port_id,
3955 struct rte_eth_rss_reta_entry64 *reta_conf,
3956 uint16_t reta_size);
3959 * Updates unicast hash table for receiving packet with the given destination
3960 * MAC address, and the packet is routed to all VFs for which the RX mode is
3961 * accept packets that match the unicast hash table.
3964 * The port identifier of the Ethernet device.
3966 * Unicast MAC address.
3968 * 1 - Set an unicast hash bit for receiving packets with the MAC address.
3969 * 0 - Clear an unicast hash bit.
3971 * - (0) if successful.
3972 * - (-ENOTSUP) if hardware doesn't support.
3973 * - (-ENODEV) if *port_id* invalid.
3974 * - (-EIO) if device is removed.
3975 * - (-EINVAL) if bad parameter.
3977 int rte_eth_dev_uc_hash_table_set(uint16_t port_id, struct rte_ether_addr *addr,
3981 * Updates all unicast hash bitmaps for receiving packet with any Unicast
3982 * Ethernet MAC addresses,the packet is routed to all VFs for which the RX
3983 * mode is accept packets that match the unicast hash table.
3986 * The port identifier of the Ethernet device.
3988 * 1 - Set all unicast hash bitmaps for receiving all the Ethernet
3990 * 0 - Clear all unicast hash bitmaps
3992 * - (0) if successful.
3993 * - (-ENOTSUP) if hardware doesn't support.
3994 * - (-ENODEV) if *port_id* invalid.
3995 * - (-EIO) if device is removed.
3996 * - (-EINVAL) if bad parameter.
3998 int rte_eth_dev_uc_all_hash_table_set(uint16_t port_id, uint8_t on);
4001 * Set the rate limitation for a queue on an Ethernet device.
4004 * The port identifier of the Ethernet device.
4008 * The tx rate in Mbps. Allocated from the total port link speed.
4010 * - (0) if successful.
4011 * - (-ENOTSUP) if hardware doesn't support this feature.
4012 * - (-ENODEV) if *port_id* invalid.
4013 * - (-EIO) if device is removed.
4014 * - (-EINVAL) if bad parameter.
4016 int rte_eth_set_queue_rate_limit(uint16_t port_id, uint16_t queue_idx,
4020 * Configuration of Receive Side Scaling hash computation of Ethernet device.
4023 * The port identifier of the Ethernet device.
4025 * The new configuration to use for RSS hash computation on the port.
4027 * - (0) if successful.
4028 * - (-ENODEV) if port identifier is invalid.
4029 * - (-EIO) if device is removed.
4030 * - (-ENOTSUP) if hardware doesn't support.
4031 * - (-EINVAL) if bad parameter.
4033 int rte_eth_dev_rss_hash_update(uint16_t port_id,
4034 struct rte_eth_rss_conf *rss_conf);
4037 * Retrieve current configuration of Receive Side Scaling hash computation
4038 * of Ethernet device.
4041 * The port identifier of the Ethernet device.
4043 * Where to store the current RSS hash configuration of the Ethernet device.
4045 * - (0) if successful.
4046 * - (-ENODEV) if port identifier is invalid.
4047 * - (-EIO) if device is removed.
4048 * - (-ENOTSUP) if hardware doesn't support RSS.
4049 * - (-EINVAL) if bad parameter.
4052 rte_eth_dev_rss_hash_conf_get(uint16_t port_id,
4053 struct rte_eth_rss_conf *rss_conf);
4056 * Add UDP tunneling port for a type of tunnel.
4058 * Some NICs may require such configuration to properly parse a tunnel
4059 * with any standard or custom UDP port.
4060 * The packets with this UDP port will be parsed for this type of tunnel.
4061 * The device parser will also check the rest of the tunnel headers
4062 * before classifying the packet.
4064 * With some devices, this API will affect packet classification, i.e.:
4065 * - mbuf.packet_type reported on Rx
4066 * - rte_flow rules with tunnel items
4069 * The port identifier of the Ethernet device.
4071 * UDP tunneling configuration.
4074 * - (0) if successful.
4075 * - (-ENODEV) if port identifier is invalid.
4076 * - (-EIO) if device is removed.
4077 * - (-ENOTSUP) if hardware doesn't support tunnel type.
4080 rte_eth_dev_udp_tunnel_port_add(uint16_t port_id,
4081 struct rte_eth_udp_tunnel *tunnel_udp);
4084 * Delete UDP tunneling port for a type of tunnel.
4086 * The packets with this UDP port will not be classified as this type of tunnel
4087 * anymore if the device use such mapping for tunnel packet classification.
4089 * @see rte_eth_dev_udp_tunnel_port_add
4092 * The port identifier of the Ethernet device.
4094 * UDP tunneling configuration.
4097 * - (0) if successful.
4098 * - (-ENODEV) if port identifier is invalid.
4099 * - (-EIO) if device is removed.
4100 * - (-ENOTSUP) if hardware doesn't support tunnel type.
4103 rte_eth_dev_udp_tunnel_port_delete(uint16_t port_id,
4104 struct rte_eth_udp_tunnel *tunnel_udp);
4107 * Get DCB information on an Ethernet device.
4110 * The port identifier of the Ethernet device.
4114 * - (0) if successful.
4115 * - (-ENODEV) if port identifier is invalid.
4116 * - (-EIO) if device is removed.
4117 * - (-ENOTSUP) if hardware doesn't support.
4118 * - (-EINVAL) if bad parameter.
4120 int rte_eth_dev_get_dcb_info(uint16_t port_id,
4121 struct rte_eth_dcb_info *dcb_info);
4123 struct rte_eth_rxtx_callback;
4126 * Add a callback to be called on packet RX on a given port and queue.
4128 * This API configures a function to be called for each burst of
4129 * packets received on a given NIC port queue. The return value is a pointer
4130 * that can be used to later remove the callback using
4131 * rte_eth_remove_rx_callback().
4133 * Multiple functions are called in the order that they are added.
4136 * The port identifier of the Ethernet device.
4138 * The queue on the Ethernet device on which the callback is to be added.
4140 * The callback function
4142 * A generic pointer parameter which will be passed to each invocation of the
4143 * callback function on this port and queue. Inter-thread synchronization
4144 * of any user data changes is the responsibility of the user.
4148 * On success, a pointer value which can later be used to remove the callback.
4150 const struct rte_eth_rxtx_callback *
4151 rte_eth_add_rx_callback(uint16_t port_id, uint16_t queue_id,
4152 rte_rx_callback_fn fn, void *user_param);
4155 * Add a callback that must be called first on packet RX on a given port
4158 * This API configures a first function to be called for each burst of
4159 * packets received on a given NIC port queue. The return value is a pointer
4160 * that can be used to later remove the callback using
4161 * rte_eth_remove_rx_callback().
4163 * Multiple functions are called in the order that they are added.
4166 * The port identifier of the Ethernet device.
4168 * The queue on the Ethernet device on which the callback is to be added.
4170 * The callback function
4172 * A generic pointer parameter which will be passed to each invocation of the
4173 * callback function on this port and queue. Inter-thread synchronization
4174 * of any user data changes is the responsibility of the user.
4178 * On success, a pointer value which can later be used to remove the callback.
4180 const struct rte_eth_rxtx_callback *
4181 rte_eth_add_first_rx_callback(uint16_t port_id, uint16_t queue_id,
4182 rte_rx_callback_fn fn, void *user_param);
4185 * Add a callback to be called on packet TX on a given port and queue.
4187 * This API configures a function to be called for each burst of
4188 * packets sent on a given NIC port queue. The return value is a pointer
4189 * that can be used to later remove the callback using
4190 * rte_eth_remove_tx_callback().
4192 * Multiple functions are called in the order that they are added.
4195 * The port identifier of the Ethernet device.
4197 * The queue on the Ethernet device on which the callback is to be added.
4199 * The callback function
4201 * A generic pointer parameter which will be passed to each invocation of the
4202 * callback function on this port and queue. Inter-thread synchronization
4203 * of any user data changes is the responsibility of the user.
4207 * On success, a pointer value which can later be used to remove the callback.
4209 const struct rte_eth_rxtx_callback *
4210 rte_eth_add_tx_callback(uint16_t port_id, uint16_t queue_id,
4211 rte_tx_callback_fn fn, void *user_param);
4214 * Remove an RX packet callback from a given port and queue.
4216 * This function is used to removed callbacks that were added to a NIC port
4217 * queue using rte_eth_add_rx_callback().
4219 * Note: the callback is removed from the callback list but it isn't freed
4220 * since the it may still be in use. The memory for the callback can be
4221 * subsequently freed back by the application by calling rte_free():
4223 * - Immediately - if the port is stopped, or the user knows that no
4224 * callbacks are in flight e.g. if called from the thread doing RX/TX
4227 * - After a short delay - where the delay is sufficient to allow any
4228 * in-flight callbacks to complete. Alternately, the RCU mechanism can be
4229 * used to detect when data plane threads have ceased referencing the
4233 * The port identifier of the Ethernet device.
4235 * The queue on the Ethernet device from which the callback is to be removed.
4237 * User supplied callback created via rte_eth_add_rx_callback().
4240 * - 0: Success. Callback was removed.
4241 * - -ENODEV: If *port_id* is invalid.
4242 * - -ENOTSUP: Callback support is not available.
4243 * - -EINVAL: The queue_id is out of range, or the callback
4244 * is NULL or not found for the port/queue.
4246 int rte_eth_remove_rx_callback(uint16_t port_id, uint16_t queue_id,
4247 const struct rte_eth_rxtx_callback *user_cb);
4250 * Remove a TX packet callback from a given port and queue.
4252 * This function is used to removed callbacks that were added to a NIC port
4253 * queue using rte_eth_add_tx_callback().
4255 * Note: the callback is removed from the callback list but it isn't freed
4256 * since the it may still be in use. The memory for the callback can be
4257 * subsequently freed back by the application by calling rte_free():
4259 * - Immediately - if the port is stopped, or the user knows that no
4260 * callbacks are in flight e.g. if called from the thread doing RX/TX
4263 * - After a short delay - where the delay is sufficient to allow any
4264 * in-flight callbacks to complete. Alternately, the RCU mechanism can be
4265 * used to detect when data plane threads have ceased referencing the
4269 * The port identifier of the Ethernet device.
4271 * The queue on the Ethernet device from which the callback is to be removed.
4273 * User supplied callback created via rte_eth_add_tx_callback().
4276 * - 0: Success. Callback was removed.
4277 * - -ENODEV: If *port_id* is invalid.
4278 * - -ENOTSUP: Callback support is not available.
4279 * - -EINVAL: The queue_id is out of range, or the callback
4280 * is NULL or not found for the port/queue.
4282 int rte_eth_remove_tx_callback(uint16_t port_id, uint16_t queue_id,
4283 const struct rte_eth_rxtx_callback *user_cb);
4286 * Retrieve information about given port's RX queue.
4289 * The port identifier of the Ethernet device.
4291 * The RX queue on the Ethernet device for which information
4292 * will be retrieved.
4294 * A pointer to a structure of type *rte_eth_rxq_info_info* to be filled with
4295 * the information of the Ethernet device.
4299 * - -ENODEV: If *port_id* is invalid.
4300 * - -ENOTSUP: routine is not supported by the device PMD.
4301 * - -EINVAL: The queue_id is out of range, or the queue
4304 int rte_eth_rx_queue_info_get(uint16_t port_id, uint16_t queue_id,
4305 struct rte_eth_rxq_info *qinfo);
4308 * Retrieve information about given port's TX queue.
4311 * The port identifier of the Ethernet device.
4313 * The TX queue on the Ethernet device for which information
4314 * will be retrieved.
4316 * A pointer to a structure of type *rte_eth_txq_info_info* to be filled with
4317 * the information of the Ethernet device.
4321 * - -ENODEV: If *port_id* is invalid.
4322 * - -ENOTSUP: routine is not supported by the device PMD.
4323 * - -EINVAL: The queue_id is out of range, or the queue
4326 int rte_eth_tx_queue_info_get(uint16_t port_id, uint16_t queue_id,
4327 struct rte_eth_txq_info *qinfo);
4330 * Retrieve information about the Rx packet burst mode.
4333 * The port identifier of the Ethernet device.
4335 * The Rx queue on the Ethernet device for which information
4336 * will be retrieved.
4338 * A pointer to a structure of type *rte_eth_burst_mode* to be filled
4339 * with the information of the packet burst mode.
4343 * - -ENODEV: If *port_id* is invalid.
4344 * - -ENOTSUP: routine is not supported by the device PMD.
4345 * - -EINVAL: The queue_id is out of range.
4347 int rte_eth_rx_burst_mode_get(uint16_t port_id, uint16_t queue_id,
4348 struct rte_eth_burst_mode *mode);
4351 * Retrieve information about the Tx packet burst mode.
4354 * The port identifier of the Ethernet device.
4356 * The Tx queue on the Ethernet device for which information
4357 * will be retrieved.
4359 * A pointer to a structure of type *rte_eth_burst_mode* to be filled
4360 * with the information of the packet burst mode.
4364 * - -ENODEV: If *port_id* is invalid.
4365 * - -ENOTSUP: routine is not supported by the device PMD.
4366 * - -EINVAL: The queue_id is out of range.
4368 int rte_eth_tx_burst_mode_get(uint16_t port_id, uint16_t queue_id,
4369 struct rte_eth_burst_mode *mode);
4373 * @b EXPERIMENTAL: this API may change without prior notice.
4375 * Retrieve the monitor condition for a given receive queue.
4378 * The port identifier of the Ethernet device.
4380 * The Rx queue on the Ethernet device for which information
4381 * will be retrieved.
4383 * The pointer to power-optimized monitoring condition structure.
4387 * -ENOTSUP: Operation not supported.
4388 * -EINVAL: Invalid parameters.
4389 * -ENODEV: Invalid port ID.
4392 int rte_eth_get_monitor_addr(uint16_t port_id, uint16_t queue_id,
4393 struct rte_power_monitor_cond *pmc);
4396 * Retrieve device registers and register attributes (number of registers and
4400 * The port identifier of the Ethernet device.
4402 * Pointer to rte_dev_reg_info structure to fill in. If info->data is
4403 * NULL the function fills in the width and length fields. If non-NULL
4404 * the registers are put into the buffer pointed at by the data field.
4406 * - (0) if successful.
4407 * - (-ENOTSUP) if hardware doesn't support.
4408 * - (-EINVAL) if bad parameter.
4409 * - (-ENODEV) if *port_id* invalid.
4410 * - (-EIO) if device is removed.
4411 * - others depends on the specific operations implementation.
4413 int rte_eth_dev_get_reg_info(uint16_t port_id, struct rte_dev_reg_info *info);
4416 * Retrieve size of device EEPROM
4419 * The port identifier of the Ethernet device.
4421 * - (>=0) EEPROM size if successful.
4422 * - (-ENOTSUP) if hardware doesn't support.
4423 * - (-ENODEV) if *port_id* invalid.
4424 * - (-EIO) if device is removed.
4425 * - others depends on the specific operations implementation.
4427 int rte_eth_dev_get_eeprom_length(uint16_t port_id);
4430 * Retrieve EEPROM and EEPROM attribute
4433 * The port identifier of the Ethernet device.
4435 * The template includes buffer for return EEPROM data and
4436 * EEPROM attributes to be filled.
4438 * - (0) if successful.
4439 * - (-ENOTSUP) if hardware doesn't support.
4440 * - (-EINVAL) if bad parameter.
4441 * - (-ENODEV) if *port_id* invalid.
4442 * - (-EIO) if device is removed.
4443 * - others depends on the specific operations implementation.
4445 int rte_eth_dev_get_eeprom(uint16_t port_id, struct rte_dev_eeprom_info *info);
4448 * Program EEPROM with provided data
4451 * The port identifier of the Ethernet device.
4453 * The template includes EEPROM data for programming and
4454 * EEPROM attributes to be filled
4456 * - (0) if successful.
4457 * - (-ENOTSUP) if hardware doesn't support.
4458 * - (-ENODEV) if *port_id* invalid.
4459 * - (-EINVAL) if bad parameter.
4460 * - (-EIO) if device is removed.
4461 * - others depends on the specific operations implementation.
4463 int rte_eth_dev_set_eeprom(uint16_t port_id, struct rte_dev_eeprom_info *info);
4467 * @b EXPERIMENTAL: this API may change without prior notice.
4469 * Retrieve the type and size of plugin module EEPROM
4472 * The port identifier of the Ethernet device.
4474 * The type and size of plugin module EEPROM.
4476 * - (0) if successful.
4477 * - (-ENOTSUP) if hardware doesn't support.
4478 * - (-ENODEV) if *port_id* invalid.
4479 * - (-EINVAL) if bad parameter.
4480 * - (-EIO) if device is removed.
4481 * - others depends on the specific operations implementation.
4485 rte_eth_dev_get_module_info(uint16_t port_id,
4486 struct rte_eth_dev_module_info *modinfo);
4490 * @b EXPERIMENTAL: this API may change without prior notice.
4492 * Retrieve the data of plugin module EEPROM
4495 * The port identifier of the Ethernet device.
4497 * The template includes the plugin module EEPROM attributes, and the
4498 * buffer for return plugin module EEPROM data.
4500 * - (0) if successful.
4501 * - (-ENOTSUP) if hardware doesn't support.
4502 * - (-EINVAL) if bad parameter.
4503 * - (-ENODEV) if *port_id* invalid.
4504 * - (-EIO) if device is removed.
4505 * - others depends on the specific operations implementation.
4509 rte_eth_dev_get_module_eeprom(uint16_t port_id,
4510 struct rte_dev_eeprom_info *info);
4513 * Set the list of multicast addresses to filter on an Ethernet device.
4516 * The port identifier of the Ethernet device.
4517 * @param mc_addr_set
4518 * The array of multicast addresses to set. Equal to NULL when the function
4519 * is invoked to flush the set of filtered addresses.
4521 * The number of multicast addresses in the *mc_addr_set* array. Equal to 0
4522 * when the function is invoked to flush the set of filtered addresses.
4524 * - (0) if successful.
4525 * - (-ENODEV) if *port_id* invalid.
4526 * - (-EIO) if device is removed.
4527 * - (-ENOTSUP) if PMD of *port_id* doesn't support multicast filtering.
4528 * - (-ENOSPC) if *port_id* has not enough multicast filtering resources.
4529 * - (-EINVAL) if bad parameter.
4531 int rte_eth_dev_set_mc_addr_list(uint16_t port_id,
4532 struct rte_ether_addr *mc_addr_set,
4533 uint32_t nb_mc_addr);
4536 * Enable IEEE1588/802.1AS timestamping for an Ethernet device.
4539 * The port identifier of the Ethernet device.
4543 * - -ENODEV: The port ID is invalid.
4544 * - -EIO: if device is removed.
4545 * - -ENOTSUP: The function is not supported by the Ethernet driver.
4547 int rte_eth_timesync_enable(uint16_t port_id);
4550 * Disable IEEE1588/802.1AS timestamping for an Ethernet device.
4553 * The port identifier of the Ethernet device.
4557 * - -ENODEV: The port ID is invalid.
4558 * - -EIO: if device is removed.
4559 * - -ENOTSUP: The function is not supported by the Ethernet driver.
4561 int rte_eth_timesync_disable(uint16_t port_id);
4564 * Read an IEEE1588/802.1AS RX timestamp from an Ethernet device.
4567 * The port identifier of the Ethernet device.
4569 * Pointer to the timestamp struct.
4571 * Device specific flags. Used to pass the RX timesync register index to
4572 * i40e. Unused in igb/ixgbe, pass 0 instead.
4576 * - -EINVAL: No timestamp is available.
4577 * - -ENODEV: The port ID is invalid.
4578 * - -EIO: if device is removed.
4579 * - -ENOTSUP: The function is not supported by the Ethernet driver.
4581 int rte_eth_timesync_read_rx_timestamp(uint16_t port_id,
4582 struct timespec *timestamp, uint32_t flags);
4585 * Read an IEEE1588/802.1AS TX timestamp from an Ethernet device.
4588 * The port identifier of the Ethernet device.
4590 * Pointer to the timestamp struct.
4594 * - -EINVAL: No timestamp is available.
4595 * - -ENODEV: The port ID is invalid.
4596 * - -EIO: if device is removed.
4597 * - -ENOTSUP: The function is not supported by the Ethernet driver.
4599 int rte_eth_timesync_read_tx_timestamp(uint16_t port_id,
4600 struct timespec *timestamp);
4603 * Adjust the timesync clock on an Ethernet device.
4605 * This is usually used in conjunction with other Ethdev timesync functions to
4606 * synchronize the device time using the IEEE1588/802.1AS protocol.
4609 * The port identifier of the Ethernet device.
4611 * The adjustment in nanoseconds.
4615 * - -ENODEV: The port ID is invalid.
4616 * - -EIO: if device is removed.
4617 * - -ENOTSUP: The function is not supported by the Ethernet driver.
4619 int rte_eth_timesync_adjust_time(uint16_t port_id, int64_t delta);
4622 * Read the time from the timesync clock on an Ethernet device.
4624 * This is usually used in conjunction with other Ethdev timesync functions to
4625 * synchronize the device time using the IEEE1588/802.1AS protocol.
4628 * The port identifier of the Ethernet device.
4630 * Pointer to the timespec struct that holds the time.
4634 * - -EINVAL: Bad parameter.
4636 int rte_eth_timesync_read_time(uint16_t port_id, struct timespec *time);
4639 * Set the time of the timesync clock on an Ethernet device.
4641 * This is usually used in conjunction with other Ethdev timesync functions to
4642 * synchronize the device time using the IEEE1588/802.1AS protocol.
4645 * The port identifier of the Ethernet device.
4647 * Pointer to the timespec struct that holds the time.
4651 * - -EINVAL: No timestamp is available.
4652 * - -ENODEV: The port ID is invalid.
4653 * - -EIO: if device is removed.
4654 * - -ENOTSUP: The function is not supported by the Ethernet driver.
4656 int rte_eth_timesync_write_time(uint16_t port_id, const struct timespec *time);
4660 * @b EXPERIMENTAL: this API may change without prior notice.
4662 * Read the current clock counter of an Ethernet device
4664 * This returns the current raw clock value of an Ethernet device. It is
4665 * a raw amount of ticks, with no given time reference.
4666 * The value returned here is from the same clock than the one
4667 * filling timestamp field of Rx packets when using hardware timestamp
4668 * offload. Therefore it can be used to compute a precise conversion of
4669 * the device clock to the real time.
4671 * E.g, a simple heuristic to derivate the frequency would be:
4672 * uint64_t start, end;
4673 * rte_eth_read_clock(port, start);
4674 * rte_delay_ms(100);
4675 * rte_eth_read_clock(port, end);
4676 * double freq = (end - start) * 10;
4678 * Compute a common reference with:
4679 * uint64_t base_time_sec = current_time();
4680 * uint64_t base_clock;
4681 * rte_eth_read_clock(port, base_clock);
4683 * Then, convert the raw mbuf timestamp with:
4684 * base_time_sec + (double)(*timestamp_dynfield(mbuf) - base_clock) / freq;
4686 * This simple example will not provide a very good accuracy. One must
4687 * at least measure multiple times the frequency and do a regression.
4688 * To avoid deviation from the system time, the common reference can
4689 * be repeated from time to time. The integer division can also be
4690 * converted by a multiplication and a shift for better performance.
4693 * The port identifier of the Ethernet device.
4695 * Pointer to the uint64_t that holds the raw clock value.
4699 * - -ENODEV: The port ID is invalid.
4700 * - -ENOTSUP: The function is not supported by the Ethernet driver.
4701 * - -EINVAL: if bad parameter.
4705 rte_eth_read_clock(uint16_t port_id, uint64_t *clock);
4708 * Get the port id from device name. The device name should be specified
4710 * - PCIe address (Domain:Bus:Device.Function), for example- 0000:2:00.0
4711 * - SoC device name, for example- fsl-gmac0
4712 * - vdev dpdk name, for example- net_[pcap0|null0|tap0]
4715 * pci address or name of the device
4717 * pointer to port identifier of the device
4719 * - (0) if successful and port_id is filled.
4720 * - (-ENODEV or -EINVAL) on failure.
4723 rte_eth_dev_get_port_by_name(const char *name, uint16_t *port_id);
4726 * Get the device name from port id. The device name is specified as below:
4727 * - PCIe address (Domain:Bus:Device.Function), for example- 0000:02:00.0
4728 * - SoC device name, for example- fsl-gmac0
4729 * - vdev dpdk name, for example- net_[pcap0|null0|tun0|tap0]
4732 * Port identifier of the device.
4734 * Buffer of size RTE_ETH_NAME_MAX_LEN to store the name.
4736 * - (0) if successful.
4737 * - (-ENODEV) if *port_id* is invalid.
4738 * - (-EINVAL) on failure.
4741 rte_eth_dev_get_name_by_port(uint16_t port_id, char *name);
4744 * Check that numbers of Rx and Tx descriptors satisfy descriptors limits from
4745 * the ethernet device information, otherwise adjust them to boundaries.
4748 * The port identifier of the Ethernet device.
4750 * A pointer to a uint16_t where the number of receive
4751 * descriptors stored.
4753 * A pointer to a uint16_t where the number of transmit
4754 * descriptors stored.
4756 * - (0) if successful.
4757 * - (-ENOTSUP, -ENODEV or -EINVAL) on failure.
4759 int rte_eth_dev_adjust_nb_rx_tx_desc(uint16_t port_id,
4760 uint16_t *nb_rx_desc,
4761 uint16_t *nb_tx_desc);
4764 * Test if a port supports specific mempool ops.
4767 * Port identifier of the Ethernet device.
4769 * The name of the pool operations to test.
4771 * - 0: best mempool ops choice for this port.
4772 * - 1: mempool ops are supported for this port.
4773 * - -ENOTSUP: mempool ops not supported for this port.
4774 * - -ENODEV: Invalid port Identifier.
4775 * - -EINVAL: Pool param is null.
4778 rte_eth_dev_pool_ops_supported(uint16_t port_id, const char *pool);
4781 * Get the security context for the Ethernet device.
4784 * Port identifier of the Ethernet device
4787 * - pointer to security context on success.
4790 rte_eth_dev_get_sec_ctx(uint16_t port_id);
4794 * @b EXPERIMENTAL: this API may change, or be removed, without prior notice
4796 * Query the device hairpin capabilities.
4799 * The port identifier of the Ethernet device.
4801 * Pointer to a structure that will hold the hairpin capabilities.
4803 * - (0) if successful.
4804 * - (-ENOTSUP) if hardware doesn't support.
4805 * - (-EINVAL) if bad parameter.
4808 int rte_eth_dev_hairpin_capability_get(uint16_t port_id,
4809 struct rte_eth_hairpin_cap *cap);
4813 * @b EXPERIMENTAL: this structure may change without prior notice.
4815 * ethernet device representor ID range entry
4817 struct rte_eth_representor_range {
4818 enum rte_eth_representor_type type; /**< Representor type */
4819 int controller; /**< Controller index */
4820 int pf; /**< Physical function index */
4823 int vf; /**< VF start index */
4824 int sf; /**< SF start index */
4826 uint32_t id_base; /**< Representor ID start index */
4827 uint32_t id_end; /**< Representor ID end index */
4828 char name[RTE_DEV_NAME_MAX_LEN]; /**< Representor name */
4833 * @b EXPERIMENTAL: this structure may change without prior notice.
4835 * Ethernet device representor information
4837 struct rte_eth_representor_info {
4838 uint16_t controller; /**< Controller ID of caller device. */
4839 uint16_t pf; /**< Physical function ID of caller device. */
4840 uint32_t nb_ranges_alloc; /**< Size of the ranges array. */
4841 uint32_t nb_ranges; /**< Number of initialized ranges. */
4842 struct rte_eth_representor_range ranges[];/**< Representor ID range. */
4846 * Retrieve the representor info of the device.
4848 * Get device representor info to be able to calculate a unique
4849 * representor ID. @see rte_eth_representor_id_get helper.
4852 * The port identifier of the device.
4854 * A pointer to a representor info structure.
4855 * NULL to return number of range entries and allocate memory
4856 * for next call to store detail.
4857 * The number of ranges that were written into this structure
4858 * will be placed into its nb_ranges field. This number cannot be
4859 * larger than the nb_ranges_alloc that by the user before calling
4860 * this function. It can be smaller than the value returned by the
4861 * function, however.
4863 * - (-ENOTSUP) if operation is not supported.
4864 * - (-ENODEV) if *port_id* invalid.
4865 * - (-EIO) if device is removed.
4866 * - (>=0) number of available representor range entries.
4869 int rte_eth_representor_info_get(uint16_t port_id,
4870 struct rte_eth_representor_info *info);
4872 /** The NIC is able to deliver flag (if set) with packets to the PMD. */
4873 #define RTE_ETH_RX_METADATA_USER_FLAG (UINT64_C(1) << 0)
4875 /** The NIC is able to deliver mark ID with packets to the PMD. */
4876 #define RTE_ETH_RX_METADATA_USER_MARK (UINT64_C(1) << 1)
4878 /** The NIC is able to deliver tunnel ID with packets to the PMD. */
4879 #define RTE_ETH_RX_METADATA_TUNNEL_ID (UINT64_C(1) << 2)
4883 * @b EXPERIMENTAL: this API may change without prior notice
4885 * Negotiate the NIC's ability to deliver specific kinds of metadata to the PMD.
4887 * Invoke this API before the first rte_eth_dev_configure() invocation
4888 * to let the PMD make preparations that are inconvenient to do later.
4890 * The negotiation process is as follows:
4892 * - the application requests features intending to use at least some of them;
4893 * - the PMD responds with the guaranteed subset of the requested feature set;
4894 * - the application can retry negotiation with another set of features;
4895 * - the application can pass zero to clear the negotiation result;
4896 * - the last negotiated result takes effect upon
4897 * the ethdev configure and start.
4900 * The PMD is supposed to first consider enabling the requested feature set
4901 * in its entirety. Only if it fails to do so, does it have the right to
4902 * respond with a smaller set of the originally requested features.
4905 * Return code (-ENOTSUP) does not necessarily mean that the requested
4906 * features are unsupported. In this case, the application should just
4907 * assume that these features can be used without prior negotiations.
4910 * Port (ethdev) identifier
4912 * @param[inout] features
4913 * Feature selection buffer
4916 * - (-EBUSY) if the port can't handle this in its current state;
4917 * - (-ENOTSUP) if the method itself is not supported by the PMD;
4918 * - (-ENODEV) if *port_id* is invalid;
4919 * - (-EINVAL) if *features* is NULL;
4920 * - (-EIO) if the device is removed;
4924 int rte_eth_rx_metadata_negotiate(uint16_t port_id, uint64_t *features);
4926 #include <rte_ethdev_core.h>
4930 * Helper routine for rte_eth_rx_burst().
4931 * Should be called at exit from PMD's rte_eth_rx_bulk implementation.
4932 * Does necessary post-processing - invokes Rx callbacks if any, etc.
4935 * The port identifier of the Ethernet device.
4937 * The index of the receive queue from which to retrieve input packets.
4939 * The address of an array of pointers to *rte_mbuf* structures that
4940 * have been retrieved from the device.
4942 * The number of packets that were retrieved from the device.
4944 * The number of elements in @p rx_pkts array.
4946 * Opaque pointer of Rx queue callback related data.
4949 * The number of packets effectively supplied to the @p rx_pkts array.
4951 uint16_t rte_eth_call_rx_callbacks(uint16_t port_id, uint16_t queue_id,
4952 struct rte_mbuf **rx_pkts, uint16_t nb_rx, uint16_t nb_pkts,
4957 * Retrieve a burst of input packets from a receive queue of an Ethernet
4958 * device. The retrieved packets are stored in *rte_mbuf* structures whose
4959 * pointers are supplied in the *rx_pkts* array.
4961 * The rte_eth_rx_burst() function loops, parsing the RX ring of the
4962 * receive queue, up to *nb_pkts* packets, and for each completed RX
4963 * descriptor in the ring, it performs the following operations:
4965 * - Initialize the *rte_mbuf* data structure associated with the
4966 * RX descriptor according to the information provided by the NIC into
4967 * that RX descriptor.
4969 * - Store the *rte_mbuf* data structure into the next entry of the
4972 * - Replenish the RX descriptor with a new *rte_mbuf* buffer
4973 * allocated from the memory pool associated with the receive queue at
4974 * initialization time.
4976 * When retrieving an input packet that was scattered by the controller
4977 * into multiple receive descriptors, the rte_eth_rx_burst() function
4978 * appends the associated *rte_mbuf* buffers to the first buffer of the
4981 * The rte_eth_rx_burst() function returns the number of packets
4982 * actually retrieved, which is the number of *rte_mbuf* data structures
4983 * effectively supplied into the *rx_pkts* array.
4984 * A return value equal to *nb_pkts* indicates that the RX queue contained
4985 * at least *rx_pkts* packets, and this is likely to signify that other
4986 * received packets remain in the input queue. Applications implementing
4987 * a "retrieve as much received packets as possible" policy can check this
4988 * specific case and keep invoking the rte_eth_rx_burst() function until
4989 * a value less than *nb_pkts* is returned.
4991 * This receive method has the following advantages:
4993 * - It allows a run-to-completion network stack engine to retrieve and
4994 * to immediately process received packets in a fast burst-oriented
4995 * approach, avoiding the overhead of unnecessary intermediate packet
4996 * queue/dequeue operations.
4998 * - Conversely, it also allows an asynchronous-oriented processing
4999 * method to retrieve bursts of received packets and to immediately
5000 * queue them for further parallel processing by another logical core,
5001 * for instance. However, instead of having received packets being
5002 * individually queued by the driver, this approach allows the caller
5003 * of the rte_eth_rx_burst() function to queue a burst of retrieved
5004 * packets at a time and therefore dramatically reduce the cost of
5005 * enqueue/dequeue operations per packet.
5007 * - It allows the rte_eth_rx_burst() function of the driver to take
5008 * advantage of burst-oriented hardware features (CPU cache,
5009 * prefetch instructions, and so on) to minimize the number of CPU
5010 * cycles per packet.
5012 * To summarize, the proposed receive API enables many
5013 * burst-oriented optimizations in both synchronous and asynchronous
5014 * packet processing environments with no overhead in both cases.
5017 * Some drivers using vector instructions require that *nb_pkts* is
5018 * divisible by 4 or 8, depending on the driver implementation.
5020 * The rte_eth_rx_burst() function does not provide any error
5021 * notification to avoid the corresponding overhead. As a hint, the
5022 * upper-level application might check the status of the device link once
5023 * being systematically returned a 0 value for a given number of tries.
5026 * The port identifier of the Ethernet device.
5028 * The index of the receive queue from which to retrieve input packets.
5029 * The value must be in the range [0, nb_rx_queue - 1] previously supplied
5030 * to rte_eth_dev_configure().
5032 * The address of an array of pointers to *rte_mbuf* structures that
5033 * must be large enough to store *nb_pkts* pointers in it.
5035 * The maximum number of packets to retrieve.
5036 * The value must be divisible by 8 in order to work with any driver.
5038 * The number of packets actually retrieved, which is the number
5039 * of pointers to *rte_mbuf* structures effectively supplied to the
5042 static inline uint16_t
5043 rte_eth_rx_burst(uint16_t port_id, uint16_t queue_id,
5044 struct rte_mbuf **rx_pkts, const uint16_t nb_pkts)
5047 struct rte_eth_fp_ops *p;
5050 #ifdef RTE_ETHDEV_DEBUG_RX
5051 if (port_id >= RTE_MAX_ETHPORTS ||
5052 queue_id >= RTE_MAX_QUEUES_PER_PORT) {
5054 "Invalid port_id=%u or queue_id=%u\n",
5060 /* fetch pointer to queue data */
5061 p = &rte_eth_fp_ops[port_id];
5062 qd = p->rxq.data[queue_id];
5064 #ifdef RTE_ETHDEV_DEBUG_RX
5065 RTE_ETH_VALID_PORTID_OR_ERR_RET(port_id, 0);
5068 RTE_ETHDEV_LOG(ERR, "Invalid RX queue_id=%u for port_id=%u\n",
5074 nb_rx = p->rx_pkt_burst(qd, rx_pkts, nb_pkts);
5076 #ifdef RTE_ETHDEV_RXTX_CALLBACKS
5080 /* __ATOMIC_RELEASE memory order was used when the
5081 * call back was inserted into the list.
5082 * Since there is a clear dependency between loading
5083 * cb and cb->fn/cb->next, __ATOMIC_ACQUIRE memory order is
5086 cb = __atomic_load_n((void **)&p->rxq.clbk[queue_id],
5088 if (unlikely(cb != NULL))
5089 nb_rx = rte_eth_call_rx_callbacks(port_id, queue_id,
5090 rx_pkts, nb_rx, nb_pkts, cb);
5094 rte_ethdev_trace_rx_burst(port_id, queue_id, (void **)rx_pkts, nb_rx);
5099 * Get the number of used descriptors of a rx queue
5102 * The port identifier of the Ethernet device.
5104 * The queue id on the specific port.
5106 * The number of used descriptors in the specific queue, or:
5107 * - (-ENODEV) if *port_id* is invalid.
5108 * (-EINVAL) if *queue_id* is invalid
5109 * (-ENOTSUP) if the device does not support this function
5112 rte_eth_rx_queue_count(uint16_t port_id, uint16_t queue_id)
5114 struct rte_eth_fp_ops *p;
5117 if (port_id >= RTE_MAX_ETHPORTS ||
5118 queue_id >= RTE_MAX_QUEUES_PER_PORT) {
5120 "Invalid port_id=%u or queue_id=%u\n",
5125 /* fetch pointer to queue data */
5126 p = &rte_eth_fp_ops[port_id];
5127 qd = p->rxq.data[queue_id];
5129 RTE_ETH_VALID_PORTID_OR_ERR_RET(port_id, -ENODEV);
5130 RTE_FUNC_PTR_OR_ERR_RET(*p->rx_queue_count, -ENOTSUP);
5134 return (int)(*p->rx_queue_count)(qd);
5137 /**@{@name Rx hardware descriptor states
5138 * @see rte_eth_rx_descriptor_status
5140 #define RTE_ETH_RX_DESC_AVAIL 0 /**< Desc available for hw. */
5141 #define RTE_ETH_RX_DESC_DONE 1 /**< Desc done, filled by hw. */
5142 #define RTE_ETH_RX_DESC_UNAVAIL 2 /**< Desc used by driver or hw. */
5146 * Check the status of a Rx descriptor in the queue
5148 * It should be called in a similar context than the Rx function:
5149 * - on a dataplane core
5150 * - not concurrently on the same queue
5152 * Since it's a dataplane function, no check is performed on port_id and
5153 * queue_id. The caller must therefore ensure that the port is enabled
5154 * and the queue is configured and running.
5156 * Note: accessing to a random descriptor in the ring may trigger cache
5157 * misses and have a performance impact.
5160 * A valid port identifier of the Ethernet device which.
5162 * A valid Rx queue identifier on this port.
5164 * The offset of the descriptor starting from tail (0 is the next
5165 * packet to be received by the driver).
5168 * - (RTE_ETH_RX_DESC_AVAIL): Descriptor is available for the hardware to
5170 * - (RTE_ETH_RX_DESC_DONE): Descriptor is done, it is filled by hw, but
5171 * not yet processed by the driver (i.e. in the receive queue).
5172 * - (RTE_ETH_RX_DESC_UNAVAIL): Descriptor is unavailable, either hold by
5173 * the driver and not yet returned to hw, or reserved by the hw.
5174 * - (-EINVAL) bad descriptor offset.
5175 * - (-ENOTSUP) if the device does not support this function.
5176 * - (-ENODEV) bad port or queue (only if compiled with debug).
5179 rte_eth_rx_descriptor_status(uint16_t port_id, uint16_t queue_id,
5182 struct rte_eth_fp_ops *p;
5185 #ifdef RTE_ETHDEV_DEBUG_RX
5186 if (port_id >= RTE_MAX_ETHPORTS ||
5187 queue_id >= RTE_MAX_QUEUES_PER_PORT) {
5189 "Invalid port_id=%u or queue_id=%u\n",
5195 /* fetch pointer to queue data */
5196 p = &rte_eth_fp_ops[port_id];
5197 qd = p->rxq.data[queue_id];
5199 #ifdef RTE_ETHDEV_DEBUG_RX
5200 RTE_ETH_VALID_PORTID_OR_ERR_RET(port_id, -ENODEV);
5204 RTE_FUNC_PTR_OR_ERR_RET(*p->rx_descriptor_status, -ENOTSUP);
5205 return (*p->rx_descriptor_status)(qd, offset);
5208 /**@{@name Tx hardware descriptor states
5209 * @see rte_eth_tx_descriptor_status
5211 #define RTE_ETH_TX_DESC_FULL 0 /**< Desc filled for hw, waiting xmit. */
5212 #define RTE_ETH_TX_DESC_DONE 1 /**< Desc done, packet is transmitted. */
5213 #define RTE_ETH_TX_DESC_UNAVAIL 2 /**< Desc used by driver or hw. */
5217 * Check the status of a Tx descriptor in the queue.
5219 * It should be called in a similar context than the Tx function:
5220 * - on a dataplane core
5221 * - not concurrently on the same queue
5223 * Since it's a dataplane function, no check is performed on port_id and
5224 * queue_id. The caller must therefore ensure that the port is enabled
5225 * and the queue is configured and running.
5227 * Note: accessing to a random descriptor in the ring may trigger cache
5228 * misses and have a performance impact.
5231 * A valid port identifier of the Ethernet device which.
5233 * A valid Tx queue identifier on this port.
5235 * The offset of the descriptor starting from tail (0 is the place where
5236 * the next packet will be send).
5239 * - (RTE_ETH_TX_DESC_FULL) Descriptor is being processed by the hw, i.e.
5240 * in the transmit queue.
5241 * - (RTE_ETH_TX_DESC_DONE) Hardware is done with this descriptor, it can
5242 * be reused by the driver.
5243 * - (RTE_ETH_TX_DESC_UNAVAIL): Descriptor is unavailable, reserved by the
5244 * driver or the hardware.
5245 * - (-EINVAL) bad descriptor offset.
5246 * - (-ENOTSUP) if the device does not support this function.
5247 * - (-ENODEV) bad port or queue (only if compiled with debug).
5249 static inline int rte_eth_tx_descriptor_status(uint16_t port_id,
5250 uint16_t queue_id, uint16_t offset)
5252 struct rte_eth_fp_ops *p;
5255 #ifdef RTE_ETHDEV_DEBUG_TX
5256 if (port_id >= RTE_MAX_ETHPORTS ||
5257 queue_id >= RTE_MAX_QUEUES_PER_PORT) {
5259 "Invalid port_id=%u or queue_id=%u\n",
5265 /* fetch pointer to queue data */
5266 p = &rte_eth_fp_ops[port_id];
5267 qd = p->txq.data[queue_id];
5269 #ifdef RTE_ETHDEV_DEBUG_TX
5270 RTE_ETH_VALID_PORTID_OR_ERR_RET(port_id, -ENODEV);
5274 RTE_FUNC_PTR_OR_ERR_RET(*p->tx_descriptor_status, -ENOTSUP);
5275 return (*p->tx_descriptor_status)(qd, offset);
5280 * Helper routine for rte_eth_tx_burst().
5281 * Should be called before entry PMD's rte_eth_tx_bulk implementation.
5282 * Does necessary pre-processing - invokes Tx callbacks if any, etc.
5285 * The port identifier of the Ethernet device.
5287 * The index of the transmit queue through which output packets must be
5290 * The address of an array of *nb_pkts* pointers to *rte_mbuf* structures
5291 * which contain the output packets.
5293 * The maximum number of packets to transmit.
5295 * The number of output packets to transmit.
5297 uint16_t rte_eth_call_tx_callbacks(uint16_t port_id, uint16_t queue_id,
5298 struct rte_mbuf **tx_pkts, uint16_t nb_pkts, void *opaque);
5301 * Send a burst of output packets on a transmit queue of an Ethernet device.
5303 * The rte_eth_tx_burst() function is invoked to transmit output packets
5304 * on the output queue *queue_id* of the Ethernet device designated by its
5306 * The *nb_pkts* parameter is the number of packets to send which are
5307 * supplied in the *tx_pkts* array of *rte_mbuf* structures, each of them
5308 * allocated from a pool created with rte_pktmbuf_pool_create().
5309 * The rte_eth_tx_burst() function loops, sending *nb_pkts* packets,
5310 * up to the number of transmit descriptors available in the TX ring of the
5312 * For each packet to send, the rte_eth_tx_burst() function performs
5313 * the following operations:
5315 * - Pick up the next available descriptor in the transmit ring.
5317 * - Free the network buffer previously sent with that descriptor, if any.
5319 * - Initialize the transmit descriptor with the information provided
5320 * in the *rte_mbuf data structure.
5322 * In the case of a segmented packet composed of a list of *rte_mbuf* buffers,
5323 * the rte_eth_tx_burst() function uses several transmit descriptors
5326 * The rte_eth_tx_burst() function returns the number of packets it
5327 * actually sent. A return value equal to *nb_pkts* means that all packets
5328 * have been sent, and this is likely to signify that other output packets
5329 * could be immediately transmitted again. Applications that implement a
5330 * "send as many packets to transmit as possible" policy can check this
5331 * specific case and keep invoking the rte_eth_tx_burst() function until
5332 * a value less than *nb_pkts* is returned.
5334 * It is the responsibility of the rte_eth_tx_burst() function to
5335 * transparently free the memory buffers of packets previously sent.
5336 * This feature is driven by the *tx_free_thresh* value supplied to the
5337 * rte_eth_dev_configure() function at device configuration time.
5338 * When the number of free TX descriptors drops below this threshold, the
5339 * rte_eth_tx_burst() function must [attempt to] free the *rte_mbuf* buffers
5340 * of those packets whose transmission was effectively completed.
5342 * If the PMD is DEV_TX_OFFLOAD_MT_LOCKFREE capable, multiple threads can
5343 * invoke this function concurrently on the same tx queue without SW lock.
5344 * @see rte_eth_dev_info_get, struct rte_eth_txconf::offloads
5346 * @see rte_eth_tx_prepare to perform some prior checks or adjustments
5350 * The port identifier of the Ethernet device.
5352 * The index of the transmit queue through which output packets must be
5354 * The value must be in the range [0, nb_tx_queue - 1] previously supplied
5355 * to rte_eth_dev_configure().
5357 * The address of an array of *nb_pkts* pointers to *rte_mbuf* structures
5358 * which contain the output packets.
5360 * The maximum number of packets to transmit.
5362 * The number of output packets actually stored in transmit descriptors of
5363 * the transmit ring. The return value can be less than the value of the
5364 * *tx_pkts* parameter when the transmit ring is full or has been filled up.
5366 static inline uint16_t
5367 rte_eth_tx_burst(uint16_t port_id, uint16_t queue_id,
5368 struct rte_mbuf **tx_pkts, uint16_t nb_pkts)
5370 struct rte_eth_fp_ops *p;
5373 #ifdef RTE_ETHDEV_DEBUG_TX
5374 if (port_id >= RTE_MAX_ETHPORTS ||
5375 queue_id >= RTE_MAX_QUEUES_PER_PORT) {
5377 "Invalid port_id=%u or queue_id=%u\n",
5383 /* fetch pointer to queue data */
5384 p = &rte_eth_fp_ops[port_id];
5385 qd = p->txq.data[queue_id];
5387 #ifdef RTE_ETHDEV_DEBUG_TX
5388 RTE_ETH_VALID_PORTID_OR_ERR_RET(port_id, 0);
5391 RTE_ETHDEV_LOG(ERR, "Invalid TX queue_id=%u for port_id=%u\n",
5397 #ifdef RTE_ETHDEV_RXTX_CALLBACKS
5401 /* __ATOMIC_RELEASE memory order was used when the
5402 * call back was inserted into the list.
5403 * Since there is a clear dependency between loading
5404 * cb and cb->fn/cb->next, __ATOMIC_ACQUIRE memory order is
5407 cb = __atomic_load_n((void **)&p->txq.clbk[queue_id],
5409 if (unlikely(cb != NULL))
5410 nb_pkts = rte_eth_call_tx_callbacks(port_id, queue_id,
5411 tx_pkts, nb_pkts, cb);
5415 nb_pkts = p->tx_pkt_burst(qd, tx_pkts, nb_pkts);
5417 rte_ethdev_trace_tx_burst(port_id, queue_id, (void **)tx_pkts, nb_pkts);
5422 * Process a burst of output packets on a transmit queue of an Ethernet device.
5424 * The rte_eth_tx_prepare() function is invoked to prepare output packets to be
5425 * transmitted on the output queue *queue_id* of the Ethernet device designated
5427 * The *nb_pkts* parameter is the number of packets to be prepared which are
5428 * supplied in the *tx_pkts* array of *rte_mbuf* structures, each of them
5429 * allocated from a pool created with rte_pktmbuf_pool_create().
5430 * For each packet to send, the rte_eth_tx_prepare() function performs
5431 * the following operations:
5433 * - Check if packet meets devices requirements for tx offloads.
5435 * - Check limitations about number of segments.
5437 * - Check additional requirements when debug is enabled.
5439 * - Update and/or reset required checksums when tx offload is set for packet.
5441 * Since this function can modify packet data, provided mbufs must be safely
5442 * writable (e.g. modified data cannot be in shared segment).
5444 * The rte_eth_tx_prepare() function returns the number of packets ready to be
5445 * sent. A return value equal to *nb_pkts* means that all packets are valid and
5446 * ready to be sent, otherwise stops processing on the first invalid packet and
5447 * leaves the rest packets untouched.
5449 * When this functionality is not implemented in the driver, all packets are
5450 * are returned untouched.
5453 * The port identifier of the Ethernet device.
5454 * The value must be a valid port id.
5456 * The index of the transmit queue through which output packets must be
5458 * The value must be in the range [0, nb_tx_queue - 1] previously supplied
5459 * to rte_eth_dev_configure().
5461 * The address of an array of *nb_pkts* pointers to *rte_mbuf* structures
5462 * which contain the output packets.
5464 * The maximum number of packets to process.
5466 * The number of packets correct and ready to be sent. The return value can be
5467 * less than the value of the *tx_pkts* parameter when some packet doesn't
5468 * meet devices requirements with rte_errno set appropriately:
5469 * - EINVAL: offload flags are not correctly set
5470 * - ENOTSUP: the offload feature is not supported by the hardware
5471 * - ENODEV: if *port_id* is invalid (with debug enabled only)
5475 #ifndef RTE_ETHDEV_TX_PREPARE_NOOP
5477 static inline uint16_t
5478 rte_eth_tx_prepare(uint16_t port_id, uint16_t queue_id,
5479 struct rte_mbuf **tx_pkts, uint16_t nb_pkts)
5481 struct rte_eth_fp_ops *p;
5484 #ifdef RTE_ETHDEV_DEBUG_TX
5485 if (port_id >= RTE_MAX_ETHPORTS ||
5486 queue_id >= RTE_MAX_QUEUES_PER_PORT) {
5488 "Invalid port_id=%u or queue_id=%u\n",
5495 /* fetch pointer to queue data */
5496 p = &rte_eth_fp_ops[port_id];
5497 qd = p->txq.data[queue_id];
5499 #ifdef RTE_ETHDEV_DEBUG_TX
5500 if (!rte_eth_dev_is_valid_port(port_id)) {
5501 RTE_ETHDEV_LOG(ERR, "Invalid TX port_id=%u\n", port_id);
5506 RTE_ETHDEV_LOG(ERR, "Invalid TX queue_id=%u for port_id=%u\n",
5513 if (!p->tx_pkt_prepare)
5516 return p->tx_pkt_prepare(qd, tx_pkts, nb_pkts);
5522 * Native NOOP operation for compilation targets which doesn't require any
5523 * preparations steps, and functional NOOP may introduce unnecessary performance
5526 * Generally this is not a good idea to turn it on globally and didn't should
5527 * be used if behavior of tx_preparation can change.
5530 static inline uint16_t
5531 rte_eth_tx_prepare(__rte_unused uint16_t port_id,
5532 __rte_unused uint16_t queue_id,
5533 __rte_unused struct rte_mbuf **tx_pkts, uint16_t nb_pkts)
5541 * Send any packets queued up for transmission on a port and HW queue
5543 * This causes an explicit flush of packets previously buffered via the
5544 * rte_eth_tx_buffer() function. It returns the number of packets successfully
5545 * sent to the NIC, and calls the error callback for any unsent packets. Unless
5546 * explicitly set up otherwise, the default callback simply frees the unsent
5547 * packets back to the owning mempool.
5550 * The port identifier of the Ethernet device.
5552 * The index of the transmit queue through which output packets must be
5554 * The value must be in the range [0, nb_tx_queue - 1] previously supplied
5555 * to rte_eth_dev_configure().
5557 * Buffer of packets to be transmit.
5559 * The number of packets successfully sent to the Ethernet device. The error
5560 * callback is called for any packets which could not be sent.
5562 static inline uint16_t
5563 rte_eth_tx_buffer_flush(uint16_t port_id, uint16_t queue_id,
5564 struct rte_eth_dev_tx_buffer *buffer)
5567 uint16_t to_send = buffer->length;
5572 sent = rte_eth_tx_burst(port_id, queue_id, buffer->pkts, to_send);
5576 /* All packets sent, or to be dealt with by callback below */
5577 if (unlikely(sent != to_send))
5578 buffer->error_callback(&buffer->pkts[sent],
5579 (uint16_t)(to_send - sent),
5580 buffer->error_userdata);
5586 * Buffer a single packet for future transmission on a port and queue
5588 * This function takes a single mbuf/packet and buffers it for later
5589 * transmission on the particular port and queue specified. Once the buffer is
5590 * full of packets, an attempt will be made to transmit all the buffered
5591 * packets. In case of error, where not all packets can be transmitted, a
5592 * callback is called with the unsent packets as a parameter. If no callback
5593 * is explicitly set up, the unsent packets are just freed back to the owning
5594 * mempool. The function returns the number of packets actually sent i.e.
5595 * 0 if no buffer flush occurred, otherwise the number of packets successfully
5599 * The port identifier of the Ethernet device.
5601 * The index of the transmit queue through which output packets must be
5603 * The value must be in the range [0, nb_tx_queue - 1] previously supplied
5604 * to rte_eth_dev_configure().
5606 * Buffer used to collect packets to be sent.
5608 * Pointer to the packet mbuf to be sent.
5610 * 0 = packet has been buffered for later transmission
5611 * N > 0 = packet has been buffered, and the buffer was subsequently flushed,
5612 * causing N packets to be sent, and the error callback to be called for
5615 static __rte_always_inline uint16_t
5616 rte_eth_tx_buffer(uint16_t port_id, uint16_t queue_id,
5617 struct rte_eth_dev_tx_buffer *buffer, struct rte_mbuf *tx_pkt)
5619 buffer->pkts[buffer->length++] = tx_pkt;
5620 if (buffer->length < buffer->size)
5623 return rte_eth_tx_buffer_flush(port_id, queue_id, buffer);
5630 #endif /* _RTE_ETHDEV_H_ */