1 .. SPDX-License-Identifier: BSD-3-Clause
2 Copyright(c) 2010-2015 Intel Corporation.
9 The DPDK includes 1 Gigabit, 10 Gigabit and 40 Gigabit and para virtualized virtio Poll Mode Drivers.
11 A Poll Mode Driver (PMD) consists of APIs, provided through the BSD driver running in user space,
12 to configure the devices and their respective queues.
13 In addition, a PMD accesses the RX and TX descriptors directly without any interrupts
14 (with the exception of Link Status Change interrupts) to quickly receive,
15 process and deliver packets in the user's application.
16 This section describes the requirements of the PMDs,
17 their global design principles and proposes a high-level architecture and a generic external API for the Ethernet PMDs.
19 Requirements and Assumptions
20 ----------------------------
22 The DPDK environment for packet processing applications allows for two models, run-to-completion and pipe-line:
24 * In the *run-to-completion* model, a specific port's RX descriptor ring is polled for packets through an API.
25 Packets are then processed on the same core and placed on a port's TX descriptor ring through an API for transmission.
27 * In the *pipe-line* model, one core polls one or more port's RX descriptor ring through an API.
28 Packets are received and passed to another core via a ring.
29 The other core continues to process the packet which then may be placed on a port's TX descriptor ring through an API for transmission.
31 In a synchronous run-to-completion model,
32 each logical core assigned to the DPDK executes a packet processing loop that includes the following steps:
34 * Retrieve input packets through the PMD receive API
36 * Process each received packet one at a time, up to its forwarding
38 * Send pending output packets through the PMD transmit API
40 Conversely, in an asynchronous pipe-line model, some logical cores may be dedicated to the retrieval of received packets and
41 other logical cores to the processing of previously received packets.
42 Received packets are exchanged between logical cores through rings.
43 The loop for packet retrieval includes the following steps:
45 * Retrieve input packets through the PMD receive API
47 * Provide received packets to processing lcores through packet queues
49 The loop for packet processing includes the following steps:
51 * Retrieve the received packet from the packet queue
53 * Process the received packet, up to its retransmission if forwarded
55 To avoid any unnecessary interrupt processing overhead, the execution environment must not use any asynchronous notification mechanisms.
56 Whenever needed and appropriate, asynchronous communication should be introduced as much as possible through the use of rings.
58 Avoiding lock contention is a key issue in a multi-core environment.
59 To address this issue, PMDs are designed to work with per-core private resources as much as possible.
60 For example, a PMD maintains a separate transmit queue per-core, per-port, if the PMD is not ``RTE_ETH_TX_OFFLOAD_MT_LOCKFREE`` capable.
61 In the same way, every receive queue of a port is assigned to and polled by a single logical core (lcore).
63 To comply with Non-Uniform Memory Access (NUMA), memory management is designed to assign to each logical core
64 a private buffer pool in local memory to minimize remote memory access.
65 The configuration of packet buffer pools should take into account the underlying physical memory architecture in terms of DIMMS,
67 The application must ensure that appropriate parameters are given at memory pool creation time.
68 See :ref:`Mempool Library <Mempool_Library>`.
73 The API and architecture of the Ethernet* PMDs are designed with the following guidelines in mind.
75 PMDs must help global policy-oriented decisions to be enforced at the upper application level.
76 Conversely, NIC PMD functions should not impede the benefits expected by upper-level global policies,
77 or worse prevent such policies from being applied.
79 For instance, both the receive and transmit functions of a PMD have a maximum number of packets/descriptors to poll.
80 This allows a run-to-completion processing stack to statically fix or
81 to dynamically adapt its overall behavior through different global loop policies, such as:
83 * Receive, process immediately and transmit packets one at a time in a piecemeal fashion.
85 * Receive as many packets as possible, then process all received packets, transmitting them immediately.
87 * Receive a given maximum number of packets, process the received packets, accumulate them and finally send all accumulated packets to transmit.
89 To achieve optimal performance, overall software design choices and pure software optimization techniques must be considered and
90 balanced against available low-level hardware-based optimization features (CPU cache properties, bus speed, NIC PCI bandwidth, and so on).
91 The case of packet transmission is an example of this software/hardware tradeoff issue when optimizing burst-oriented network packet processing engines.
92 In the initial case, the PMD could export only an rte_eth_tx_one function to transmit one packet at a time on a given queue.
93 On top of that, one can easily build an rte_eth_tx_burst function that loops invoking the rte_eth_tx_one function to transmit several packets at a time.
94 However, an rte_eth_tx_burst function is effectively implemented by the PMD to minimize the driver-level transmit cost per packet through the following optimizations:
96 * Share among multiple packets the un-amortized cost of invoking the rte_eth_tx_one function.
98 * Enable the rte_eth_tx_burst function to take advantage of burst-oriented hardware features (prefetch data in cache, use of NIC head/tail registers)
99 to minimize the number of CPU cycles per packet, for example by avoiding unnecessary read memory accesses to ring transmit descriptors,
100 or by systematically using arrays of pointers that exactly fit cache line boundaries and sizes.
102 * Apply burst-oriented software optimization techniques to remove operations that would otherwise be unavoidable, such as ring index wrap back management.
104 Burst-oriented functions are also introduced via the API for services that are intensively used by the PMD.
105 This applies in particular to buffer allocators used to populate NIC rings, which provide functions to allocate/free several buffers at a time.
106 For example, an mbuf_multiple_alloc function returning an array of pointers to rte_mbuf buffers which speeds up the receive poll function of the PMD when
107 replenishing multiple descriptors of the receive ring.
109 Logical Cores, Memory and NIC Queues Relationships
110 --------------------------------------------------
112 The DPDK supports NUMA allowing for better performance when a processor's logical cores and interfaces utilize its local memory.
113 Therefore, mbuf allocation associated with local PCIe* interfaces should be allocated from memory pools created in the local memory.
114 The buffers should, if possible, remain on the local processor to obtain the best performance results and RX and TX buffer descriptors
115 should be populated with mbufs allocated from a mempool allocated from local memory.
117 The run-to-completion model also performs better if packet or data manipulation is in local memory instead of a remote processors memory.
118 This is also true for the pipe-line model provided all logical cores used are located on the same processor.
120 Multiple logical cores should never share receive or transmit queues for interfaces since this would require global locks and hinder performance.
122 If the PMD is ``RTE_ETH_TX_OFFLOAD_MT_LOCKFREE`` capable, multiple threads can invoke ``rte_eth_tx_burst()``
123 concurrently on the same tx queue without SW lock. This PMD feature found in some NICs and useful in the following use cases:
125 * Remove explicit spinlock in some applications where lcores are not mapped to Tx queues with 1:1 relation.
127 * In the eventdev use case, avoid dedicating a separate TX core for transmitting and thus
128 enables more scaling as all workers can send the packets.
130 See `Hardware Offload`_ for ``RTE_ETH_TX_OFFLOAD_MT_LOCKFREE`` capability probing details.
132 Device Identification, Ownership and Configuration
133 --------------------------------------------------
135 Device Identification
136 ~~~~~~~~~~~~~~~~~~~~~
138 Each NIC port is uniquely designated by its (bus/bridge, device, function) PCI
139 identifiers assigned by the PCI probing/enumeration function executed at DPDK initialization.
140 Based on their PCI identifier, NIC ports are assigned two other identifiers:
142 * A port index used to designate the NIC port in all functions exported by the PMD API.
144 * A port name used to designate the port in console messages, for administration or debugging purposes.
145 For ease of use, the port name includes the port index.
150 The Ethernet devices ports can be owned by a single DPDK entity (application, library, PMD, process, etc).
151 The ownership mechanism is controlled by ethdev APIs and allows to set/remove/get a port owner by DPDK entities.
152 It prevents Ethernet ports to be managed by different entities.
156 It is the DPDK entity responsibility to set the port owner before using it and to manage the port usage synchronization between different threads or processes.
158 It is recommended to set port ownership early,
159 like during the probing notification ``RTE_ETH_EVENT_NEW``.
164 The configuration of each NIC port includes the following operations:
166 * Allocate PCI resources
168 * Reset the hardware (issue a Global Reset) to a well-known default state
170 * Set up the PHY and the link
172 * Initialize statistics counters
174 The PMD API must also export functions to start/stop the all-multicast feature of a port and functions to set/unset the port in promiscuous mode.
176 Some hardware offload features must be individually configured at port initialization through specific configuration parameters.
177 This is the case for the Receive Side Scaling (RSS) and Data Center Bridging (DCB) features for example.
179 On-the-Fly Configuration
180 ~~~~~~~~~~~~~~~~~~~~~~~~
182 All device features that can be started or stopped "on the fly" (that is, without stopping the device) do not require the PMD API to export dedicated functions for this purpose.
184 All that is required is the mapping address of the device PCI registers to implement the configuration of these features in specific functions outside of the drivers.
187 the PMD API exports a function that provides all the information associated with a device that can be used to set up a given device feature outside of the driver.
188 This includes the PCI vendor identifier, the PCI device identifier, the mapping address of the PCI device registers, and the name of the driver.
190 The main advantage of this approach is that it gives complete freedom on the choice of the API used to configure, to start, and to stop such features.
192 As an example, refer to the configuration of the IEEE1588 feature for the Intel® 82576 Gigabit Ethernet Controller and
193 the Intel® 82599 10 Gigabit Ethernet Controller controllers in the testpmd application.
195 Other features such as the L3/L4 5-Tuple packet filtering feature of a port can be configured in the same way.
196 Ethernet* flow control (pause frame) can be configured on the individual port.
197 Refer to the testpmd source code for details.
198 Also, L4 (UDP/TCP/ SCTP) checksum offload by the NIC can be enabled for an individual packet as long as the packet mbuf is set up correctly. See `Hardware Offload`_ for details.
200 Configuration of Transmit Queues
201 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
203 Each transmit queue is independently configured with the following information:
205 * The number of descriptors of the transmit ring
207 * The socket identifier used to identify the appropriate DMA memory zone from which to allocate the transmit ring in NUMA architectures
209 * The values of the Prefetch, Host and Write-Back threshold registers of the transmit queue
211 * The *minimum* transmit packets to free threshold (tx_free_thresh).
212 When the number of descriptors used to transmit packets exceeds this threshold, the network adaptor should be checked to see if it has written back descriptors.
213 A value of 0 can be passed during the TX queue configuration to indicate the default value should be used.
214 The default value for tx_free_thresh is 32.
215 This ensures that the PMD does not search for completed descriptors until at least 32 have been processed by the NIC for this queue.
217 * The *minimum* RS bit threshold. The minimum number of transmit descriptors to use before setting the Report Status (RS) bit in the transmit descriptor.
218 Note that this parameter may only be valid for Intel 10 GbE network adapters.
219 The RS bit is set on the last descriptor used to transmit a packet if the number of descriptors used since the last RS bit setting,
220 up to the first descriptor used to transmit the packet, exceeds the transmit RS bit threshold (tx_rs_thresh).
221 In short, this parameter controls which transmit descriptors are written back to host memory by the network adapter.
222 A value of 0 can be passed during the TX queue configuration to indicate that the default value should be used.
223 The default value for tx_rs_thresh is 32.
224 This ensures that at least 32 descriptors are used before the network adapter writes back the most recently used descriptor.
225 This saves upstream PCIe* bandwidth resulting from TX descriptor write-backs.
226 It is important to note that the TX Write-back threshold (TX wthresh) should be set to 0 when tx_rs_thresh is greater than 1.
227 Refer to the Intel® 82599 10 Gigabit Ethernet Controller Datasheet for more details.
229 The following constraints must be satisfied for tx_free_thresh and tx_rs_thresh:
231 * tx_rs_thresh must be greater than 0.
233 * tx_rs_thresh must be less than the size of the ring minus 2.
235 * tx_rs_thresh must be less than or equal to tx_free_thresh.
237 * tx_free_thresh must be greater than 0.
239 * tx_free_thresh must be less than the size of the ring minus 3.
241 * For optimal performance, TX wthresh should be set to 0 when tx_rs_thresh is greater than 1.
243 One descriptor in the TX ring is used as a sentinel to avoid a hardware race condition, hence the maximum threshold constraints.
247 When configuring for DCB operation, at port initialization, both the number of transmit queues and the number of receive queues must be set to 128.
249 Free Tx mbuf on Demand
250 ~~~~~~~~~~~~~~~~~~~~~~
252 Many of the drivers do not release the mbuf back to the mempool, or local cache,
253 immediately after the packet has been transmitted.
254 Instead, they leave the mbuf in their Tx ring and
255 either perform a bulk release when the ``tx_rs_thresh`` has been crossed
256 or free the mbuf when a slot in the Tx ring is needed.
258 An application can request the driver to release used mbufs with the ``rte_eth_tx_done_cleanup()`` API.
259 This API requests the driver to release mbufs that are no longer in use,
260 independent of whether or not the ``tx_rs_thresh`` has been crossed.
261 There are two scenarios when an application may want the mbuf released immediately:
263 * When a given packet needs to be sent to multiple destination interfaces
264 (either for Layer 2 flooding or Layer 3 multi-cast).
265 One option is to make a copy of the packet or a copy of the header portion that needs to be manipulated.
266 A second option is to transmit the packet and then poll the ``rte_eth_tx_done_cleanup()`` API
267 until the reference count on the packet is decremented.
268 Then the same packet can be transmitted to the next destination interface.
269 The application is still responsible for managing any packet manipulations needed
270 between the different destination interfaces, but a packet copy can be avoided.
271 This API is independent of whether the packet was transmitted or dropped,
272 only that the mbuf is no longer in use by the interface.
274 * Some applications are designed to make multiple runs, like a packet generator.
275 For performance reasons and consistency between runs,
276 the application may want to reset back to an initial state
277 between each run, where all mbufs are returned to the mempool.
278 In this case, it can call the ``rte_eth_tx_done_cleanup()`` API
279 for each destination interface it has been using
280 to request it to release of all its used mbufs.
282 To determine if a driver supports this API, check for the *Free Tx mbuf on demand* feature
283 in the *Network Interface Controller Drivers* document.
288 Depending on driver capabilities advertised by
289 ``rte_eth_dev_info_get()``, the PMD may support hardware offloading
290 feature like checksumming, TCP segmentation, VLAN insertion or
291 lockfree multithreaded TX burst on the same TX queue.
293 The support of these offload features implies the addition of dedicated
294 status bit(s) and value field(s) into the rte_mbuf data structure, along
295 with their appropriate handling by the receive/transmit functions
296 exported by each PMD. The list of flags and their precise meaning is
297 described in the mbuf API documentation and in the in :ref:`Mbuf Library
298 <Mbuf_Library>`, section "Meta Information".
300 Per-Port and Per-Queue Offloads
301 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
303 In the DPDK offload API, offloads are divided into per-port and per-queue offloads as follows:
305 * A per-queue offloading can be enabled on a queue and disabled on another queue at the same time.
306 * A pure per-port offload is the one supported by device but not per-queue type.
307 * A pure per-port offloading can't be enabled on a queue and disabled on another queue at the same time.
308 * A pure per-port offloading must be enabled or disabled on all queues at the same time.
309 * Any offloading is per-queue or pure per-port type, but can't be both types at same devices.
310 * Port capabilities = per-queue capabilities + pure per-port capabilities.
311 * Any supported offloading can be enabled on all queues.
313 The different offloads capabilities can be queried using ``rte_eth_dev_info_get()``.
314 The ``dev_info->[rt]x_queue_offload_capa`` returned from ``rte_eth_dev_info_get()`` includes all per-queue offloading capabilities.
315 The ``dev_info->[rt]x_offload_capa`` returned from ``rte_eth_dev_info_get()`` includes all pure per-port and per-queue offloading capabilities.
316 Supported offloads can be either per-port or per-queue.
318 Offloads are enabled using the existing ``RTE_ETH_TX_OFFLOAD_*`` or ``RTE_ETH_RX_OFFLOAD_*`` flags.
319 Any requested offloading by an application must be within the device capabilities.
320 Any offloading is disabled by default if it is not set in the parameter
321 ``dev_conf->[rt]xmode.offloads`` to ``rte_eth_dev_configure()`` and
322 ``[rt]x_conf->offloads`` to ``rte_eth_[rt]x_queue_setup()``.
324 If any offloading is enabled in ``rte_eth_dev_configure()`` by an application,
325 it is enabled on all queues no matter whether it is per-queue or
326 per-port type and no matter whether it is set or cleared in
327 ``[rt]x_conf->offloads`` to ``rte_eth_[rt]x_queue_setup()``.
329 If a per-queue offloading hasn't been enabled in ``rte_eth_dev_configure()``,
330 it can be enabled or disabled in ``rte_eth_[rt]x_queue_setup()`` for individual queue.
331 A newly added offloads in ``[rt]x_conf->offloads`` to ``rte_eth_[rt]x_queue_setup()`` input by application
332 is the one which hasn't been enabled in ``rte_eth_dev_configure()`` and is requested to be enabled
333 in ``rte_eth_[rt]x_queue_setup()``. It must be per-queue type, otherwise trigger an error log.
341 By default, all functions exported by a PMD are lock-free functions that are assumed
342 not to be invoked in parallel on different logical cores to work on the same target object.
343 For instance, a PMD receive function cannot be invoked in parallel on two logical cores to poll the same RX queue of the same port.
344 Of course, this function can be invoked in parallel by different logical cores on different RX queues.
345 It is the responsibility of the upper-level application to enforce this rule.
347 If needed, parallel accesses by multiple logical cores to shared queues can be explicitly protected by dedicated inline lock-aware functions
348 built on top of their corresponding lock-free functions of the PMD API.
350 Generic Packet Representation
351 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
353 A packet is represented by an rte_mbuf structure, which is a generic metadata structure containing all necessary housekeeping information.
354 This includes fields and status bits corresponding to offload hardware features, such as checksum computation of IP headers or VLAN tags.
356 The rte_mbuf data structure includes specific fields to represent, in a generic way, the offload features provided by network controllers.
357 For an input packet, most fields of the rte_mbuf structure are filled in by the PMD receive function with the information contained in the receive descriptor.
358 Conversely, for output packets, most fields of rte_mbuf structures are used by the PMD transmit function to initialize transmit descriptors.
360 The mbuf structure is fully described in the :ref:`Mbuf Library <Mbuf_Library>` chapter.
365 The Ethernet device API exported by the Ethernet PMDs is described in the *DPDK API Reference*.
367 .. _ethernet_device_standard_device_arguments:
369 Ethernet Device Standard Device Arguments
370 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
372 Standard Ethernet device arguments allow for a set of commonly used arguments/
373 parameters which are applicable to all Ethernet devices to be available to for
374 specification of specific device and for passing common configuration
375 parameters to those ports.
377 * ``representor`` for a device which supports the creation of representor ports
378 this argument allows user to specify which switch ports to enable port
379 representors for. Multiple representors in one device argument is invalid::
381 -a DBDF,representor=vf0
382 -a DBDF,representor=vf[0,4,6,9]
383 -a DBDF,representor=vf[0-31]
384 -a DBDF,representor=vf[0,2-4,7,9-11]
385 -a DBDF,representor=sf0
386 -a DBDF,representor=sf[1,3,5]
387 -a DBDF,representor=sf[0-1023]
388 -a DBDF,representor=sf[0,2-4,7,9-11]
389 -a DBDF,representor=pf1vf0
390 -a DBDF,representor=pf[0-1]sf[0-127]
391 -a DBDF,representor=pf1
393 Note: PMDs are not required to support the standard device arguments and users
394 should consult the relevant PMD documentation to see support devargs.
396 Extended Statistics API
397 ~~~~~~~~~~~~~~~~~~~~~~~
399 The extended statistics API allows a PMD to expose all statistics that are
400 available to it, including statistics that are unique to the device.
401 Each statistic has three properties ``name``, ``id`` and ``value``:
403 * ``name``: A human readable string formatted by the scheme detailed below.
404 * ``id``: An integer that represents only that statistic.
405 * ``value``: A unsigned 64-bit integer that is the value of the statistic.
407 Note that extended statistic identifiers are
408 driver-specific, and hence might not be the same for different ports.
409 The API consists of various ``rte_eth_xstats_*()`` functions, and allows an
410 application to be flexible in how it retrieves statistics.
412 Scheme for Human Readable Names
413 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
415 A naming scheme exists for the strings exposed to clients of the API. This is
416 to allow scraping of the API for statistics of interest. The naming scheme uses
417 strings split by a single underscore ``_``. The scheme is as follows:
425 Examples of common statistics xstats strings, formatted to comply to the scheme
430 * ``tx_multicast_packets``
432 The scheme, although quite simple, allows flexibility in presenting and reading
433 information from the statistic strings. The following example illustrates the
434 naming scheme:``rx_packets``. In this example, the string is split into two
435 components. The first component ``rx`` indicates that the statistic is
436 associated with the receive side of the NIC. The second component ``packets``
437 indicates that the unit of measure is packets.
439 A more complicated example: ``tx_size_128_to_255_packets``. In this example,
440 ``tx`` indicates transmission, ``size`` is the first detail, ``128`` etc are
441 more details, and ``packets`` indicates that this is a packet counter.
443 Some additions in the metadata scheme are as follows:
445 * If the first part does not match ``rx`` or ``tx``, the statistic does not
446 have an affinity with either receive of transmit.
448 * If the first letter of the second part is ``q`` and this ``q`` is followed
449 by a number, this statistic is part of a specific queue.
451 An example where queue numbers are used is as follows: ``tx_q7_bytes`` which
452 indicates this statistic applies to queue number 7, and represents the number
453 of transmitted bytes on that queue.
458 The xstats API uses the ``name``, ``id``, and ``value`` to allow performant
459 lookup of specific statistics. Performant lookup means two things;
461 * No string comparisons with the ``name`` of the statistic in fast-path
462 * Allow requesting of only the statistics of interest
464 The API ensures these requirements are met by mapping the ``name`` of the
465 statistic to a unique ``id``, which is used as a key for lookup in the fast-path.
466 The API allows applications to request an array of ``id`` values, so that the
467 PMD only performs the required calculations. Expected usage is that the
468 application scans the ``name`` of each statistic, and caches the ``id``
469 if it has an interest in that statistic. On the fast-path, the integer can be used
470 to retrieve the actual ``value`` of the statistic that the ``id`` represents.
475 The API is built out of a small number of functions, which can be used to
476 retrieve the number of statistics and the names, IDs and values of those
479 * ``rte_eth_xstats_get_names_by_id()``: returns the names of the statistics. When given a
480 ``NULL`` parameter the function returns the number of statistics that are available.
482 * ``rte_eth_xstats_get_id_by_name()``: Searches for the statistic ID that matches
483 ``xstat_name``. If found, the ``id`` integer is set.
485 * ``rte_eth_xstats_get_by_id()``: Fills in an array of ``uint64_t`` values
486 with matching the provided ``ids`` array. If the ``ids`` array is NULL, it
487 returns all statistics that are available.
493 Imagine an application that wants to view the dropped packet count. If no
494 packets are dropped, the application does not read any other metrics for
495 performance reasons. If packets are dropped, the application has a particular
496 set of statistics that it requests. This "set" of statistics allows the app to
497 decide what next steps to perform. The following code-snippets show how the
498 xstats API can be used to achieve this goal.
500 First step is to get all statistics names and list them:
504 struct rte_eth_xstat_name *xstats_names;
508 /* Get number of stats */
509 len = rte_eth_xstats_get_names_by_id(port_id, NULL, NULL, 0);
511 printf("Cannot get xstats count\n");
515 xstats_names = malloc(sizeof(struct rte_eth_xstat_name) * len);
516 if (xstats_names == NULL) {
517 printf("Cannot allocate memory for xstat names\n");
521 /* Retrieve xstats names, passing NULL for IDs to return all statistics */
522 if (len != rte_eth_xstats_get_names_by_id(port_id, xstats_names, NULL, len)) {
523 printf("Cannot get xstat names\n");
527 values = malloc(sizeof(values) * len);
528 if (values == NULL) {
529 printf("Cannot allocate memory for xstats\n");
533 /* Getting xstats values */
534 if (len != rte_eth_xstats_get_by_id(port_id, NULL, values, len)) {
535 printf("Cannot get xstat values\n");
539 /* Print all xstats names and values */
540 for (i = 0; i < len; i++) {
541 printf("%s: %"PRIu64"\n", xstats_names[i].name, values[i]);
544 The application has access to the names of all of the statistics that the PMD
545 exposes. The application can decide which statistics are of interest, cache the
546 ids of those statistics by looking up the name as follows:
552 const char *xstat_name = "rx_errors";
554 if(!rte_eth_xstats_get_id_by_name(port_id, xstat_name, &id)) {
555 rte_eth_xstats_get_by_id(port_id, &id, &value, 1);
556 printf("%s: %"PRIu64"\n", xstat_name, value);
559 printf("Cannot find xstats with a given name\n");
563 The API provides flexibility to the application so that it can look up multiple
564 statistics using an array containing multiple ``id`` numbers. This reduces the
565 function call overhead of retrieving statistics, and makes lookup of multiple
566 statistics simpler for the application.
570 #define APP_NUM_STATS 4
571 /* application cached these ids previously; see above */
572 uint64_t ids_array[APP_NUM_STATS] = {3,4,7,21};
573 uint64_t value_array[APP_NUM_STATS];
575 /* Getting multiple xstats values from array of IDs */
576 rte_eth_xstats_get_by_id(port_id, ids_array, value_array, APP_NUM_STATS);
579 for(i = 0; i < APP_NUM_STATS; i++) {
580 printf("%d: %"PRIu64"\n", ids_array[i], value_array[i]);
584 This array lookup API for xstats allows the application create multiple
585 "groups" of statistics, and look up the values of those IDs using a single API
586 call. As an end result, the application is able to achieve its goal of
587 monitoring a single statistic ("rx_errors" in this case), and if that shows
588 packets being dropped, it can easily retrieve a "set" of statistics using the
589 IDs array parameter to ``rte_eth_xstats_get_by_id`` function.
596 int rte_eth_dev_reset(uint16_t port_id);
598 Sometimes a port has to be reset passively. For example when a PF is
599 reset, all its VFs should also be reset by the application to make them
600 consistent with the PF. A DPDK application also can call this function
601 to trigger a port reset. Normally, a DPDK application would invokes this
602 function when an RTE_ETH_EVENT_INTR_RESET event is detected.
604 It is the duty of the PMD to trigger RTE_ETH_EVENT_INTR_RESET events and
605 the application should register a callback function to handle these
606 events. When a PMD needs to trigger a reset, it can trigger an
607 RTE_ETH_EVENT_INTR_RESET event. On receiving an RTE_ETH_EVENT_INTR_RESET
608 event, applications can handle it as follows: Stop working queues, stop
609 calling Rx and Tx functions, and then call rte_eth_dev_reset(). For
610 thread safety all these operations should be called from the same thread.
612 For example when PF is reset, the PF sends a message to notify VFs of
613 this event and also trigger an interrupt to VFs. Then in the interrupt
614 service routine the VFs detects this notification message and calls
615 rte_eth_dev_callback_process(dev, RTE_ETH_EVENT_INTR_RESET, NULL).
616 This means that a PF reset triggers an RTE_ETH_EVENT_INTR_RESET
617 event within VFs. The function rte_eth_dev_callback_process() will
618 call the registered callback function. The callback function can trigger
619 the application to handle all operations the VF reset requires including
620 stopping Rx/Tx queues and calling rte_eth_dev_reset().
622 The rte_eth_dev_reset() itself is a generic function which only does
623 some hardware reset operations through calling dev_unint() and
624 dev_init(), and itself does not handle synchronization, which is handled
627 The PMD itself should not call rte_eth_dev_reset(). The PMD can trigger
628 the application to handle reset event. It is duty of application to
629 handle all synchronization before it calls rte_eth_dev_reset().