1 .. SPDX-License-Identifier: BSD-3-Clause
2 Copyright 2016 The DPDK contributors
10 * **Introduce ARMv7 and ARMv8 architectures.**
12 * It is now possible to build DPDK for the ARMv7 and ARMv8 platforms.
13 * ARMv7 can be tested with virtual PMDs.
14 * ARMv8 can be tested with virtual and physical PMDs.
16 * **Enabled freeing of ring.**
18 A new function ``rte_ring_free()`` has been added to allow the user to free
19 a ring if it was created with ``rte_ring_create()``.
21 * **Added keepalive support to EAL and example application.**
23 * **Added experimental cryptodev API**
25 The cryptographic processing of packets is provided as a preview
28 * Intel QuickAssist devices
29 * Intel AES-NI multi-buffer library
31 Due to its experimental state, the API may change without prior notice.
33 * **Added ethdev APIs for additional IEEE1588 support.**
35 Added functions to read, write and adjust system time in the NIC.
36 Added client slave sample application to demonstrate the IEEE1588
39 * **Extended Statistics.**
41 Defined an extended statistics naming scheme to store metadata in the name
42 string of each statistic. Refer to the Extended Statistics section of the
43 Programmers Guide for more details.
45 Implemented the extended statistics API for the following PMDs:
54 * **Added API in ethdev to retrieve RX/TX queue information.**
56 * Added the ability for the upper layer to query RX/TX queue information.
57 * Added new fields in ``rte_eth_dev_info`` to represent information about
58 RX/TX descriptors min/max/align numbers, per queue, for the device.
60 * **Added RSS dynamic configuration to bonding.**
62 * **Updated the e1000 base driver.**
64 The e1000 base driver was updated with several features including the
67 * Added new i218 devices
68 * Allowed both ULP and EEE in Sx state
69 * Initialized 88E1543 (Marvell 1543) PHY
70 * Added flags to set EEE advertisement modes
71 * Supported inverted format ETrackId
72 * Added bit to disable packetbuffer read
73 * Added defaults for i210 RX/TX PBSIZE
74 * Check more errors for ESB2 init and reset
75 * Check more NVM read errors
76 * Return code after setting receive address register
77 * Removed all NAHUM6LP_HW tags
79 * **Added e1000 RX interrupt support.**
81 * **Added igb TSO support for both PF and VF.**
83 * **Added RSS enhancements to Intel x550 NIC.**
85 * Added support for 512 entry RSS redirection table.
86 * Added support for per VF RSS redirection table.
88 * **Added Flow director enhancements on Intel x550 NIC.**
90 * Added 2 new flow director modes on x550.
91 One is MAC VLAN mode, the other is tunnel mode.
93 * **Updated the i40e base driver.**
95 The i40e base driver was updated with several changes including the
98 * Added promiscuous on VLAN support
99 * Added a workaround to drop all flow control frames
100 * Added VF capabilities to virtual channel interface
101 * Added TX Scheduling related AQ commands
102 * Added additional PCTYPES supported for FortPark RSS
103 * Added parsing for CEE DCBX TLVs
104 * Added FortPark specific registers
105 * Added AQ functions to handle RSS Key and LUT programming
106 * Increased PF reset max loop limit
108 * **Added i40e vector RX/TX.**
110 * **Added i40e RX interrupt support.**
112 * **Added i40e flow control support.**
114 * **Added DCB support to i40e PF driver.**
116 * **Added RSS/FD input set granularity on Intel X710/XL710.**
118 * **Added different GRE key length for input set on Intel X710/XL710.**
120 * **Added flow director support in i40e VF.**
122 * **Added i40e support of early X722 series.**
124 Added early X722 support, for evaluation only, as the hardware is alpha.
126 * **Added fm10k vector RX/TX.**
128 * **Added fm10k TSO support for both PF and VF.**
130 * **Added fm10k VMDQ support.**
132 * **New NIC Boulder Rapid support.**
134 Added support for the Boulder Rapid variant of Intel's fm10k NIC family.
136 * **Enhanced support for the Chelsio CXGBE driver.**
138 * Added support for Jumbo Frames.
139 * Optimized forwarding performance for Chelsio T5 40GbE cards.
141 * **Improved enic TX packet rate.**
143 Reduced frequency of TX tail pointer updates to the NIC.
145 * **Added support for link status interrupts in mlx4.**
147 * **Added partial support (TX only) for secondary processes in mlx4.**
149 * **Added support for Mellanox ConnectX-4 adapters (mlx5).**
151 The mlx5 poll-mode driver implements support for Mellanox ConnectX-4 EN
152 and Mellanox ConnectX-4 Lx EN families of 10/25/40/50/100 Gb/s adapters.
154 Like mlx4, this PMD is only available for Linux and is disabled by default
155 due to external dependencies (libibverbs and libmlx5).
157 * **Added driver for Netronome nfp-6xxx card.**
159 Support for using Netronome nfp-6xxx with PCI VFs.
161 * **Added virtual szedata2 driver for COMBO cards.**
163 Added virtual PMD for COMBO-100G and COMBO-80G cards.
164 PMD is disabled in default configuration.
166 * **Enhanced support for virtio driver.**
168 * Virtio ring layout optimization (fixed avail ring)
172 * **Added vhost-user multiple queue support.**
174 * **Added port hotplug support to vmxnet3.**
176 * **Added port hotplug support to xenvirt.**
178 * **Added ethtool shim and sample application.**
180 * **Added experimental performance thread example application.**
182 The new sample application demonstrates L3 forwarding with different threading
183 models: pthreads, cgroups, or lightweight threads. The example includes
184 a simple cooperative scheduler.
186 Due to its experimental state this application may change without notice.
187 The application is supported only for Linux x86_64.
189 * **Enhancements to the IP pipeline application.**
191 The following features have been added to the ``ip_pipeline``
194 * Added Multiple Producers/Multiple Consumers (MPSC)
195 and fragmentation/reassembly support to software rings.
197 * Added a dynamic pipeline reconfiguration feature that
198 allows binding a pipeline to other threads at runtime
201 * Added enable/disable of ``promisc`` mode from ``ip_pipeline``
204 * Added check on RX queues and TX queues of each link
205 whether they are used correctly in the ``ip_pipeline``
208 * Added flow id parameters to the flow-classification
211 * Added more functions to the routing pipeline:
212 ARP table enable/disable, Q-in-Q and MPLS encapsulation,
213 add color (traffic-class for QoS) to the MPLS tag.
215 * Added flow-actions pipeline for traffic metering/marking
216 (for e.g. Two Rate Three Color Marker (trTCM)), policer etc.
218 * Modified the pass-through pipeline's actions-handler to
219 implement a generic approach to extract fields from the
220 packet's header and copy them to packet metadata.
229 * **eal/linux: Fixed epoll timeout.**
231 Fixed issue where the ``rte_epoll_wait()`` function didn't return when the
232 underlying call to ``epoll_wait()`` timed out.
238 * **e1000/base: Synchronize PHY interface on non-ME systems.**
240 On power up, the MAC - PHY interface needs to be set to PCIe, even if the
241 cable is disconnected. In ME systems, the ME handles this on exit from the
242 Sx (Sticky mode) state. In non-ME, the driver handles it. Added a check for
243 non-ME system to the driver code that handles it.
245 * **e1000/base: Increased timeout of reset check.**
247 Previously, in ``check_reset_block`` RSPCIPHY was polled for 100 ms before
248 determining that the ME veto was set. This was not enough and it was
251 * **e1000/base: Disabled IPv6 extension header parsing on 82575.**
253 Disabled IPv6 options as per hardware limitation.
255 * **e1000/base: Prevent ULP flow if cable connected.**
257 Enabling ULP on link down when the cable is connected caused an infinite
258 loop of link up/down indications in the NDIS driver.
259 The driver now enables ULP only when the cable is disconnected.
261 * **e1000/base: Support different EEARBC for i210.**
263 EEARBC has changed on i210. It means EEARBC has a different address on
264 i210 than on other NICs. So, add a new entity named EEARBC_I210 to the
265 register list and make sure the right one is being used on i210.
267 * **e1000/base: Fix K1 configuration.**
269 Added fix for the following updates to the K1 configurations:
270 TX idle period for entering K1 should be 128 ns.
271 Minimum TX idle period in K1 should be 256 ns.
273 * **e1000/base: Fix link detect flow.**
275 Fix link detect flow in case where auto-negotiate is not enabled, by calling
276 ``e1000_setup_copper_link_generic`` instead of ``e1000_phy_setup_autoneg``.
278 * **e1000/base: Fix link check for i354 M88E1112 PHY.**
280 The ``e1000_check_for_link_media_swap()`` function is supposed to check PHY
281 page 0 for copper and PHY page 1 for "other" (fiber) links. The driver
282 switched back from page 1 to page 0 too soon, before
283 ``e1000_check_for_link_82575()`` is executed and was never finding the link
284 on the fiber (other).
286 If the link is copper, as the M88E1112 page address is set to 1, it should be
287 set back to 0 before checking this link.
289 * **e1000/base: Fix beacon duration for i217.**
291 Fix for I217 Packet Loss issue - The Management Engine sets the FEXTNVM4
292 Beacon Duration incorrectly. This fix ensures that the correct value will
293 always be set. Correct value for this field is 8 usec.
295 * **e1000/base: Fix TIPG for non 10 half duplex mode.**
297 TIPG value is increased when setting speed to 10 half duplex to prevent
298 packet loss. However, it was never decreased again when speed
299 changed. This caused performance issues in the NDIS driver.
300 Fix this to restore TIPG to default value on non 10 half duplex.
302 * **e1000/base: Fix reset of DH89XXCC SGMII.**
304 For DH89XXCC_SGMII, a write flush leaves registers of this device trashed
305 (0xFFFFFFFF). Add check for this device.
307 Also, after both Port SW Reset and Device Reset case, the platform should
308 wait at least 3ms before reading any registers. Remove this condition since
309 waiting is conditionally executed only for Device Reset.
311 * **e1000/base: Fix redundant PHY power down for i210.**
313 Bit 11 of PHYREG 0 is used to power down PHY. The use of PHYREG 16 is
316 * **e1000/base: fix jumbo frame CRC failures.**
318 Change the value of register 776.20[11:2] for jumbo mode from 0x1A to 0x1F.
319 This is to enlarge the gap between read and write pointers in the TX FIFO.
321 * **e1000/base: Fix link flap on 82579.**
323 Several customers have reported a link flap issue on 82579. The symptoms
324 are random and intermittent link losses when 82579 is connected to specific
325 switches. the Issue was root caused as an interoperability problem between
326 the NIC and at least some Broadcom PHYs in the Energy Efficient Ethernet
329 To fix the issue, we are disabling the Phase Locked Loop shutdown in 100M
330 Low Power Idle. This solution will cause an increase of power in 100M EEE
331 link. It may cost an additional 28mW in this specific mode.
333 * **igb: Fixed IEEE1588 frame identification in I210.**
335 Fixed issue where the flag ``PKT_RX_IEEE1588_PTP`` was not being set
336 in the Intel I210 NIC, as the EtherType in RX descriptor is in bits 8:10 of
337 Packet Type and not in the default bits 0:2.
339 * **igb: Fixed VF start with PF stopped.**
341 VF needs the PF interrupt support initialized even if not started.
343 * **igb: Fixed VF MAC address when using with DPDK PF.**
345 Assign a random MAC address in VF when not assigned by PF.
347 * **igb: Removed CRC bytes from byte counter statistics.**
349 * **ixgbe: Fixed issue with X550 DCB.**
351 Fixed a DCB issue with x550 where for 8 TCs (Traffic Classes), if a packet
352 with user priority 6 or 7 was injected to the NIC, then the NIC would only
353 put 3 packets into the queue. There was also a similar issue for 4 TCs.
355 * **ixgbe: Removed burst size restriction of vector RX.**
357 Fixed issue where a burst size less than 32 didn't receive anything.
359 * **ixgbe: Fixed VF start with PF stopped.**
361 VF needs the PF interrupt support initialized even if not started.
363 * **ixgbe: Fixed TX hang when RS distance exceeds HW limit.**
365 Fixed an issue where the TX queue can hang when a lot of highly fragmented
366 packets have to be sent. As part of that fix, ``tx_rs_thresh`` for ixgbe PMD
367 is not allowed to be greater then to 32 to comply with HW restrictions.
369 * **ixgbe: Fixed rx error statistic counter.**
371 Fixed an issue that the rx error counter of ixgbe was not accurate. The
372 mac short packet discard count (mspdc) was added to the counter. Mac local
373 faults and mac remote faults are removed as they do not count packets but
374 errors, and jabber errors were removed as they are already accounted for
375 by the CRC error counter. Finally the XEC (l3 / l4 checksum error) counter
376 was removed due to errata, see commit 256ff05a9cae for details.
378 * **ixgbe: Removed CRC bytes from byte counter statistics.**
380 * **i40e: Fixed base driver allocation when not using first numa node.**
382 Fixed i40e issue that occurred when a DPDK application didn't initialize
383 ports if memory wasn't available on socket 0.
385 * **i40e: Fixed maximum of 64 queues per port.**
387 Fixed an issue in i40e where it would not support more than 64 queues per
388 port, even though the hardware actually supports it. The real number of
389 queues may vary, as long as the total number of queues used in PF, VFs, VMDq
390 and FD does not exceeds the hardware maximum.
392 * **i40e: Fixed statistics of packets.**
394 Added discarding packets on VSI to the stats and rectify the old statistics.
396 * **i40e: Fixed issue of not freeing memzone.**
398 Fixed an issue of not freeing a memzone in the call to free the memory for
401 * **i40e: Removed CRC bytes from byte counter statistics.**
403 * **mlx: Fixed driver loading.**
405 The mlx drivers were unable to load when built as a shared library,
406 due to a missing symbol in the mempool library.
408 * **mlx4: Performance improvements.**
410 Fixed bugs in TX and RX flows that improves mlx4 performance.
412 * **mlx4: Fixed TX loss after initialization.**
414 * **mlx4: Fixed scattered TX with too many segments.**
416 * **mlx4: Fixed memory registration for indirect mbuf data.**
418 * **vhost: Fixed Qemu shutdown.**
420 Fixed issue with libvirt ``virsh destroy`` not killing the VM.
422 * **virtio: Fixed crash after changing link state.**
424 Fixed IO permission in the interrupt handler.
426 * **virtio: Fixed crash when releasing queue.**
428 Fixed issue when releasing null control queue.
434 * **hash: Fixed memory allocation of Cuckoo Hash key table.**
436 Fixed issue where an incorrect Cuckoo Hash key table size could be
437 calculated limiting the size to 4GB.
439 * **hash: Fixed incorrect lookup if key is all zero.**
441 Fixed issue in hash library that occurred if an all zero
442 key was not added to the table and the key was looked up,
443 resulting in an incorrect hit.
445 * **hash: Fixed thread scaling by reducing contention.**
447 Fixed issue in the hash library where, using multiple cores with
448 hardware transactional memory support, thread scaling did not work,
449 due to the global ring that is shared by all cores.
455 * **l3fwd: Fixed crash with IPv6.**
457 * **vhost_xen: Fixed compile error.**
463 * This release drops compatibility with Linux kernel 2.6.33. The minimum
464 kernel requirement is now 2.6.34.
470 * Some drivers do not fill in the packet type when receiving.
471 As the l3fwd example application requires this info, the i40e vector
472 driver must be disabled to benefit of the packet type with i40e.
474 * Some (possibly all) VF drivers (e.g. i40evf) do not handle any PF reset
475 events/requests in the VF driver. This means that the VF driver may not work
476 after a PF reset in the host side. The workaround is to avoid triggering any
477 PF reset events/requests on the host side.
479 * 100G link report support is missing.
481 * **Mellanox PMDs (mlx4 & mlx5):**
483 * PMDs do not support CONFIG_RTE_BUILD_COMBINE_LIBS and
484 CONFIG_RTE_BUILD_SHARED_LIB simultaneously.
486 * There is performance degradation for small packets when the PMD is
487 compiled with ``SGE_WR_N = 4`` compared to the performance when ``SGE_WR_N
488 = 1``. If scattered packets are not used it is recommended to compile the
489 PMD with ``SGE_WR_N = 1``.
491 * When a Multicast or Broadcast packet is sent to the SR-IOV mlx4 VF,
492 it is returned back to the port.
494 * PMDs report "bad" L4 checksum when IP packet is received.
496 * mlx5 PMD reports "bad" checksum although the packet has "good" checksum.
497 Will be fixed in upcoming MLNX_OFED release.
503 * The deprecated flow director API is removed.
504 It was replaced by ``rte_eth_dev_filter_ctrl()``.
506 * The ``dcb_queue`` is renamed to ``dcb_tc`` in following dcb configuration
507 structures: ``rte_eth_dcb_rx_conf``, ``rte_eth_dcb_tx_conf``,
508 ``rte_eth_vmdq_dcb_conf``, ``rte_eth_vmdq_dcb_tx_conf``.
510 * The ``rte_eth_rx_queue_count()`` function now returns "int" instead of
511 "uint32_t" to allow the use of negative values as error codes on return.
513 * The function ``rte_eal_pci_close_one()`` is removed.
514 It was replaced by ``rte_eal_pci_detach()``.
516 * The deprecated ACL API ``ipv4vlan`` is removed.
518 * The deprecated hash function ``rte_jhash2()`` is removed.
519 It was replaced by ``rte_jhash_32b()``.
521 * The deprecated KNI functions are removed:
522 ``rte_kni_create()``, ``rte_kni_get_port_id()`` and ``rte_kni_info_get()``.
524 * The deprecated ring PMD functions are removed:
525 ``rte_eth_ring_pair_create()`` and ``rte_eth_ring_pair_attach()``.
527 * The devargs union field ``virtual`` is renamed to ``virt`` for C++
534 * The EAL and ethdev structures ``rte_intr_handle`` and ``rte_eth_conf`` were
535 changed to support RX interrupt. This was already included in 2.1 under the
536 ``CONFIG_RTE_NEXT_ABI`` #define.
538 * The ethdev flow director entries for SCTP were changed.
539 This was already included in 2.1 under the ``CONFIG_RTE_NEXT_ABI`` #define.
541 * The ethdev flow director structure ``rte_eth_fdir_flow_ext`` structure was
542 changed. New fields were added to support flow director filtering in VF.
544 * The size of the ethdev structure ``rte_eth_hash_filter_info`` is changed
545 by adding a new element ``rte_eth_input_set_conf`` in a union.
547 * New fields ``rx_desc_lim`` and ``tx_desc_lim`` are added into
548 ``rte_eth_dev_info`` structure.
550 * For debug builds, the functions ``rte_eth_rx_burst()``, ``rte_eth_tx_burst()``
551 ``rte_eth_rx_descriptor_done()`` and ``rte_eth_rx_queue_count()`` will
552 no longer be separate functions in the DPDK libraries. Instead, they will
553 only be present in the ``rte_ethdev.h`` header file.
555 * The maximum number of queues per port ``CONFIG_RTE_MAX_QUEUES_PER_PORT`` is
558 * The mbuf structure was changed to support the unified packet type.
559 This was already included in 2.1 under the ``CONFIG_RTE_NEXT_ABI`` #define.
561 * The dummy malloc library is removed. The content was moved into EAL in 2.1.
563 * The LPM structure is changed. The deprecated field ``mem_location`` is
566 * librte_table LPM: A new parameter to hold the table name will be added to
567 the LPM table parameter structure.
569 * librte_table hash: The key mask parameter is added to the hash table
570 parameter structure for 8-byte key and 16-byte key extendable bucket
573 * librte_port: Macros to access the packet meta-data stored within the packet
574 buffer has been adjusted to cover the packet mbuf structure.
576 * librte_cfgfile: Allow longer names and values by increasing the constants
577 ``CFG_NAME_LEN`` and ``CFG_VALUE_LEN`` to 64 and 256 respectively.
579 * vhost: a new field enabled is added to the ``vhost_virtqueue`` structure.
581 * vhost: a new field ``virt_qp_nb`` is added to ``virtio_net`` structure, and
582 the ``virtqueue`` field is moved to the end of virtio_net structure.
584 * vhost: a new operation ``vring_state_changed`` is added to
585 ``virtio_net_device_ops`` structure.
587 * vhost: a few spaces are reserved both at ``vhost_virtqueue`` and
588 ``virtio_net`` structure for future extension.
591 Shared Library Versions
592 -----------------------
594 The libraries prepended with a plus sign were incremented in this version.
600 + librte_cfgfile.so.2
602 librte_distributor.so.1
614 + librte_pipeline.so.2
616 + librte_pmd_ring.so.2