7 * **Introduce ARMv7 and ARMv8 architectures**
9 * It is now possible to build DPDK for the ARMv7 and ARMv8 platforms.
10 * ARMv7 can be tested with virtual PMD drivers.
11 * ARMv8 can be tested with virtual and physical PMD drivers.
13 * **Enabled freeing of ring.**
15 New function rte_ring_free() allows the user to free a ring
16 if it was created with rte_ring_create().
18 * **Added keepalive support to EAL and example application.**
20 * **Added experimental cryptodev API**
22 The cryptographic processing of packet is provided as a preview
25 * Intel QuickAssist devices
26 * Intel AES-NI multi-buffer library
28 Due to its experimental state, the API may change without prior notice.
30 * **Added ethdev API to support IEEE1588.**
32 Added functions to read, write and adjust system time in the NIC.
34 * **Extended Statistics**
36 Define extended statistics naming scheme to store metadata in the name
37 string name of each statistic, refer to the Extended Statistics section
38 of the programmers guide. Implemented the extended stats API for these
48 * **Added API in ethdev to retrieve RX/TX queue information.**
50 * Add the ability for the upper layer to query RX/TX queue information.
51 * Add into rte_eth_dev_info new fields to represent information about
52 RX/TX descriptors min/max/align numbers per queue for the device.
54 * **Added RSS dynamic configuration to bonding.**
56 * **Updated the e1000 base driver.**
58 The e1000 base driver was updated with several features including the
61 * Add new i218 devices
62 * Allow both ULP and EEE in Sx state
63 * Initialize 88E1543 (Marvell 1543) PHY
64 * Add flags to set EEE advertisement modes
65 * Support inverted format ETrackId
66 * Add bit to disable packetbuffer read
67 * Add defaults for i210 Rx/Tx PBSIZE
68 * Check more errors for ESB2 init and reset
69 * Check more NVM read errors
70 * Return code after setting receive address register
71 * Remove all NAHUM6LP_HW tags
73 * **Added e1000 Rx interrupt support.**
75 * **Added igb TSO support for both PF and VF.**
77 * **RSS enhancement on Intel x550 NIC**
79 * Support 512 entries RSS redirection table.
80 * Support per VF RSS redirection table.
82 * **Flow director enhancement on Intel x550 NIC**
84 * Add 2 new flow director modes on x550.
85 * One is MAC VLAN mode, the other is tunnel mode.
87 * **Updated the i40e base driver.**
89 The i40e base driver was updated with several changes including the
92 * Added promiscuous on VLAN support
93 * Added a workaround to drop all flow control frames
94 * Added VF capabilities to virtual channel interface
95 * Added Tx Scheduling related AQ commands
96 * Added additional PCTYPES supported for FortPark RSS
97 * Added parsing for CEE DCBX TLVs
98 * Added FortPark specific registers
99 * Added AQ functions to handle RSS Key and LUT programming
100 * Increased pf reset max loop limit
102 * **Added i40e vector RX/TX.**
104 * **Added i40e Rx interrupt support.**
106 * **Added i40e flow control support.**
108 * **Added DCB support to i40e PF driver.**
110 * **Added RSS/FD input set granularity on Intel X710/XL710.**
112 * **Added different GRE key length for input set on Intel X710/XL710.**
114 * **Added flow director support in i40e VF.**
116 * **Added i40e support of early X722 series.**
118 Add early X722 support for evaluation only, as the hardware is in A0.
120 * **Added fm10k vector RX/TX.**
122 * **Added fm10k TSO support for both PF and VF.**
124 * **Added fm10k VMDQ support.**
126 * **New NIC Boulder Rapid support.**
128 Boulder Rapid is a new NIC of Intel's fm10k family.
130 * **Enhanced support for the Chelsio CXGBE driver.**
132 * Added support for Jumbo Frames.
133 * Optimize forwarding performance for Chelsio T5 40GbE cards.
135 * **Improved enic Tx packet rate.**
137 Reduced frequency of Tx tail pointer updates to the NIC.
139 * **Added support for link status interrupts in mlx4.**
141 * **Added partial support (TX only) for secondary processes in mlx4.**
143 * **Added support for Mellanox ConnectX-4 adapters (mlx5).**
145 The mlx5 poll-mode driver implements support for Mellanox ConnectX-4 EN
146 and Mellanox ConnectX-4 Lx EN families of 10/25/40/50/100 Gb/s adapters.
148 Like mlx4, this PMD is only available for Linux and is disabled by default
149 due to external dependencies (libibverbs and libmlx5).
151 * **Added driver for Netronome nfp-6xxx card.**
153 Support for using Netronome nfp-6xxx with PCI VFs.
155 * **Added virtual szedata2 driver for COMBO cards.**
157 Added virtual PMD for COMBO-100G and COMBO-80G cards.
158 PMD is disabled in default configuration.
160 * **Enhanced support for virtio driver.**
162 * Virtio ring layout optimization (fixed avail ring)
166 * **Added vhost-user multiple queue support.**
168 * **Added port hotplug support to vmxnet3.**
170 * **Added port hotplug support to xenvirt.**
172 * **Added ethtool shim and sample application.**
174 * **Added experimental performance thread example application.**
176 The application demonstrates L3 fowarding with different threading
177 models: pthreads, cgroups, or lighweight threads. The example inludes
178 a simple cooperative scheduler.
180 Due to its experimental state this application may change without notice.
181 The application is supported only for Linux x86_64.
190 * **eal/linux: Fixed epoll timeout.**
192 Fixed issue where the ``rte_epoll_wait()`` function didn't return when the
193 underlying call to ``epoll_wait()`` timed out.
199 * **e1000/base: Synchronize PHY interface on non-ME systems.**
201 On power up, the MAC - PHY interface needs to be set to PCIe, even if
202 cable is disconnected. In ME systems, the ME handles this on exit from
203 Sx(Sticky mode) state. In non-ME, the driver handles it. Add a check
204 for non-ME system to the driver code that handles that.
206 * **e1000/base: Increase timeout of reset check.**
208 Previously, in check_reset_block RSPCIPHY was polled for 100 ms before
209 determining that the ME veto is set. It's not enough. It need to be increased
212 * **e1000/base: Disable IPv6 extension header parsing on 82575.**
214 Disable IPv6 options as per hardware limitation.
216 * **e1000/base: Prevent ULP flow if cable connected.**
218 Enabling ulp on link down when cable is connect caused an infinite
219 loop of linkup/down indications in the NDIS driver.
220 After discussed, correct flow is to enable ULP only when cable is
223 * **e1000/base: Support different EEARBC for i210.**
225 EEARBC has changed on i210. It means EEARBC has a different address on
226 i210 than on other NICs. So, add a new entity named EEARBC_I210 to the
227 register list and make sure the right one is being used on i210.
229 * **e1000/base: Fix K1 configuration**
231 This patch is for the following updates to the K1 configurations:
232 Tx idle period for entering K1 should be 128 ns.
233 Minimum Tx idle period in K1 should be 256 ns.
235 * **e1000/base: Fix link detect flow**
237 In case that auto-negotiate is not enabled, call
238 e1000_setup_copper_link_generic instead of e1000_phy_setup_autoneg.
240 * **e1000/base: Fix link check for i354 M88E1112 PHY**
242 e1000_check_for_link_media_swap() is supposed to check PHY page 0 for
243 copper and PHY page 1 for "other" (fiber) link. We switched back from
244 page 1 to page 0 too soon, before e1000_check_for_link_82575() is
245 executed and we were never finding link on fiber (other).
247 If the link is copper, as the M88E1112 page address is set to 1, it should be
248 set back to 0 before checking this link.
250 * **e1000/base: Fix beacon duration for i217**
252 Fix for I217 Packet Loss issue - The Management Engine sets the FEXTNVM4
253 Beacon Duration incorrectly. This fix ensures that the correct value will
254 always be set. Correct value for this field is 8 usec.
256 * **e1000/base: Fix TIPG for non 10 half duplex mode**
258 TIPG value is increased when setting speed to 10 half to prevent
259 packet loss. However, it was never decreased again when speed
260 changes. This caused performance issues in the NDIS driver.
261 Fix this to restore TIPG to default value on non 10 half.
263 * **e1000/base: Fix reset of DH89XXCC SGMII**
265 For DH89XXCC_SGMII, write flush leaves registers of this device trashed
266 (0xFFFFFFFF). Add check for this device.
267 Also, after both for Port SW Reset and Device Reset case, platform should
268 wait at least 3ms before reading any registers. Since waiting is
269 conditionally executed only for Device Reset - remove the condition.
271 * **e1000/base: Fix redundant PHY power down for i210**
273 Bit 11 of PHYREG 0 is used to power down PHY. The use of PHYREG 16 is
274 unnecessary any more.
276 * **e1000/base: fix jumbo frame CRC failures**
278 Change the value of register 776.20[11:2] for jumbo mode from 0x1A to 0x1F.
279 This is to enlarge the gap between read and write pointers in the TX Fifo.
280 And replace the magic number with a macro by the way.
282 * **e1000/base: Fix link flap on 82579**
284 Several customers have reported a link flap issue on 82579. The symptoms
285 are random and intermittent link losses when 82579 is connected to specific
286 switches. Issue has been root caused as interoperability problem between
287 the NIC and at least some Broadcom PHYs in the Energy Efficient Ethernet
289 To fix the issue, we are disabling the Phase Locked Loop shutdown in 100M
290 Low Power Idle. This solution will cause an increase of power in 100M EEE
291 link. It may cost additional 28mW in this specific mode.
293 * **igb: Fixed IEEE1588 frame identification in I210.**
295 Fixed issue where the flag PKT_RX_IEEE1588_PTP was not being set
296 in Intel I210 NIC, as EtherType in RX descriptor is in bits 8:10 of
297 Packet Type and not in the default bits 0:2.
299 * **igb: Fixed VF start with PF stopped.**
301 VF needs the PF interrupt support initialized even if not started.
303 * **igb: Fixed VF MAC address when using with DPDK PF.**
305 Assign a random MAC address in VF when not assigned by PF.
307 * **ixgbe: Fixed issue with X550 DCB.**
309 Fixed a DCB issue with x550 where for 8 TCs (Traffic Classes), if a packet
310 with user priority 6 or 7 was injected to the NIC, then the NIC would only
311 put 3 packets into the queue. There was also a similar issue for 4 TCs.
313 * **ixgbe: Removed burst size restriction of vector RX.**
315 Fixed issue where a burst size less than 32 didn't receive anything.
317 * **ixgbe: Fixed VF start with PF stopped.**
319 VF needs the PF interrupt support initialized even if not started.
321 * **ixgbe: Fixed TX hang when RS distance exceeds HW limit.**
323 Fixed an issue when TX queue can hang when a lot of highly fragmented
324 packets have to be sent.
325 As part of that fix, tx_rs_thresh for ixgbe PMD is not allowed to be greater
326 then to 32 to comply with HW restrictions.
328 * **i40e: Fixed base driver allocation when not using first numa node.**
330 Fixed i40e issue that occurred when a DPDK application didn't initialize
331 ports if memory wasn't available on socket 0.
333 * **i40e: Fixed maximum of 64 queues per port.**
335 Fixed the issue in i40e of cannot supporting more than 64 queues per port,
336 though hardware actually supports that. The real number of queues may vary,
337 as long as the total number of queues used in PF, VFs, VMDq and FD does not
338 exceeds the hardware maximum.
340 * **i40e: Fixed statistics of packets.**
342 Added discarding packets on VSI to the stats and rectify the old statistics.
344 * **i40e: Fixed issue of not freeing memzone.**
346 Fixed the issue of not freeing memzone in the call to free the memory for
349 * **mlx: Fixed driver loading.**
351 The mlx drivers were unable to load when built as a shared library,
352 due to a missing symbol in mempool library.
354 * **mlx4: Performance improvements.**
356 Fixed bugs in TX and RX flows that improves mlx4 performance.
358 * **mlx4: Fixed Tx loss after initialization.**
360 * **mlx4: Fixed scattered Tx with too many segments.**
362 * **mlx4: Fixed memory registration for indirect mbuf data.**
364 * **vhost: Fixed Qemu shutdown.**
366 Fixed issue with libvirt ``virsh destroy`` not killing the VM.
368 * **virtio: Fixed crash after changing link state.**
370 Fixed io permission in the interrupt handler.
372 * **virtio: Fixed crash when releasing queue.**
374 Fixed issue when releasing null control queue.
380 * **hash: Fixed memory allocation of Cuckoo Hash key table.**
382 Fixed issue where an incorrect Cuckoo Hash key table size could be
383 calculated limiting the size to 4GB.
385 * **hash: Fixed incorrect lookup if key is all zero.**
387 Fixed issue in hash library that occurred if an all zero
388 key was not added in the table and the key was looked up,
389 resulting in an incorrect hit.
391 * **hash: Fixed thread scaling by reducing contention.**
393 Fixed issue in hash library where, using multiple cores with
394 hardware transactional memory support, thread scaling did not work,
395 due to the global ring that is shared by all cores.
401 * **l3fwd: Fixed crash with IPv6.**
403 * **vhost_xen: Fixed compile error.**
409 * This release drops compatibility with Linux kernel 2.6.33. The minimum
410 kernel requirement is now 2.6.34.
416 * Some drivers do not fill the packet type when receiving.
417 As the l3fwd example application requires this info, the i40e vector
418 driver must be disabled to benefit of the packet type with i40e.
420 * **Mellanox PMDs (mlx4 & mlx5):**
422 * PMDs do not support DPDK integrated shared library.
424 * There is performance degradation for small packets when PMD
425 is compiled with SGE_WR_N = 4 compared to the performance when SGE_WR_N = 1.
426 If scattered packets are not used it is recomended
427 to compile PMD with SGE_WR_N = 1.
429 * When a Multicast or Broadcast packet is sent to the SR-IOV mlx4 VF,
430 it is returned back to the port.
436 * The deprecated flow director API is removed.
437 It was replaced by rte_eth_dev_filter_ctrl().
439 * The dcb_queue is renamed to dcb_tc in following dcb configuration
440 structures: rte_eth_dcb_rx_conf, rte_eth_dcb_tx_conf,
441 rte_eth_vmdq_dcb_conf, rte_eth_vmdq_dcb_tx_conf.
443 * The rte_eth_rx_queue_count() function now returns "int" instead of "uint32_t"
444 to allow the use of negative values as error codes on return.
446 * The function rte_eal_pci_close_one() is removed.
447 It was replaced by rte_eal_pci_detach().
449 * The deprecated ACL API ipv4vlan is removed.
451 * The deprecated hash function rte_jhash2() is removed.
452 It was replaced by rte_jhash_32b().
454 * The deprecated KNI functions are removed:
455 rte_kni_create(), rte_kni_get_port_id() and rte_kni_info_get().
457 * The deprecated ring PMD functions are removed:
458 rte_eth_ring_pair_create() and rte_eth_ring_pair_attach().
460 * The devargs union field virtual is renamed to virt for C++ compatibility.
465 * The EAL and ethdev structures rte_intr_handle and rte_eth_conf were changed
466 to support Rx interrupt. It was already done in 2.1 for CONFIG_RTE_NEXT_ABI.
468 * The ethdev flow director entries for SCTP were changed.
469 It was already done in 2.1 for CONFIG_RTE_NEXT_ABI.
471 * The ethdev flow director structure rte_eth_fdir_flow_ext structure is changed.
472 New fields are added to support flow director filtering in VF.
474 * The size of the ethdev structure rte_eth_hash_filter_info is changed
475 by adding a new element rte_eth_input_set_conf in an union.
477 * The new fields rx_desc_lim and tx_desc_lim are added into rte_eth_dev_info
480 * For debug builds, the functions rte_eth_rx_burst(), rte_eth_tx_burst()
481 rte_eth_rx_descriptor_done() and rte_eth_rx_queue_count() will
482 no longer be separate functions in the DPDK libraries. Instead, they will
483 only be present in the rte_ethdev.h header file.
485 * The maximum number of queues per port CONFIG_RTE_MAX_QUEUES_PER_PORT is
488 * The mbuf structure was changed to support unified packet type.
489 It was already done in 2.1 for CONFIG_RTE_NEXT_ABI.
491 * The dummy malloc library is removed. The content was moved into EAL in 2.1.
493 * The LPM structure is changed. The deprecated field mem_location is removed.
495 * librte_table LPM: A new parameter to hold the table name will be added to
496 the LPM table parameter structure.
498 * librte_table hash: The key mask parameter is added to the hash table
499 parameter structure for 8-byte key and 16-byte key extendible bucket
502 * librte_port: Macros to access the packet meta-data stored within the packet
503 buffer has been adjusted to cover the packet mbuf structure.
505 * librte_cfgfile: Allow longer names and values by increasing the constants
506 CFG_NAME_LEN and CFG_VALUE_LEN to 64 and 256 respectively.
509 Shared Library Versions
510 -----------------------
512 The libraries prepended with a plus sign were incremented in this version.
518 + librte_cfgfile.so.2
520 librte_distributor.so.1
532 + librte_pipeline.so.2
534 + librte_pmd_ring.so.2