1.. SPDX-License-Identifier: BSD-3-Clause 2 Copyright 2016 The DPDK contributors 3 4DPDK Release 16.07 5================== 6 7New Features 8------------ 9 10* **Removed the mempool cache memory if caching is not being used.** 11 12 The size of the mempool structure is reduced if the per-lcore cache is disabled. 13 14* **Added mempool external cache for non-EAL thread.** 15 16 Added new functions to create, free or flush a user-owned mempool 17 cache for non-EAL threads. Previously the caching was always disabled 18 on these threads. 19 20* **Changed the memory allocation scheme in the mempool library.** 21 22 * Added the ability to allocate a large mempool in fragmented virtual memory. 23 * Added new APIs to populate a mempool with memory. 24 * Added an API to free a mempool. 25 * Modified the API of the ``rte_mempool_obj_iter()`` function. 26 * Dropped the specific Xen Dom0 code. 27 * Dropped the specific anonymous mempool code in testpmd. 28 29* **Added a new driver for Broadcom NetXtreme-C devices.** 30 31 Added the new bnxt driver for Broadcom NetXtreme-C devices. See the 32 "Network Interface Controller Drivers" document for more details on this 33 new driver. 34 35* **Added a new driver for ThunderX nicvf devices.** 36 37 Added the new thunderx net driver for ThunderX nicvf devices. See the 38 "Network Interface Controller Drivers" document for more details on this new 39 driver. 40 41* **Added mailbox interrupt support for ixgbe and igb VFs.** 42 43 When the physical NIC link comes up or down, the PF driver will send a 44 mailbox message to notify each VF. To handle this link up/down event, 45 support have been added for a mailbox interrupt to receive the message and 46 allow the application to register a callback for it. 47 48* **Updated the ixgbe base driver.** 49 50 The ixgbe base driver was updated with changes including the 51 following: 52 53 * Added sgmii link for X550. 54 * Added MAC link setup for X550a SFP and SFP+. 55 * Added KR support for X550em_a. 56 * Added new PHY definitions for M88E1500. 57 * Added support for the VLVF to be bypassed when adding/removing a VFTA entry. 58 * Added X550a flow control auto negotiation support. 59 60* **Updated the i40e base driver.** 61 62 Updated the i40e base driver including support for new devices IDs. 63 64* **Updated the enic driver.** 65 66 The enic driver was updated with changes including the following: 67 68 * Optimized the Tx function. 69 * Added Scattered Rx capability. 70 * Improved packet type identification. 71 * Added MTU update in non Scattered Rx mode and enabled MTU of up to 9208 72 with UCS Software release 2.2 on 1300 series VICs. 73 74* **Updated the mlx5 driver.** 75 76 The mlx5 driver was updated with changes including the following: 77 78 * Data path was refactored to bypass Verbs to improve RX and TX performance. 79 * Removed compilation parameters for inline send, ``MLX5_MAX_INLINE``, and 80 added command line parameter instead, ``txq_inline``. 81 * Improved TX scatter gather support: 82 Removed compilation parameter ``MLX5_PMD_SGE_WR_N``. 83 Scatter-gather elements is set to the maximum value the NIC supports. 84 Removed linearization logic, this decreases the memory consumption of the PMD. 85 * Improved jumbo frames support, by dynamically setting RX scatter gather elements 86 according to the MTU and mbuf size, 87 no need for compilation parameter ``MLX5_PMD_SGE_WR_N`` 88 89* **Added support for virtio on IBM POWER8.** 90 91 The ioports are mapped in memory when using Linux UIO. 92 93* **Added support for Virtio in containers.** 94 95 Add a new virtual device, named virtio_user, to support virtio for containers. 96 97 Known limitations: 98 99 * Control queue and multi-queue are not supported yet. 100 * Doesn't work with ``--huge-unlink``. 101 * Doesn't work with ``--no-huge``. 102 * Doesn't work when there are more than ``VHOST_MEMORY_MAX_NREGIONS(8)`` hugepages. 103 * Root privilege is required for sorting hugepages by physical address. 104 * Can only be used with the vhost user backend. 105 106* **Added vhost-user client mode.** 107 108 DPDK vhost-user now supports client mode as well as server mode. Client mode 109 is enabled when the ``RTE_VHOST_USER_CLIENT`` flag is set while calling 110 ``rte_vhost_driver_register``. 111 112 When DPDK vhost-user restarts from an normal or abnormal exit (such as a 113 crash), the client mode allows DPDK to establish the connection again. Note 114 that QEMU version v2.7 or above is required for this feature. 115 116 DPDK vhost-user will also try to reconnect by default when: 117 118 * The first connect fails (for example when QEMU is not started yet). 119 * The connection is broken (for example when QEMU restarts). 120 121 It can be turned off by setting the ``RTE_VHOST_USER_NO_RECONNECT`` flag. 122 123* **Added NSH packet recognition in i40e.** 124 125* **Added AES-CTR support to AESNI MB PMD.** 126 127 Now AESNI MB PMD supports 128/192/256-bit counter mode AES encryption and 128 decryption. 129 130* **Added AES counter mode support for Intel QuickAssist devices.** 131 132 Enabled support for the AES CTR algorithm for Intel QuickAssist devices. 133 Provided support for algorithm-chaining operations. 134 135* **Added KASUMI SW PMD.** 136 137 A new Crypto PMD has been added, which provides KASUMI F8 (UEA1) ciphering 138 and KASUMI F9 (UIA1) hashing. 139 140* **Added multi-writer support for RTE Hash with Intel TSX.** 141 142 The following features/modifications have been added to rte_hash library: 143 144 * Enabled application developers to use an extra flag for ``rte_hash`` 145 creation to specify default behavior (multi-thread safe/unsafe) with the 146 ``rte_hash_add_key`` function. 147 * Changed the Cuckoo Hash Search algorithm to breadth first search for 148 multi-writer routines and split Cuckoo Hash Search and Move operations in 149 order to reduce transactional code region and improve TSX performance. 150 * Added a hash multi-writer test case to the test app. 151 152* **Improved IP Pipeline Application.** 153 154 The following features have been added to the ip_pipeline application: 155 156 * Configure the MAC address in the routing pipeline and automatic route 157 updates with change in link state. 158 * Enable RSS per network interface through the configuration file. 159 * Streamline the CLI code. 160 161* **Added keepalive enhancements.** 162 163 Added support for reporting of core states other than "dead" to 164 monitoring applications, enabling the support of broader liveness 165 reporting to external processes. 166 167* **Added packet capture framework.** 168 169 * A new library ``librte_pdump`` is added to provide a packet capture API. 170 * A new ``app/pdump`` tool is added to demonstrate capture packets in DPDK. 171 172 173* **Added floating VEB support for i40e PF driver.** 174 175 A "floating VEB" is a special Virtual Ethernet Bridge (VEB) which does not 176 have an upload port, but instead is used for switching traffic between 177 virtual functions (VFs) on a port. 178 179 For information on this feature, please see the "I40E Poll Mode Driver" 180 section of the "Network Interface Controller Drivers" document. 181 182* **Added support for live migration of a VM with SRIOV VF.** 183 184 Live migration of a VM with Virtio and VF PMD's using the bonding PMD. 185 186 187Resolved Issues 188--------------- 189 190EAL 191~~~ 192 193* **igb_uio: Fixed possible mmap failure for Linux >= 4.5.** 194 195 The mmapping of the iomem range of the PCI device fails for kernels that 196 enabled the ``CONFIG_IO_STRICT_DEVMEM`` option. The error seen by the 197 user is as similar to the following:: 198 199 EAL: pci_map_resource(): 200 201 cannot mmap(39, 0x7f1c51800000, 0x100000, 0x0): 202 Invalid argument (0xffffffffffffffff) 203 204 The ``CONFIG_IO_STRICT_DEVMEM`` kernel option was introduced in Linux v4.5. 205 206 The issues was resolve by updating ``igb_uio`` to stop reserving PCI memory 207 resources. From the kernel point of view the iomem region looks like idle 208 and mmap works again. This matches the ``uio_pci_generic`` usage. 209 210 211Drivers 212~~~~~~~ 213 214* **i40e: Fixed vlan stripping from inner header.** 215 216 Previously, for tunnel packets, such as VXLAN/NVGRE, the vlan 217 tags of the inner header will be stripped without putting vlan 218 info to descriptor. 219 Now this issue is fixed by disabling vlan stripping from inner header. 220 221* **i40e: Fixed the type issue of a single VLAN type.** 222 223 Currently, if a single VLAN header is added in a packet, it's treated 224 as inner VLAN. But generally, a single VLAN header is treated as the 225 outer VLAN header. 226 This issue is fixed by changing corresponding register for single VLAN. 227 228* **enic: Fixed several issues when stopping then restarting ports and queues.** 229 230 Fixed several crashes related to stopping then restarting ports and queues. 231 Fixed possible crash when re-configuring the number of Rx queue descriptors. 232 233* **enic: Fixed Rx data mis-alignment if mbuf data offset modified.** 234 235 Fixed possible Rx corruption when mbufs were returned to a pool with data 236 offset other than RTE_PKTMBUF_HEADROOM. 237 238* **enic: Fixed Tx IP/UDP/TCP checksum offload and VLAN insertion.** 239 240* **enic: Fixed Rx error and missed counters.** 241 242 243Libraries 244~~~~~~~~~ 245 246* **mbuf: Fixed refcnt update when detaching.** 247 248 Fix the ``rte_pktmbuf_detach()`` function to decrement the direct mbuf's 249 reference counter. The previous behavior was not to affect the reference 250 counter. This lead to a memory leak of the direct mbuf. 251 252 253API Changes 254----------- 255 256* The following counters are removed from the ``rte_eth_stats`` structure: 257 258 * ``ibadcrc`` 259 * ``ibadlen`` 260 * ``imcasts`` 261 * ``fdirmatch`` 262 * ``fdirmiss`` 263 * ``tx_pause_xon`` 264 * ``rx_pause_xon`` 265 * ``tx_pause_xoff`` 266 * ``rx_pause_xoff`` 267 268* The extended statistics are fetched by ids with ``rte_eth_xstats_get`` 269 after a lookup by name ``rte_eth_xstats_get_names``. 270 271* The function ``rte_eth_dev_info_get`` fill the new fields ``nb_rx_queues`` 272 and ``nb_tx_queues`` in the structure ``rte_eth_dev_info``. 273 274* The vhost function ``rte_vring_available_entries`` is renamed to 275 ``rte_vhost_avail_entries``. 276 277* All existing vhost APIs and callbacks with ``virtio_net`` struct pointer 278 as the parameter have been changed due to the ABI refactoring described 279 below. It is replaced by ``int vid``. 280 281* The function ``rte_vhost_enqueue_burst`` no longer supports concurrent enqueuing 282 packets to the same queue. 283 284* The function ``rte_eth_dev_set_mtu`` adds a new return value ``-EBUSY``, which 285 indicates the operation is forbidden because the port is running. 286 287* The script ``dpdk_nic_bind.py`` is renamed to ``dpdk-devbind.py``. 288 And the script ``setup.sh`` is renamed to ``dpdk-setup.sh``. 289 290 291ABI Changes 292----------- 293 294* The ``rte_port_source_params`` structure has new fields to support PCAP files. 295 It was already in release 16.04 with ``RTE_NEXT_ABI`` flag. 296 297* The ``rte_eth_dev_info`` structure has new fields ``nb_rx_queues`` and ``nb_tx_queues`` 298 to support the number of queues configured by software. 299 300* A Vhost ABI refactoring has been made: the ``virtio_net`` structure is no 301 longer exported directly to the application. Instead, a handle, ``vid``, has 302 been used to represent this structure internally. 303 304 305Shared Library Versions 306----------------------- 307 308The libraries prepended with a plus sign were incremented in this version. 309 310.. code-block:: diff 311 312 + libethdev.so.4 313 librte_acl.so.2 314 librte_cfgfile.so.2 315 librte_cmdline.so.2 316 librte_cryptodev.so.1 317 librte_distributor.so.1 318 librte_eal.so.2 319 librte_hash.so.2 320 librte_ip_frag.so.1 321 librte_ivshmem.so.1 322 librte_jobstats.so.1 323 librte_kni.so.2 324 librte_kvargs.so.1 325 librte_lpm.so.2 326 librte_mbuf.so.2 327 + librte_mempool.so.2 328 librte_meter.so.1 329 librte_pdump.so.1 330 librte_pipeline.so.3 331 librte_pmd_bond.so.1 332 librte_pmd_ring.so.2 333 + librte_port.so.3 334 librte_power.so.1 335 librte_reorder.so.1 336 librte_ring.so.1 337 librte_sched.so.1 338 librte_table.so.2 339 librte_timer.so.1 340 + librte_vhost.so.3 341 342 343Tested Platforms 344---------------- 345 346#. SuperMicro 1U 347 348 - BIOS: 1.0c 349 - Processor: Intel(R) Atom(TM) CPU C2758 @ 2.40GHz 350 351#. SuperMicro 1U 352 353 - BIOS: 1.0a 354 - Processor: Intel(R) Xeon(R) CPU D-1540 @ 2.00GHz 355 - Onboard NIC: Intel(R) X552/X557-AT (2x10G) 356 357 - Firmware-version: 0x800001cf 358 - Device ID (PF/VF): 8086:15ad /8086:15a8 359 360 - kernel driver version: 4.2.5 (ixgbe) 361 362#. SuperMicro 2U 363 364 - BIOS: 1.0a 365 - Processor: Intel(R) Xeon(R) CPU E5-4667 v3 @ 2.00GHz 366 367#. Intel(R) Server board S2600GZ 368 369 - BIOS: SE5C600.86B.02.02.0002.122320131210 370 - Processor: Intel(R) Xeon(R) CPU E5-2680 v2 @ 2.80GHz 371 372#. Intel(R) Server board W2600CR 373 374 - BIOS: SE5C600.86B.02.01.0002.082220131453 375 - Processor: Intel(R) Xeon(R) CPU E5-2680 v2 @ 2.80GHz 376 377#. Intel(R) Server board S2600CWT 378 379 - BIOS: SE5C610.86B.01.01.0009.060120151350 380 - Processor: Intel(R) Xeon(R) CPU E5-2699 v3 @ 2.30GHz 381 382#. Intel(R) Server board S2600WTT 383 384 - BIOS: SE5C610.86B.01.01.0005.101720141054 385 - Processor: Intel(R) Xeon(R) CPU E5-2699 v3 @ 2.30GHz 386 387#. Intel(R) Server board S2600WTT 388 389 - BIOS: SE5C610.86B.11.01.0044.090120151156 390 - Processor: Intel(R) Xeon(R) CPU E5-2695 v4 @ 2.10GHz 391 392 393Tested NICs 394----------- 395 396#. Intel(R) Ethernet Controller X540-AT2 397 398 - Firmware version: 0x80000389 399 - Device id (pf): 8086:1528 400 - Driver version: 3.23.2 (ixgbe) 401 402#. Intel(R) 82599ES 10 Gigabit Ethernet Controller 403 404 - Firmware version: 0x61bf0001 405 - Device id (pf/vf): 8086:10fb / 8086:10ed 406 - Driver version: 4.0.1-k (ixgbe) 407 408#. Intel(R) Corporation Ethernet Connection X552/X557-AT 10GBASE-T 409 410 - Firmware version: 0x800001cf 411 - Device id (pf/vf): 8086:15ad / 8086:15a8 412 - Driver version: 4.2.5 (ixgbe) 413 414#. Intel(R) Ethernet Converged Network Adapter X710-DA4 (4x10G) 415 416 - Firmware version: 5.04 417 - Device id (pf/vf): 8086:1572 / 8086:154c 418 - Driver version: 1.4.26 (i40e) 419 420#. Intel(R) Ethernet Converged Network Adapter X710-DA2 (2x10G) 421 422 - Firmware version: 5.04 423 - Device id (pf/vf): 8086:1572 / 8086:154c 424 - Driver version: 1.4.25 (i40e) 425 426#. Intel(R) Ethernet Converged Network Adapter XL710-QDA1 (1x40G) 427 428 - Firmware version: 5.04 429 - Device id (pf/vf): 8086:1584 / 8086:154c 430 - Driver version: 1.4.25 (i40e) 431 432#. Intel(R) Ethernet Converged Network Adapter XL710-QDA2 (2X40G) 433 434 - Firmware version: 5.04 435 - Device id (pf/vf): 8086:1583 / 8086:154c 436 - Driver version: 1.4.25 (i40e) 437 438#. Intel(R) Corporation I350 Gigabit Network Connection 439 440 - Firmware version: 1.48, 0x800006e7 441 - Device id (pf/vf): 8086:1521 / 8086:1520 442 - Driver version: 5.2.13-k (igb) 443 444#. Intel(R) Ethernet Multi-host Controller FM10000 445 446 - Firmware version: N/A 447 - Device id (pf/vf): 8086:15d0 448 - Driver version: 0.17.0.9 (fm10k) 449 450 451Tested OSes 452----------- 453 454- CentOS 7.0 455- Fedora 23 456- Fedora 24 457- FreeBSD 10.3 458- Red Hat Enterprise Linux 7.2 459- SUSE Enterprise Linux 12 460- Ubuntu 15.10 461- Ubuntu 16.04 LTS 462- Wind River Linux 8 463