1 .. SPDX-License-Identifier: BSD-3-Clause
2 Copyright 2018 The DPDK contributors
4 ABI and API Deprecation
5 =======================
7 See the guidelines document for details of the :doc:`ABI policy
8 <../contributing/abi_policy>`. API and ABI deprecation notices are to be posted
14 * meson: The minimum supported version of meson for configuring and building
15 DPDK will be increased to v0.47.1 (from 0.41) from DPDK 19.05 onwards. For
16 those users with a version earlier than 0.47.1, an updated copy of meson
17 can be got using the ``pip``, or ``pip3``, tool for downloading python
20 * build macros: The macros defining RTE_MACHINE_CPUFLAG_* will be removed
21 from the build. The information provided by these macros is available
22 through standard compiler macros. For example, RTE_MACHINE_CPUFLAG_SSE3
23 duplicates the compiler-provided macro __SSE3__.
25 * kvargs: The function ``rte_kvargs_process`` will get a new parameter
26 for returning key match count. It will ease handling of no-match case.
28 * eal: To be more inclusive in choice of naming, the DPDK project
29 will replace uses of master/slave in the API's and command line arguments.
31 References to master/slave in relation to lcore will be renamed
32 to initial/worker. The function ``rte_get_master_lcore()``
33 will be renamed to ``rte_get_initial_lcore()``.
34 For the 20.11 release, both names will be present and the
35 old function will be marked with the deprecated tag.
36 The old function will be removed in a future version.
38 The iterator for worker lcores will also change:
39 ``RTE_LCORE_FOREACH_SLAVE`` will be replaced with
40 ``RTE_LCORE_FOREACH_WORKER``.
42 The ``master-lcore`` argument to testpmd will be replaced
43 with ``initial-lcore``. The old ``master-lcore`` argument
44 will produce a runtime notification in 20.11 release, and
45 be removed completely in a future release.
47 * eal: The terms blacklist and whitelist to describe devices used
48 by DPDK will be replaced in the 20.11 relase.
49 This will apply to command line arguments as well as macros.
51 The macro ``RTE_DEV_BLACKLISTED`` will be replaced with ``RTE_DEV_EXCLUDED``
52 and ``RTE_DEV_WHITELISTED`` will be replaced with ``RTE_DEV_INCLUDED``
53 ``RTE_BUS_SCAN_BLACKLIST`` and ``RTE_BUS_SCAN_WHITELIST`` will be
54 replaced with ``RTE_BUS_SCAN_EXCLUDED`` and ``RTE_BUS_SCAN_INCLUDED``
55 respectively. Likewise ``RTE_DEVTYPE_BLACKLISTED_PCI`` and
56 ``RTE_DEVTYPE_WHITELISTED_PCI`` will be replaced with
57 ``RTE_DEVTYPE_EXCLUDED`` and ``RTE_DEVTYPE_INCLUDED``.
59 The old macros will be marked as deprecated in 20.11 and any
60 usage will cause a compile warning. They will be removed in
63 The command line arguments to ``rte_eal_init`` will change from
64 ``-b, --pci-blacklist`` to ``-x, --exclude`` and
65 ``-w, --pci-whitelist`` to ``-i, --include``.
66 The old command line arguments will continue to be accepted in 20.11
67 but will cause a runtime warning message. The old arguments will
68 be removed in a future release.
70 * eal: The function ``rte_eal_remote_launch`` will return new error codes
71 after read or write error on the pipe, instead of calling ``rte_panic``.
73 * eal: The ``rte_logs`` struct and global symbol will be made private to
74 remove it from the externally visible ABI and allow it to be updated in the
77 * eal: The ``rte_dev_event`` structure will be made private to the EAL as no
78 public API makes use of it.
80 * rte_atomicNN_xxx: These APIs do not take memory order parameter. This does
81 not allow for writing optimized code for all the CPU architectures supported
82 in DPDK. DPDK will adopt C11 atomic operations semantics and provide wrappers
83 using C11 atomic built-ins. These wrappers must be used for patches that
84 need to be merged in 20.08 onwards. This change will not introduce any
85 performance degradation.
87 * rte_smp_*mb: These APIs provide full barrier functionality. However, many
88 use cases do not require full barriers. To support such use cases, DPDK will
89 adopt C11 barrier semantics and provide wrappers using C11 atomic built-ins.
90 These wrappers must be used for patches that need to be merged in 20.08
91 onwards. This change will not introduce any performance degradation.
93 * rte_cio_*mb: Since the IO barriers for ARMv8 platforms are relaxed from DSB
94 to DMB, rte_cio_*mb APIs provide the same functionality as rte_io_*mb
95 APIs (taking all platforms into consideration). rte_io_*mb APIs should be
96 used in the place of rte_cio_*mb APIs. The rte_cio_*mb APIs will be
97 deprecated in 20.11 release.
99 * igb_uio: In the view of reducing the kernel dependency from the main tree,
100 as a first step, the Technical Board decided to move ``igb_uio``
101 kernel module to the dpdk-kmods repository in the /linux/igb_uio/ directory
103 Minutes of Technical Board Meeting of `2019-11-06
104 <https://mails.dpdk.org/archives/dev/2019-November/151763.html>`_.
106 * lib: will fix extending some enum/define breaking the ABI. There are multiple
107 samples in DPDK that enum/define terminated with a ``.*MAX.*`` value which is
108 used by iterators, and arrays holding these values are sized with this
109 ``.*MAX.*`` value. So extending this enum/define increases the ``.*MAX.*``
110 value which increases the size of the array and depending on how/where the
111 array is used this may break the ABI.
112 ``RTE_ETH_FLOW_MAX`` is one sample of the mentioned case, adding a new flow
113 type will break the ABI because of ``flex_mask[RTE_ETH_FLOW_MAX]`` array
114 usage in following public struct hierarchy:
115 ``rte_eth_fdir_flex_conf -> rte_fdir_conf -> rte_eth_conf (in the middle)``.
116 Need to identify this kind of usages and fix in 20.11, otherwise this blocks
117 us extending existing enum/define.
118 One solution can be using a fixed size array instead of ``.*MAX.*`` value.
120 * pci: The PCI resources map API (``pci_map_resource`` and
121 ``pci_unmap_resource``) was not abstracting the Unix mmap flags (see the
122 workaround for Windows support implemented in the commit
123 9d2b24593724 ("pci: keep API compatibility with mmap values")).
124 This API will be removed from the public API in 20.11 and moved to the PCI
125 bus driver along with the PCI resources lists and associated structures
126 (``pci_map``, ``pci_msix_table``, ``mapped_pci_resource`` and
127 ``mapped_pci_res_list``).
128 With this removal, there won't be a need for the mentioned workaround which
131 * pci: The ``rte_kernel_driver`` enum defined in rte_dev.h will be made private
132 to the PCI subsystem as it is used only by the PCI bus driver and PCI
134 The associated field ``kdrv`` in the ethdev ``rte_eth_dev_data`` structure
135 will be removed as it gave no useful abstracted information to the
136 applications and had no user (neither internal nor external).
138 * mbuf: Some fields will be converted to dynamic API in DPDK 20.11
139 in order to reserve more space for the dynamic fields, as explained in
140 `this presentation <https://www.youtube.com/watch?v=Ttl6MlhmzWY>`_.
141 The following static fields will be moved as dynamic:
144 - ``userdata`` / ``udata64``
147 As a consequence, the layout of the ``struct rte_mbuf`` will be re-arranged,
148 avoiding impact on vectorized implementation of the driver datapaths,
149 while evaluating performance gains of a better use of the first cache line.
151 The deprecated unioned fields ``buf_physaddr`` and ``refcnt_atomic``
152 (as explained below) will be removed in DPDK 20.11.
154 * mbuf: ``refcnt_atomic`` member in structures ``rte_mbuf`` and
155 ``rte_mbuf_ext_shared_info`` is of type ``rte_atomic16_t``.
156 Due to adoption of C11 atomic builtins, the field ``refcnt_atomic``
157 will be replaced with ``refcnt`` of type ``uint16_t`` in DPDK 20.11.
159 * ethdev: Split the ``struct eth_dev_ops`` struct to hide it as much as possible
160 will be done in 20.11.
161 Currently the ``struct eth_dev_ops`` struct is accessible by the application
162 because some inline functions, like ``rte_eth_tx_descriptor_status()``,
163 access the struct directly.
164 The struct will be separate in two, the ops used by inline functions will be
165 moved next to Rx/Tx burst functions, rest of the ``struct eth_dev_ops`` struct
166 will be moved to header file for drivers to hide it from applications.
168 * ethdev: the legacy filter API, including
169 ``rte_eth_dev_filter_supported()``, ``rte_eth_dev_filter_ctrl()`` as well
170 as filter types MACVLAN, ETHERTYPE, FLEXIBLE, SYN, NTUPLE, TUNNEL, FDIR,
171 HASH and L2_TUNNEL, is superseded by the generic flow API (rte_flow) in
172 PMDs that implement the latter.
173 The legacy API will be removed in DPDK 20.11.
175 * ethdev: The flow director API, including ``rte_eth_conf.fdir_conf`` field,
176 and the related structures (``rte_fdir_*`` and ``rte_eth_fdir_*``),
177 will be removed in DPDK 20.11.
179 * ethdev: The legacy L2 tunnel filtering API is deprecated as the rest of
180 the legacy filtering API.
181 The functions ``rte_eth_dev_l2_tunnel_eth_type_conf`` and
182 ``rte_eth_dev_l2_tunnel_offload_set`` which were not marked as deprecated,
183 will be removed in DPDK 20.11.
185 * ethdev: Update API functions returning ``void`` to return ``int`` with
186 negative errno values to indicate various error conditions (e.g.
187 invalid port ID, unsupported operation, failed operation):
189 - ``rte_eth_dev_stop``
190 - ``rte_eth_dev_close``
192 * ethdev: The temporary flag RTE_ETH_DEV_CLOSE_REMOVE will be removed in 20.11.
193 As a consequence, the new behaviour introduced in 18.11 will be effective
194 for all drivers: generic port resources are freed on close operation.
195 Private resources are expected to be released in the ``dev_close`` callback.
196 More details in http://inbox.dpdk.org/dev/5248162.j6AOsuQRmx@thomas/
198 * ethdev: New offload flags ``DEV_RX_OFFLOAD_FLOW_MARK`` will be added in 19.11.
199 This will allow application to enable or disable PMDs from updating
200 ``rte_mbuf::hash::fdir``.
201 This scheme will allow PMDs to avoid writes to ``rte_mbuf`` fields on Rx and
202 thereby improve Rx performance if application wishes do so.
203 In 19.11 PMDs will still update the field even when the offload is not
206 * ethdev: Add new fields to ``rte_eth_rxconf`` to configure the receiving
207 queues to split ingress packets into multiple segments according to the
208 specified lengths into the buffers allocated from the specified
209 memory pools. The backward compatibility to existing API is preserved.
211 * ethdev: The ``struct rte_eth_rxq_info`` will be modified to include
212 a new optional field, indicating the buffer size used in receiving packets
213 for HW. This change is planned for 20.11. For more details:
214 https://mails.dpdk.org/archives/dev/2020-July/176135.html.
216 * ethdev: ``rx_descriptor_done`` dev_ops and ``rte_eth_rx_descriptor_done``
217 will be deprecated in 20.11 and will be removed in 21.11.
218 Existing ``rte_eth_rx_descriptor_status`` and ``rte_eth_tx_descriptor_status``
219 APIs can be used as replacement.
221 * ethdev: The port mirroring API can be replaced with a more fine grain flow API.
222 The structs ``rte_eth_mirror_conf``, ``rte_eth_vlan_mirror`` and the functions
223 ``rte_eth_mirror_rule_set``, ``rte_eth_mirror_rule_reset`` will be marked
224 as deprecated in DPDK 20.11, along with the associated macros ``ETH_MIRROR_*``.
225 This API will be fully removed in DPDK 21.11.
227 * ethdev: The ``struct rte_flow_item_eth`` and ``struct rte_flow_item_vlan``
228 structs will be modified, to include an additional value, indicating existence
229 or absence of a VLAN header following the current header, as proposed in RFC
230 https://mails.dpdk.org/archives/dev/2020-August/177536.html.
232 * ethdev: The ``struct rte_flow_item_ipv6`` struct will be modified to include
233 additional values, indicating existence or absence of IPv6 extension headers
234 following the IPv6 header, as proposed in RFC
235 https://mails.dpdk.org/archives/dev/2020-August/177257.html.
237 * ethdev: Some internal APIs for driver usage are exported in the .map file.
238 Now DPDK has ``__rte_internal`` marker so we can mark internal APIs and move
239 them to the INTERNAL block in .map. Although these APIs are internal it will
240 break the ABI checks, that is why change is planned for 20.11.
241 The list of internal APIs are mainly ones listed in ``rte_ethdev_driver.h``.
243 * traffic manager: All traffic manager API's in ``rte_tm.h`` were mistakenly made
244 ABI stable in the v19.11 release. The TM maintainer and other contributors have
245 agreed to keep the TM APIs as experimental in expectation of additional spec
246 improvements. Therefore, all APIs in ``rte_tm.h`` will be marked back as
247 experimental in v20.11 DPDK release. For more details, please see `the thread
248 <https://mails.dpdk.org/archives/dev/2020-April/164970.html>`_.
250 * pmd_dpaa: The API ``rte_pmd_dpaa_set_tx_loopback`` will have extended
251 ``port_id`` definition from ``uint8_t`` to ``uint16_t``.
253 * vhost: Vhost-user dequeue zero-copy support will be removed in 20.11.
254 The only known user is OVS where the feature is still experimental,
255 and has not received any update for 2.5 years.
256 This feature faces reliability issues and is often conflicting with
257 new features being implemented.
259 * security: The API ``rte_security_session_create`` takes only single mempool
260 for session and session private data. So the application need to create
261 mempool for twice the number of sessions needed and will also lead to
262 wastage of memory as session private data need more memory compared to session.
263 Hence the API will be modified to take two mempool pointers - one for session
264 and one for private data.
266 * cryptodev: ``RTE_CRYPTO_AEAD_LIST_END`` from ``enum rte_crypto_aead_algorithm``,
267 ``RTE_CRYPTO_CIPHER_LIST_END`` from ``enum rte_crypto_cipher_algorithm`` and
268 ``RTE_CRYPTO_AUTH_LIST_END`` from ``enum rte_crypto_auth_algorithm``
271 * cryptodev: support for using IV with all sizes is added, J0 still can
272 be used but only when IV length in following structs ``rte_crypto_auth_xform``,
273 ``rte_crypto_aead_xform`` is set to zero. When IV length is greater or equal
274 to one it means it represents IV, when is set to zero it means J0 is used
275 directly, in this case 16 bytes of J0 need to be passed.
277 * scheduler: The functions ``rte_cryptodev_scheduler_slave_attach``,
278 ``rte_cryptodev_scheduler_slave_detach`` and
279 ``rte_cryptodev_scheduler_slaves_get`` will be replaced in 20.11 by
280 ``rte_cryptodev_scheduler_worker_attach``,
281 ``rte_cryptodev_scheduler_worker_detach`` and
282 ``rte_cryptodev_scheduler_workers_get`` accordingly.
284 * eventdev: Following structures will be modified to support DLB PMD
285 and future extensions:
287 - ``rte_event_dev_info``
288 - ``rte_event_dev_config``
289 - ``rte_event_port_conf``
291 Patches containing justification, documentation, and proposed modifications
294 - https://patches.dpdk.org/patch/71457/
295 - https://patches.dpdk.org/patch/71456/
297 * acl: ``RTE_ACL_CLASSIFY_NUM`` enum value will be removed.
298 This enum value is not used inside DPDK, while it prevents to add new
299 classify algorithms without causing an ABI breakage.
301 * sched: To allow more traffic classes, flexible mapping of pipe queues to
302 traffic classes, and subport level configuration of pipes and queues
303 changes will be made to macros, data structures and API functions defined
304 in "rte_sched.h". These changes are aligned to improvements suggested in the
305 RFC https://mails.dpdk.org/archives/dev/2018-November/120035.html.
307 * sched: To allow dynamic configuration of the subport bandwidth profile,
308 changes will be made to data structures ``rte_sched_subport_params``,
309 ``rte_sched_port_params`` and new data structure, API functions will be
310 defined in ``rte_sched.h``. These changes are aligned as suggested in the
311 RFC https://mails.dpdk.org/archives/dev/2020-July/175161.html
313 * metrics: The function ``rte_metrics_init`` will have a non-void return
314 in order to notify errors instead of calling ``rte_exit``.
316 * power: ``rte_power_set_env`` function will no longer return 0 on attempt
317 to set new power environment if power environment was already initialized.
318 In this case the function will return -1 unless the environment is unset first
319 (using ``rte_power_unset_env``). Other function usage scenarios will not change.
321 * python: Since the beginning of 2020, Python 2 has officially reached
322 end-of-support: https://www.python.org/doc/sunset-python-2/.
323 Python 2 support will be completely removed in 20.11.
324 In 20.08, explicit deprecation warnings will be displayed when running
325 scripts with Python 2.
327 * dpdk-setup.sh: This old script relies on deprecated stuff, and especially
328 ``make``. Given environments are too much variables for such a simple script,
329 it will be removed in DPDK 20.11.
330 Some useful parts may be converted into specific scripts.