xref: /dpdk/doc/guides/prog_guide/cryptodev_lib.rst (revision 00e57b0e550b7df2047e6d0bde8965c7ae17d203)
1..  SPDX-License-Identifier: BSD-3-Clause
2    Copyright(c) 2016-2020 Intel Corporation.
3
4Cryptography Device Library
5===========================
6
7The cryptodev library provides a Crypto device framework for management and
8provisioning of hardware and software Crypto poll mode drivers, defining generic
9APIs which support a number of different Crypto operations. The framework
10currently only supports cipher, authentication, chained cipher/authentication
11and AEAD symmetric and asymmetric Crypto operations.
12
13
14Design Principles
15-----------------
16
17The cryptodev library follows the same basic principles as those used in DPDK's
18Ethernet Device framework. The Crypto framework provides a generic Crypto device
19framework which supports both physical (hardware) and virtual (software) Crypto
20devices as well as a generic Crypto API which allows Crypto devices to be
21managed and configured and supports Crypto operations to be provisioned on
22Crypto poll mode driver.
23
24
25Device Management
26-----------------
27
28Device Creation
29~~~~~~~~~~~~~~~
30
31Physical Crypto devices are discovered during the PCI probe/enumeration of the
32EAL function which is executed at DPDK initialization, based on
33their PCI device identifier, each unique PCI BDF (bus/bridge, device,
34function). Specific physical Crypto devices, like other physical devices in DPDK
35can be listed using the EAL command line options.
36
37Virtual devices can be created by two mechanisms, either using the EAL command
38line options or from within the application using an EAL API directly.
39
40From the command line using the --vdev EAL option
41
42.. code-block:: console
43
44   --vdev  'crypto_aesni_mb0,max_nb_queue_pairs=2,socket_id=0'
45
46.. Note::
47
48   * If DPDK application requires multiple software crypto PMD devices then required
49     number of ``--vdev`` with appropriate libraries are to be added.
50
51   * An Application with crypto PMD instances sharing the same library requires unique ID.
52
53   Example: ``--vdev  'crypto_aesni_mb0' --vdev  'crypto_aesni_mb1'``
54
55Or using the rte_vdev_init API within the application code.
56
57.. code-block:: c
58
59   rte_vdev_init("crypto_aesni_mb",
60                     "max_nb_queue_pairs=2,socket_id=0")
61
62All virtual Crypto devices support the following initialization parameters:
63
64* ``max_nb_queue_pairs`` - maximum number of queue pairs supported by the device.
65* ``socket_id`` - socket on which to allocate the device resources on.
66
67
68Device Identification
69~~~~~~~~~~~~~~~~~~~~~
70
71Each device, whether virtual or physical is uniquely designated by two
72identifiers:
73
74- A unique device index used to designate the Crypto device in all functions
75  exported by the cryptodev API.
76
77- A device name used to designate the Crypto device in console messages, for
78  administration or debugging purposes. For ease of use, the port name includes
79  the port index.
80
81
82Device Configuration
83~~~~~~~~~~~~~~~~~~~~
84
85The configuration of each Crypto device includes the following operations:
86
87- Allocation of resources, including hardware resources if a physical device.
88- Resetting the device into a well-known default state.
89- Initialization of statistics counters.
90
91The rte_cryptodev_configure API is used to configure a Crypto device.
92
93.. code-block:: c
94
95   int rte_cryptodev_configure(uint8_t dev_id,
96                               struct rte_cryptodev_config *config)
97
98The ``rte_cryptodev_config`` structure is used to pass the configuration
99parameters for socket selection and number of queue pairs.
100
101.. literalinclude:: ../../../lib/cryptodev/rte_cryptodev.h
102   :language: c
103   :start-after: Structure rte_cryptodev_config 8<
104   :end-before: >8 End of structure rte_cryptodev_config.
105
106
107Configuration of Queue Pairs
108~~~~~~~~~~~~~~~~~~~~~~~~~~~~
109
110Each Crypto devices queue pair is individually configured through the
111``rte_cryptodev_queue_pair_setup`` API.
112Each queue pairs resources may be allocated on a specified socket.
113
114.. code-block:: c
115
116    int rte_cryptodev_queue_pair_setup(uint8_t dev_id, uint16_t queue_pair_id,
117                const struct rte_cryptodev_qp_conf *qp_conf,
118                int socket_id)
119
120
121.. literalinclude:: ../../../lib/cryptodev/rte_cryptodev.h
122   :language: c
123   :start-after: Structure rte_cryptodev_qp_conf 8<
124   :end-before: >8 End of structure rte_cryptodev_qp_conf.
125
126
127The field ``mp_session`` is used for creating temporary session to process
128the crypto operations in the session-less mode.
129They can be the same other different mempools. Please note not all Cryptodev
130PMDs supports session-less mode.
131
132
133Logical Cores, Memory and Queues Pair Relationships
134~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
135
136The Crypto device Library as the Poll Mode Driver library support NUMA for when
137a processor’s logical cores and interfaces utilize its local memory. Therefore
138Crypto operations, and in the case of symmetric Crypto operations, the session
139and the mbuf being operated on, should be allocated from memory pools created
140in the local memory. The buffers should, if possible, remain on the local
141processor to obtain the best performance results and buffer descriptors should
142be populated with mbufs allocated from a mempool allocated from local memory.
143
144The run-to-completion model also performs better, especially in the case of
145virtual Crypto devices, if the Crypto operation and session and data buffer is
146in local memory instead of a remote processor's memory. This is also true for
147the pipe-line model provided all logical cores used are located on the same
148processor.
149
150Multiple logical cores should never share the same queue pair for enqueuing
151operations or dequeuing operations on the same Crypto device since this would
152require global locks and hinder performance. It is however possible to use a
153different logical core to dequeue an operation on a queue pair from the logical
154core which it was enqueued on. This means that a crypto burst enqueue/dequeue
155APIs are a logical place to transition from one logical core to another in a
156packet processing pipeline.
157
158
159Device Features and Capabilities
160---------------------------------
161
162Crypto devices define their functionality through two mechanisms, global device
163features and algorithm capabilities. Global devices features identify device
164wide level features which are applicable to the whole device such as
165the device having hardware acceleration or supporting symmetric and/or asymmetric
166Crypto operations.
167
168The capabilities mechanism defines the individual algorithms/functions which
169the device supports, such as a specific symmetric Crypto cipher,
170authentication operation or Authenticated Encryption with Associated Data
171(AEAD) operation.
172
173
174Device Features
175~~~~~~~~~~~~~~~
176
177Currently the following Crypto device features are defined:
178
179* Symmetric Crypto operations
180* Asymmetric Crypto operations
181* Chaining of symmetric Crypto operations
182* SSE accelerated SIMD vector operations
183* AVX accelerated SIMD vector operations
184* AVX2 accelerated SIMD vector operations
185* AESNI accelerated instructions
186* Hardware off-load processing
187
188
189Device Operation Capabilities
190~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
191
192Crypto capabilities which identify particular algorithm which the Crypto PMD
193supports are  defined by the operation type, the operation transform, the
194transform identifier and then the particulars of the transform. For the full
195scope of the Crypto capability see the definition of the structure in the
196*DPDK API Reference*.
197
198.. code-block:: c
199
200   struct rte_cryptodev_capabilities;
201
202Each Crypto poll mode driver defines its own private array of capabilities
203for the operations it supports. Below is an example of the capabilities for a
204PMD which supports the authentication algorithm SHA1_HMAC and the cipher
205algorithm AES_CBC.
206
207.. code-block:: c
208
209    static const struct rte_cryptodev_capabilities pmd_capabilities[] = {
210        {    /* SHA1 HMAC */
211            .op = RTE_CRYPTO_OP_TYPE_SYMMETRIC,
212            .sym = {
213                .xform_type = RTE_CRYPTO_SYM_XFORM_AUTH,
214                .auth = {
215                    .algo = RTE_CRYPTO_AUTH_SHA1_HMAC,
216                    .block_size = 64,
217                    .key_size = {
218                        .min = 64,
219                        .max = 64,
220                        .increment = 0
221                    },
222                    .digest_size = {
223                        .min = 12,
224                        .max = 12,
225                        .increment = 0
226                    },
227                    .aad_size = { 0 },
228                    .iv_size = { 0 }
229                }
230            }
231        },
232        {    /* AES CBC */
233            .op = RTE_CRYPTO_OP_TYPE_SYMMETRIC,
234            .sym = {
235                .xform_type = RTE_CRYPTO_SYM_XFORM_CIPHER,
236                .cipher = {
237                    .algo = RTE_CRYPTO_CIPHER_AES_CBC,
238                    .block_size = 16,
239                    .key_size = {
240                        .min = 16,
241                        .max = 32,
242                        .increment = 8
243                    },
244                    .iv_size = {
245                        .min = 16,
246                        .max = 16,
247                        .increment = 0
248                    }
249                }
250            }
251        }
252    }
253
254
255Capabilities Discovery
256~~~~~~~~~~~~~~~~~~~~~~
257
258Discovering the features and capabilities of a Crypto device poll mode driver
259is achieved through the ``rte_cryptodev_info_get`` function.
260
261.. code-block:: c
262
263   void rte_cryptodev_info_get(uint8_t dev_id,
264                               struct rte_cryptodev_info *dev_info);
265
266This allows the user to query a specific Crypto PMD and get all the device
267features and capabilities. The ``rte_cryptodev_info`` structure contains all the
268relevant information for the device.
269
270.. literalinclude:: ../../../lib/cryptodev/rte_cryptodev.h
271   :language: c
272   :start-after: Structure rte_cryptodev_info 8<
273   :end-before: >8 End of structure rte_cryptodev_info.
274
275
276Operation Processing
277--------------------
278
279Scheduling of Crypto operations on DPDK's application data path is
280performed using a burst oriented asynchronous API set. A queue pair on a Crypto
281device accepts a burst of Crypto operations using enqueue burst API. On physical
282Crypto devices the enqueue burst API will place the operations to be processed
283on the devices hardware input queue, for virtual devices the processing of the
284Crypto operations is usually completed during the enqueue call to the Crypto
285device. The dequeue burst API will retrieve any processed operations available
286from the queue pair on the Crypto device, from physical devices this is usually
287directly from the devices processed queue, and for virtual device's from a
288``rte_ring`` where processed operations are placed after being processed on the
289enqueue call.
290
291
292Private data
293~~~~~~~~~~~~
294For session-based operations, the set and get API provides a mechanism for an
295application to store and retrieve the private user data information stored along
296with the crypto session.
297
298For example, suppose an application is submitting a crypto operation with a session
299associated and wants to indicate private user data information which is required to be
300used after completion of the crypto operation. In this case, the application can use
301the set API to set the user data and retrieve it using get API.
302
303.. code-block:: c
304
305	int rte_cryptodev_sym_session_set_user_data(
306		struct rte_cryptodev_sym_session *sess,	void *data, uint16_t size);
307
308	void * rte_cryptodev_sym_session_get_user_data(
309		struct rte_cryptodev_sym_session *sess);
310
311Please note the ``size`` passed to set API cannot be bigger than the predefined
312``user_data_sz`` when creating the session header mempool, otherwise the
313function will return error. Also when ``user_data_sz`` was defined as ``0`` when
314creating the session header mempool, the get API will always return ``NULL``.
315
316For session-less mode, the private user data information can be placed along with the
317``struct rte_crypto_op``. The ``rte_crypto_op::private_data_offset`` indicates the
318start of private data information. The offset is counted from the start of the
319rte_crypto_op including other crypto information such as the IVs (since there can
320be an IV also for authentication).
321
322User callback APIs
323~~~~~~~~~~~~~~~~~~
324The add APIs configures a user callback function to be called for each burst of crypto
325ops received/sent on a given crypto device queue pair. The return value is a pointer
326that can be used later to remove the callback using remove API. Application is expected
327to register a callback function of type ``rte_cryptodev_callback_fn``. Multiple callback
328functions can be added for a given queue pair. API does not restrict on maximum number of
329callbacks.
330
331Callbacks registered by application would not survive ``rte_cryptodev_configure`` as it
332reinitializes the callback list. It is user responsibility to remove all installed
333callbacks before calling ``rte_cryptodev_configure`` to avoid possible memory leakage.
334
335So, the application is expected to add user callback after ``rte_cryptodev_configure``.
336The callbacks can also be added at the runtime. These callbacks get executed when
337``rte_cryptodev_enqueue_burst``/``rte_cryptodev_dequeue_burst`` is called.
338
339.. code-block:: c
340
341	struct rte_cryptodev_cb *
342		rte_cryptodev_add_enq_callback(uint8_t dev_id, uint16_t qp_id,
343					       rte_cryptodev_callback_fn cb_fn,
344					       void *cb_arg);
345
346	struct rte_cryptodev_cb *
347		rte_cryptodev_add_deq_callback(uint8_t dev_id, uint16_t qp_id,
348					       rte_cryptodev_callback_fn cb_fn,
349					       void *cb_arg);
350
351	uint16_t (* rte_cryptodev_callback_fn)(uint16_t dev_id, uint16_t qp_id,
352					       struct rte_crypto_op **ops,
353					       uint16_t nb_ops, void *user_param);
354
355The remove API removes a callback function added by
356``rte_cryptodev_add_enq_callback``/``rte_cryptodev_add_deq_callback``.
357
358.. code-block:: c
359
360	int rte_cryptodev_remove_enq_callback(uint8_t dev_id, uint16_t qp_id,
361					      struct rte_cryptodev_cb *cb);
362
363	int rte_cryptodev_remove_deq_callback(uint8_t dev_id, uint16_t qp_id,
364					      struct rte_cryptodev_cb *cb);
365
366
367Enqueue / Dequeue Burst APIs
368~~~~~~~~~~~~~~~~~~~~~~~~~~~~
369
370The burst enqueue API uses a Crypto device identifier and a queue pair
371identifier to specify the Crypto device queue pair to schedule the processing on.
372The ``nb_ops`` parameter is the number of operations to process which are
373supplied in the ``ops`` array of ``rte_crypto_op`` structures.
374The enqueue function returns the number of operations it actually enqueued for
375processing, a return value equal to ``nb_ops`` means that all packets have been
376enqueued.
377
378.. code-block:: c
379
380   uint16_t rte_cryptodev_enqueue_burst(uint8_t dev_id, uint16_t qp_id,
381                                        struct rte_crypto_op **ops, uint16_t nb_ops)
382
383The dequeue API uses the same format as the enqueue API of processed but
384the ``nb_ops`` and ``ops`` parameters are now used to specify the max processed
385operations the user wishes to retrieve and the location in which to store them.
386The API call returns the actual number of processed operations returned, this
387can never be larger than ``nb_ops``.
388
389.. code-block:: c
390
391   uint16_t rte_cryptodev_dequeue_burst(uint8_t dev_id, uint16_t qp_id,
392                                        struct rte_crypto_op **ops, uint16_t nb_ops)
393
394
395Operation Representation
396~~~~~~~~~~~~~~~~~~~~~~~~
397
398An Crypto operation is represented by an rte_crypto_op structure, which is a
399generic metadata container for all necessary information required for the
400Crypto operation to be processed on a particular Crypto device poll mode driver.
401
402.. figure:: img/crypto_op.*
403
404The operation structure includes the operation type, the operation status
405and the session type (session-based/less), a reference to the operation
406specific data, which can vary in size and content depending on the operation
407being provisioned. It also contains the source mempool for the operation,
408if it allocated from a mempool.
409
410If Crypto operations are allocated from a Crypto operation mempool, see next
411section, there is also the ability to allocate private memory with the
412operation for applications purposes.
413
414Application software is responsible for specifying all the operation specific
415fields in the ``rte_crypto_op`` structure which are then used by the Crypto PMD
416to process the requested operation.
417
418
419Operation Management and Allocation
420~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
421
422The cryptodev library provides an API set for managing Crypto operations which
423utilize the Mempool Library to allocate operation buffers. Therefore, it ensures
424that the crypto operation is interleaved optimally across the channels and
425ranks for optimal processing.
426A ``rte_crypto_op`` contains a field indicating the pool that it originated from.
427When calling ``rte_crypto_op_free(op)``, the operation returns to its original pool.
428
429.. code-block:: c
430
431   extern struct rte_mempool *
432   rte_crypto_op_pool_create(const char *name, enum rte_crypto_op_type type,
433                             unsigned nb_elts, unsigned cache_size, uint16_t priv_size,
434                             int socket_id);
435
436During pool creation ``rte_crypto_op_init()`` is called as a constructor to
437initialize each Crypto operation which subsequently calls
438``__rte_crypto_op_reset()`` to configure any operation type specific fields based
439on the type parameter.
440
441
442``rte_crypto_op_alloc()`` and ``rte_crypto_op_bulk_alloc()`` are used to allocate
443Crypto operations of a specific type from a given Crypto operation mempool.
444``__rte_crypto_op_reset()`` is called on each operation before being returned to
445allocate to a user so the operation is always in a good known state before use
446by the application.
447
448.. code-block:: c
449
450   struct rte_crypto_op *rte_crypto_op_alloc(struct rte_mempool *mempool,
451                                             enum rte_crypto_op_type type)
452
453   unsigned rte_crypto_op_bulk_alloc(struct rte_mempool *mempool,
454                                     enum rte_crypto_op_type type,
455                                     struct rte_crypto_op **ops, uint16_t nb_ops)
456
457``rte_crypto_op_free()`` is called by the application to return an operation to
458its allocating pool.
459
460.. code-block:: c
461
462   void rte_crypto_op_free(struct rte_crypto_op *op)
463
464
465Symmetric Cryptography Support
466------------------------------
467
468The cryptodev library currently provides support for the following symmetric
469Crypto operations; cipher, authentication, including chaining of these
470operations, as well as also supporting AEAD operations.
471
472
473Session and Session Management
474~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
475
476Sessions are used in symmetric cryptographic processing to store the immutable
477data defined in a cryptographic transform which is used in the operation
478processing of a packet flow. Sessions are used to manage information such as
479expand cipher keys and HMAC IPADs and OPADs, which need to be calculated for a
480particular Crypto operation, but are immutable on a packet to packet basis for
481a flow. Crypto sessions cache this immutable data in a optimal way for the
482underlying PMD and this allows further acceleration of the offload of
483Crypto workloads.
484
485The Crypto device framework provides APIs to create session mempool and allocate
486and initialize sessions for crypto devices, where sessions are mempool objects.
487The application has to use ``rte_cryptodev_sym_session_pool_create()`` to
488create the session mempool header and the private data with the size specified
489by the user through the ``elt_size`` parameter in the function.
490The session private data is for the driver to initialize and access
491during crypto operations, hence the ``elt_size`` should be big enough
492for all drivers that will share this mempool.
493To obtain the proper session private data size of a crypto device,
494the user can call ``rte_cryptodev_sym_get_private_session_size()`` function.
495In case of heterogeneous crypto devices which will share the same session mempool,
496the maximum session private data size of them should be passed.
497
498Once the session mempools have been created, ``rte_cryptodev_sym_session_create()``
499is used to allocate and initialize the session from the given mempool.
500The created session can ONLY be used by the crypto devices sharing the same driver ID
501as the device ID passed into the function as the parameter.
502In addition, a symmetric transform chain is used to specify the operation and its parameters.
503See the section below for details on transforms.
504
505When a session is no longer used, user must call ``rte_cryptodev_sym_session_free()``
506to uninitialize the session data and return the session
507back to the mempool it belongs.
508
509
510Transforms and Transform Chaining
511~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
512
513Symmetric Crypto transforms (``rte_crypto_sym_xform``) are the mechanism used
514to specify the details of the Crypto operation. For chaining of symmetric
515operations such as cipher encrypt and authentication generate, the next pointer
516allows transform to be chained together. Crypto devices which support chaining
517must publish the chaining of symmetric Crypto operations feature flag. Allocation of the
518xform structure is in the application domain. To allow future API extensions in a
519backwardly compatible manner, e.g. addition of a new parameter, the application should
520zero the full xform struct before populating it.
521
522Currently there are three transforms types cipher, authentication and AEAD.
523Also it is important to note that the order in which the
524transforms are passed indicates the order of the chaining.
525
526.. literalinclude:: ../../../lib/cryptodev/rte_crypto_sym.h
527   :language: c
528   :start-after: Structure rte_crypto_sym_xform 8<
529   :end-before: >8 End of structure rte_crypto_sym_xform.
530
531The API does not place a limit on the number of transforms that can be chained
532together but this will be limited by the underlying Crypto device poll mode
533driver which is processing the operation.
534
535.. figure:: img/crypto_xform_chain.*
536
537
538Symmetric Operations
539~~~~~~~~~~~~~~~~~~~~
540
541The symmetric Crypto operation structure contains all the mutable data relating
542to performing symmetric cryptographic processing on a referenced mbuf data
543buffer. It is used for either cipher, authentication, AEAD and chained
544operations.
545
546As a minimum the symmetric operation must have a source data buffer (``m_src``),
547a valid session (or transform chain if in session-less mode) and the minimum
548authentication/ cipher/ AEAD parameters required depending on the type of operation
549specified in the session or the transform
550chain.
551
552.. literalinclude:: ../../../lib/cryptodev/rte_crypto_sym.h
553   :language: c
554   :start-after: Structure rte_crypto_sym_op 8<
555   :end-before: >8 End of structure rte_crypto_sym_op.
556
557
558Synchronous mode
559----------------
560
561Some cryptodevs support synchronous mode alongside with a standard asynchronous
562mode. In that case operations are performed directly when calling
563``rte_cryptodev_sym_cpu_crypto_process`` method instead of enqueuing and
564dequeuing an operation before. This mode of operation allows cryptodevs which
565utilize CPU cryptographic acceleration to have significant performance boost
566comparing to standard asynchronous approach. Cryptodevs supporting synchronous
567mode have ``RTE_CRYPTODEV_FF_SYM_CPU_CRYPTO`` feature flag set.
568
569To perform a synchronous operation a call to
570``rte_cryptodev_sym_cpu_crypto_process`` has to be made with vectorized
571operation descriptor (``struct rte_crypto_sym_vec``) containing:
572
573- ``num`` - number of operations to perform,
574- pointer to an array of size ``num`` containing a scatter-gather list
575  descriptors of performed operations (``struct rte_crypto_sgl``). Each instance
576  of ``struct rte_crypto_sgl`` consists of a number of segments and a pointer to
577  an array of segment descriptors ``struct rte_crypto_vec``;
578- pointers to arrays of size ``num`` containing IV, AAD and digest information
579  in the ``cpu_crypto`` sub-structure,
580- pointer to an array of size ``num`` where status information will be stored
581  for each operation.
582
583Function returns a number of successfully completed operations and sets
584appropriate status number for each operation in the status array provided as
585a call argument. Status different than zero must be treated as error.
586
587For more details, e.g. how to convert an mbuf to an SGL, please refer to an
588example usage in the IPsec library implementation.
589
590Cryptodev Raw Data-path APIs
591~~~~~~~~~~~~~~~~~~~~~~~~~~~~
592
593The Crypto Raw data-path APIs are a set of APIs designed to enable external
594libraries/applications to leverage the cryptographic processing provided by
595DPDK crypto PMDs through the cryptodev API but in a manner that is not
596dependent on native DPDK data structures (eg. rte_mbuf, rte_crypto_op, ... etc)
597in their data-path implementation.
598
599The raw data-path APIs have the following advantages:
600
601- External data structure friendly design. The new APIs uses the operation
602  descriptor ``struct rte_crypto_sym_vec`` that supports raw data pointer and
603  IOVA addresses as input. Moreover, the APIs does not require the user to
604  allocate the descriptor from mempool, nor requiring mbufs to describe input
605  data's virtual and IOVA addresses. All these features made the translation
606  from user's own data structure into the descriptor easier and more efficient.
607
608- Flexible enqueue and dequeue operation. The raw data-path APIs gives the
609  user more control to the enqueue and dequeue operations, including the
610  capability of precious enqueue/dequeue count, abandoning enqueue or dequeue
611  at any time, and operation status translation and set on the fly.
612
613Cryptodev PMDs which support the raw data-path APIs will have
614``RTE_CRYPTODEV_FF_SYM_RAW_DP`` feature flag presented. To use this feature,
615the user shall create a local ``struct rte_crypto_raw_dp_ctx`` buffer and
616extend to at least the length returned by ``rte_cryptodev_get_raw_dp_ctx_size``
617function call. The created buffer is then initialized using
618``rte_cryptodev_configure_raw_dp_ctx`` function with the ``is_update``
619parameter as 0. The library and the crypto device driver will then set the
620buffer and attach either the cryptodev sym session, the rte_security session,
621or the cryptodev xform for session-less operation into the ctx buffer, and
622set the corresponding enqueue and dequeue function handlers based on the
623algorithm information stored in the session or xform. When the ``is_update``
624parameter passed into ``rte_cryptodev_configure_raw_dp_ctx`` is 1, the driver
625will not initialize the buffer but only update the session or xform and
626the function handlers accordingly.
627
628After the ``struct rte_crypto_raw_dp_ctx`` buffer is initialized, it is now
629ready for enqueue and dequeue operation. There are two different enqueue
630functions: ``rte_cryptodev_raw_enqueue`` to enqueue single raw data
631operation, and ``rte_cryptodev_raw_enqueue_burst`` to enqueue a descriptor
632with multiple operations. In case of the application uses similar approach to
633``struct rte_crypto_sym_vec`` to manage its data burst but with different
634data structure, using the ``rte_cryptodev_raw_enqueue_burst`` function may be
635less efficient as this is a situation where the application has to loop over
636all crypto operations to assemble the ``struct rte_crypto_sym_vec`` descriptor
637from its own data structure, and then the driver will loop over them again to
638translate every operation in the descriptor to the driver's specific queue data.
639The ``rte_cryptodev_raw_enqueue`` should be used to save one loop for each data
640burst instead.
641
642The ``rte_cryptodev_raw_enqueue`` and ``rte_cryptodev_raw_enqueue_burst``
643functions will return or set the enqueue status. ``rte_cryptodev_raw_enqueue``
644will return the status directly, ``rte_cryptodev_raw_enqueue_burst`` will
645return the number of operations enqueued or stored (explained as follows) and
646set the ``enqueue_status`` buffer provided by the user. The possible
647enqueue status values are:
648
649- ``1``: the operation(s) is/are enqueued successfully.
650- ``0``: the operation(s) is/are cached successfully in the crypto device queue
651  but is not actually enqueued. The user shall call
652  ``rte_cryptodev_raw_enqueue_done`` function after the expected operations
653  are stored. The crypto device will then start enqueuing all of them at
654  once.
655- The negative integer: error occurred during enqueue.
656
657Calling ``rte_cryptodev_configure_raw_dp_ctx`` with the parameter ``is_update``
658set as 0 twice without the enqueue function returning or setting enqueue status
659to 1 or ``rte_cryptodev_raw_enqueue_done`` function being called in between will
660invalidate any operation stored in the device queue but not enqueued. This
661feature is useful when the user wants to abandon partially enqueued operations
662for a failed enqueue burst operation and try enqueuing in a whole later.
663
664Similar as enqueue, there are two dequeue functions:
665``rte_cryptodev_raw_dequeue`` for dequeuing single operation, and
666``rte_cryptodev_raw_dequeue_burst`` for dequeuing a burst of operations (e.g.
667all operations in a ``struct rte_crypto_sym_vec`` descriptor). The
668``rte_cryptodev_raw_dequeue_burst`` function allows the user to provide callback
669functions to retrieve dequeue count from the enqueued user data and write the
670expected status value to the user data on the fly. The dequeue functions also
671set the dequeue status:
672
673- ``1``: the operation(s) is/are dequeued successfully.
674- ``0``: the operation(s) is/are completed but is not actually dequeued (hence
675  still kept in the device queue). The user shall call the
676  ``rte_cryptodev_raw_dequeue_done`` function after the expected number of
677  operations (e.g. all operations in a descriptor) are dequeued. The crypto
678  device driver will then free them from the queue at once.
679- The negative integer: error occurred during dequeue.
680
681Calling ``rte_cryptodev_configure_raw_dp_ctx`` with the parameter ``is_update``
682set as 0 twice without the dequeue functions execution changed dequeue_status
683to 1 or ``rte_cryptodev_raw_dequeue_done`` function being called in between will
684revert the crypto device queue's dequeue effort to the moment when the
685``struct rte_crypto_raw_dp_ctx`` buffer is initialized. This feature is useful
686when the user wants to abandon partially dequeued data and try dequeuing again
687later in a whole.
688
689There are a few limitations to the raw data path APIs:
690
691* Only support in-place operations.
692* APIs are NOT thread-safe.
693* CANNOT mix the raw data-path API's enqueue with rte_cryptodev_enqueue_burst,
694  or vice versa.
695
696See *DPDK API Reference* for details on each API definitions.
697
698Sample code
699-----------
700
701There are various sample applications that show how to use the cryptodev library,
702such as the L2fwd with Crypto sample application (L2fwd-crypto) and
703the IPsec Security Gateway application (ipsec-secgw).
704
705While these applications demonstrate how an application can be created to perform
706generic crypto operation, the required complexity hides the basic steps of
707how to use the cryptodev APIs.
708
709The following sample code shows the basic steps to encrypt several buffers
710with AES-CBC (although performing other crypto operations is similar),
711using one of the crypto PMDs available in DPDK.
712
713.. code-block:: c
714
715    /*
716     * Simple example to encrypt several buffers with AES-CBC using
717     * the Cryptodev APIs.
718     */
719
720    #define MAX_SESSIONS         1024
721    #define NUM_MBUFS            1024
722    #define POOL_CACHE_SIZE      128
723    #define BURST_SIZE           32
724    #define BUFFER_SIZE          1024
725    #define AES_CBC_IV_LENGTH    16
726    #define AES_CBC_KEY_LENGTH   16
727    #define IV_OFFSET            (sizeof(struct rte_crypto_op) + \
728                                 sizeof(struct rte_crypto_sym_op))
729
730    struct rte_mempool *mbuf_pool, *crypto_op_pool;
731    struct rte_mempool *session_pool, *session_priv_pool;
732    unsigned int session_size;
733    int ret;
734
735    /* Initialize EAL. */
736    ret = rte_eal_init(argc, argv);
737    if (ret < 0)
738        rte_exit(EXIT_FAILURE, "Invalid EAL arguments\n");
739
740    uint8_t socket_id = rte_socket_id();
741
742    /* Create the mbuf pool. */
743    mbuf_pool = rte_pktmbuf_pool_create("mbuf_pool",
744                                    NUM_MBUFS,
745                                    POOL_CACHE_SIZE,
746                                    0,
747                                    RTE_MBUF_DEFAULT_BUF_SIZE,
748                                    socket_id);
749    if (mbuf_pool == NULL)
750        rte_exit(EXIT_FAILURE, "Cannot create mbuf pool\n");
751
752    /*
753     * The IV is always placed after the crypto operation,
754     * so some private data is required to be reserved.
755     */
756    unsigned int crypto_op_private_data = AES_CBC_IV_LENGTH;
757
758    /* Create crypto operation pool. */
759    crypto_op_pool = rte_crypto_op_pool_create("crypto_op_pool",
760                                            RTE_CRYPTO_OP_TYPE_SYMMETRIC,
761                                            NUM_MBUFS,
762                                            POOL_CACHE_SIZE,
763                                            crypto_op_private_data,
764                                            socket_id);
765    if (crypto_op_pool == NULL)
766        rte_exit(EXIT_FAILURE, "Cannot create crypto op pool\n");
767
768    /* Create the virtual crypto device. */
769    char args[128];
770    const char *crypto_name = "crypto_aesni_mb0";
771    snprintf(args, sizeof(args), "socket_id=%d", socket_id);
772    ret = rte_vdev_init(crypto_name, args);
773    if (ret != 0)
774        rte_exit(EXIT_FAILURE, "Cannot create virtual device");
775
776    uint8_t cdev_id = rte_cryptodev_get_dev_id(crypto_name);
777
778    /* Get private session data size. */
779    session_size = rte_cryptodev_sym_get_private_session_size(cdev_id);
780
781    #ifdef USE_TWO_MEMPOOLS
782    /* Create session mempool for the session header. */
783    session_pool = rte_cryptodev_sym_session_pool_create("session_pool",
784                                    MAX_SESSIONS,
785                                    0,
786                                    POOL_CACHE_SIZE,
787                                    0,
788                                    socket_id);
789
790    /*
791     * Create session private data mempool for the
792     * private session data for the crypto device.
793     */
794    session_priv_pool = rte_mempool_create("session_pool",
795                                    MAX_SESSIONS,
796                                    session_size,
797                                    POOL_CACHE_SIZE,
798                                    0, NULL, NULL, NULL,
799                                    NULL, socket_id,
800                                    0);
801
802    #else
803    /* Use of the same mempool for session header and private data */
804	session_pool = rte_cryptodev_sym_session_pool_create("session_pool",
805                                    MAX_SESSIONS * 2,
806                                    session_size,
807                                    POOL_CACHE_SIZE,
808                                    0,
809                                    socket_id);
810
811	session_priv_pool = session_pool;
812
813    #endif
814
815    /* Configure the crypto device. */
816    struct rte_cryptodev_config conf = {
817        .nb_queue_pairs = 1,
818        .socket_id = socket_id
819    };
820
821    struct rte_cryptodev_qp_conf qp_conf = {
822        .nb_descriptors = 2048,
823        .mp_session = session_pool,
824        .mp_session_private = session_priv_pool
825    };
826
827    if (rte_cryptodev_configure(cdev_id, &conf) < 0)
828        rte_exit(EXIT_FAILURE, "Failed to configure cryptodev %u", cdev_id);
829
830    if (rte_cryptodev_queue_pair_setup(cdev_id, 0, &qp_conf, socket_id) < 0)
831        rte_exit(EXIT_FAILURE, "Failed to setup queue pair\n");
832
833    if (rte_cryptodev_start(cdev_id) < 0)
834        rte_exit(EXIT_FAILURE, "Failed to start device\n");
835
836    /* Create the crypto transform. */
837    uint8_t cipher_key[16] = {0};
838    struct rte_crypto_sym_xform cipher_xform = {
839        .next = NULL,
840        .type = RTE_CRYPTO_SYM_XFORM_CIPHER,
841        .cipher = {
842            .op = RTE_CRYPTO_CIPHER_OP_ENCRYPT,
843            .algo = RTE_CRYPTO_CIPHER_AES_CBC,
844            .key = {
845                .data = cipher_key,
846                .length = AES_CBC_KEY_LENGTH
847            },
848            .iv = {
849                .offset = IV_OFFSET,
850                .length = AES_CBC_IV_LENGTH
851            }
852        }
853    };
854
855    /* Create crypto session and initialize it for the crypto device. */
856    struct rte_cryptodev_sym_session *session;
857    session = rte_cryptodev_sym_session_create(cdev_id, &cipher_xform,
858                    session_pool);
859    if (session == NULL)
860        rte_exit(EXIT_FAILURE, "Session could not be created\n");
861
862    /* Get a burst of crypto operations. */
863    struct rte_crypto_op *crypto_ops[BURST_SIZE];
864    if (rte_crypto_op_bulk_alloc(crypto_op_pool,
865                            RTE_CRYPTO_OP_TYPE_SYMMETRIC,
866                            crypto_ops, BURST_SIZE) == 0)
867        rte_exit(EXIT_FAILURE, "Not enough crypto operations available\n");
868
869    /* Get a burst of mbufs. */
870    struct rte_mbuf *mbufs[BURST_SIZE];
871    if (rte_pktmbuf_alloc_bulk(mbuf_pool, mbufs, BURST_SIZE) < 0)
872        rte_exit(EXIT_FAILURE, "Not enough mbufs available");
873
874    /* Initialize the mbufs and append them to the crypto operations. */
875    unsigned int i;
876    for (i = 0; i < BURST_SIZE; i++) {
877        if (rte_pktmbuf_append(mbufs[i], BUFFER_SIZE) == NULL)
878            rte_exit(EXIT_FAILURE, "Not enough room in the mbuf\n");
879        crypto_ops[i]->sym->m_src = mbufs[i];
880    }
881
882    /* Set up the crypto operations. */
883    for (i = 0; i < BURST_SIZE; i++) {
884        struct rte_crypto_op *op = crypto_ops[i];
885        /* Modify bytes of the IV at the end of the crypto operation */
886        uint8_t *iv_ptr = rte_crypto_op_ctod_offset(op, uint8_t *,
887                                                IV_OFFSET);
888
889        generate_random_bytes(iv_ptr, AES_CBC_IV_LENGTH);
890
891        op->sym->cipher.data.offset = 0;
892        op->sym->cipher.data.length = BUFFER_SIZE;
893
894        /* Attach the crypto session to the operation */
895        rte_crypto_op_attach_sym_session(op, session);
896    }
897
898    /* Enqueue the crypto operations in the crypto device. */
899    uint16_t num_enqueued_ops = rte_cryptodev_enqueue_burst(cdev_id, 0,
900                                            crypto_ops, BURST_SIZE);
901
902    /*
903     * Dequeue the crypto operations until all the operations
904     * are processed in the crypto device.
905     */
906    uint16_t num_dequeued_ops, total_num_dequeued_ops = 0;
907    do {
908        struct rte_crypto_op *dequeued_ops[BURST_SIZE];
909        num_dequeued_ops = rte_cryptodev_dequeue_burst(cdev_id, 0,
910                                        dequeued_ops, BURST_SIZE);
911        total_num_dequeued_ops += num_dequeued_ops;
912
913        /* Check if operation was processed successfully */
914        for (i = 0; i < num_dequeued_ops; i++) {
915            if (dequeued_ops[i]->status != RTE_CRYPTO_OP_STATUS_SUCCESS)
916                rte_exit(EXIT_FAILURE,
917                        "Some operations were not processed correctly");
918        }
919
920        rte_mempool_put_bulk(crypto_op_pool, (void **)dequeued_ops,
921                                            num_dequeued_ops);
922    } while (total_num_dequeued_ops < num_enqueued_ops);
923
924Asymmetric Cryptography
925-----------------------
926
927The cryptodev library currently provides support for the following asymmetric
928Crypto operations; RSA, Modular exponentiation and inversion, Diffie-Hellman and
929Elliptic Curve Diffie-Hellman public and/or private key generation and shared
930secret compute, DSA Signature generation and verification.
931
932Session and Session Management
933~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
934
935Sessions are used in asymmetric cryptographic processing to store the immutable
936data defined in asymmetric cryptographic transform which is further used in the
937operation processing. Sessions typically stores information, such as, public
938and private key information or domain params or prime modulus data i.e. immutable
939across data sets. Crypto sessions cache this immutable data in a optimal way for the
940underlying PMD and this allows further acceleration of the offload of Crypto workloads.
941
942Like symmetric, the Crypto device framework provides APIs to allocate and initialize
943asymmetric sessions for crypto devices, where sessions are mempool objects.
944It is the application's responsibility to create and manage the session mempools.
945Application using both symmetric and asymmetric sessions should allocate and maintain
946different sessions pools for each type.
947
948An application can use ``rte_cryptodev_asym_session_pool_create()`` to create a mempool
949with a specified number of elements. The element size will allow for the session header,
950and the max private session size.
951The max private session size is chosen based on available crypto devices,
952the biggest private session size is used. This means any of those devices can be used,
953and the mempool element will have available space for its private session data.
954
955Once the session mempools have been created, ``rte_cryptodev_asym_session_create()``
956is used to allocate and initialize an asymmetric session from the given mempool.
957An asymmetric transform chain is used to specify the operation and its parameters.
958See the section below for details on transforms.
959
960When a session is no longer used, user must call ``rte_cryptodev_asym_session_clear()``
961for each of the crypto devices that are using the session, to free all driver
962private asymmetric session data. Once this is done, session should be freed using
963``rte_cryptodev_asym_session_free()`` which returns them to their mempool.
964
965Asymmetric Sessionless Support
966~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
967
968Asymmetric crypto framework supports session-less operations as well.
969
970Fields that should be set by user are:
971
972Member xform of struct rte_crypto_asym_op should point to the user created rte_crypto_asym_xform.
973Note that rte_crypto_asym_xform should be immutable for the lifetime of associated crypto_op.
974
975Member sess_type of rte_crypto_op should also be set to RTE_CRYPTO_OP_SESSIONLESS.
976
977Transforms and Transform Chaining
978~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
979
980Asymmetric Crypto transforms (``rte_crypto_asym_xform``) are the mechanism used
981to specify the details of the asymmetric Crypto operation. Next pointer within
982xform allows transform to be chained together. Also it is important to note that
983the order in which the transforms are passed indicates the order of the chaining. Allocation
984of the xform structure is in the application domain. To allow future API extensions in a
985backwardly compatible manner, e.g. addition of a new parameter, the application should
986zero the full xform struct before populating it.
987
988Not all asymmetric crypto xforms are supported for chaining. Currently supported
989asymmetric crypto chaining is Diffie-Hellman private key generation followed by
990public generation. Also, currently API does not support chaining of symmetric and
991asymmetric crypto xforms.
992
993Each xform defines specific asymmetric crypto algo. Currently supported are:
994* RSA
995* Modular operations (Exponentiation and Inverse)
996* Diffie-Hellman
997* DSA
998* Elliptic Curve Diffie-Hellman
999* None - special case where PMD may support a passthrough mode. More for diagnostic purpose
1000
1001See *DPDK API Reference* for details on each rte_crypto_xxx_xform struct
1002
1003Asymmetric Operations
1004~~~~~~~~~~~~~~~~~~~~~
1005
1006The asymmetric Crypto operation structure contains all the mutable data relating
1007to asymmetric cryptographic processing on an input data buffer. It uses either
1008RSA, Modular, Diffie-Hellman or DSA operations depending upon session it is attached
1009to.
1010
1011Every operation must carry a valid session handle which further carries information
1012on xform or xform-chain to be performed on op. Every xform type defines its own set
1013of operational params in their respective rte_crypto_xxx_op_param struct. Depending
1014on xform information within session, PMD picks up and process respective op_param
1015struct.
1016Unlike symmetric, asymmetric operations do not use mbufs for input/output.
1017They operate on data buffer of type ``rte_crypto_param``.
1018
1019See *DPDK API Reference* for details on each rte_crypto_xxx_op_param struct
1020
1021Private user data
1022~~~~~~~~~~~~~~~~~
1023
1024Similar to symmetric above, asymmetric also has a set and get API that provides a
1025mechanism for an application to store and retrieve the private user data information
1026stored along with the crypto session.
1027
1028.. code-block:: c
1029
1030	int rte_cryptodev_asym_session_set_user_data(void *sess,
1031		void *data, uint16_t size);
1032
1033	void * rte_cryptodev_asym_session_get_user_data(void *sess);
1034
1035Please note the ``size`` passed to set API cannot be bigger than the predefined
1036``user_data_sz`` when creating the session mempool, otherwise the function will
1037return an error. Also when ``user_data_sz`` was defined as ``0`` when
1038creating the session mempool, the get API will always return ``NULL``.
1039
1040Asymmetric crypto Sample code
1041-----------------------------
1042
1043There's a unit test application test_cryptodev_asym.c inside unit test framework that
1044show how to setup and process asymmetric operations using cryptodev library.
1045
1046The following code samples are taken from the test application mentioned above,
1047and show basic steps to compute modular exponentiation using an openssl PMD
1048available in DPDK (performing other crypto operations is similar except change
1049to respective op and xform setup).
1050
1051.. note::
1052   The following code snippets are taken from multiple functions, so variable
1053   names may differ slightly between sections.
1054
1055Configure the virtual device, queue pairs, crypto op pool and session mempool.
1056
1057.. literalinclude:: ../../../app/test/test_cryptodev_asym.c
1058   :language: c
1059   :start-after: Device, op pool and session configuration for asymmetric crypto. 8<
1060   :end-before: >8 End of device, op pool and session configuration for asymmetric crypto section.
1061   :dedent: 1
1062
1063Create MODEX data vectors.
1064
1065.. literalinclude:: ../../../app/test/test_cryptodev_mod_test_vectors.h
1066   :language: c
1067   :start-after: MODEX data. 8<
1068   :end-before: >8 End of MODEX data.
1069
1070Setup crypto xform to do modular exponentiation using data vectors.
1071
1072.. literalinclude:: ../../../app/test/test_cryptodev_mod_test_vectors.h
1073   :language: c
1074   :start-after: MODEX vector. 8<
1075   :end-before: >8 End of MODEX vector.
1076
1077Generate crypto op, create and attach a session, then process packets.
1078
1079.. literalinclude:: ../../../app/test/test_cryptodev_asym.c
1080   :language: c
1081   :start-after: Create op, create session, and process packets. 8<
1082   :end-before: >8 End of create op, create session, and process packets section.
1083   :dedent: 1
1084
1085.. note::
1086   The ``rte_cryptodev_asym_session`` struct is hidden from the application.
1087   The ``sess`` pointer used above is a void pointer.
1088
1089
1090Asymmetric Crypto Device API
1091~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1092
1093The cryptodev Library API is described in the
1094`DPDK API Reference <https://doc.dpdk.org/api/>`_
1095
1096
1097Device Statistics
1098-----------------
1099
1100The Cryptodev library has support for displaying Crypto device information
1101through the Telemetry interface. Telemetry commands that can be used
1102are shown below.
1103
1104#. Get the list of available Crypto devices by ID::
1105
1106     --> /cryptodev/list
1107     {"/cryptodev/list": [0, 1, 2, 3]}
1108
1109#. Get general information from a Crypto device::
1110
1111     --> /cryptodev/info,0
1112     {"/cryptodev/info": {"device_name": "0000:1c:01.0_qat_sym",
1113     "max_nb_queue_pairs": 2}}
1114
1115#. Get the statistics for a particular Crypto device::
1116
1117     --> /cryptodev/stats,0
1118     {"/cryptodev/stats": {"enqueued_count": 0, "dequeued_count": 0,
1119     "enqueue_err_count": 0, "dequeue_err_count": 0}}
1120
1121#. Get the capabilities of a particular Crypto device::
1122
1123     --> /cryptodev/caps,0
1124     {"/cryptodev/caps": {"crypto_caps": [<array of serialized bytes of
1125     capabilities>], "crypto_caps_n": <number of capabilities>}}
1126
1127For more information on how to use the Telemetry interface, see
1128the :doc:`../howto/telemetry`.
1129