Stephan Mueller

smueller@chronox.de

Marek Vasut

marek@denx.de

2014 Stephan Mueller This documentation is free software; you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software Foundation; either version 2 of the License, or (at your option) any later version. This program is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for more details. You should have received a copy of the GNU General Public License along with this program; if not, write to the Free Software Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA For more details see the file COPYING in the source distribution of Linux. The kernel crypto API offers a rich set of cryptographic ciphers as well as other data transformation mechanisms and methods to invoke these. This document contains a description of the API and provides example code. To understand and properly use the kernel crypto API a brief explanation of its structure is given. Based on the architecture, the API can be separated into different components. Following the architecture specification, hints to developers of ciphers are provided. Pointers to the API function call documentation are given at the end. The kernel crypto API refers to all algorithms as "transformations". Therefore, a cipher handle variable usually has the name "tfm". Besides cryptographic operations, the kernel crypto API also knows compression transformations and handles them the same way as ciphers. The kernel crypto API serves the following entity types: consumers requesting cryptographic services data transformation implementations (typically ciphers) that can be called by consumers using the kernel crypto API This specification is intended for consumers of the kernel crypto API as well as for developers implementing ciphers. This API specification, however, does not discuss all API calls available to data transformation implementations (i.e. implementations of ciphers and other transformations (such as CRC or even compression algorithms) that can register with the kernel crypto API). Note: The terms "transformation" and cipher algorithm are used interchangeably. The transformation implementation is an actual code or interface to hardware which implements a certain transformation with precisely defined behavior. The transformation object (TFM) is an instance of a transformation implementation. There can be multiple transformation objects associated with a single transformation implementation. Each of those transformation objects is held by a crypto API consumer or another transformation. Transformation object is allocated when a crypto API consumer requests a transformation implementation. The consumer is then provided with a structure, which contains a transformation object (TFM). The structure that contains transformation objects may also be referred to as a "cipher handle". Such a cipher handle is always subject to the following phases that are reflected in the API calls applicable to such a cipher handle: Initialization of a cipher handle. Execution of all intended cipher operations applicable for the handle where the cipher handle must be furnished to every API call. Destruction of a cipher handle. When using the initialization API calls, a cipher handle is created and returned to the consumer. Therefore, please refer to all initialization API calls that refer to the data structure type a consumer is expected to receive and subsequently to use. The initialization API calls have all the same naming conventions of crypto_alloc_*. The transformation context is private data associated with the transformation object. The kernel crypto API provides different API calls for the following cipher types: Symmetric ciphers AEAD ciphers Message digest, including keyed message digest Random number generation User space interface The kernel crypto API provides implementations of single block ciphers and message digests. In addition, the kernel crypto API provides numerous "templates" that can be used in conjunction with the single block ciphers and message digests. Templates include all types of block chaining mode, the HMAC mechanism, etc. Single block ciphers and message digests can either be directly used by a caller or invoked together with a template to form multi-block ciphers or keyed message digests. A single block cipher may even be called with multiple templates. However, templates cannot be used without a single cipher. See /proc/crypto and search for "name". For example: aes ecb(aes) cmac(aes) ccm(aes) rfc4106(gcm(aes)) sha1 hmac(sha1) authenc(hmac(sha1),cbc(aes)) In these examples, "aes" and "sha1" are the ciphers and all others are the templates. The kernel crypto API provides synchronous and asynchronous API operations. When using the synchronous API operation, the caller invokes a cipher operation which is performed synchronously by the kernel crypto API. That means, the caller waits until the cipher operation completes. Therefore, the kernel crypto API calls work like regular function calls. For synchronous operation, the set of API calls is small and conceptually similar to any other crypto library. Asynchronous operation is provided by the kernel crypto API which implies that the invocation of a cipher operation will complete almost instantly. That invocation triggers the cipher operation but it does not signal its completion. Before invoking a cipher operation, the caller must provide a callback function the kernel crypto API can invoke to signal the completion of the cipher operation. Furthermore, the caller must ensure it can handle such asynchronous events by applying appropriate locking around its data. The kernel crypto API does not perform any special serialization operation to protect the caller's data integrity. A cipher is referenced by the caller with a string. That string has the following semantics: template(single block cipher) where "template" and "single block cipher" is the aforementioned template and single block cipher, respectively. If applicable, additional templates may enclose other templates, such as template1(template2(single block cipher))) The kernel crypto API may provide multiple implementations of a template or a single block cipher. For example, AES on newer Intel hardware has the following implementations: AES-NI, assembler implementation, or straight C. Now, when using the string "aes" with the kernel crypto API, which cipher implementation is used? The answer to that question is the priority number assigned to each cipher implementation by the kernel crypto API. When a caller uses the string to refer to a cipher during initialization of a cipher handle, the kernel crypto API looks up all implementations providing an implementation with that name and selects the implementation with the highest priority. Now, a caller may have the need to refer to a specific cipher implementation and thus does not want to rely on the priority-based selection. To accommodate this scenario, the kernel crypto API allows the cipher implementation to register a unique name in addition to common names. When using that unique name, a caller is therefore always sure to refer to the intended cipher implementation. The list of available ciphers is given in /proc/crypto. However, that list does not specify all possible permutations of templates and ciphers. Each block listed in /proc/crypto may contain the following information -- if one of the components listed as follows are not applicable to a cipher, it is not displayed: name: the generic name of the cipher that is subject to the priority-based selection -- this name can be used by the cipher allocation API calls (all names listed above are examples for such generic names) driver: the unique name of the cipher -- this name can be used by the cipher allocation API calls module: the kernel module providing the cipher implementation (or "kernel" for statically linked ciphers) priority: the priority value of the cipher implementation refcnt: the reference count of the respective cipher (i.e. the number of current consumers of this cipher) selftest: specification whether the self test for the cipher passed type: blkcipher for synchronous block ciphers ablkcipher for asynchronous block ciphers cipher for single block ciphers that may be used with an additional template shash for synchronous message digest ahash for asynchronous message digest aead for AEAD cipher type compression for compression type transformations rng for random number generator givcipher for cipher with associated IV generator (see the geniv entry below for the specification of the IV generator type used by the cipher implementation) blocksize: blocksize of cipher in bytes keysize: key size in bytes ivsize: IV size in bytes seedsize: required size of seed data for random number generator digestsize: output size of the message digest geniv: IV generation type: eseqiv for encrypted sequence number based IV generation seqiv for sequence number based IV generation chainiv for chain iv generation <builtin> is a marker that the cipher implements IV generation and handling as it is specific to the given cipher When allocating a cipher handle, the caller only specifies the cipher type. Symmetric ciphers, however, typically support multiple key sizes (e.g. AES-128 vs. AES-192 vs. AES-256). These key sizes are determined with the length of the provided key. Thus, the kernel crypto API does not provide a separate way to select the particular symmetric cipher key size. The different cipher handle allocation functions allow the specification of a type and mask flag. Both parameters have the following meaning (and are therefore not covered in the subsequent sections). The type flag specifies the type of the cipher algorithm. The caller usually provides a 0 when the caller wants the default handling. Otherwise, the caller may provide the following selections which match the the aforementioned cipher types: CRYPTO_ALG_TYPE_CIPHER Single block cipher CRYPTO_ALG_TYPE_COMPRESS Compression CRYPTO_ALG_TYPE_AEAD Authenticated Encryption with Associated Data (MAC) CRYPTO_ALG_TYPE_BLKCIPHER Synchronous multi-block cipher CRYPTO_ALG_TYPE_ABLKCIPHER Asynchronous multi-block cipher CRYPTO_ALG_TYPE_GIVCIPHER Asynchronous multi-block cipher packed together with an IV generator (see geniv field in the /proc/crypto listing for the known IV generators) CRYPTO_ALG_TYPE_DIGEST Raw message digest CRYPTO_ALG_TYPE_HASH Alias for CRYPTO_ALG_TYPE_DIGEST CRYPTO_ALG_TYPE_SHASH Synchronous multi-block hash CRYPTO_ALG_TYPE_AHASH Asynchronous multi-block hash CRYPTO_ALG_TYPE_RNG Random Number Generation CRYPTO_ALG_TYPE_PCOMPRESS Enhanced version of CRYPTO_ALG_TYPE_COMPRESS allowing for segmented compression / decompression instead of performing the operation on one segment only. CRYPTO_ALG_TYPE_PCOMPRESS is intended to replace CRYPTO_ALG_TYPE_COMPRESS once existing consumers are converted. The mask flag restricts the type of cipher. The only allowed flag is CRYPTO_ALG_ASYNC to restrict the cipher lookup function to asynchronous ciphers. Usually, a caller provides a 0 for the mask flag. When the caller provides a mask and type specification, the caller limits the search the kernel crypto API can perform for a suitable cipher implementation for the given cipher name. That means, even when a caller uses a cipher name that exists during its initialization call, the kernel crypto API may not select it due to the used type and mask field. The kernel crypto API has an internal structure where a cipher implementation may use many layers and indirections. This section shall help to clarify how the kernel crypto API uses various components to implement the complete cipher. The following subsections explain the internal structure based on existing cipher implementations. The first section addresses the most complex scenario where all other scenarios form a logical subset. The following ASCII art decomposes the kernel crypto API layers when using the AEAD cipher with the automated IV generation. The shown example is used by the IPSEC layer. For other use cases of AEAD ciphers, the ASCII art applies as well, but the caller may not use the AEAD cipher with a separate IV generator. In this case, the caller must generate the IV. The depicted example decomposes the AEAD cipher of GCM(AES) based on the generic C implementations (gcm.c, aes-generic.c, ctr.c, ghash-generic.c, seqiv.c). The generic implementation serves as an example showing the complete logic of the kernel crypto API. It is possible that some streamlined cipher implementations (like AES-NI) provide implementations merging aspects which in the view of the kernel crypto API cannot be decomposed into layers any more. In case of the AES-NI implementation, the CTR mode, the GHASH implementation and the AES cipher are all merged into one cipher implementation registered with the kernel crypto API. In this case, the concept described by the following ASCII art applies too. However, the decomposition of GCM into the individual sub-components by the kernel crypto API is not done any more. Each block in the following ASCII art is an independent cipher instance obtained from the kernel crypto API. Each block is accessed by the caller or by other blocks using the API functions defined by the kernel crypto API for the cipher implementation type. The blocks below indicate the cipher type as well as the specific logic implemented in the cipher. The ASCII art picture also indicates the call structure, i.e. who calls which component. The arrows point to the invoked block where the caller uses the API applicable to the cipher type specified for the block. The following call sequence is applicable when the IPSEC layer triggers an encryption operation with the esp_output function. During configuration, the administrator set up the use of rfc4106(gcm(aes)) as the cipher for ESP. The following call sequence is now depicted in the ASCII art above: esp_output() invokes crypto_aead_encrypt() to trigger an encryption operation of the AEAD cipher with IV generator. In case of GCM, the SEQIV implementation is registered as GIVCIPHER in crypto_rfc4106_alloc(). The SEQIV performs its operation to generate an IV where the core function is seqiv_geniv(). Now, SEQIV uses the AEAD API function calls to invoke the associated AEAD cipher. In our case, during the instantiation of SEQIV, the cipher handle for GCM is provided to SEQIV. This means that SEQIV invokes AEAD cipher operations with the GCM cipher handle. During instantiation of the GCM handle, the CTR(AES) and GHASH ciphers are instantiated. The cipher handles for CTR(AES) and GHASH are retained for later use. The GCM implementation is responsible to invoke the CTR mode AES and the GHASH cipher in the right manner to implement the GCM specification. The GCM AEAD cipher type implementation now invokes the ABLKCIPHER API with the instantiated CTR(AES) cipher handle. During instantiation of the CTR(AES) cipher, the CIPHER type implementation of AES is instantiated. The cipher handle for AES is retained. That means that the ABLKCIPHER implementation of CTR(AES) only implements the CTR block chaining mode. After performing the block chaining operation, the CIPHER implementation of AES is invoked. The ABLKCIPHER of CTR(AES) now invokes the CIPHER API with the AES cipher handle to encrypt one block. The GCM AEAD implementation also invokes the GHASH cipher implementation via the AHASH API. When the IPSEC layer triggers the esp_input() function, the same call sequence is followed with the only difference that the operation starts with step (2). Generic block ciphers follow the same concept as depicted with the ASCII art picture above. For example, CBC(AES) is implemented with cbc.c, and aes-generic.c. The ASCII art picture above applies as well with the difference that only step (4) is used and the ABLKCIPHER block chaining mode is CBC. Keyed message digest implementations again follow the same concept as depicted in the ASCII art picture above. For example, HMAC(SHA256) is implemented with hmac.c and sha256_generic.c. The following ASCII art illustrates the implementation: The following call sequence is applicable when a caller triggers an HMAC operation: The AHASH API functions are invoked by the caller. The HMAC implementation performs its operation as needed. During initialization of the HMAC cipher, the SHASH cipher type of SHA256 is instantiated. The cipher handle for the SHA256 instance is retained. At one time, the HMAC implementation requires a SHA256 operation where the SHA256 cipher handle is used. The HMAC instance now invokes the SHASH API with the SHA256 cipher handle to calculate the message digest. There are three distinct types of registration functions in the Crypto API. One is used to register a generic cryptographic transformation, while the other two are specific to HASH transformations and COMPRESSion. We will discuss the latter two in a separate chapter, here we will only look at the generic ones. Before discussing the register functions, the data structure to be filled with each, struct crypto_alg, must be considered -- see below for a description of this data structure. The generic registration functions can be found in include/linux/crypto.h and their definition can be seen below. The former function registers a single transformation, while the latter works on an array of transformation descriptions. The latter is useful when registering transformations in bulk. int crypto_register_alg(struct crypto_alg *alg); int crypto_register_algs(struct crypto_alg *algs, int count); The counterparts to those functions are listed below. int crypto_unregister_alg(struct crypto_alg *alg); int crypto_unregister_algs(struct crypto_alg *algs, int count); Notice that both registration and unregistration functions do return a value, so make sure to handle errors. A return code of zero implies success. Any return code < 0 implies an error. The bulk registration / unregistration functions require that struct crypto_alg is an array of count size. These functions simply loop over that array and register / unregister each individual algorithm. If an error occurs, the loop is terminated at the offending algorithm definition. That means, the algorithms prior to the offending algorithm are successfully registered. Note, the caller has no way of knowing which cipher implementations have successfully registered. If this is important to know, the caller should loop through the different implementations using the single instance *_alg functions for each individual implementation. Example of transformations: aes, arc4, ... This section describes the simplest of all transformation implementations, that being the CIPHER type used for symmetric ciphers. The CIPHER type is used for transformations which operate on exactly one block at a time and there are no dependencies between blocks at all. The registration of [CIPHER] algorithm is specific in that struct crypto_alg field .cra_type is empty. The .cra_u.cipher has to be filled in with proper callbacks to implement this transformation. See struct cipher_alg below. Struct cipher_alg defines a single block cipher. Here are schematics of how these functions are called when operated from other part of the kernel. Note that the .cia_setkey() call might happen before or after any of these schematics happen, but must not happen during any of these are in-flight. KEY ---. PLAINTEXT ---. v v .cia_setkey() -> .cia_encrypt() | '-----> CIPHERTEXT Please note that a pattern where .cia_setkey() is called multiple times is also valid: KEY1 --. PLAINTEXT1 --. KEY2 --. PLAINTEXT2 --. v v v v .cia_setkey() -> .cia_encrypt() -> .cia_setkey() -> .cia_encrypt() | | '---> CIPHERTEXT1 '---> CIPHERTEXT2 Example of transformations: cbc(aes), ecb(arc4), ... This section describes the multi-block cipher transformation implementations for both synchronous [BLKCIPHER] and asynchronous [ABLKCIPHER] case. The multi-block ciphers are used for transformations which operate on scatterlists of data supplied to the transformation functions. They output the result into a scatterlist of data as well. The registration of [BLKCIPHER] or [ABLKCIPHER] algorithms is one of the most standard procedures throughout the crypto API. Note, if a cipher implementation requires a proper alignment of data, the caller should use the functions of crypto_blkcipher_alignmask() or crypto_ablkcipher_alignmask() respectively to identify a memory alignment mask. The kernel crypto API is able to process requests that are unaligned. This implies, however, additional overhead as the kernel crypto API needs to perform the realignment of the data which may imply moving of data. Struct blkcipher_alg defines a synchronous block cipher whereas struct ablkcipher_alg defines an asynchronous block cipher. Please refer to the single block cipher description for schematics of the block cipher usage. The usage patterns are exactly the same for [ABLKCIPHER] and [BLKCIPHER] as they are for plain [CIPHER]. There are a couple of specifics to the [ABLKCIPHER] interface. First of all, some of the drivers will want to use the Generic ScatterWalk in case the hardware needs to be fed separate chunks of the scatterlist which contains the plaintext and will contain the ciphertext. Please refer to the ScatterWalk interface offered by the Linux kernel scatter / gather list implementation. Example of transformations: crc32, md5, sha1, sha256,... There are multiple ways to register a HASH transformation, depending on whether the transformation is synchronous [SHASH] or asynchronous [AHASH] and the amount of HASH transformations we are registering. You can find the prototypes defined in include/crypto/internal/hash.h: int crypto_register_ahash(struct ahash_alg *alg); int crypto_register_shash(struct shash_alg *alg); int crypto_register_shashes(struct shash_alg *algs, int count); The respective counterparts for unregistering the HASH transformation are as follows: int crypto_unregister_ahash(struct ahash_alg *alg); int crypto_unregister_shash(struct shash_alg *alg); int crypto_unregister_shashes(struct shash_alg *algs, int count); Here are schematics of how these functions are called when operated from other part of the kernel. Note that the .setkey() call might happen before or after any of these schematics happen, but must not happen during any of these are in-flight. Please note that calling .init() followed immediately by .finish() is also a perfectly valid transformation. I) DATA -----------. v .init() -> .update() -> .final() ! .update() might not be called ^ | | at all in this scenario. '----' '---> HASH II) DATA -----------.-----------. v v .init() -> .update() -> .finup() ! .update() may not be called ^ | | at all in this scenario. '----' '---> HASH III) DATA -----------. v .digest() ! The entire process is handled | by the .digest() call. '---------------> HASH Here is a schematic of how the .export()/.import() functions are called when used from another part of the kernel. KEY--. DATA--. v v ! .update() may not be called .setkey() -> .init() -> .update() -> .export() at all in this scenario. ^ | | '-----' '--> PARTIAL_HASH ----------- other transformations happen here ----------- PARTIAL_HASH--. DATA1--. v v .import -> .update() -> .final() ! .update() may not be called ^ | | at all in this scenario. '----' '--> HASH1 PARTIAL_HASH--. DATA2-. v v .import -> .finup() | '---------------> HASH2 Some of the drivers will want to use the Generic ScatterWalk in case the implementation needs to be fed separate chunks of the scatterlist which contains the input data. The buffer containing the resulting hash will always be properly aligned to .cra_alignmask so there is no need to worry about this. The concepts of the kernel crypto API visible to kernel space is fully applicable to the user space interface as well. Therefore, the kernel crypto API high level discussion for the in-kernel use cases applies here as well. The major difference, however, is that user space can only act as a consumer and never as a provider of a transformation or cipher algorithm. The following covers the user space interface exported by the kernel crypto API. A working example of this description is libkcapi that can be obtained from [1]. That library can be used by user space applications that require cryptographic services from the kernel. Some details of the in-kernel kernel crypto API aspects do not apply to user space, however. This includes the difference between synchronous and asynchronous invocations. The user space API call is fully synchronous. [1] http://www.chronox.de/libkcapi.html The kernel crypto API is accessible from user space. Currently, the following ciphers are accessible: Message digest including keyed message digest (HMAC, CMAC) Symmetric ciphers AEAD ciphers Random Number Generators The interface is provided via socket type using the type AF_ALG. In addition, the setsockopt option type is SOL_ALG. In case the user space header files do not export these flags yet, use the following macros: #ifndef AF_ALG #define AF_ALG 38 #endif #ifndef SOL_ALG #define SOL_ALG 279 #endif A cipher is accessed with the same name as done for the in-kernel API calls. This includes the generic vs. unique naming schema for ciphers as well as the enforcement of priorities for generic names. To interact with the kernel crypto API, a socket must be created by the user space application. User space invokes the cipher operation with the send()/write() system call family. The result of the cipher operation is obtained with the read()/recv() system call family. The following API calls assume that the socket descriptor is already opened by the user space application and discusses only the kernel crypto API specific invocations. To initialize the socket interface, the following sequence has to be performed by the consumer: Create a socket of type AF_ALG with the struct sockaddr_alg parameter specified below for the different cipher types. Invoke bind with the socket descriptor Invoke accept with the socket descriptor. The accept system call returns a new file descriptor that is to be used to interact with the particular cipher instance. When invoking send/write or recv/read system calls to send data to the kernel or obtain data from the kernel, the file descriptor returned by accept must be used. Just like the in-kernel operation of the kernel crypto API, the user space interface allows the cipher operation in-place. That means that the input buffer used for the send/write system call and the output buffer used by the read/recv system call may be one and the same. This is of particular interest for symmetric cipher operations where a copying of the output data to its final destination can be avoided. If a consumer on the other hand wants to maintain the plaintext and the ciphertext in different memory locations, all a consumer needs to do is to provide different memory pointers for the encryption and decryption operation. The message digest type to be used for the cipher operation is selected when invoking the bind syscall. bind requires the caller to provide a filled struct sockaddr data structure. This data structure must be filled as follows: struct sockaddr_alg sa = { .salg_family = AF_ALG, .salg_type = "hash", /* this selects the hash logic in the kernel */ .salg_name = "sha1" /* this is the cipher name */ }; The salg_type value "hash" applies to message digests and keyed message digests. Though, a keyed message digest is referenced by the appropriate salg_name. Please see below for the setsockopt interface that explains how the key can be set for a keyed message digest. Using the send() system call, the application provides the data that should be processed with the message digest. The send system call allows the following flags to be specified: MSG_MORE: If this flag is set, the send system call acts like a message digest update function where the final hash is not yet calculated. If the flag is not set, the send system call calculates the final message digest immediately. With the recv() system call, the application can read the message digest from the kernel crypto API. If the buffer is too small for the message digest, the flag MSG_TRUNC is set by the kernel. In order to set a message digest key, the calling application must use the setsockopt() option of ALG_SET_KEY. If the key is not set the HMAC operation is performed without the initial HMAC state change caused by the key. The operation is very similar to the message digest discussion. During initialization, the struct sockaddr data structure must be filled as follows: struct sockaddr_alg sa = { .salg_family = AF_ALG, .salg_type = "skcipher", /* this selects the symmetric cipher */ .salg_name = "cbc(aes)" /* this is the cipher name */ }; Before data can be sent to the kernel using the write/send system call family, the consumer must set the key. The key setting is described with the setsockopt invocation below. Using the sendmsg() system call, the application provides the data that should be processed for encryption or decryption. In addition, the IV is specified with the data structure provided by the sendmsg() system call. The sendmsg system call parameter of struct msghdr is embedded into the struct cmsghdr data structure. See recv(2) and cmsg(3) for more information on how the cmsghdr data structure is used together with the send/recv system call family. That cmsghdr data structure holds the following information specified with a separate header instances: specification of the cipher operation type with one of these flags: ALG_OP_ENCRYPT - encryption of data ALG_OP_DECRYPT - decryption of data specification of the IV information marked with the flag ALG_SET_IV The send system call family allows the following flag to be specified: MSG_MORE: If this flag is set, the send system call acts like a cipher update function where more input data is expected with a subsequent invocation of the send system call. Note: The kernel reports -EINVAL for any unexpected data. The caller must make sure that all data matches the constraints given in /proc/crypto for the selected cipher. With the recv() system call, the application can read the result of the cipher operation from the kernel crypto API. The output buffer must be at least as large as to hold all blocks of the encrypted or decrypted data. If the output data size is smaller, only as many blocks are returned that fit into that output buffer size. The operation is very similar to the symmetric cipher discussion. During initialization, the struct sockaddr data structure must be filled as follows: struct sockaddr_alg sa = { .salg_family = AF_ALG, .salg_type = "aead", /* this selects the symmetric cipher */ .salg_name = "gcm(aes)" /* this is the cipher name */ }; Before data can be sent to the kernel using the write/send system call family, the consumer must set the key. The key setting is described with the setsockopt invocation below. In addition, before data can be sent to the kernel using the write/send system call family, the consumer must set the authentication tag size. To set the authentication tag size, the caller must use the setsockopt invocation described below. Using the sendmsg() system call, the application provides the data that should be processed for encryption or decryption. In addition, the IV is specified with the data structure provided by the sendmsg() system call. The sendmsg system call parameter of struct msghdr is embedded into the struct cmsghdr data structure. See recv(2) and cmsg(3) for more information on how the cmsghdr data structure is used together with the send/recv system call family. That cmsghdr data structure holds the following information specified with a separate header instances: specification of the cipher operation type with one of these flags: ALG_OP_ENCRYPT - encryption of data ALG_OP_DECRYPT - decryption of data specification of the IV information marked with the flag ALG_SET_IV specification of the associated authentication data (AAD) with the flag ALG_SET_AEAD_ASSOCLEN. The AAD is sent to the kernel together with the plaintext / ciphertext. See below for the memory structure. The send system call family allows the following flag to be specified: MSG_MORE: If this flag is set, the send system call acts like a cipher update function where more input data is expected with a subsequent invocation of the send system call. Note: The kernel reports -EINVAL for any unexpected data. The caller must make sure that all data matches the constraints given in /proc/crypto for the selected cipher. With the recv() system call, the application can read the result of the cipher operation from the kernel crypto API. The output buffer must be at least as large as defined with the memory structure below. If the output data size is smaller, the cipher operation is not performed. The authenticated decryption operation may indicate an integrity error. Such breach in integrity is marked with the -EBADMSG error code. The AEAD cipher operates with the following information that is communicated between user and kernel space as one data stream: plaintext or ciphertext associated authentication data (AAD) authentication tag The sizes of the AAD and the authentication tag are provided with the sendmsg and setsockopt calls (see there). As the kernel knows the size of the entire data stream, the kernel is now able to calculate the right offsets of the data components in the data stream. The user space caller must arrange the aforementioned information in the following order: AEAD encryption input: AAD || plaintext AEAD decryption input: AAD || ciphertext || authentication tag The output buffer the user space caller provides must be at least as large to hold the following data: AEAD encryption output: ciphertext || authentication tag AEAD decryption output: plaintext Again, the operation is very similar to the other APIs. During initialization, the struct sockaddr data structure must be filled as follows: struct sockaddr_alg sa = { .salg_family = AF_ALG, .salg_type = "rng", /* this selects the symmetric cipher */ .salg_name = "drbg_nopr_sha256" /* this is the cipher name */ }; Depending on the RNG type, the RNG must be seeded. The seed is provided using the setsockopt interface to set the key. For example, the ansi_cprng requires a seed. The DRBGs do not require a seed, but may be seeded. Using the read()/recvmsg() system calls, random numbers can be obtained. The kernel generates at most 128 bytes in one call. If user space requires more data, multiple calls to read()/recvmsg() must be made. WARNING: The user space caller may invoke the initially mentioned accept system call multiple times. In this case, the returned file descriptors have the same state. In addition to the send/write/read/recv system call family, the AF_ALG interface can be accessed with the zero-copy interface of splice/vmsplice. As the name indicates, the kernel tries to avoid a copy operation into kernel space. The zero-copy operation requires data to be aligned at the page boundary. Non-aligned data can be used as well, but may require more operations of the kernel which would defeat the speed gains obtained from the zero-copy interface. The system-interent limit for the size of one zero-copy operation is 16 pages. If more data is to be sent to AF_ALG, user space must slice the input into segments with a maximum size of 16 pages. Zero-copy can be used with the following code example (a complete working example is provided with libkcapi): int pipes[2]; pipe(pipes); /* input data in iov */ vmsplice(pipes[1], iov, iovlen, SPLICE_F_GIFT); /* opfd is the file descriptor returned from accept() system call */ splice(pipes[0], NULL, opfd, NULL, ret, 0); read(opfd, out, outlen); In addition to the read/recv and send/write system call handling to send and retrieve data subject to the cipher operation, a consumer also needs to set the additional information for the cipher operation. This additional information is set using the setsockopt system call that must be invoked with the file descriptor of the open cipher (i.e. the file descriptor returned by the accept system call). Each setsockopt invocation must use the level SOL_ALG. The setsockopt interface allows setting the following data using the mentioned optname: ALG_SET_KEY -- Setting the key. Key setting is applicable to: the skcipher cipher type (symmetric ciphers) the hash cipher type (keyed message digests) the AEAD cipher type the RNG cipher type to provide the seed ALG_SET_AEAD_AUTHSIZE -- Setting the authentication tag size for AEAD ciphers. For a encryption operation, the authentication tag of the given size will be generated. For a decryption operation, the provided ciphertext is assumed to contain an authentication tag of the given size (see section about AEAD memory layout below). Please see [1] for libkcapi which provides an easy-to-use wrapper around the aforementioned Netlink kernel interface. [1] also contains a test application that invokes all libkcapi API calls. [1] http://www.chronox.de/libkcapi.html Please note that the kernel crypto API contains the AEAD givcrypt API (crypto_aead_giv* and aead_givcrypt_* function calls in include/crypto/aead.h). This API is obsolete and will be removed in the future. To obtain the functionality of an AEAD cipher with internal IV generation, use the IV generator as a regular cipher. For example, rfc4106(gcm(aes)) is the AEAD cipher with external IV generation and seqniv(rfc4106(gcm(aes))) implies that the kernel crypto API generates the IV. Different IV generators are available. !Pinclude/linux/crypto.h Block Cipher Context Data Structures !Finclude/crypto/aead.h aead_request !Pinclude/linux/crypto.h Block Cipher Algorithm Definitions !Finclude/linux/crypto.h crypto_alg !Finclude/linux/crypto.h ablkcipher_alg !Finclude/crypto/aead.h aead_alg !Finclude/linux/crypto.h blkcipher_alg !Finclude/linux/crypto.h cipher_alg !Finclude/crypto/rng.h rng_alg !Pinclude/linux/crypto.h Asynchronous Block Cipher API !Finclude/linux/crypto.h crypto_alloc_ablkcipher !Finclude/linux/crypto.h crypto_free_ablkcipher !Finclude/linux/crypto.h crypto_has_ablkcipher !Finclude/linux/crypto.h crypto_ablkcipher_ivsize !Finclude/linux/crypto.h crypto_ablkcipher_blocksize !Finclude/linux/crypto.h crypto_ablkcipher_setkey !Finclude/linux/crypto.h crypto_ablkcipher_reqtfm !Finclude/linux/crypto.h crypto_ablkcipher_encrypt !Finclude/linux/crypto.h crypto_ablkcipher_decrypt !Pinclude/linux/crypto.h Asynchronous Cipher Request Handle !Finclude/linux/crypto.h crypto_ablkcipher_reqsize !Finclude/linux/crypto.h ablkcipher_request_set_tfm !Finclude/linux/crypto.h ablkcipher_request_alloc !Finclude/linux/crypto.h ablkcipher_request_free !Finclude/linux/crypto.h ablkcipher_request_set_callback !Finclude/linux/crypto.h ablkcipher_request_set_crypt !Pinclude/crypto/aead.h Authenticated Encryption With Associated Data (AEAD) Cipher API !Finclude/crypto/aead.h crypto_alloc_aead !Finclude/crypto/aead.h crypto_free_aead !Finclude/crypto/aead.h crypto_aead_ivsize !Finclude/crypto/aead.h crypto_aead_authsize !Finclude/crypto/aead.h crypto_aead_blocksize !Finclude/crypto/aead.h crypto_aead_setkey !Finclude/crypto/aead.h crypto_aead_setauthsize !Finclude/crypto/aead.h crypto_aead_encrypt !Finclude/crypto/aead.h crypto_aead_decrypt !Pinclude/crypto/aead.h Asynchronous AEAD Request Handle !Finclude/crypto/aead.h crypto_aead_reqsize !Finclude/crypto/aead.h aead_request_set_tfm !Finclude/crypto/aead.h aead_request_alloc !Finclude/crypto/aead.h aead_request_free !Finclude/crypto/aead.h aead_request_set_callback !Finclude/crypto/aead.h aead_request_set_crypt !Finclude/crypto/aead.h aead_request_set_assoc !Finclude/crypto/aead.h aead_request_set_ad !Pinclude/linux/crypto.h Synchronous Block Cipher API !Finclude/linux/crypto.h crypto_alloc_blkcipher !Finclude/linux/crypto.h crypto_free_blkcipher !Finclude/linux/crypto.h crypto_has_blkcipher !Finclude/linux/crypto.h crypto_blkcipher_name !Finclude/linux/crypto.h crypto_blkcipher_ivsize !Finclude/linux/crypto.h crypto_blkcipher_blocksize !Finclude/linux/crypto.h crypto_blkcipher_setkey !Finclude/linux/crypto.h crypto_blkcipher_encrypt !Finclude/linux/crypto.h crypto_blkcipher_encrypt_iv !Finclude/linux/crypto.h crypto_blkcipher_decrypt !Finclude/linux/crypto.h crypto_blkcipher_decrypt_iv !Finclude/linux/crypto.h crypto_blkcipher_set_iv !Finclude/linux/crypto.h crypto_blkcipher_get_iv !Pinclude/linux/crypto.h Single Block Cipher API !Finclude/linux/crypto.h crypto_alloc_cipher !Finclude/linux/crypto.h crypto_free_cipher !Finclude/linux/crypto.h crypto_has_cipher !Finclude/linux/crypto.h crypto_cipher_blocksize !Finclude/linux/crypto.h crypto_cipher_setkey !Finclude/linux/crypto.h crypto_cipher_encrypt_one !Finclude/linux/crypto.h crypto_cipher_decrypt_one !Pinclude/crypto/hash.h Message Digest Algorithm Definitions !Finclude/crypto/hash.h hash_alg_common !Finclude/crypto/hash.h ahash_alg !Finclude/crypto/hash.h shash_alg !Pinclude/crypto/hash.h Asynchronous Message Digest API !Finclude/crypto/hash.h crypto_alloc_ahash !Finclude/crypto/hash.h crypto_free_ahash !Finclude/crypto/hash.h crypto_ahash_init !Finclude/crypto/hash.h crypto_ahash_digestsize !Finclude/crypto/hash.h crypto_ahash_reqtfm !Finclude/crypto/hash.h crypto_ahash_reqsize !Finclude/crypto/hash.h crypto_ahash_setkey !Finclude/crypto/hash.h crypto_ahash_finup !Finclude/crypto/hash.h crypto_ahash_final !Finclude/crypto/hash.h crypto_ahash_digest !Finclude/crypto/hash.h crypto_ahash_export !Finclude/crypto/hash.h crypto_ahash_import !Pinclude/crypto/hash.h Asynchronous Hash Request Handle !Finclude/crypto/hash.h ahash_request_set_tfm !Finclude/crypto/hash.h ahash_request_alloc !Finclude/crypto/hash.h ahash_request_free !Finclude/crypto/hash.h ahash_request_set_callback !Finclude/crypto/hash.h ahash_request_set_crypt !Pinclude/crypto/hash.h Synchronous Message Digest API !Finclude/crypto/hash.h crypto_alloc_shash !Finclude/crypto/hash.h crypto_free_shash !Finclude/crypto/hash.h crypto_shash_blocksize !Finclude/crypto/hash.h crypto_shash_digestsize !Finclude/crypto/hash.h crypto_shash_descsize !Finclude/crypto/hash.h crypto_shash_setkey !Finclude/crypto/hash.h crypto_shash_digest !Finclude/crypto/hash.h crypto_shash_export !Finclude/crypto/hash.h crypto_shash_import !Finclude/crypto/hash.h crypto_shash_init !Finclude/crypto/hash.h crypto_shash_update !Finclude/crypto/hash.h crypto_shash_final !Finclude/crypto/hash.h crypto_shash_finup !Pinclude/crypto/rng.h Random number generator API !Finclude/crypto/rng.h crypto_alloc_rng !Finclude/crypto/rng.h crypto_rng_alg !Finclude/crypto/rng.h crypto_free_rng !Finclude/crypto/rng.h crypto_rng_get_bytes !Finclude/crypto/rng.h crypto_rng_reset !Finclude/crypto/rng.h crypto_rng_seedsize !Cinclude/crypto/rng.h struct tcrypt_result { struct completion completion; int err; }; /* tie all data structures together */ struct ablkcipher_def { struct scatterlist sg; struct crypto_ablkcipher *tfm; struct ablkcipher_request *req; struct tcrypt_result result; }; /* Callback function */ static void test_ablkcipher_cb(struct crypto_async_request *req, int error) { struct tcrypt_result *result = req->data; if (error == -EINPROGRESS) return; result->err = error; complete(&result->completion); pr_info("Encryption finished successfully\n"); } /* Perform cipher operation */ static unsigned int test_ablkcipher_encdec(struct ablkcipher_def *ablk, int enc) { int rc = 0; if (enc) rc = crypto_ablkcipher_encrypt(ablk->req); else rc = crypto_ablkcipher_decrypt(ablk->req); switch (rc) { case 0: break; case -EINPROGRESS: case -EBUSY: rc = wait_for_completion_interruptible( &ablk->result.completion); if (!rc && !ablk->result.err) { reinit_completion(&ablk->result.completion); break; } default: pr_info("ablkcipher encrypt returned with %d result %d\n", rc, ablk->result.err); break; } init_completion(&ablk->result.completion); return rc; } /* Initialize and trigger cipher operation */ static int test_ablkcipher(void) { struct ablkcipher_def ablk; struct crypto_ablkcipher *ablkcipher = NULL; struct ablkcipher_request *req = NULL; char *scratchpad = NULL; char *ivdata = NULL; unsigned char key[32]; int ret = -EFAULT; ablkcipher = crypto_alloc_ablkcipher("cbc-aes-aesni", 0, 0); if (IS_ERR(ablkcipher)) { pr_info("could not allocate ablkcipher handle\n"); return PTR_ERR(ablkcipher); } req = ablkcipher_request_alloc(ablkcipher, GFP_KERNEL); if (IS_ERR(req)) { pr_info("could not allocate request queue\n"); ret = PTR_ERR(req); goto out; } ablkcipher_request_set_callback(req, CRYPTO_TFM_REQ_MAY_BACKLOG, test_ablkcipher_cb, &ablk.result); /* AES 256 with random key */ get_random_bytes(&key, 32); if (crypto_ablkcipher_setkey(ablkcipher, key, 32)) { pr_info("key could not be set\n"); ret = -EAGAIN; goto out; } /* IV will be random */ ivdata = kmalloc(16, GFP_KERNEL); if (!ivdata) { pr_info("could not allocate ivdata\n"); goto out; } get_random_bytes(ivdata, 16); /* Input data will be random */ scratchpad = kmalloc(16, GFP_KERNEL); if (!scratchpad) { pr_info("could not allocate scratchpad\n"); goto out; } get_random_bytes(scratchpad, 16); ablk.tfm = ablkcipher; ablk.req = req; /* We encrypt one block */ sg_init_one(&ablk.sg, scratchpad, 16); ablkcipher_request_set_crypt(req, &ablk.sg, &ablk.sg, 16, ivdata); init_completion(&ablk.result.completion); /* encrypt data */ ret = test_ablkcipher_encdec(&ablk, 1); if (ret) goto out; pr_info("Encryption triggered successfully\n"); out: if (ablkcipher) crypto_free_ablkcipher(ablkcipher); if (req) ablkcipher_request_free(req); if (ivdata) kfree(ivdata); if (scratchpad) kfree(scratchpad); return ret; } static int test_blkcipher(void) { struct crypto_blkcipher *blkcipher = NULL; char *cipher = "cbc(aes)"; // AES 128 charkey = "\x12\x34\x56\x78\x90\xab\xcd\xef\x12\x34\x56\x78\x90\xab\xcd\xef"; chariv = "\x12\x34\x56\x78\x90\xab\xcd\xef\x12\x34\x56\x78\x90\xab\xcd\xef"; unsigned int ivsize = 0; char *scratchpad = NULL; // holds plaintext and ciphertext struct scatterlist sg; struct blkcipher_desc desc; int ret = -EFAULT; blkcipher = crypto_alloc_blkcipher(cipher, 0, 0); if (IS_ERR(blkcipher)) { printk("could not allocate blkcipher handle for %s\n", cipher); return -PTR_ERR(blkcipher); } if (crypto_blkcipher_setkey(blkcipher, key, strlen(key))) { printk("key could not be set\n"); ret = -EAGAIN; goto out; } ivsize = crypto_blkcipher_ivsize(blkcipher); if (ivsize) { if (ivsize != strlen(iv)) printk("IV length differs from expected length\n"); crypto_blkcipher_set_iv(blkcipher, iv, ivsize); } scratchpad = kmalloc(crypto_blkcipher_blocksize(blkcipher), GFP_KERNEL); if (!scratchpad) { printk("could not allocate scratchpad for %s\n", cipher); goto out; } /* get some random data that we want to encrypt */ get_random_bytes(scratchpad, crypto_blkcipher_blocksize(blkcipher)); desc.flags = 0; desc.tfm = blkcipher; sg_init_one(&sg, scratchpad, crypto_blkcipher_blocksize(blkcipher)); /* encrypt data in place */ crypto_blkcipher_encrypt(&desc, &sg, &sg, crypto_blkcipher_blocksize(blkcipher)); /* decrypt data in place * crypto_blkcipher_decrypt(&desc, &sg, &sg, */ crypto_blkcipher_blocksize(blkcipher)); printk("Cipher operation completed\n"); return 0; out: if (blkcipher) crypto_free_blkcipher(blkcipher); if (scratchpad) kzfree(scratchpad); return ret; } struct sdesc { struct shash_desc shash; char ctx[]; }; static struct sdescinit_sdesc(struct crypto_shash *alg) { struct sdescsdesc; int size; size = sizeof(struct shash_desc) + crypto_shash_descsize(alg); sdesc = kmalloc(size, GFP_KERNEL); if (!sdesc) return ERR_PTR(-ENOMEM); sdesc->shash.tfm = alg; sdesc->shash.flags = 0x0; return sdesc; } static int calc_hash(struct crypto_shashalg, const unsigned chardata, unsigned int datalen, unsigned chardigest) { struct sdescsdesc; int ret; sdesc = init_sdesc(alg); if (IS_ERR(sdesc)) { pr_info("trusted_key: can't alloc %s\n", hash_alg); return PTR_ERR(sdesc); } ret = crypto_shash_digest(&sdesc->shash, data, datalen, digest); kfree(sdesc); return ret; } static int get_random_numbers(u8 *buf, unsigned int len) { struct crypto_rngrng = NULL; chardrbg = "drbg_nopr_sha256"; /* Hash DRBG with SHA-256, no PR */ int ret; if (!buf || !len) { pr_debug("No output buffer provided\n"); return -EINVAL; } rng = crypto_alloc_rng(drbg, 0, 0); if (IS_ERR(rng)) { pr_debug("could not allocate RNG handle for %s\n", drbg); return -PTR_ERR(rng); } ret = crypto_rng_get_bytes(rng, buf, len); if (ret < 0) pr_debug("generation of random numbers failed\n"); else if (ret == 0) pr_debug("RNG returned no data"); else pr_debug("RNG returned %d bytes of data\n", ret); out: crypto_free_rng(rng); return ret; }