freebsd-src/sys/opencrypto/cryptodev.c

Ignoring revisions in .git-blame-ignore-revs. Click here to bypass and see the normal blame view.

1304 lines
31 KiB
C
Raw Normal View History

/* $OpenBSD: cryptodev.c,v 1.52 2002/06/19 07:22:46 deraadt Exp $ */
/*-
* Copyright (c) 2001 Theo de Raadt
* Copyright (c) 2002-2006 Sam Leffler, Errno Consulting
* Copyright (c) 2014-2021 The FreeBSD Foundation
2014-12-12 19:56:36 +00:00
* All rights reserved.
*
* Portions of this software were developed by John-Mark Gurney
* under sponsorship of the FreeBSD Foundation and
* Rubicon Communications, LLC (Netgate).
*
* Portions of this software were developed by Ararat River
* Consulting, LLC under sponsorship of the FreeBSD Foundation.
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions
* are met:
*
* 1. Redistributions of source code must retain the above copyright
* notice, this list of conditions and the following disclaimer.
* 2. Redistributions in binary form must reproduce the above copyright
* notice, this list of conditions and the following disclaimer in the
* documentation and/or other materials provided with the distribution.
* 3. The name of the author may not be used to endorse or promote products
* derived from this software without specific prior written permission.
*
* THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR
* IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
* OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
* IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT,
* INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
* NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
* DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
* THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
* (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF
* THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
*
* Effort sponsored in part by the Defense Advanced Research Projects
* Agency (DARPA) and Air Force Research Laboratory, Air Force
* Materiel Command, USAF, under agreement number F30602-01-2-0537.
*/
#include <sys/param.h>
#include <sys/systm.h>
#include <sys/malloc.h>
#include <sys/mbuf.h>
#include <sys/lock.h>
#include <sys/mutex.h>
#include <sys/proc.h>
#include <sys/sysctl.h>
#include <sys/errno.h>
#include <sys/random.h>
#include <sys/conf.h>
#include <sys/kernel.h>
2004-05-30 20:27:19 +00:00
#include <sys/module.h>
#include <sys/fcntl.h>
#include <sys/bus.h>
2014-12-12 19:56:36 +00:00
#include <sys/sdt.h>
#include <sys/syscallsubr.h>
#include <opencrypto/cryptodev.h>
#include <opencrypto/xform.h>
2014-12-12 19:56:36 +00:00
SDT_PROVIDER_DECLARE(opencrypto);
SDT_PROBE_DEFINE1(opencrypto, dev, ioctl, error, "int"/*line number*/);
#ifdef COMPAT_FREEBSD12
/*
* Previously, most ioctls were performed against a cloned descriptor
* of /dev/crypto obtained via CRIOGET. Now all ioctls are performed
* against /dev/crypto directly.
*/
#define CRIOGET _IOWR('c', 100, uint32_t)
#endif
/* the following are done against the cloned descriptor */
#ifdef COMPAT_FREEBSD32
#include <sys/mount.h>
#include <compat/freebsd32/freebsd32.h>
struct session_op32 {
uint32_t cipher;
uint32_t mac;
uint32_t keylen;
uint32_t key;
int mackeylen;
uint32_t mackey;
uint32_t ses;
};
struct session2_op32 {
uint32_t cipher;
uint32_t mac;
uint32_t keylen;
uint32_t key;
int mackeylen;
uint32_t mackey;
uint32_t ses;
int crid;
int ivlen;
int maclen;
int pad[2];
};
struct crypt_op32 {
uint32_t ses;
uint16_t op;
uint16_t flags;
u_int len;
uint32_t src, dst;
uint32_t mac;
uint32_t iv;
};
struct crypt_aead32 {
uint32_t ses;
uint16_t op;
uint16_t flags;
u_int len;
u_int aadlen;
u_int ivlen;
uint32_t src;
uint32_t dst;
uint32_t aad;
uint32_t tag;
uint32_t iv;
};
#define CIOCGSESSION32 _IOWR('c', 101, struct session_op32)
#define CIOCCRYPT32 _IOWR('c', 103, struct crypt_op32)
#define CIOCGSESSION232 _IOWR('c', 106, struct session2_op32)
#define CIOCCRYPTAEAD32 _IOWR('c', 109, struct crypt_aead32)
static void
session_op_from_32(const struct session_op32 *from, struct session2_op *to)
{
memset(to, 0, sizeof(*to));
CP(*from, *to, cipher);
CP(*from, *to, mac);
CP(*from, *to, keylen);
PTRIN_CP(*from, *to, key);
CP(*from, *to, mackeylen);
PTRIN_CP(*from, *to, mackey);
CP(*from, *to, ses);
to->crid = CRYPTOCAP_F_HARDWARE;
}
static void
session2_op_from_32(const struct session2_op32 *from, struct session2_op *to)
{
session_op_from_32((const struct session_op32 *)from, to);
CP(*from, *to, crid);
CP(*from, *to, ivlen);
CP(*from, *to, maclen);
}
static void
session_op_to_32(const struct session2_op *from, struct session_op32 *to)
{
CP(*from, *to, cipher);
CP(*from, *to, mac);
CP(*from, *to, keylen);
PTROUT_CP(*from, *to, key);
CP(*from, *to, mackeylen);
PTROUT_CP(*from, *to, mackey);
CP(*from, *to, ses);
}
static void
session2_op_to_32(const struct session2_op *from, struct session2_op32 *to)
{
session_op_to_32(from, (struct session_op32 *)to);
CP(*from, *to, crid);
}
static void
crypt_op_from_32(const struct crypt_op32 *from, struct crypt_op *to)
{
CP(*from, *to, ses);
CP(*from, *to, op);
CP(*from, *to, flags);
CP(*from, *to, len);
PTRIN_CP(*from, *to, src);
PTRIN_CP(*from, *to, dst);
PTRIN_CP(*from, *to, mac);
PTRIN_CP(*from, *to, iv);
}
static void
crypt_op_to_32(const struct crypt_op *from, struct crypt_op32 *to)
{
CP(*from, *to, ses);
CP(*from, *to, op);
CP(*from, *to, flags);
CP(*from, *to, len);
PTROUT_CP(*from, *to, src);
PTROUT_CP(*from, *to, dst);
PTROUT_CP(*from, *to, mac);
PTROUT_CP(*from, *to, iv);
}
static void
crypt_aead_from_32(const struct crypt_aead32 *from, struct crypt_aead *to)
{
CP(*from, *to, ses);
CP(*from, *to, op);
CP(*from, *to, flags);
CP(*from, *to, len);
CP(*from, *to, aadlen);
CP(*from, *to, ivlen);
PTRIN_CP(*from, *to, src);
PTRIN_CP(*from, *to, dst);
PTRIN_CP(*from, *to, aad);
PTRIN_CP(*from, *to, tag);
PTRIN_CP(*from, *to, iv);
}
static void
crypt_aead_to_32(const struct crypt_aead *from, struct crypt_aead32 *to)
{
CP(*from, *to, ses);
CP(*from, *to, op);
CP(*from, *to, flags);
CP(*from, *to, len);
CP(*from, *to, aadlen);
CP(*from, *to, ivlen);
PTROUT_CP(*from, *to, src);
PTROUT_CP(*from, *to, dst);
PTROUT_CP(*from, *to, aad);
PTROUT_CP(*from, *to, tag);
PTROUT_CP(*from, *to, iv);
}
#endif
static void
session2_op_from_op(const struct session_op *from, struct session2_op *to)
{
memset(to, 0, sizeof(*to));
memcpy(to, from, sizeof(*from));
to->crid = CRYPTOCAP_F_HARDWARE;
}
static void
session2_op_to_op(const struct session2_op *from, struct session_op *to)
{
memcpy(to, from, sizeof(*to));
}
struct csession {
TAILQ_ENTRY(csession) next;
crypto_session_t cses;
volatile u_int refs;
uint32_t ses;
struct mtx lock; /* for op submission */
u_int blocksize;
Refactor driver and consumer interfaces for OCF (in-kernel crypto). - The linked list of cryptoini structures used in session initialization is replaced with a new flat structure: struct crypto_session_params. This session includes a new mode to define how the other fields should be interpreted. Available modes include: - COMPRESS (for compression/decompression) - CIPHER (for simply encryption/decryption) - DIGEST (computing and verifying digests) - AEAD (combined auth and encryption such as AES-GCM and AES-CCM) - ETA (combined auth and encryption using encrypt-then-authenticate) Additional modes could be added in the future (e.g. if we wanted to support TLS MtE for AES-CBC in the kernel we could add a new mode for that. TLS modes might also affect how AAD is interpreted, etc.) The flat structure also includes the key lengths and algorithms as before. However, code doesn't have to walk the linked list and switch on the algorithm to determine which key is the auth key vs encryption key. The 'csp_auth_*' fields are always used for auth keys and settings and 'csp_cipher_*' for cipher. (Compression algorithms are stored in csp_cipher_alg.) - Drivers no longer register a list of supported algorithms. This doesn't quite work when you factor in modes (e.g. a driver might support both AES-CBC and SHA2-256-HMAC separately but not combined for ETA). Instead, a new 'crypto_probesession' method has been added to the kobj interface for symmteric crypto drivers. This method returns a negative value on success (similar to how device_probe works) and the crypto framework uses this value to pick the "best" driver. There are three constants for hardware (e.g. ccr), accelerated software (e.g. aesni), and plain software (cryptosoft) that give preference in that order. One effect of this is that if you request only hardware when creating a new session, you will no longer get a session using accelerated software. Another effect is that the default setting to disallow software crypto via /dev/crypto now disables accelerated software. Once a driver is chosen, 'crypto_newsession' is invoked as before. - Crypto operations are now solely described by the flat 'cryptop' structure. The linked list of descriptors has been removed. A separate enum has been added to describe the type of data buffer in use instead of using CRYPTO_F_* flags to make it easier to add more types in the future if needed (e.g. wired userspace buffers for zero-copy). It will also make it easier to re-introduce separate input and output buffers (in-kernel TLS would benefit from this). Try to make the flags related to IV handling less insane: - CRYPTO_F_IV_SEPARATE means that the IV is stored in the 'crp_iv' member of the operation structure. If this flag is not set, the IV is stored in the data buffer at the 'crp_iv_start' offset. - CRYPTO_F_IV_GENERATE means that a random IV should be generated and stored into the data buffer. This cannot be used with CRYPTO_F_IV_SEPARATE. If a consumer wants to deal with explicit vs implicit IVs, etc. it can always generate the IV however it needs and store partial IVs in the buffer and the full IV/nonce in crp_iv and set CRYPTO_F_IV_SEPARATE. The layout of the buffer is now described via fields in cryptop. crp_aad_start and crp_aad_length define the boundaries of any AAD. Previously with GCM and CCM you defined an auth crd with this range, but for ETA your auth crd had to span both the AAD and plaintext (and they had to be adjacent). crp_payload_start and crp_payload_length define the boundaries of the plaintext/ciphertext. Modes that only do a single operation (COMPRESS, CIPHER, DIGEST) should only use this region and leave the AAD region empty. If a digest is present (or should be generated), it's starting location is marked by crp_digest_start. Instead of using the CRD_F_ENCRYPT flag to determine the direction of the operation, cryptop now includes an 'op' field defining the operation to perform. For digests I've added a new VERIFY digest mode which assumes a digest is present in the input and fails the request with EBADMSG if it doesn't match the internally-computed digest. GCM and CCM already assumed this, and the new AEAD mode requires this for decryption. The new ETA mode now also requires this for decryption, so IPsec and GELI no longer do their own authentication verification. Simple DIGEST operations can also do this, though there are no in-tree consumers. To eventually support some refcounting to close races, the session cookie is now passed to crypto_getop() and clients should no longer set crp_sesssion directly. - Assymteric crypto operation structures should be allocated via crypto_getkreq() and freed via crypto_freekreq(). This permits the crypto layer to track open asym requests and close races with a driver trying to unregister while asym requests are in flight. - crypto_copyback, crypto_copydata, crypto_apply, and crypto_contiguous_subsegment now accept the 'crp' object as the first parameter instead of individual members. This makes it easier to deal with different buffer types in the future as well as separate input and output buffers. It's also simpler for driver writers to use. - bus_dmamap_load_crp() loads a DMA mapping for a crypto buffer. This understands the various types of buffers so that drivers that use DMA do not have to be aware of different buffer types. - Helper routines now exist to build an auth context for HMAC IPAD and OPAD. This reduces some duplicated work among drivers. - Key buffers are now treated as const throughout the framework and in device drivers. However, session key buffers provided when a session is created are expected to remain alive for the duration of the session. - GCM and CCM sessions now only specify a cipher algorithm and a cipher key. The redundant auth information is not needed or used. - For cryptosoft, split up the code a bit such that the 'process' callback now invokes a function pointer in the session. This function pointer is set based on the mode (in effect) though it simplifies a few edge cases that would otherwise be in the switch in 'process'. It does split up GCM vs CCM which I think is more readable even if there is some duplication. - I changed /dev/crypto to support GMAC requests using CRYPTO_AES_NIST_GMAC as an auth algorithm and updated cryptocheck to work with it. - Combined cipher and auth sessions via /dev/crypto now always use ETA mode. The COP_F_CIPHER_FIRST flag is now a no-op that is ignored. This was actually documented as being true in crypto(4) before, but the code had not implemented this before I added the CIPHER_FIRST flag. - I have not yet updated /dev/crypto to be aware of explicit modes for sessions. I will probably do that at some point in the future as well as teach it about IV/nonce and tag lengths for AEAD so we can support all of the NIST KAT tests for GCM and CCM. - I've split up the exising crypto.9 manpage into several pages of which many are written from scratch. - I have converted all drivers and consumers in the tree and verified that they compile, but I have not tested all of them. I have tested the following drivers: - cryptosoft - aesni (AES only) - blake2 - ccr and the following consumers: - cryptodev - IPsec - ktls_ocf - GELI (lightly) I have not tested the following: - ccp - aesni with sha - hifn - kgssapi_krb5 - ubsec - padlock - safe - armv8_crypto (aarch64) - glxsb (i386) - sec (ppc) - cesa (armv7) - cryptocteon (mips64) - nlmsec (mips64) Discussed with: cem Relnotes: yes Sponsored by: Chelsio Communications Differential Revision: https://reviews.freebsd.org/D23677
2020-03-27 18:25:23 +00:00
int hashsize;
int ivsize;
Refactor driver and consumer interfaces for OCF (in-kernel crypto). - The linked list of cryptoini structures used in session initialization is replaced with a new flat structure: struct crypto_session_params. This session includes a new mode to define how the other fields should be interpreted. Available modes include: - COMPRESS (for compression/decompression) - CIPHER (for simply encryption/decryption) - DIGEST (computing and verifying digests) - AEAD (combined auth and encryption such as AES-GCM and AES-CCM) - ETA (combined auth and encryption using encrypt-then-authenticate) Additional modes could be added in the future (e.g. if we wanted to support TLS MtE for AES-CBC in the kernel we could add a new mode for that. TLS modes might also affect how AAD is interpreted, etc.) The flat structure also includes the key lengths and algorithms as before. However, code doesn't have to walk the linked list and switch on the algorithm to determine which key is the auth key vs encryption key. The 'csp_auth_*' fields are always used for auth keys and settings and 'csp_cipher_*' for cipher. (Compression algorithms are stored in csp_cipher_alg.) - Drivers no longer register a list of supported algorithms. This doesn't quite work when you factor in modes (e.g. a driver might support both AES-CBC and SHA2-256-HMAC separately but not combined for ETA). Instead, a new 'crypto_probesession' method has been added to the kobj interface for symmteric crypto drivers. This method returns a negative value on success (similar to how device_probe works) and the crypto framework uses this value to pick the "best" driver. There are three constants for hardware (e.g. ccr), accelerated software (e.g. aesni), and plain software (cryptosoft) that give preference in that order. One effect of this is that if you request only hardware when creating a new session, you will no longer get a session using accelerated software. Another effect is that the default setting to disallow software crypto via /dev/crypto now disables accelerated software. Once a driver is chosen, 'crypto_newsession' is invoked as before. - Crypto operations are now solely described by the flat 'cryptop' structure. The linked list of descriptors has been removed. A separate enum has been added to describe the type of data buffer in use instead of using CRYPTO_F_* flags to make it easier to add more types in the future if needed (e.g. wired userspace buffers for zero-copy). It will also make it easier to re-introduce separate input and output buffers (in-kernel TLS would benefit from this). Try to make the flags related to IV handling less insane: - CRYPTO_F_IV_SEPARATE means that the IV is stored in the 'crp_iv' member of the operation structure. If this flag is not set, the IV is stored in the data buffer at the 'crp_iv_start' offset. - CRYPTO_F_IV_GENERATE means that a random IV should be generated and stored into the data buffer. This cannot be used with CRYPTO_F_IV_SEPARATE. If a consumer wants to deal with explicit vs implicit IVs, etc. it can always generate the IV however it needs and store partial IVs in the buffer and the full IV/nonce in crp_iv and set CRYPTO_F_IV_SEPARATE. The layout of the buffer is now described via fields in cryptop. crp_aad_start and crp_aad_length define the boundaries of any AAD. Previously with GCM and CCM you defined an auth crd with this range, but for ETA your auth crd had to span both the AAD and plaintext (and they had to be adjacent). crp_payload_start and crp_payload_length define the boundaries of the plaintext/ciphertext. Modes that only do a single operation (COMPRESS, CIPHER, DIGEST) should only use this region and leave the AAD region empty. If a digest is present (or should be generated), it's starting location is marked by crp_digest_start. Instead of using the CRD_F_ENCRYPT flag to determine the direction of the operation, cryptop now includes an 'op' field defining the operation to perform. For digests I've added a new VERIFY digest mode which assumes a digest is present in the input and fails the request with EBADMSG if it doesn't match the internally-computed digest. GCM and CCM already assumed this, and the new AEAD mode requires this for decryption. The new ETA mode now also requires this for decryption, so IPsec and GELI no longer do their own authentication verification. Simple DIGEST operations can also do this, though there are no in-tree consumers. To eventually support some refcounting to close races, the session cookie is now passed to crypto_getop() and clients should no longer set crp_sesssion directly. - Assymteric crypto operation structures should be allocated via crypto_getkreq() and freed via crypto_freekreq(). This permits the crypto layer to track open asym requests and close races with a driver trying to unregister while asym requests are in flight. - crypto_copyback, crypto_copydata, crypto_apply, and crypto_contiguous_subsegment now accept the 'crp' object as the first parameter instead of individual members. This makes it easier to deal with different buffer types in the future as well as separate input and output buffers. It's also simpler for driver writers to use. - bus_dmamap_load_crp() loads a DMA mapping for a crypto buffer. This understands the various types of buffers so that drivers that use DMA do not have to be aware of different buffer types. - Helper routines now exist to build an auth context for HMAC IPAD and OPAD. This reduces some duplicated work among drivers. - Key buffers are now treated as const throughout the framework and in device drivers. However, session key buffers provided when a session is created are expected to remain alive for the duration of the session. - GCM and CCM sessions now only specify a cipher algorithm and a cipher key. The redundant auth information is not needed or used. - For cryptosoft, split up the code a bit such that the 'process' callback now invokes a function pointer in the session. This function pointer is set based on the mode (in effect) though it simplifies a few edge cases that would otherwise be in the switch in 'process'. It does split up GCM vs CCM which I think is more readable even if there is some duplication. - I changed /dev/crypto to support GMAC requests using CRYPTO_AES_NIST_GMAC as an auth algorithm and updated cryptocheck to work with it. - Combined cipher and auth sessions via /dev/crypto now always use ETA mode. The COP_F_CIPHER_FIRST flag is now a no-op that is ignored. This was actually documented as being true in crypto(4) before, but the code had not implemented this before I added the CIPHER_FIRST flag. - I have not yet updated /dev/crypto to be aware of explicit modes for sessions. I will probably do that at some point in the future as well as teach it about IV/nonce and tag lengths for AEAD so we can support all of the NIST KAT tests for GCM and CCM. - I've split up the exising crypto.9 manpage into several pages of which many are written from scratch. - I have converted all drivers and consumers in the tree and verified that they compile, but I have not tested all of them. I have tested the following drivers: - cryptosoft - aesni (AES only) - blake2 - ccr and the following consumers: - cryptodev - IPsec - ktls_ocf - GELI (lightly) I have not tested the following: - ccp - aesni with sha - hifn - kgssapi_krb5 - ubsec - padlock - safe - armv8_crypto (aarch64) - glxsb (i386) - sec (ppc) - cesa (armv7) - cryptocteon (mips64) - nlmsec (mips64) Discussed with: cem Relnotes: yes Sponsored by: Chelsio Communications Differential Revision: https://reviews.freebsd.org/D23677
2020-03-27 18:25:23 +00:00
void *key;
void *mackey;
};
struct cryptop_data {
struct csession *cse;
Refactor driver and consumer interfaces for OCF (in-kernel crypto). - The linked list of cryptoini structures used in session initialization is replaced with a new flat structure: struct crypto_session_params. This session includes a new mode to define how the other fields should be interpreted. Available modes include: - COMPRESS (for compression/decompression) - CIPHER (for simply encryption/decryption) - DIGEST (computing and verifying digests) - AEAD (combined auth and encryption such as AES-GCM and AES-CCM) - ETA (combined auth and encryption using encrypt-then-authenticate) Additional modes could be added in the future (e.g. if we wanted to support TLS MtE for AES-CBC in the kernel we could add a new mode for that. TLS modes might also affect how AAD is interpreted, etc.) The flat structure also includes the key lengths and algorithms as before. However, code doesn't have to walk the linked list and switch on the algorithm to determine which key is the auth key vs encryption key. The 'csp_auth_*' fields are always used for auth keys and settings and 'csp_cipher_*' for cipher. (Compression algorithms are stored in csp_cipher_alg.) - Drivers no longer register a list of supported algorithms. This doesn't quite work when you factor in modes (e.g. a driver might support both AES-CBC and SHA2-256-HMAC separately but not combined for ETA). Instead, a new 'crypto_probesession' method has been added to the kobj interface for symmteric crypto drivers. This method returns a negative value on success (similar to how device_probe works) and the crypto framework uses this value to pick the "best" driver. There are three constants for hardware (e.g. ccr), accelerated software (e.g. aesni), and plain software (cryptosoft) that give preference in that order. One effect of this is that if you request only hardware when creating a new session, you will no longer get a session using accelerated software. Another effect is that the default setting to disallow software crypto via /dev/crypto now disables accelerated software. Once a driver is chosen, 'crypto_newsession' is invoked as before. - Crypto operations are now solely described by the flat 'cryptop' structure. The linked list of descriptors has been removed. A separate enum has been added to describe the type of data buffer in use instead of using CRYPTO_F_* flags to make it easier to add more types in the future if needed (e.g. wired userspace buffers for zero-copy). It will also make it easier to re-introduce separate input and output buffers (in-kernel TLS would benefit from this). Try to make the flags related to IV handling less insane: - CRYPTO_F_IV_SEPARATE means that the IV is stored in the 'crp_iv' member of the operation structure. If this flag is not set, the IV is stored in the data buffer at the 'crp_iv_start' offset. - CRYPTO_F_IV_GENERATE means that a random IV should be generated and stored into the data buffer. This cannot be used with CRYPTO_F_IV_SEPARATE. If a consumer wants to deal with explicit vs implicit IVs, etc. it can always generate the IV however it needs and store partial IVs in the buffer and the full IV/nonce in crp_iv and set CRYPTO_F_IV_SEPARATE. The layout of the buffer is now described via fields in cryptop. crp_aad_start and crp_aad_length define the boundaries of any AAD. Previously with GCM and CCM you defined an auth crd with this range, but for ETA your auth crd had to span both the AAD and plaintext (and they had to be adjacent). crp_payload_start and crp_payload_length define the boundaries of the plaintext/ciphertext. Modes that only do a single operation (COMPRESS, CIPHER, DIGEST) should only use this region and leave the AAD region empty. If a digest is present (or should be generated), it's starting location is marked by crp_digest_start. Instead of using the CRD_F_ENCRYPT flag to determine the direction of the operation, cryptop now includes an 'op' field defining the operation to perform. For digests I've added a new VERIFY digest mode which assumes a digest is present in the input and fails the request with EBADMSG if it doesn't match the internally-computed digest. GCM and CCM already assumed this, and the new AEAD mode requires this for decryption. The new ETA mode now also requires this for decryption, so IPsec and GELI no longer do their own authentication verification. Simple DIGEST operations can also do this, though there are no in-tree consumers. To eventually support some refcounting to close races, the session cookie is now passed to crypto_getop() and clients should no longer set crp_sesssion directly. - Assymteric crypto operation structures should be allocated via crypto_getkreq() and freed via crypto_freekreq(). This permits the crypto layer to track open asym requests and close races with a driver trying to unregister while asym requests are in flight. - crypto_copyback, crypto_copydata, crypto_apply, and crypto_contiguous_subsegment now accept the 'crp' object as the first parameter instead of individual members. This makes it easier to deal with different buffer types in the future as well as separate input and output buffers. It's also simpler for driver writers to use. - bus_dmamap_load_crp() loads a DMA mapping for a crypto buffer. This understands the various types of buffers so that drivers that use DMA do not have to be aware of different buffer types. - Helper routines now exist to build an auth context for HMAC IPAD and OPAD. This reduces some duplicated work among drivers. - Key buffers are now treated as const throughout the framework and in device drivers. However, session key buffers provided when a session is created are expected to remain alive for the duration of the session. - GCM and CCM sessions now only specify a cipher algorithm and a cipher key. The redundant auth information is not needed or used. - For cryptosoft, split up the code a bit such that the 'process' callback now invokes a function pointer in the session. This function pointer is set based on the mode (in effect) though it simplifies a few edge cases that would otherwise be in the switch in 'process'. It does split up GCM vs CCM which I think is more readable even if there is some duplication. - I changed /dev/crypto to support GMAC requests using CRYPTO_AES_NIST_GMAC as an auth algorithm and updated cryptocheck to work with it. - Combined cipher and auth sessions via /dev/crypto now always use ETA mode. The COP_F_CIPHER_FIRST flag is now a no-op that is ignored. This was actually documented as being true in crypto(4) before, but the code had not implemented this before I added the CIPHER_FIRST flag. - I have not yet updated /dev/crypto to be aware of explicit modes for sessions. I will probably do that at some point in the future as well as teach it about IV/nonce and tag lengths for AEAD so we can support all of the NIST KAT tests for GCM and CCM. - I've split up the exising crypto.9 manpage into several pages of which many are written from scratch. - I have converted all drivers and consumers in the tree and verified that they compile, but I have not tested all of them. I have tested the following drivers: - cryptosoft - aesni (AES only) - blake2 - ccr and the following consumers: - cryptodev - IPsec - ktls_ocf - GELI (lightly) I have not tested the following: - ccp - aesni with sha - hifn - kgssapi_krb5 - ubsec - padlock - safe - armv8_crypto (aarch64) - glxsb (i386) - sec (ppc) - cesa (armv7) - cryptocteon (mips64) - nlmsec (mips64) Discussed with: cem Relnotes: yes Sponsored by: Chelsio Communications Differential Revision: https://reviews.freebsd.org/D23677
2020-03-27 18:25:23 +00:00
char *buf;
char *obuf;
char *aad;
bool done;
};
struct fcrypt {
TAILQ_HEAD(csessionlist, csession) csessions;
int sesn;
struct mtx lock;
};
static bool use_outputbuffers;
SYSCTL_BOOL(_kern_crypto, OID_AUTO, cryptodev_use_output, CTLFLAG_RW,
&use_outputbuffers, 0,
"Use separate output buffers for /dev/crypto requests.");
static bool use_separate_aad;
SYSCTL_BOOL(_kern_crypto, OID_AUTO, cryptodev_separate_aad, CTLFLAG_RW,
&use_separate_aad, 0,
"Use separate AAD buffer for /dev/crypto requests.");
static MALLOC_DEFINE(M_CRYPTODEV, "cryptodev", "/dev/crypto data buffers");
/*
* Check a crypto identifier to see if it requested
* a software device/driver. This can be done either
* by device name/class or through search constraints.
*/
static int
2014-12-12 19:56:36 +00:00
checkforsoftware(int *cridp)
{
2014-12-12 19:56:36 +00:00
int crid;
crid = *cridp;
if (!crypto_devallowsoft) {
2014-12-12 19:56:36 +00:00
if (crid & CRYPTOCAP_F_SOFTWARE) {
if (crid & CRYPTOCAP_F_HARDWARE) {
*cridp = CRYPTOCAP_F_HARDWARE;
return 0;
}
return EINVAL;
}
if ((crid & CRYPTOCAP_F_HARDWARE) == 0 &&
(crypto_getcaps(crid) & CRYPTOCAP_F_HARDWARE) == 0)
2014-12-12 19:56:36 +00:00
return EINVAL;
}
return 0;
}
static int
cse_create(struct fcrypt *fcr, struct session2_op *sop)
{
struct crypto_session_params csp;
struct csession *cse;
const struct enc_xform *txform;
const struct auth_hash *thash;
void *key = NULL;
void *mackey = NULL;
crypto_session_t cses;
int crid, error, mac;
mac = sop->mac;
#ifdef COMPAT_FREEBSD12
switch (sop->mac) {
case CRYPTO_AES_128_NIST_GMAC:
case CRYPTO_AES_192_NIST_GMAC:
case CRYPTO_AES_256_NIST_GMAC:
/* Should always be paired with GCM. */
if (sop->cipher != CRYPTO_AES_NIST_GCM_16) {
CRYPTDEB("GMAC without GCM");
SDT_PROBE1(opencrypto, dev, ioctl, error, __LINE__);
return (EINVAL);
}
if (sop->keylen != sop->mackeylen) {
SDT_PROBE1(opencrypto, dev, ioctl, error, __LINE__);
return (EINVAL);
}
mac = 0;
break;
case CRYPTO_AES_CCM_CBC_MAC:
/* Should always be paired with CCM. */
if (sop->cipher != CRYPTO_AES_CCM_16) {
CRYPTDEB("CBC-MAC without CCM");
SDT_PROBE1(opencrypto, dev, ioctl, error, __LINE__);
return (EINVAL);
}
if (sop->keylen != sop->mackeylen) {
SDT_PROBE1(opencrypto, dev, ioctl, error, __LINE__);
return (EINVAL);
}
mac = 0;
break;
}
#endif
memset(&csp, 0, sizeof(csp));
if (use_outputbuffers)
csp.csp_flags |= CSP_F_SEPARATE_OUTPUT;
if (mac != 0) {
csp.csp_auth_alg = mac;
csp.csp_auth_klen = sop->mackeylen;
}
if (sop->cipher != 0) {
csp.csp_cipher_alg = sop->cipher;
csp.csp_cipher_klen = sop->keylen;
}
thash = crypto_auth_hash(&csp);
txform = crypto_cipher(&csp);
if (txform != NULL && txform->macsize != 0) {
if (mac != 0) {
SDT_PROBE1(opencrypto, dev, ioctl, error, __LINE__);
return (EINVAL);
}
csp.csp_mode = CSP_MODE_AEAD;
} else if (txform != NULL && thash != NULL) {
csp.csp_mode = CSP_MODE_ETA;
} else if (txform != NULL) {
csp.csp_mode = CSP_MODE_CIPHER;
} else if (thash != NULL) {
csp.csp_mode = CSP_MODE_DIGEST;
} else {
SDT_PROBE1(opencrypto, dev, ioctl, error, __LINE__);
return (EINVAL);
}
switch (csp.csp_mode) {
case CSP_MODE_AEAD:
case CSP_MODE_ETA:
if (use_separate_aad)
csp.csp_flags |= CSP_F_SEPARATE_AAD;
break;
}
if (txform != NULL) {
if (sop->keylen > txform->maxkey ||
sop->keylen < txform->minkey) {
CRYPTDEB("invalid cipher parameters");
error = EINVAL;
SDT_PROBE1(opencrypto, dev, ioctl, error, __LINE__);
goto bail;
}
key = malloc(csp.csp_cipher_klen, M_CRYPTODEV, M_WAITOK);
error = copyin(sop->key, key, csp.csp_cipher_klen);
if (error) {
CRYPTDEB("invalid key");
SDT_PROBE1(opencrypto, dev, ioctl, error, __LINE__);
goto bail;
}
csp.csp_cipher_key = key;
csp.csp_ivlen = txform->ivsize;
}
if (thash != NULL) {
if (sop->mackeylen > thash->keysize || sop->mackeylen < 0) {
CRYPTDEB("invalid mac key length");
error = EINVAL;
SDT_PROBE1(opencrypto, dev, ioctl, error, __LINE__);
goto bail;
}
if (csp.csp_auth_klen != 0) {
mackey = malloc(csp.csp_auth_klen, M_CRYPTODEV,
M_WAITOK);
error = copyin(sop->mackey, mackey, csp.csp_auth_klen);
if (error) {
CRYPTDEB("invalid mac key");
SDT_PROBE1(opencrypto, dev, ioctl, error,
__LINE__);
goto bail;
}
csp.csp_auth_key = mackey;
}
if (csp.csp_auth_alg == CRYPTO_AES_NIST_GMAC)
csp.csp_ivlen = AES_GCM_IV_LEN;
if (csp.csp_auth_alg == CRYPTO_AES_CCM_CBC_MAC)
csp.csp_ivlen = AES_CCM_IV_LEN;
}
if (sop->ivlen != 0) {
if (csp.csp_ivlen == 0) {
CRYPTDEB("does not support an IV");
error = EINVAL;
SDT_PROBE1(opencrypto, dev, ioctl, error, __LINE__);
goto bail;
}
csp.csp_ivlen = sop->ivlen;
}
if (sop->maclen != 0) {
if (!(thash != NULL || csp.csp_mode == CSP_MODE_AEAD)) {
CRYPTDEB("does not support a MAC");
error = EINVAL;
SDT_PROBE1(opencrypto, dev, ioctl, error, __LINE__);
goto bail;
}
csp.csp_auth_mlen = sop->maclen;
}
crid = sop->crid;
error = checkforsoftware(&crid);
if (error) {
CRYPTDEB("checkforsoftware");
SDT_PROBE1(opencrypto, dev, ioctl, error, __LINE__);
goto bail;
}
error = crypto_newsession(&cses, &csp, crid);
if (error) {
CRYPTDEB("crypto_newsession");
SDT_PROBE1(opencrypto, dev, ioctl, error, __LINE__);
goto bail;
}
cse = malloc(sizeof(struct csession), M_CRYPTODEV, M_WAITOK | M_ZERO);
mtx_init(&cse->lock, "cryptodev", "crypto session lock", MTX_DEF);
refcount_init(&cse->refs, 1);
cse->key = key;
cse->mackey = mackey;
cse->cses = cses;
if (sop->maclen != 0)
cse->hashsize = sop->maclen;
else if (thash != NULL)
cse->hashsize = thash->hashsize;
else if (csp.csp_mode == CSP_MODE_AEAD)
cse->hashsize = txform->macsize;
cse->ivsize = csp.csp_ivlen;
/*
* NB: This isn't necessarily the block size of the underlying
* MAC or cipher but is instead a restriction on valid input
* sizes.
*/
if (txform != NULL)
cse->blocksize = txform->blocksize;
else
cse->blocksize = 1;
mtx_lock(&fcr->lock);
TAILQ_INSERT_TAIL(&fcr->csessions, cse, next);
cse->ses = fcr->sesn++;
mtx_unlock(&fcr->lock);
sop->ses = cse->ses;
/* return hardware/driver id */
sop->crid = crypto_ses2hid(cse->cses);
bail:
if (error) {
free(key, M_CRYPTODEV);
free(mackey, M_CRYPTODEV);
}
return (error);
}
static struct csession *
cse_find(struct fcrypt *fcr, u_int ses)
{
struct csession *cse;
mtx_lock(&fcr->lock);
TAILQ_FOREACH(cse, &fcr->csessions, next) {
if (cse->ses == ses) {
refcount_acquire(&cse->refs);
mtx_unlock(&fcr->lock);
return (cse);
}
}
mtx_unlock(&fcr->lock);
return (NULL);
}
static void
cse_free(struct csession *cse)
{
if (!refcount_release(&cse->refs))
return;
crypto_freesession(cse->cses);
mtx_destroy(&cse->lock);
if (cse->key)
free(cse->key, M_CRYPTODEV);
if (cse->mackey)
free(cse->mackey, M_CRYPTODEV);
free(cse, M_CRYPTODEV);
}
static bool
cse_delete(struct fcrypt *fcr, u_int ses)
{
struct csession *cse;
mtx_lock(&fcr->lock);
TAILQ_FOREACH(cse, &fcr->csessions, next) {
if (cse->ses == ses) {
TAILQ_REMOVE(&fcr->csessions, cse, next);
mtx_unlock(&fcr->lock);
cse_free(cse);
return (true);
}
}
mtx_unlock(&fcr->lock);
return (false);
}
static struct cryptop_data *
cod_alloc(struct csession *cse, size_t aad_len, size_t len)
{
struct cryptop_data *cod;
cod = malloc(sizeof(struct cryptop_data), M_CRYPTODEV, M_WAITOK |
M_ZERO);
cod->cse = cse;
if (crypto_get_params(cse->cses)->csp_flags & CSP_F_SEPARATE_AAD) {
if (aad_len != 0)
cod->aad = malloc(aad_len, M_CRYPTODEV, M_WAITOK);
cod->buf = malloc(len, M_CRYPTODEV, M_WAITOK);
} else
cod->buf = malloc(aad_len + len, M_CRYPTODEV, M_WAITOK);
if (crypto_get_params(cse->cses)->csp_flags & CSP_F_SEPARATE_OUTPUT)
cod->obuf = malloc(len, M_CRYPTODEV, M_WAITOK);
return (cod);
}
static void
cod_free(struct cryptop_data *cod)
{
free(cod->aad, M_CRYPTODEV);
free(cod->obuf, M_CRYPTODEV);
free(cod->buf, M_CRYPTODEV);
free(cod, M_CRYPTODEV);
}
static int
cryptodev_cb(struct cryptop *crp)
{
struct cryptop_data *cod = crp->crp_opaque;
/*
* Lock to ensure the wakeup() is not missed by the loops
* waiting on cod->done in cryptodev_op() and
* cryptodev_aead().
*/
mtx_lock(&cod->cse->lock);
cod->done = true;
mtx_unlock(&cod->cse->lock);
wakeup(cod);
return (0);
}
static int
cryptodev_op(struct csession *cse, const struct crypt_op *cop)
{
const struct crypto_session_params *csp;
struct cryptop_data *cod = NULL;
struct cryptop *crp = NULL;
char *dst;
int error;
2014-12-12 19:56:36 +00:00
if (cop->len > 256*1024-4) {
SDT_PROBE1(opencrypto, dev, ioctl, error, __LINE__);
return (E2BIG);
2014-12-12 19:56:36 +00:00
}
if ((cop->len % cse->blocksize) != 0) {
SDT_PROBE1(opencrypto, dev, ioctl, error, __LINE__);
return (EINVAL);
}
Refactor driver and consumer interfaces for OCF (in-kernel crypto). - The linked list of cryptoini structures used in session initialization is replaced with a new flat structure: struct crypto_session_params. This session includes a new mode to define how the other fields should be interpreted. Available modes include: - COMPRESS (for compression/decompression) - CIPHER (for simply encryption/decryption) - DIGEST (computing and verifying digests) - AEAD (combined auth and encryption such as AES-GCM and AES-CCM) - ETA (combined auth and encryption using encrypt-then-authenticate) Additional modes could be added in the future (e.g. if we wanted to support TLS MtE for AES-CBC in the kernel we could add a new mode for that. TLS modes might also affect how AAD is interpreted, etc.) The flat structure also includes the key lengths and algorithms as before. However, code doesn't have to walk the linked list and switch on the algorithm to determine which key is the auth key vs encryption key. The 'csp_auth_*' fields are always used for auth keys and settings and 'csp_cipher_*' for cipher. (Compression algorithms are stored in csp_cipher_alg.) - Drivers no longer register a list of supported algorithms. This doesn't quite work when you factor in modes (e.g. a driver might support both AES-CBC and SHA2-256-HMAC separately but not combined for ETA). Instead, a new 'crypto_probesession' method has been added to the kobj interface for symmteric crypto drivers. This method returns a negative value on success (similar to how device_probe works) and the crypto framework uses this value to pick the "best" driver. There are three constants for hardware (e.g. ccr), accelerated software (e.g. aesni), and plain software (cryptosoft) that give preference in that order. One effect of this is that if you request only hardware when creating a new session, you will no longer get a session using accelerated software. Another effect is that the default setting to disallow software crypto via /dev/crypto now disables accelerated software. Once a driver is chosen, 'crypto_newsession' is invoked as before. - Crypto operations are now solely described by the flat 'cryptop' structure. The linked list of descriptors has been removed. A separate enum has been added to describe the type of data buffer in use instead of using CRYPTO_F_* flags to make it easier to add more types in the future if needed (e.g. wired userspace buffers for zero-copy). It will also make it easier to re-introduce separate input and output buffers (in-kernel TLS would benefit from this). Try to make the flags related to IV handling less insane: - CRYPTO_F_IV_SEPARATE means that the IV is stored in the 'crp_iv' member of the operation structure. If this flag is not set, the IV is stored in the data buffer at the 'crp_iv_start' offset. - CRYPTO_F_IV_GENERATE means that a random IV should be generated and stored into the data buffer. This cannot be used with CRYPTO_F_IV_SEPARATE. If a consumer wants to deal with explicit vs implicit IVs, etc. it can always generate the IV however it needs and store partial IVs in the buffer and the full IV/nonce in crp_iv and set CRYPTO_F_IV_SEPARATE. The layout of the buffer is now described via fields in cryptop. crp_aad_start and crp_aad_length define the boundaries of any AAD. Previously with GCM and CCM you defined an auth crd with this range, but for ETA your auth crd had to span both the AAD and plaintext (and they had to be adjacent). crp_payload_start and crp_payload_length define the boundaries of the plaintext/ciphertext. Modes that only do a single operation (COMPRESS, CIPHER, DIGEST) should only use this region and leave the AAD region empty. If a digest is present (or should be generated), it's starting location is marked by crp_digest_start. Instead of using the CRD_F_ENCRYPT flag to determine the direction of the operation, cryptop now includes an 'op' field defining the operation to perform. For digests I've added a new VERIFY digest mode which assumes a digest is present in the input and fails the request with EBADMSG if it doesn't match the internally-computed digest. GCM and CCM already assumed this, and the new AEAD mode requires this for decryption. The new ETA mode now also requires this for decryption, so IPsec and GELI no longer do their own authentication verification. Simple DIGEST operations can also do this, though there are no in-tree consumers. To eventually support some refcounting to close races, the session cookie is now passed to crypto_getop() and clients should no longer set crp_sesssion directly. - Assymteric crypto operation structures should be allocated via crypto_getkreq() and freed via crypto_freekreq(). This permits the crypto layer to track open asym requests and close races with a driver trying to unregister while asym requests are in flight. - crypto_copyback, crypto_copydata, crypto_apply, and crypto_contiguous_subsegment now accept the 'crp' object as the first parameter instead of individual members. This makes it easier to deal with different buffer types in the future as well as separate input and output buffers. It's also simpler for driver writers to use. - bus_dmamap_load_crp() loads a DMA mapping for a crypto buffer. This understands the various types of buffers so that drivers that use DMA do not have to be aware of different buffer types. - Helper routines now exist to build an auth context for HMAC IPAD and OPAD. This reduces some duplicated work among drivers. - Key buffers are now treated as const throughout the framework and in device drivers. However, session key buffers provided when a session is created are expected to remain alive for the duration of the session. - GCM and CCM sessions now only specify a cipher algorithm and a cipher key. The redundant auth information is not needed or used. - For cryptosoft, split up the code a bit such that the 'process' callback now invokes a function pointer in the session. This function pointer is set based on the mode (in effect) though it simplifies a few edge cases that would otherwise be in the switch in 'process'. It does split up GCM vs CCM which I think is more readable even if there is some duplication. - I changed /dev/crypto to support GMAC requests using CRYPTO_AES_NIST_GMAC as an auth algorithm and updated cryptocheck to work with it. - Combined cipher and auth sessions via /dev/crypto now always use ETA mode. The COP_F_CIPHER_FIRST flag is now a no-op that is ignored. This was actually documented as being true in crypto(4) before, but the code had not implemented this before I added the CIPHER_FIRST flag. - I have not yet updated /dev/crypto to be aware of explicit modes for sessions. I will probably do that at some point in the future as well as teach it about IV/nonce and tag lengths for AEAD so we can support all of the NIST KAT tests for GCM and CCM. - I've split up the exising crypto.9 manpage into several pages of which many are written from scratch. - I have converted all drivers and consumers in the tree and verified that they compile, but I have not tested all of them. I have tested the following drivers: - cryptosoft - aesni (AES only) - blake2 - ccr and the following consumers: - cryptodev - IPsec - ktls_ocf - GELI (lightly) I have not tested the following: - ccp - aesni with sha - hifn - kgssapi_krb5 - ubsec - padlock - safe - armv8_crypto (aarch64) - glxsb (i386) - sec (ppc) - cesa (armv7) - cryptocteon (mips64) - nlmsec (mips64) Discussed with: cem Relnotes: yes Sponsored by: Chelsio Communications Differential Revision: https://reviews.freebsd.org/D23677
2020-03-27 18:25:23 +00:00
if (cop->mac && cse->hashsize == 0) {
2014-12-12 19:56:36 +00:00
SDT_PROBE1(opencrypto, dev, ioctl, error, __LINE__);
return (EINVAL);
}
Refactor driver and consumer interfaces for OCF (in-kernel crypto). - The linked list of cryptoini structures used in session initialization is replaced with a new flat structure: struct crypto_session_params. This session includes a new mode to define how the other fields should be interpreted. Available modes include: - COMPRESS (for compression/decompression) - CIPHER (for simply encryption/decryption) - DIGEST (computing and verifying digests) - AEAD (combined auth and encryption such as AES-GCM and AES-CCM) - ETA (combined auth and encryption using encrypt-then-authenticate) Additional modes could be added in the future (e.g. if we wanted to support TLS MtE for AES-CBC in the kernel we could add a new mode for that. TLS modes might also affect how AAD is interpreted, etc.) The flat structure also includes the key lengths and algorithms as before. However, code doesn't have to walk the linked list and switch on the algorithm to determine which key is the auth key vs encryption key. The 'csp_auth_*' fields are always used for auth keys and settings and 'csp_cipher_*' for cipher. (Compression algorithms are stored in csp_cipher_alg.) - Drivers no longer register a list of supported algorithms. This doesn't quite work when you factor in modes (e.g. a driver might support both AES-CBC and SHA2-256-HMAC separately but not combined for ETA). Instead, a new 'crypto_probesession' method has been added to the kobj interface for symmteric crypto drivers. This method returns a negative value on success (similar to how device_probe works) and the crypto framework uses this value to pick the "best" driver. There are three constants for hardware (e.g. ccr), accelerated software (e.g. aesni), and plain software (cryptosoft) that give preference in that order. One effect of this is that if you request only hardware when creating a new session, you will no longer get a session using accelerated software. Another effect is that the default setting to disallow software crypto via /dev/crypto now disables accelerated software. Once a driver is chosen, 'crypto_newsession' is invoked as before. - Crypto operations are now solely described by the flat 'cryptop' structure. The linked list of descriptors has been removed. A separate enum has been added to describe the type of data buffer in use instead of using CRYPTO_F_* flags to make it easier to add more types in the future if needed (e.g. wired userspace buffers for zero-copy). It will also make it easier to re-introduce separate input and output buffers (in-kernel TLS would benefit from this). Try to make the flags related to IV handling less insane: - CRYPTO_F_IV_SEPARATE means that the IV is stored in the 'crp_iv' member of the operation structure. If this flag is not set, the IV is stored in the data buffer at the 'crp_iv_start' offset. - CRYPTO_F_IV_GENERATE means that a random IV should be generated and stored into the data buffer. This cannot be used with CRYPTO_F_IV_SEPARATE. If a consumer wants to deal with explicit vs implicit IVs, etc. it can always generate the IV however it needs and store partial IVs in the buffer and the full IV/nonce in crp_iv and set CRYPTO_F_IV_SEPARATE. The layout of the buffer is now described via fields in cryptop. crp_aad_start and crp_aad_length define the boundaries of any AAD. Previously with GCM and CCM you defined an auth crd with this range, but for ETA your auth crd had to span both the AAD and plaintext (and they had to be adjacent). crp_payload_start and crp_payload_length define the boundaries of the plaintext/ciphertext. Modes that only do a single operation (COMPRESS, CIPHER, DIGEST) should only use this region and leave the AAD region empty. If a digest is present (or should be generated), it's starting location is marked by crp_digest_start. Instead of using the CRD_F_ENCRYPT flag to determine the direction of the operation, cryptop now includes an 'op' field defining the operation to perform. For digests I've added a new VERIFY digest mode which assumes a digest is present in the input and fails the request with EBADMSG if it doesn't match the internally-computed digest. GCM and CCM already assumed this, and the new AEAD mode requires this for decryption. The new ETA mode now also requires this for decryption, so IPsec and GELI no longer do their own authentication verification. Simple DIGEST operations can also do this, though there are no in-tree consumers. To eventually support some refcounting to close races, the session cookie is now passed to crypto_getop() and clients should no longer set crp_sesssion directly. - Assymteric crypto operation structures should be allocated via crypto_getkreq() and freed via crypto_freekreq(). This permits the crypto layer to track open asym requests and close races with a driver trying to unregister while asym requests are in flight. - crypto_copyback, crypto_copydata, crypto_apply, and crypto_contiguous_subsegment now accept the 'crp' object as the first parameter instead of individual members. This makes it easier to deal with different buffer types in the future as well as separate input and output buffers. It's also simpler for driver writers to use. - bus_dmamap_load_crp() loads a DMA mapping for a crypto buffer. This understands the various types of buffers so that drivers that use DMA do not have to be aware of different buffer types. - Helper routines now exist to build an auth context for HMAC IPAD and OPAD. This reduces some duplicated work among drivers. - Key buffers are now treated as const throughout the framework and in device drivers. However, session key buffers provided when a session is created are expected to remain alive for the duration of the session. - GCM and CCM sessions now only specify a cipher algorithm and a cipher key. The redundant auth information is not needed or used. - For cryptosoft, split up the code a bit such that the 'process' callback now invokes a function pointer in the session. This function pointer is set based on the mode (in effect) though it simplifies a few edge cases that would otherwise be in the switch in 'process'. It does split up GCM vs CCM which I think is more readable even if there is some duplication. - I changed /dev/crypto to support GMAC requests using CRYPTO_AES_NIST_GMAC as an auth algorithm and updated cryptocheck to work with it. - Combined cipher and auth sessions via /dev/crypto now always use ETA mode. The COP_F_CIPHER_FIRST flag is now a no-op that is ignored. This was actually documented as being true in crypto(4) before, but the code had not implemented this before I added the CIPHER_FIRST flag. - I have not yet updated /dev/crypto to be aware of explicit modes for sessions. I will probably do that at some point in the future as well as teach it about IV/nonce and tag lengths for AEAD so we can support all of the NIST KAT tests for GCM and CCM. - I've split up the exising crypto.9 manpage into several pages of which many are written from scratch. - I have converted all drivers and consumers in the tree and verified that they compile, but I have not tested all of them. I have tested the following drivers: - cryptosoft - aesni (AES only) - blake2 - ccr and the following consumers: - cryptodev - IPsec - ktls_ocf - GELI (lightly) I have not tested the following: - ccp - aesni with sha - hifn - kgssapi_krb5 - ubsec - padlock - safe - armv8_crypto (aarch64) - glxsb (i386) - sec (ppc) - cesa (armv7) - cryptocteon (mips64) - nlmsec (mips64) Discussed with: cem Relnotes: yes Sponsored by: Chelsio Communications Differential Revision: https://reviews.freebsd.org/D23677
2020-03-27 18:25:23 +00:00
/*
* The COP_F_CIPHER_FIRST flag predates explicit session
* modes, but the only way it was used was for EtA so allow it
* as long as it is consistent with EtA.
*/
if (cop->flags & COP_F_CIPHER_FIRST) {
if (cop->op != COP_ENCRYPT) {
SDT_PROBE1(opencrypto, dev, ioctl, error, __LINE__);
return (EINVAL);
}
}
cod = cod_alloc(cse, 0, cop->len + cse->hashsize);
dst = cop->dst;
Refactor driver and consumer interfaces for OCF (in-kernel crypto). - The linked list of cryptoini structures used in session initialization is replaced with a new flat structure: struct crypto_session_params. This session includes a new mode to define how the other fields should be interpreted. Available modes include: - COMPRESS (for compression/decompression) - CIPHER (for simply encryption/decryption) - DIGEST (computing and verifying digests) - AEAD (combined auth and encryption such as AES-GCM and AES-CCM) - ETA (combined auth and encryption using encrypt-then-authenticate) Additional modes could be added in the future (e.g. if we wanted to support TLS MtE for AES-CBC in the kernel we could add a new mode for that. TLS modes might also affect how AAD is interpreted, etc.) The flat structure also includes the key lengths and algorithms as before. However, code doesn't have to walk the linked list and switch on the algorithm to determine which key is the auth key vs encryption key. The 'csp_auth_*' fields are always used for auth keys and settings and 'csp_cipher_*' for cipher. (Compression algorithms are stored in csp_cipher_alg.) - Drivers no longer register a list of supported algorithms. This doesn't quite work when you factor in modes (e.g. a driver might support both AES-CBC and SHA2-256-HMAC separately but not combined for ETA). Instead, a new 'crypto_probesession' method has been added to the kobj interface for symmteric crypto drivers. This method returns a negative value on success (similar to how device_probe works) and the crypto framework uses this value to pick the "best" driver. There are three constants for hardware (e.g. ccr), accelerated software (e.g. aesni), and plain software (cryptosoft) that give preference in that order. One effect of this is that if you request only hardware when creating a new session, you will no longer get a session using accelerated software. Another effect is that the default setting to disallow software crypto via /dev/crypto now disables accelerated software. Once a driver is chosen, 'crypto_newsession' is invoked as before. - Crypto operations are now solely described by the flat 'cryptop' structure. The linked list of descriptors has been removed. A separate enum has been added to describe the type of data buffer in use instead of using CRYPTO_F_* flags to make it easier to add more types in the future if needed (e.g. wired userspace buffers for zero-copy). It will also make it easier to re-introduce separate input and output buffers (in-kernel TLS would benefit from this). Try to make the flags related to IV handling less insane: - CRYPTO_F_IV_SEPARATE means that the IV is stored in the 'crp_iv' member of the operation structure. If this flag is not set, the IV is stored in the data buffer at the 'crp_iv_start' offset. - CRYPTO_F_IV_GENERATE means that a random IV should be generated and stored into the data buffer. This cannot be used with CRYPTO_F_IV_SEPARATE. If a consumer wants to deal with explicit vs implicit IVs, etc. it can always generate the IV however it needs and store partial IVs in the buffer and the full IV/nonce in crp_iv and set CRYPTO_F_IV_SEPARATE. The layout of the buffer is now described via fields in cryptop. crp_aad_start and crp_aad_length define the boundaries of any AAD. Previously with GCM and CCM you defined an auth crd with this range, but for ETA your auth crd had to span both the AAD and plaintext (and they had to be adjacent). crp_payload_start and crp_payload_length define the boundaries of the plaintext/ciphertext. Modes that only do a single operation (COMPRESS, CIPHER, DIGEST) should only use this region and leave the AAD region empty. If a digest is present (or should be generated), it's starting location is marked by crp_digest_start. Instead of using the CRD_F_ENCRYPT flag to determine the direction of the operation, cryptop now includes an 'op' field defining the operation to perform. For digests I've added a new VERIFY digest mode which assumes a digest is present in the input and fails the request with EBADMSG if it doesn't match the internally-computed digest. GCM and CCM already assumed this, and the new AEAD mode requires this for decryption. The new ETA mode now also requires this for decryption, so IPsec and GELI no longer do their own authentication verification. Simple DIGEST operations can also do this, though there are no in-tree consumers. To eventually support some refcounting to close races, the session cookie is now passed to crypto_getop() and clients should no longer set crp_sesssion directly. - Assymteric crypto operation structures should be allocated via crypto_getkreq() and freed via crypto_freekreq(). This permits the crypto layer to track open asym requests and close races with a driver trying to unregister while asym requests are in flight. - crypto_copyback, crypto_copydata, crypto_apply, and crypto_contiguous_subsegment now accept the 'crp' object as the first parameter instead of individual members. This makes it easier to deal with different buffer types in the future as well as separate input and output buffers. It's also simpler for driver writers to use. - bus_dmamap_load_crp() loads a DMA mapping for a crypto buffer. This understands the various types of buffers so that drivers that use DMA do not have to be aware of different buffer types. - Helper routines now exist to build an auth context for HMAC IPAD and OPAD. This reduces some duplicated work among drivers. - Key buffers are now treated as const throughout the framework and in device drivers. However, session key buffers provided when a session is created are expected to remain alive for the duration of the session. - GCM and CCM sessions now only specify a cipher algorithm and a cipher key. The redundant auth information is not needed or used. - For cryptosoft, split up the code a bit such that the 'process' callback now invokes a function pointer in the session. This function pointer is set based on the mode (in effect) though it simplifies a few edge cases that would otherwise be in the switch in 'process'. It does split up GCM vs CCM which I think is more readable even if there is some duplication. - I changed /dev/crypto to support GMAC requests using CRYPTO_AES_NIST_GMAC as an auth algorithm and updated cryptocheck to work with it. - Combined cipher and auth sessions via /dev/crypto now always use ETA mode. The COP_F_CIPHER_FIRST flag is now a no-op that is ignored. This was actually documented as being true in crypto(4) before, but the code had not implemented this before I added the CIPHER_FIRST flag. - I have not yet updated /dev/crypto to be aware of explicit modes for sessions. I will probably do that at some point in the future as well as teach it about IV/nonce and tag lengths for AEAD so we can support all of the NIST KAT tests for GCM and CCM. - I've split up the exising crypto.9 manpage into several pages of which many are written from scratch. - I have converted all drivers and consumers in the tree and verified that they compile, but I have not tested all of them. I have tested the following drivers: - cryptosoft - aesni (AES only) - blake2 - ccr and the following consumers: - cryptodev - IPsec - ktls_ocf - GELI (lightly) I have not tested the following: - ccp - aesni with sha - hifn - kgssapi_krb5 - ubsec - padlock - safe - armv8_crypto (aarch64) - glxsb (i386) - sec (ppc) - cesa (armv7) - cryptocteon (mips64) - nlmsec (mips64) Discussed with: cem Relnotes: yes Sponsored by: Chelsio Communications Differential Revision: https://reviews.freebsd.org/D23677
2020-03-27 18:25:23 +00:00
crp = crypto_getreq(cse->cses, M_WAITOK);
Refactor driver and consumer interfaces for OCF (in-kernel crypto). - The linked list of cryptoini structures used in session initialization is replaced with a new flat structure: struct crypto_session_params. This session includes a new mode to define how the other fields should be interpreted. Available modes include: - COMPRESS (for compression/decompression) - CIPHER (for simply encryption/decryption) - DIGEST (computing and verifying digests) - AEAD (combined auth and encryption such as AES-GCM and AES-CCM) - ETA (combined auth and encryption using encrypt-then-authenticate) Additional modes could be added in the future (e.g. if we wanted to support TLS MtE for AES-CBC in the kernel we could add a new mode for that. TLS modes might also affect how AAD is interpreted, etc.) The flat structure also includes the key lengths and algorithms as before. However, code doesn't have to walk the linked list and switch on the algorithm to determine which key is the auth key vs encryption key. The 'csp_auth_*' fields are always used for auth keys and settings and 'csp_cipher_*' for cipher. (Compression algorithms are stored in csp_cipher_alg.) - Drivers no longer register a list of supported algorithms. This doesn't quite work when you factor in modes (e.g. a driver might support both AES-CBC and SHA2-256-HMAC separately but not combined for ETA). Instead, a new 'crypto_probesession' method has been added to the kobj interface for symmteric crypto drivers. This method returns a negative value on success (similar to how device_probe works) and the crypto framework uses this value to pick the "best" driver. There are three constants for hardware (e.g. ccr), accelerated software (e.g. aesni), and plain software (cryptosoft) that give preference in that order. One effect of this is that if you request only hardware when creating a new session, you will no longer get a session using accelerated software. Another effect is that the default setting to disallow software crypto via /dev/crypto now disables accelerated software. Once a driver is chosen, 'crypto_newsession' is invoked as before. - Crypto operations are now solely described by the flat 'cryptop' structure. The linked list of descriptors has been removed. A separate enum has been added to describe the type of data buffer in use instead of using CRYPTO_F_* flags to make it easier to add more types in the future if needed (e.g. wired userspace buffers for zero-copy). It will also make it easier to re-introduce separate input and output buffers (in-kernel TLS would benefit from this). Try to make the flags related to IV handling less insane: - CRYPTO_F_IV_SEPARATE means that the IV is stored in the 'crp_iv' member of the operation structure. If this flag is not set, the IV is stored in the data buffer at the 'crp_iv_start' offset. - CRYPTO_F_IV_GENERATE means that a random IV should be generated and stored into the data buffer. This cannot be used with CRYPTO_F_IV_SEPARATE. If a consumer wants to deal with explicit vs implicit IVs, etc. it can always generate the IV however it needs and store partial IVs in the buffer and the full IV/nonce in crp_iv and set CRYPTO_F_IV_SEPARATE. The layout of the buffer is now described via fields in cryptop. crp_aad_start and crp_aad_length define the boundaries of any AAD. Previously with GCM and CCM you defined an auth crd with this range, but for ETA your auth crd had to span both the AAD and plaintext (and they had to be adjacent). crp_payload_start and crp_payload_length define the boundaries of the plaintext/ciphertext. Modes that only do a single operation (COMPRESS, CIPHER, DIGEST) should only use this region and leave the AAD region empty. If a digest is present (or should be generated), it's starting location is marked by crp_digest_start. Instead of using the CRD_F_ENCRYPT flag to determine the direction of the operation, cryptop now includes an 'op' field defining the operation to perform. For digests I've added a new VERIFY digest mode which assumes a digest is present in the input and fails the request with EBADMSG if it doesn't match the internally-computed digest. GCM and CCM already assumed this, and the new AEAD mode requires this for decryption. The new ETA mode now also requires this for decryption, so IPsec and GELI no longer do their own authentication verification. Simple DIGEST operations can also do this, though there are no in-tree consumers. To eventually support some refcounting to close races, the session cookie is now passed to crypto_getop() and clients should no longer set crp_sesssion directly. - Assymteric crypto operation structures should be allocated via crypto_getkreq() and freed via crypto_freekreq(). This permits the crypto layer to track open asym requests and close races with a driver trying to unregister while asym requests are in flight. - crypto_copyback, crypto_copydata, crypto_apply, and crypto_contiguous_subsegment now accept the 'crp' object as the first parameter instead of individual members. This makes it easier to deal with different buffer types in the future as well as separate input and output buffers. It's also simpler for driver writers to use. - bus_dmamap_load_crp() loads a DMA mapping for a crypto buffer. This understands the various types of buffers so that drivers that use DMA do not have to be aware of different buffer types. - Helper routines now exist to build an auth context for HMAC IPAD and OPAD. This reduces some duplicated work among drivers. - Key buffers are now treated as const throughout the framework and in device drivers. However, session key buffers provided when a session is created are expected to remain alive for the duration of the session. - GCM and CCM sessions now only specify a cipher algorithm and a cipher key. The redundant auth information is not needed or used. - For cryptosoft, split up the code a bit such that the 'process' callback now invokes a function pointer in the session. This function pointer is set based on the mode (in effect) though it simplifies a few edge cases that would otherwise be in the switch in 'process'. It does split up GCM vs CCM which I think is more readable even if there is some duplication. - I changed /dev/crypto to support GMAC requests using CRYPTO_AES_NIST_GMAC as an auth algorithm and updated cryptocheck to work with it. - Combined cipher and auth sessions via /dev/crypto now always use ETA mode. The COP_F_CIPHER_FIRST flag is now a no-op that is ignored. This was actually documented as being true in crypto(4) before, but the code had not implemented this before I added the CIPHER_FIRST flag. - I have not yet updated /dev/crypto to be aware of explicit modes for sessions. I will probably do that at some point in the future as well as teach it about IV/nonce and tag lengths for AEAD so we can support all of the NIST KAT tests for GCM and CCM. - I've split up the exising crypto.9 manpage into several pages of which many are written from scratch. - I have converted all drivers and consumers in the tree and verified that they compile, but I have not tested all of them. I have tested the following drivers: - cryptosoft - aesni (AES only) - blake2 - ccr and the following consumers: - cryptodev - IPsec - ktls_ocf - GELI (lightly) I have not tested the following: - ccp - aesni with sha - hifn - kgssapi_krb5 - ubsec - padlock - safe - armv8_crypto (aarch64) - glxsb (i386) - sec (ppc) - cesa (armv7) - cryptocteon (mips64) - nlmsec (mips64) Discussed with: cem Relnotes: yes Sponsored by: Chelsio Communications Differential Revision: https://reviews.freebsd.org/D23677
2020-03-27 18:25:23 +00:00
error = copyin(cop->src, cod->buf, cop->len);
if (error) {
SDT_PROBE1(opencrypto, dev, ioctl, error, __LINE__);
goto bail;
}
Refactor driver and consumer interfaces for OCF (in-kernel crypto). - The linked list of cryptoini structures used in session initialization is replaced with a new flat structure: struct crypto_session_params. This session includes a new mode to define how the other fields should be interpreted. Available modes include: - COMPRESS (for compression/decompression) - CIPHER (for simply encryption/decryption) - DIGEST (computing and verifying digests) - AEAD (combined auth and encryption such as AES-GCM and AES-CCM) - ETA (combined auth and encryption using encrypt-then-authenticate) Additional modes could be added in the future (e.g. if we wanted to support TLS MtE for AES-CBC in the kernel we could add a new mode for that. TLS modes might also affect how AAD is interpreted, etc.) The flat structure also includes the key lengths and algorithms as before. However, code doesn't have to walk the linked list and switch on the algorithm to determine which key is the auth key vs encryption key. The 'csp_auth_*' fields are always used for auth keys and settings and 'csp_cipher_*' for cipher. (Compression algorithms are stored in csp_cipher_alg.) - Drivers no longer register a list of supported algorithms. This doesn't quite work when you factor in modes (e.g. a driver might support both AES-CBC and SHA2-256-HMAC separately but not combined for ETA). Instead, a new 'crypto_probesession' method has been added to the kobj interface for symmteric crypto drivers. This method returns a negative value on success (similar to how device_probe works) and the crypto framework uses this value to pick the "best" driver. There are three constants for hardware (e.g. ccr), accelerated software (e.g. aesni), and plain software (cryptosoft) that give preference in that order. One effect of this is that if you request only hardware when creating a new session, you will no longer get a session using accelerated software. Another effect is that the default setting to disallow software crypto via /dev/crypto now disables accelerated software. Once a driver is chosen, 'crypto_newsession' is invoked as before. - Crypto operations are now solely described by the flat 'cryptop' structure. The linked list of descriptors has been removed. A separate enum has been added to describe the type of data buffer in use instead of using CRYPTO_F_* flags to make it easier to add more types in the future if needed (e.g. wired userspace buffers for zero-copy). It will also make it easier to re-introduce separate input and output buffers (in-kernel TLS would benefit from this). Try to make the flags related to IV handling less insane: - CRYPTO_F_IV_SEPARATE means that the IV is stored in the 'crp_iv' member of the operation structure. If this flag is not set, the IV is stored in the data buffer at the 'crp_iv_start' offset. - CRYPTO_F_IV_GENERATE means that a random IV should be generated and stored into the data buffer. This cannot be used with CRYPTO_F_IV_SEPARATE. If a consumer wants to deal with explicit vs implicit IVs, etc. it can always generate the IV however it needs and store partial IVs in the buffer and the full IV/nonce in crp_iv and set CRYPTO_F_IV_SEPARATE. The layout of the buffer is now described via fields in cryptop. crp_aad_start and crp_aad_length define the boundaries of any AAD. Previously with GCM and CCM you defined an auth crd with this range, but for ETA your auth crd had to span both the AAD and plaintext (and they had to be adjacent). crp_payload_start and crp_payload_length define the boundaries of the plaintext/ciphertext. Modes that only do a single operation (COMPRESS, CIPHER, DIGEST) should only use this region and leave the AAD region empty. If a digest is present (or should be generated), it's starting location is marked by crp_digest_start. Instead of using the CRD_F_ENCRYPT flag to determine the direction of the operation, cryptop now includes an 'op' field defining the operation to perform. For digests I've added a new VERIFY digest mode which assumes a digest is present in the input and fails the request with EBADMSG if it doesn't match the internally-computed digest. GCM and CCM already assumed this, and the new AEAD mode requires this for decryption. The new ETA mode now also requires this for decryption, so IPsec and GELI no longer do their own authentication verification. Simple DIGEST operations can also do this, though there are no in-tree consumers. To eventually support some refcounting to close races, the session cookie is now passed to crypto_getop() and clients should no longer set crp_sesssion directly. - Assymteric crypto operation structures should be allocated via crypto_getkreq() and freed via crypto_freekreq(). This permits the crypto layer to track open asym requests and close races with a driver trying to unregister while asym requests are in flight. - crypto_copyback, crypto_copydata, crypto_apply, and crypto_contiguous_subsegment now accept the 'crp' object as the first parameter instead of individual members. This makes it easier to deal with different buffer types in the future as well as separate input and output buffers. It's also simpler for driver writers to use. - bus_dmamap_load_crp() loads a DMA mapping for a crypto buffer. This understands the various types of buffers so that drivers that use DMA do not have to be aware of different buffer types. - Helper routines now exist to build an auth context for HMAC IPAD and OPAD. This reduces some duplicated work among drivers. - Key buffers are now treated as const throughout the framework and in device drivers. However, session key buffers provided when a session is created are expected to remain alive for the duration of the session. - GCM and CCM sessions now only specify a cipher algorithm and a cipher key. The redundant auth information is not needed or used. - For cryptosoft, split up the code a bit such that the 'process' callback now invokes a function pointer in the session. This function pointer is set based on the mode (in effect) though it simplifies a few edge cases that would otherwise be in the switch in 'process'. It does split up GCM vs CCM which I think is more readable even if there is some duplication. - I changed /dev/crypto to support GMAC requests using CRYPTO_AES_NIST_GMAC as an auth algorithm and updated cryptocheck to work with it. - Combined cipher and auth sessions via /dev/crypto now always use ETA mode. The COP_F_CIPHER_FIRST flag is now a no-op that is ignored. This was actually documented as being true in crypto(4) before, but the code had not implemented this before I added the CIPHER_FIRST flag. - I have not yet updated /dev/crypto to be aware of explicit modes for sessions. I will probably do that at some point in the future as well as teach it about IV/nonce and tag lengths for AEAD so we can support all of the NIST KAT tests for GCM and CCM. - I've split up the exising crypto.9 manpage into several pages of which many are written from scratch. - I have converted all drivers and consumers in the tree and verified that they compile, but I have not tested all of them. I have tested the following drivers: - cryptosoft - aesni (AES only) - blake2 - ccr and the following consumers: - cryptodev - IPsec - ktls_ocf - GELI (lightly) I have not tested the following: - ccp - aesni with sha - hifn - kgssapi_krb5 - ubsec - padlock - safe - armv8_crypto (aarch64) - glxsb (i386) - sec (ppc) - cesa (armv7) - cryptocteon (mips64) - nlmsec (mips64) Discussed with: cem Relnotes: yes Sponsored by: Chelsio Communications Differential Revision: https://reviews.freebsd.org/D23677
2020-03-27 18:25:23 +00:00
crp->crp_payload_start = 0;
crp->crp_payload_length = cop->len;
if (cse->hashsize)
crp->crp_digest_start = cop->len;
csp = crypto_get_params(cse->cses);
switch (csp->csp_mode) {
Refactor driver and consumer interfaces for OCF (in-kernel crypto). - The linked list of cryptoini structures used in session initialization is replaced with a new flat structure: struct crypto_session_params. This session includes a new mode to define how the other fields should be interpreted. Available modes include: - COMPRESS (for compression/decompression) - CIPHER (for simply encryption/decryption) - DIGEST (computing and verifying digests) - AEAD (combined auth and encryption such as AES-GCM and AES-CCM) - ETA (combined auth and encryption using encrypt-then-authenticate) Additional modes could be added in the future (e.g. if we wanted to support TLS MtE for AES-CBC in the kernel we could add a new mode for that. TLS modes might also affect how AAD is interpreted, etc.) The flat structure also includes the key lengths and algorithms as before. However, code doesn't have to walk the linked list and switch on the algorithm to determine which key is the auth key vs encryption key. The 'csp_auth_*' fields are always used for auth keys and settings and 'csp_cipher_*' for cipher. (Compression algorithms are stored in csp_cipher_alg.) - Drivers no longer register a list of supported algorithms. This doesn't quite work when you factor in modes (e.g. a driver might support both AES-CBC and SHA2-256-HMAC separately but not combined for ETA). Instead, a new 'crypto_probesession' method has been added to the kobj interface for symmteric crypto drivers. This method returns a negative value on success (similar to how device_probe works) and the crypto framework uses this value to pick the "best" driver. There are three constants for hardware (e.g. ccr), accelerated software (e.g. aesni), and plain software (cryptosoft) that give preference in that order. One effect of this is that if you request only hardware when creating a new session, you will no longer get a session using accelerated software. Another effect is that the default setting to disallow software crypto via /dev/crypto now disables accelerated software. Once a driver is chosen, 'crypto_newsession' is invoked as before. - Crypto operations are now solely described by the flat 'cryptop' structure. The linked list of descriptors has been removed. A separate enum has been added to describe the type of data buffer in use instead of using CRYPTO_F_* flags to make it easier to add more types in the future if needed (e.g. wired userspace buffers for zero-copy). It will also make it easier to re-introduce separate input and output buffers (in-kernel TLS would benefit from this). Try to make the flags related to IV handling less insane: - CRYPTO_F_IV_SEPARATE means that the IV is stored in the 'crp_iv' member of the operation structure. If this flag is not set, the IV is stored in the data buffer at the 'crp_iv_start' offset. - CRYPTO_F_IV_GENERATE means that a random IV should be generated and stored into the data buffer. This cannot be used with CRYPTO_F_IV_SEPARATE. If a consumer wants to deal with explicit vs implicit IVs, etc. it can always generate the IV however it needs and store partial IVs in the buffer and the full IV/nonce in crp_iv and set CRYPTO_F_IV_SEPARATE. The layout of the buffer is now described via fields in cryptop. crp_aad_start and crp_aad_length define the boundaries of any AAD. Previously with GCM and CCM you defined an auth crd with this range, but for ETA your auth crd had to span both the AAD and plaintext (and they had to be adjacent). crp_payload_start and crp_payload_length define the boundaries of the plaintext/ciphertext. Modes that only do a single operation (COMPRESS, CIPHER, DIGEST) should only use this region and leave the AAD region empty. If a digest is present (or should be generated), it's starting location is marked by crp_digest_start. Instead of using the CRD_F_ENCRYPT flag to determine the direction of the operation, cryptop now includes an 'op' field defining the operation to perform. For digests I've added a new VERIFY digest mode which assumes a digest is present in the input and fails the request with EBADMSG if it doesn't match the internally-computed digest. GCM and CCM already assumed this, and the new AEAD mode requires this for decryption. The new ETA mode now also requires this for decryption, so IPsec and GELI no longer do their own authentication verification. Simple DIGEST operations can also do this, though there are no in-tree consumers. To eventually support some refcounting to close races, the session cookie is now passed to crypto_getop() and clients should no longer set crp_sesssion directly. - Assymteric crypto operation structures should be allocated via crypto_getkreq() and freed via crypto_freekreq(). This permits the crypto layer to track open asym requests and close races with a driver trying to unregister while asym requests are in flight. - crypto_copyback, crypto_copydata, crypto_apply, and crypto_contiguous_subsegment now accept the 'crp' object as the first parameter instead of individual members. This makes it easier to deal with different buffer types in the future as well as separate input and output buffers. It's also simpler for driver writers to use. - bus_dmamap_load_crp() loads a DMA mapping for a crypto buffer. This understands the various types of buffers so that drivers that use DMA do not have to be aware of different buffer types. - Helper routines now exist to build an auth context for HMAC IPAD and OPAD. This reduces some duplicated work among drivers. - Key buffers are now treated as const throughout the framework and in device drivers. However, session key buffers provided when a session is created are expected to remain alive for the duration of the session. - GCM and CCM sessions now only specify a cipher algorithm and a cipher key. The redundant auth information is not needed or used. - For cryptosoft, split up the code a bit such that the 'process' callback now invokes a function pointer in the session. This function pointer is set based on the mode (in effect) though it simplifies a few edge cases that would otherwise be in the switch in 'process'. It does split up GCM vs CCM which I think is more readable even if there is some duplication. - I changed /dev/crypto to support GMAC requests using CRYPTO_AES_NIST_GMAC as an auth algorithm and updated cryptocheck to work with it. - Combined cipher and auth sessions via /dev/crypto now always use ETA mode. The COP_F_CIPHER_FIRST flag is now a no-op that is ignored. This was actually documented as being true in crypto(4) before, but the code had not implemented this before I added the CIPHER_FIRST flag. - I have not yet updated /dev/crypto to be aware of explicit modes for sessions. I will probably do that at some point in the future as well as teach it about IV/nonce and tag lengths for AEAD so we can support all of the NIST KAT tests for GCM and CCM. - I've split up the exising crypto.9 manpage into several pages of which many are written from scratch. - I have converted all drivers and consumers in the tree and verified that they compile, but I have not tested all of them. I have tested the following drivers: - cryptosoft - aesni (AES only) - blake2 - ccr and the following consumers: - cryptodev - IPsec - ktls_ocf - GELI (lightly) I have not tested the following: - ccp - aesni with sha - hifn - kgssapi_krb5 - ubsec - padlock - safe - armv8_crypto (aarch64) - glxsb (i386) - sec (ppc) - cesa (armv7) - cryptocteon (mips64) - nlmsec (mips64) Discussed with: cem Relnotes: yes Sponsored by: Chelsio Communications Differential Revision: https://reviews.freebsd.org/D23677
2020-03-27 18:25:23 +00:00
case CSP_MODE_COMPRESS:
switch (cop->op) {
case COP_ENCRYPT:
crp->crp_op = CRYPTO_OP_COMPRESS;
break;
case COP_DECRYPT:
crp->crp_op = CRYPTO_OP_DECOMPRESS;
break;
default:
SDT_PROBE1(opencrypto, dev, ioctl, error, __LINE__);
error = EINVAL;
goto bail;
}
break;
case CSP_MODE_CIPHER:
if (cop->len == 0 ||
(cop->iv == NULL && cop->len == cse->ivsize)) {
SDT_PROBE1(opencrypto, dev, ioctl, error, __LINE__);
error = EINVAL;
goto bail;
}
Refactor driver and consumer interfaces for OCF (in-kernel crypto). - The linked list of cryptoini structures used in session initialization is replaced with a new flat structure: struct crypto_session_params. This session includes a new mode to define how the other fields should be interpreted. Available modes include: - COMPRESS (for compression/decompression) - CIPHER (for simply encryption/decryption) - DIGEST (computing and verifying digests) - AEAD (combined auth and encryption such as AES-GCM and AES-CCM) - ETA (combined auth and encryption using encrypt-then-authenticate) Additional modes could be added in the future (e.g. if we wanted to support TLS MtE for AES-CBC in the kernel we could add a new mode for that. TLS modes might also affect how AAD is interpreted, etc.) The flat structure also includes the key lengths and algorithms as before. However, code doesn't have to walk the linked list and switch on the algorithm to determine which key is the auth key vs encryption key. The 'csp_auth_*' fields are always used for auth keys and settings and 'csp_cipher_*' for cipher. (Compression algorithms are stored in csp_cipher_alg.) - Drivers no longer register a list of supported algorithms. This doesn't quite work when you factor in modes (e.g. a driver might support both AES-CBC and SHA2-256-HMAC separately but not combined for ETA). Instead, a new 'crypto_probesession' method has been added to the kobj interface for symmteric crypto drivers. This method returns a negative value on success (similar to how device_probe works) and the crypto framework uses this value to pick the "best" driver. There are three constants for hardware (e.g. ccr), accelerated software (e.g. aesni), and plain software (cryptosoft) that give preference in that order. One effect of this is that if you request only hardware when creating a new session, you will no longer get a session using accelerated software. Another effect is that the default setting to disallow software crypto via /dev/crypto now disables accelerated software. Once a driver is chosen, 'crypto_newsession' is invoked as before. - Crypto operations are now solely described by the flat 'cryptop' structure. The linked list of descriptors has been removed. A separate enum has been added to describe the type of data buffer in use instead of using CRYPTO_F_* flags to make it easier to add more types in the future if needed (e.g. wired userspace buffers for zero-copy). It will also make it easier to re-introduce separate input and output buffers (in-kernel TLS would benefit from this). Try to make the flags related to IV handling less insane: - CRYPTO_F_IV_SEPARATE means that the IV is stored in the 'crp_iv' member of the operation structure. If this flag is not set, the IV is stored in the data buffer at the 'crp_iv_start' offset. - CRYPTO_F_IV_GENERATE means that a random IV should be generated and stored into the data buffer. This cannot be used with CRYPTO_F_IV_SEPARATE. If a consumer wants to deal with explicit vs implicit IVs, etc. it can always generate the IV however it needs and store partial IVs in the buffer and the full IV/nonce in crp_iv and set CRYPTO_F_IV_SEPARATE. The layout of the buffer is now described via fields in cryptop. crp_aad_start and crp_aad_length define the boundaries of any AAD. Previously with GCM and CCM you defined an auth crd with this range, but for ETA your auth crd had to span both the AAD and plaintext (and they had to be adjacent). crp_payload_start and crp_payload_length define the boundaries of the plaintext/ciphertext. Modes that only do a single operation (COMPRESS, CIPHER, DIGEST) should only use this region and leave the AAD region empty. If a digest is present (or should be generated), it's starting location is marked by crp_digest_start. Instead of using the CRD_F_ENCRYPT flag to determine the direction of the operation, cryptop now includes an 'op' field defining the operation to perform. For digests I've added a new VERIFY digest mode which assumes a digest is present in the input and fails the request with EBADMSG if it doesn't match the internally-computed digest. GCM and CCM already assumed this, and the new AEAD mode requires this for decryption. The new ETA mode now also requires this for decryption, so IPsec and GELI no longer do their own authentication verification. Simple DIGEST operations can also do this, though there are no in-tree consumers. To eventually support some refcounting to close races, the session cookie is now passed to crypto_getop() and clients should no longer set crp_sesssion directly. - Assymteric crypto operation structures should be allocated via crypto_getkreq() and freed via crypto_freekreq(). This permits the crypto layer to track open asym requests and close races with a driver trying to unregister while asym requests are in flight. - crypto_copyback, crypto_copydata, crypto_apply, and crypto_contiguous_subsegment now accept the 'crp' object as the first parameter instead of individual members. This makes it easier to deal with different buffer types in the future as well as separate input and output buffers. It's also simpler for driver writers to use. - bus_dmamap_load_crp() loads a DMA mapping for a crypto buffer. This understands the various types of buffers so that drivers that use DMA do not have to be aware of different buffer types. - Helper routines now exist to build an auth context for HMAC IPAD and OPAD. This reduces some duplicated work among drivers. - Key buffers are now treated as const throughout the framework and in device drivers. However, session key buffers provided when a session is created are expected to remain alive for the duration of the session. - GCM and CCM sessions now only specify a cipher algorithm and a cipher key. The redundant auth information is not needed or used. - For cryptosoft, split up the code a bit such that the 'process' callback now invokes a function pointer in the session. This function pointer is set based on the mode (in effect) though it simplifies a few edge cases that would otherwise be in the switch in 'process'. It does split up GCM vs CCM which I think is more readable even if there is some duplication. - I changed /dev/crypto to support GMAC requests using CRYPTO_AES_NIST_GMAC as an auth algorithm and updated cryptocheck to work with it. - Combined cipher and auth sessions via /dev/crypto now always use ETA mode. The COP_F_CIPHER_FIRST flag is now a no-op that is ignored. This was actually documented as being true in crypto(4) before, but the code had not implemented this before I added the CIPHER_FIRST flag. - I have not yet updated /dev/crypto to be aware of explicit modes for sessions. I will probably do that at some point in the future as well as teach it about IV/nonce and tag lengths for AEAD so we can support all of the NIST KAT tests for GCM and CCM. - I've split up the exising crypto.9 manpage into several pages of which many are written from scratch. - I have converted all drivers and consumers in the tree and verified that they compile, but I have not tested all of them. I have tested the following drivers: - cryptosoft - aesni (AES only) - blake2 - ccr and the following consumers: - cryptodev - IPsec - ktls_ocf - GELI (lightly) I have not tested the following: - ccp - aesni with sha - hifn - kgssapi_krb5 - ubsec - padlock - safe - armv8_crypto (aarch64) - glxsb (i386) - sec (ppc) - cesa (armv7) - cryptocteon (mips64) - nlmsec (mips64) Discussed with: cem Relnotes: yes Sponsored by: Chelsio Communications Differential Revision: https://reviews.freebsd.org/D23677
2020-03-27 18:25:23 +00:00
switch (cop->op) {
case COP_ENCRYPT:
crp->crp_op = CRYPTO_OP_ENCRYPT;
break;
case COP_DECRYPT:
crp->crp_op = CRYPTO_OP_DECRYPT;
break;
default:
SDT_PROBE1(opencrypto, dev, ioctl, error, __LINE__);
error = EINVAL;
goto bail;
}
break;
case CSP_MODE_DIGEST:
switch (cop->op) {
case 0:
case COP_ENCRYPT:
case COP_DECRYPT:
crp->crp_op = CRYPTO_OP_COMPUTE_DIGEST;
if (cod->obuf != NULL)
crp->crp_digest_start = 0;
Refactor driver and consumer interfaces for OCF (in-kernel crypto). - The linked list of cryptoini structures used in session initialization is replaced with a new flat structure: struct crypto_session_params. This session includes a new mode to define how the other fields should be interpreted. Available modes include: - COMPRESS (for compression/decompression) - CIPHER (for simply encryption/decryption) - DIGEST (computing and verifying digests) - AEAD (combined auth and encryption such as AES-GCM and AES-CCM) - ETA (combined auth and encryption using encrypt-then-authenticate) Additional modes could be added in the future (e.g. if we wanted to support TLS MtE for AES-CBC in the kernel we could add a new mode for that. TLS modes might also affect how AAD is interpreted, etc.) The flat structure also includes the key lengths and algorithms as before. However, code doesn't have to walk the linked list and switch on the algorithm to determine which key is the auth key vs encryption key. The 'csp_auth_*' fields are always used for auth keys and settings and 'csp_cipher_*' for cipher. (Compression algorithms are stored in csp_cipher_alg.) - Drivers no longer register a list of supported algorithms. This doesn't quite work when you factor in modes (e.g. a driver might support both AES-CBC and SHA2-256-HMAC separately but not combined for ETA). Instead, a new 'crypto_probesession' method has been added to the kobj interface for symmteric crypto drivers. This method returns a negative value on success (similar to how device_probe works) and the crypto framework uses this value to pick the "best" driver. There are three constants for hardware (e.g. ccr), accelerated software (e.g. aesni), and plain software (cryptosoft) that give preference in that order. One effect of this is that if you request only hardware when creating a new session, you will no longer get a session using accelerated software. Another effect is that the default setting to disallow software crypto via /dev/crypto now disables accelerated software. Once a driver is chosen, 'crypto_newsession' is invoked as before. - Crypto operations are now solely described by the flat 'cryptop' structure. The linked list of descriptors has been removed. A separate enum has been added to describe the type of data buffer in use instead of using CRYPTO_F_* flags to make it easier to add more types in the future if needed (e.g. wired userspace buffers for zero-copy). It will also make it easier to re-introduce separate input and output buffers (in-kernel TLS would benefit from this). Try to make the flags related to IV handling less insane: - CRYPTO_F_IV_SEPARATE means that the IV is stored in the 'crp_iv' member of the operation structure. If this flag is not set, the IV is stored in the data buffer at the 'crp_iv_start' offset. - CRYPTO_F_IV_GENERATE means that a random IV should be generated and stored into the data buffer. This cannot be used with CRYPTO_F_IV_SEPARATE. If a consumer wants to deal with explicit vs implicit IVs, etc. it can always generate the IV however it needs and store partial IVs in the buffer and the full IV/nonce in crp_iv and set CRYPTO_F_IV_SEPARATE. The layout of the buffer is now described via fields in cryptop. crp_aad_start and crp_aad_length define the boundaries of any AAD. Previously with GCM and CCM you defined an auth crd with this range, but for ETA your auth crd had to span both the AAD and plaintext (and they had to be adjacent). crp_payload_start and crp_payload_length define the boundaries of the plaintext/ciphertext. Modes that only do a single operation (COMPRESS, CIPHER, DIGEST) should only use this region and leave the AAD region empty. If a digest is present (or should be generated), it's starting location is marked by crp_digest_start. Instead of using the CRD_F_ENCRYPT flag to determine the direction of the operation, cryptop now includes an 'op' field defining the operation to perform. For digests I've added a new VERIFY digest mode which assumes a digest is present in the input and fails the request with EBADMSG if it doesn't match the internally-computed digest. GCM and CCM already assumed this, and the new AEAD mode requires this for decryption. The new ETA mode now also requires this for decryption, so IPsec and GELI no longer do their own authentication verification. Simple DIGEST operations can also do this, though there are no in-tree consumers. To eventually support some refcounting to close races, the session cookie is now passed to crypto_getop() and clients should no longer set crp_sesssion directly. - Assymteric crypto operation structures should be allocated via crypto_getkreq() and freed via crypto_freekreq(). This permits the crypto layer to track open asym requests and close races with a driver trying to unregister while asym requests are in flight. - crypto_copyback, crypto_copydata, crypto_apply, and crypto_contiguous_subsegment now accept the 'crp' object as the first parameter instead of individual members. This makes it easier to deal with different buffer types in the future as well as separate input and output buffers. It's also simpler for driver writers to use. - bus_dmamap_load_crp() loads a DMA mapping for a crypto buffer. This understands the various types of buffers so that drivers that use DMA do not have to be aware of different buffer types. - Helper routines now exist to build an auth context for HMAC IPAD and OPAD. This reduces some duplicated work among drivers. - Key buffers are now treated as const throughout the framework and in device drivers. However, session key buffers provided when a session is created are expected to remain alive for the duration of the session. - GCM and CCM sessions now only specify a cipher algorithm and a cipher key. The redundant auth information is not needed or used. - For cryptosoft, split up the code a bit such that the 'process' callback now invokes a function pointer in the session. This function pointer is set based on the mode (in effect) though it simplifies a few edge cases that would otherwise be in the switch in 'process'. It does split up GCM vs CCM which I think is more readable even if there is some duplication. - I changed /dev/crypto to support GMAC requests using CRYPTO_AES_NIST_GMAC as an auth algorithm and updated cryptocheck to work with it. - Combined cipher and auth sessions via /dev/crypto now always use ETA mode. The COP_F_CIPHER_FIRST flag is now a no-op that is ignored. This was actually documented as being true in crypto(4) before, but the code had not implemented this before I added the CIPHER_FIRST flag. - I have not yet updated /dev/crypto to be aware of explicit modes for sessions. I will probably do that at some point in the future as well as teach it about IV/nonce and tag lengths for AEAD so we can support all of the NIST KAT tests for GCM and CCM. - I've split up the exising crypto.9 manpage into several pages of which many are written from scratch. - I have converted all drivers and consumers in the tree and verified that they compile, but I have not tested all of them. I have tested the following drivers: - cryptosoft - aesni (AES only) - blake2 - ccr and the following consumers: - cryptodev - IPsec - ktls_ocf - GELI (lightly) I have not tested the following: - ccp - aesni with sha - hifn - kgssapi_krb5 - ubsec - padlock - safe - armv8_crypto (aarch64) - glxsb (i386) - sec (ppc) - cesa (armv7) - cryptocteon (mips64) - nlmsec (mips64) Discussed with: cem Relnotes: yes Sponsored by: Chelsio Communications Differential Revision: https://reviews.freebsd.org/D23677
2020-03-27 18:25:23 +00:00
break;
default:
SDT_PROBE1(opencrypto, dev, ioctl, error, __LINE__);
error = EINVAL;
goto bail;
}
break;
case CSP_MODE_AEAD:
if (cse->ivsize != 0 && cop->iv == NULL) {
SDT_PROBE1(opencrypto, dev, ioctl, error, __LINE__);
error = EINVAL;
goto bail;
}
/* FALLTHROUGH */
Refactor driver and consumer interfaces for OCF (in-kernel crypto). - The linked list of cryptoini structures used in session initialization is replaced with a new flat structure: struct crypto_session_params. This session includes a new mode to define how the other fields should be interpreted. Available modes include: - COMPRESS (for compression/decompression) - CIPHER (for simply encryption/decryption) - DIGEST (computing and verifying digests) - AEAD (combined auth and encryption such as AES-GCM and AES-CCM) - ETA (combined auth and encryption using encrypt-then-authenticate) Additional modes could be added in the future (e.g. if we wanted to support TLS MtE for AES-CBC in the kernel we could add a new mode for that. TLS modes might also affect how AAD is interpreted, etc.) The flat structure also includes the key lengths and algorithms as before. However, code doesn't have to walk the linked list and switch on the algorithm to determine which key is the auth key vs encryption key. The 'csp_auth_*' fields are always used for auth keys and settings and 'csp_cipher_*' for cipher. (Compression algorithms are stored in csp_cipher_alg.) - Drivers no longer register a list of supported algorithms. This doesn't quite work when you factor in modes (e.g. a driver might support both AES-CBC and SHA2-256-HMAC separately but not combined for ETA). Instead, a new 'crypto_probesession' method has been added to the kobj interface for symmteric crypto drivers. This method returns a negative value on success (similar to how device_probe works) and the crypto framework uses this value to pick the "best" driver. There are three constants for hardware (e.g. ccr), accelerated software (e.g. aesni), and plain software (cryptosoft) that give preference in that order. One effect of this is that if you request only hardware when creating a new session, you will no longer get a session using accelerated software. Another effect is that the default setting to disallow software crypto via /dev/crypto now disables accelerated software. Once a driver is chosen, 'crypto_newsession' is invoked as before. - Crypto operations are now solely described by the flat 'cryptop' structure. The linked list of descriptors has been removed. A separate enum has been added to describe the type of data buffer in use instead of using CRYPTO_F_* flags to make it easier to add more types in the future if needed (e.g. wired userspace buffers for zero-copy). It will also make it easier to re-introduce separate input and output buffers (in-kernel TLS would benefit from this). Try to make the flags related to IV handling less insane: - CRYPTO_F_IV_SEPARATE means that the IV is stored in the 'crp_iv' member of the operation structure. If this flag is not set, the IV is stored in the data buffer at the 'crp_iv_start' offset. - CRYPTO_F_IV_GENERATE means that a random IV should be generated and stored into the data buffer. This cannot be used with CRYPTO_F_IV_SEPARATE. If a consumer wants to deal with explicit vs implicit IVs, etc. it can always generate the IV however it needs and store partial IVs in the buffer and the full IV/nonce in crp_iv and set CRYPTO_F_IV_SEPARATE. The layout of the buffer is now described via fields in cryptop. crp_aad_start and crp_aad_length define the boundaries of any AAD. Previously with GCM and CCM you defined an auth crd with this range, but for ETA your auth crd had to span both the AAD and plaintext (and they had to be adjacent). crp_payload_start and crp_payload_length define the boundaries of the plaintext/ciphertext. Modes that only do a single operation (COMPRESS, CIPHER, DIGEST) should only use this region and leave the AAD region empty. If a digest is present (or should be generated), it's starting location is marked by crp_digest_start. Instead of using the CRD_F_ENCRYPT flag to determine the direction of the operation, cryptop now includes an 'op' field defining the operation to perform. For digests I've added a new VERIFY digest mode which assumes a digest is present in the input and fails the request with EBADMSG if it doesn't match the internally-computed digest. GCM and CCM already assumed this, and the new AEAD mode requires this for decryption. The new ETA mode now also requires this for decryption, so IPsec and GELI no longer do their own authentication verification. Simple DIGEST operations can also do this, though there are no in-tree consumers. To eventually support some refcounting to close races, the session cookie is now passed to crypto_getop() and clients should no longer set crp_sesssion directly. - Assymteric crypto operation structures should be allocated via crypto_getkreq() and freed via crypto_freekreq(). This permits the crypto layer to track open asym requests and close races with a driver trying to unregister while asym requests are in flight. - crypto_copyback, crypto_copydata, crypto_apply, and crypto_contiguous_subsegment now accept the 'crp' object as the first parameter instead of individual members. This makes it easier to deal with different buffer types in the future as well as separate input and output buffers. It's also simpler for driver writers to use. - bus_dmamap_load_crp() loads a DMA mapping for a crypto buffer. This understands the various types of buffers so that drivers that use DMA do not have to be aware of different buffer types. - Helper routines now exist to build an auth context for HMAC IPAD and OPAD. This reduces some duplicated work among drivers. - Key buffers are now treated as const throughout the framework and in device drivers. However, session key buffers provided when a session is created are expected to remain alive for the duration of the session. - GCM and CCM sessions now only specify a cipher algorithm and a cipher key. The redundant auth information is not needed or used. - For cryptosoft, split up the code a bit such that the 'process' callback now invokes a function pointer in the session. This function pointer is set based on the mode (in effect) though it simplifies a few edge cases that would otherwise be in the switch in 'process'. It does split up GCM vs CCM which I think is more readable even if there is some duplication. - I changed /dev/crypto to support GMAC requests using CRYPTO_AES_NIST_GMAC as an auth algorithm and updated cryptocheck to work with it. - Combined cipher and auth sessions via /dev/crypto now always use ETA mode. The COP_F_CIPHER_FIRST flag is now a no-op that is ignored. This was actually documented as being true in crypto(4) before, but the code had not implemented this before I added the CIPHER_FIRST flag. - I have not yet updated /dev/crypto to be aware of explicit modes for sessions. I will probably do that at some point in the future as well as teach it about IV/nonce and tag lengths for AEAD so we can support all of the NIST KAT tests for GCM and CCM. - I've split up the exising crypto.9 manpage into several pages of which many are written from scratch. - I have converted all drivers and consumers in the tree and verified that they compile, but I have not tested all of them. I have tested the following drivers: - cryptosoft - aesni (AES only) - blake2 - ccr and the following consumers: - cryptodev - IPsec - ktls_ocf - GELI (lightly) I have not tested the following: - ccp - aesni with sha - hifn - kgssapi_krb5 - ubsec - padlock - safe - armv8_crypto (aarch64) - glxsb (i386) - sec (ppc) - cesa (armv7) - cryptocteon (mips64) - nlmsec (mips64) Discussed with: cem Relnotes: yes Sponsored by: Chelsio Communications Differential Revision: https://reviews.freebsd.org/D23677
2020-03-27 18:25:23 +00:00
case CSP_MODE_ETA:
switch (cop->op) {
case COP_ENCRYPT:
crp->crp_op = CRYPTO_OP_ENCRYPT |
CRYPTO_OP_COMPUTE_DIGEST;
break;
case COP_DECRYPT:
crp->crp_op = CRYPTO_OP_DECRYPT |
CRYPTO_OP_VERIFY_DIGEST;
break;
default:
SDT_PROBE1(opencrypto, dev, ioctl, error, __LINE__);
error = EINVAL;
goto bail;
}
break;
default:
SDT_PROBE1(opencrypto, dev, ioctl, error, __LINE__);
error = EINVAL;
goto bail;
}
Refactor driver and consumer interfaces for OCF (in-kernel crypto). - The linked list of cryptoini structures used in session initialization is replaced with a new flat structure: struct crypto_session_params. This session includes a new mode to define how the other fields should be interpreted. Available modes include: - COMPRESS (for compression/decompression) - CIPHER (for simply encryption/decryption) - DIGEST (computing and verifying digests) - AEAD (combined auth and encryption such as AES-GCM and AES-CCM) - ETA (combined auth and encryption using encrypt-then-authenticate) Additional modes could be added in the future (e.g. if we wanted to support TLS MtE for AES-CBC in the kernel we could add a new mode for that. TLS modes might also affect how AAD is interpreted, etc.) The flat structure also includes the key lengths and algorithms as before. However, code doesn't have to walk the linked list and switch on the algorithm to determine which key is the auth key vs encryption key. The 'csp_auth_*' fields are always used for auth keys and settings and 'csp_cipher_*' for cipher. (Compression algorithms are stored in csp_cipher_alg.) - Drivers no longer register a list of supported algorithms. This doesn't quite work when you factor in modes (e.g. a driver might support both AES-CBC and SHA2-256-HMAC separately but not combined for ETA). Instead, a new 'crypto_probesession' method has been added to the kobj interface for symmteric crypto drivers. This method returns a negative value on success (similar to how device_probe works) and the crypto framework uses this value to pick the "best" driver. There are three constants for hardware (e.g. ccr), accelerated software (e.g. aesni), and plain software (cryptosoft) that give preference in that order. One effect of this is that if you request only hardware when creating a new session, you will no longer get a session using accelerated software. Another effect is that the default setting to disallow software crypto via /dev/crypto now disables accelerated software. Once a driver is chosen, 'crypto_newsession' is invoked as before. - Crypto operations are now solely described by the flat 'cryptop' structure. The linked list of descriptors has been removed. A separate enum has been added to describe the type of data buffer in use instead of using CRYPTO_F_* flags to make it easier to add more types in the future if needed (e.g. wired userspace buffers for zero-copy). It will also make it easier to re-introduce separate input and output buffers (in-kernel TLS would benefit from this). Try to make the flags related to IV handling less insane: - CRYPTO_F_IV_SEPARATE means that the IV is stored in the 'crp_iv' member of the operation structure. If this flag is not set, the IV is stored in the data buffer at the 'crp_iv_start' offset. - CRYPTO_F_IV_GENERATE means that a random IV should be generated and stored into the data buffer. This cannot be used with CRYPTO_F_IV_SEPARATE. If a consumer wants to deal with explicit vs implicit IVs, etc. it can always generate the IV however it needs and store partial IVs in the buffer and the full IV/nonce in crp_iv and set CRYPTO_F_IV_SEPARATE. The layout of the buffer is now described via fields in cryptop. crp_aad_start and crp_aad_length define the boundaries of any AAD. Previously with GCM and CCM you defined an auth crd with this range, but for ETA your auth crd had to span both the AAD and plaintext (and they had to be adjacent). crp_payload_start and crp_payload_length define the boundaries of the plaintext/ciphertext. Modes that only do a single operation (COMPRESS, CIPHER, DIGEST) should only use this region and leave the AAD region empty. If a digest is present (or should be generated), it's starting location is marked by crp_digest_start. Instead of using the CRD_F_ENCRYPT flag to determine the direction of the operation, cryptop now includes an 'op' field defining the operation to perform. For digests I've added a new VERIFY digest mode which assumes a digest is present in the input and fails the request with EBADMSG if it doesn't match the internally-computed digest. GCM and CCM already assumed this, and the new AEAD mode requires this for decryption. The new ETA mode now also requires this for decryption, so IPsec and GELI no longer do their own authentication verification. Simple DIGEST operations can also do this, though there are no in-tree consumers. To eventually support some refcounting to close races, the session cookie is now passed to crypto_getop() and clients should no longer set crp_sesssion directly. - Assymteric crypto operation structures should be allocated via crypto_getkreq() and freed via crypto_freekreq(). This permits the crypto layer to track open asym requests and close races with a driver trying to unregister while asym requests are in flight. - crypto_copyback, crypto_copydata, crypto_apply, and crypto_contiguous_subsegment now accept the 'crp' object as the first parameter instead of individual members. This makes it easier to deal with different buffer types in the future as well as separate input and output buffers. It's also simpler for driver writers to use. - bus_dmamap_load_crp() loads a DMA mapping for a crypto buffer. This understands the various types of buffers so that drivers that use DMA do not have to be aware of different buffer types. - Helper routines now exist to build an auth context for HMAC IPAD and OPAD. This reduces some duplicated work among drivers. - Key buffers are now treated as const throughout the framework and in device drivers. However, session key buffers provided when a session is created are expected to remain alive for the duration of the session. - GCM and CCM sessions now only specify a cipher algorithm and a cipher key. The redundant auth information is not needed or used. - For cryptosoft, split up the code a bit such that the 'process' callback now invokes a function pointer in the session. This function pointer is set based on the mode (in effect) though it simplifies a few edge cases that would otherwise be in the switch in 'process'. It does split up GCM vs CCM which I think is more readable even if there is some duplication. - I changed /dev/crypto to support GMAC requests using CRYPTO_AES_NIST_GMAC as an auth algorithm and updated cryptocheck to work with it. - Combined cipher and auth sessions via /dev/crypto now always use ETA mode. The COP_F_CIPHER_FIRST flag is now a no-op that is ignored. This was actually documented as being true in crypto(4) before, but the code had not implemented this before I added the CIPHER_FIRST flag. - I have not yet updated /dev/crypto to be aware of explicit modes for sessions. I will probably do that at some point in the future as well as teach it about IV/nonce and tag lengths for AEAD so we can support all of the NIST KAT tests for GCM and CCM. - I've split up the exising crypto.9 manpage into several pages of which many are written from scratch. - I have converted all drivers and consumers in the tree and verified that they compile, but I have not tested all of them. I have tested the following drivers: - cryptosoft - aesni (AES only) - blake2 - ccr and the following consumers: - cryptodev - IPsec - ktls_ocf - GELI (lightly) I have not tested the following: - ccp - aesni with sha - hifn - kgssapi_krb5 - ubsec - padlock - safe - armv8_crypto (aarch64) - glxsb (i386) - sec (ppc) - cesa (armv7) - cryptocteon (mips64) - nlmsec (mips64) Discussed with: cem Relnotes: yes Sponsored by: Chelsio Communications Differential Revision: https://reviews.freebsd.org/D23677
2020-03-27 18:25:23 +00:00
crp->crp_flags = CRYPTO_F_CBIMM | (cop->flags & COP_F_BATCH);
Add support for optional separate output buffers to in-kernel crypto. Some crypto consumers such as GELI and KTLS for file-backed sendfile need to store their output in a separate buffer from the input. Currently these consumers copy the contents of the input buffer into the output buffer and queue an in-place crypto operation on the output buffer. Using a separate output buffer avoids this copy. - Create a new 'struct crypto_buffer' describing a crypto buffer containing a type and type-specific fields. crp_ilen is gone, instead buffers that use a flat kernel buffer have a cb_buf_len field for their length. The length of other buffer types is inferred from the backing store (e.g. uio_resid for a uio). Requests now have two such structures: crp_buf for the input buffer, and crp_obuf for the output buffer. - Consumers now use helper functions (crypto_use_*, e.g. crypto_use_mbuf()) to configure the input buffer. If an output buffer is not configured, the request still modifies the input buffer in-place. A consumer uses a second set of helper functions (crypto_use_output_*) to configure an output buffer. - Consumers must request support for separate output buffers when creating a crypto session via the CSP_F_SEPARATE_OUTPUT flag and are only permitted to queue a request with a separate output buffer on sessions with this flag set. Existing drivers already reject sessions with unknown flags, so this permits drivers to be modified to support this extension without requiring all drivers to change. - Several data-related functions now have matching versions that operate on an explicit buffer (e.g. crypto_apply_buf, crypto_contiguous_subsegment_buf, bus_dma_load_crp_buf). - Most of the existing data-related functions operate on the input buffer. However crypto_copyback always writes to the output buffer if a request uses a separate output buffer. - For the regions in input/output buffers, the following conventions are followed: - AAD and IV are always present in input only and their fields are offsets into the input buffer. - payload is always present in both buffers. If a request uses a separate output buffer, it must set a new crp_payload_start_output field to the offset of the payload in the output buffer. - digest is in the input buffer for verify operations, and in the output buffer for compute operations. crp_digest_start is relative to the appropriate buffer. - Add a crypto buffer cursor abstraction. This is a more general form of some bits in the cryptosoft driver that tried to always use uio's. However, compared to the original code, this avoids rewalking the uio iovec array for requests with multiple vectors. It also avoids allocate an iovec array for mbufs and populating it by instead walking the mbuf chain directly. - Update the cryptosoft(4) driver to support separate output buffers making use of the cursor abstraction. Sponsored by: Netflix Differential Revision: https://reviews.freebsd.org/D24545
2020-05-25 22:12:04 +00:00
crypto_use_buf(crp, cod->buf, cop->len + cse->hashsize);
if (cod->obuf)
crypto_use_output_buf(crp, cod->obuf, cop->len + cse->hashsize);
crp->crp_callback = cryptodev_cb;
crp->crp_opaque = cod;
if (cop->iv) {
Refactor driver and consumer interfaces for OCF (in-kernel crypto). - The linked list of cryptoini structures used in session initialization is replaced with a new flat structure: struct crypto_session_params. This session includes a new mode to define how the other fields should be interpreted. Available modes include: - COMPRESS (for compression/decompression) - CIPHER (for simply encryption/decryption) - DIGEST (computing and verifying digests) - AEAD (combined auth and encryption such as AES-GCM and AES-CCM) - ETA (combined auth and encryption using encrypt-then-authenticate) Additional modes could be added in the future (e.g. if we wanted to support TLS MtE for AES-CBC in the kernel we could add a new mode for that. TLS modes might also affect how AAD is interpreted, etc.) The flat structure also includes the key lengths and algorithms as before. However, code doesn't have to walk the linked list and switch on the algorithm to determine which key is the auth key vs encryption key. The 'csp_auth_*' fields are always used for auth keys and settings and 'csp_cipher_*' for cipher. (Compression algorithms are stored in csp_cipher_alg.) - Drivers no longer register a list of supported algorithms. This doesn't quite work when you factor in modes (e.g. a driver might support both AES-CBC and SHA2-256-HMAC separately but not combined for ETA). Instead, a new 'crypto_probesession' method has been added to the kobj interface for symmteric crypto drivers. This method returns a negative value on success (similar to how device_probe works) and the crypto framework uses this value to pick the "best" driver. There are three constants for hardware (e.g. ccr), accelerated software (e.g. aesni), and plain software (cryptosoft) that give preference in that order. One effect of this is that if you request only hardware when creating a new session, you will no longer get a session using accelerated software. Another effect is that the default setting to disallow software crypto via /dev/crypto now disables accelerated software. Once a driver is chosen, 'crypto_newsession' is invoked as before. - Crypto operations are now solely described by the flat 'cryptop' structure. The linked list of descriptors has been removed. A separate enum has been added to describe the type of data buffer in use instead of using CRYPTO_F_* flags to make it easier to add more types in the future if needed (e.g. wired userspace buffers for zero-copy). It will also make it easier to re-introduce separate input and output buffers (in-kernel TLS would benefit from this). Try to make the flags related to IV handling less insane: - CRYPTO_F_IV_SEPARATE means that the IV is stored in the 'crp_iv' member of the operation structure. If this flag is not set, the IV is stored in the data buffer at the 'crp_iv_start' offset. - CRYPTO_F_IV_GENERATE means that a random IV should be generated and stored into the data buffer. This cannot be used with CRYPTO_F_IV_SEPARATE. If a consumer wants to deal with explicit vs implicit IVs, etc. it can always generate the IV however it needs and store partial IVs in the buffer and the full IV/nonce in crp_iv and set CRYPTO_F_IV_SEPARATE. The layout of the buffer is now described via fields in cryptop. crp_aad_start and crp_aad_length define the boundaries of any AAD. Previously with GCM and CCM you defined an auth crd with this range, but for ETA your auth crd had to span both the AAD and plaintext (and they had to be adjacent). crp_payload_start and crp_payload_length define the boundaries of the plaintext/ciphertext. Modes that only do a single operation (COMPRESS, CIPHER, DIGEST) should only use this region and leave the AAD region empty. If a digest is present (or should be generated), it's starting location is marked by crp_digest_start. Instead of using the CRD_F_ENCRYPT flag to determine the direction of the operation, cryptop now includes an 'op' field defining the operation to perform. For digests I've added a new VERIFY digest mode which assumes a digest is present in the input and fails the request with EBADMSG if it doesn't match the internally-computed digest. GCM and CCM already assumed this, and the new AEAD mode requires this for decryption. The new ETA mode now also requires this for decryption, so IPsec and GELI no longer do their own authentication verification. Simple DIGEST operations can also do this, though there are no in-tree consumers. To eventually support some refcounting to close races, the session cookie is now passed to crypto_getop() and clients should no longer set crp_sesssion directly. - Assymteric crypto operation structures should be allocated via crypto_getkreq() and freed via crypto_freekreq(). This permits the crypto layer to track open asym requests and close races with a driver trying to unregister while asym requests are in flight. - crypto_copyback, crypto_copydata, crypto_apply, and crypto_contiguous_subsegment now accept the 'crp' object as the first parameter instead of individual members. This makes it easier to deal with different buffer types in the future as well as separate input and output buffers. It's also simpler for driver writers to use. - bus_dmamap_load_crp() loads a DMA mapping for a crypto buffer. This understands the various types of buffers so that drivers that use DMA do not have to be aware of different buffer types. - Helper routines now exist to build an auth context for HMAC IPAD and OPAD. This reduces some duplicated work among drivers. - Key buffers are now treated as const throughout the framework and in device drivers. However, session key buffers provided when a session is created are expected to remain alive for the duration of the session. - GCM and CCM sessions now only specify a cipher algorithm and a cipher key. The redundant auth information is not needed or used. - For cryptosoft, split up the code a bit such that the 'process' callback now invokes a function pointer in the session. This function pointer is set based on the mode (in effect) though it simplifies a few edge cases that would otherwise be in the switch in 'process'. It does split up GCM vs CCM which I think is more readable even if there is some duplication. - I changed /dev/crypto to support GMAC requests using CRYPTO_AES_NIST_GMAC as an auth algorithm and updated cryptocheck to work with it. - Combined cipher and auth sessions via /dev/crypto now always use ETA mode. The COP_F_CIPHER_FIRST flag is now a no-op that is ignored. This was actually documented as being true in crypto(4) before, but the code had not implemented this before I added the CIPHER_FIRST flag. - I have not yet updated /dev/crypto to be aware of explicit modes for sessions. I will probably do that at some point in the future as well as teach it about IV/nonce and tag lengths for AEAD so we can support all of the NIST KAT tests for GCM and CCM. - I've split up the exising crypto.9 manpage into several pages of which many are written from scratch. - I have converted all drivers and consumers in the tree and verified that they compile, but I have not tested all of them. I have tested the following drivers: - cryptosoft - aesni (AES only) - blake2 - ccr and the following consumers: - cryptodev - IPsec - ktls_ocf - GELI (lightly) I have not tested the following: - ccp - aesni with sha - hifn - kgssapi_krb5 - ubsec - padlock - safe - armv8_crypto (aarch64) - glxsb (i386) - sec (ppc) - cesa (armv7) - cryptocteon (mips64) - nlmsec (mips64) Discussed with: cem Relnotes: yes Sponsored by: Chelsio Communications Differential Revision: https://reviews.freebsd.org/D23677
2020-03-27 18:25:23 +00:00
if (cse->ivsize == 0) {
2014-12-12 19:56:36 +00:00
SDT_PROBE1(opencrypto, dev, ioctl, error, __LINE__);
error = EINVAL;
goto bail;
}
Refactor driver and consumer interfaces for OCF (in-kernel crypto). - The linked list of cryptoini structures used in session initialization is replaced with a new flat structure: struct crypto_session_params. This session includes a new mode to define how the other fields should be interpreted. Available modes include: - COMPRESS (for compression/decompression) - CIPHER (for simply encryption/decryption) - DIGEST (computing and verifying digests) - AEAD (combined auth and encryption such as AES-GCM and AES-CCM) - ETA (combined auth and encryption using encrypt-then-authenticate) Additional modes could be added in the future (e.g. if we wanted to support TLS MtE for AES-CBC in the kernel we could add a new mode for that. TLS modes might also affect how AAD is interpreted, etc.) The flat structure also includes the key lengths and algorithms as before. However, code doesn't have to walk the linked list and switch on the algorithm to determine which key is the auth key vs encryption key. The 'csp_auth_*' fields are always used for auth keys and settings and 'csp_cipher_*' for cipher. (Compression algorithms are stored in csp_cipher_alg.) - Drivers no longer register a list of supported algorithms. This doesn't quite work when you factor in modes (e.g. a driver might support both AES-CBC and SHA2-256-HMAC separately but not combined for ETA). Instead, a new 'crypto_probesession' method has been added to the kobj interface for symmteric crypto drivers. This method returns a negative value on success (similar to how device_probe works) and the crypto framework uses this value to pick the "best" driver. There are three constants for hardware (e.g. ccr), accelerated software (e.g. aesni), and plain software (cryptosoft) that give preference in that order. One effect of this is that if you request only hardware when creating a new session, you will no longer get a session using accelerated software. Another effect is that the default setting to disallow software crypto via /dev/crypto now disables accelerated software. Once a driver is chosen, 'crypto_newsession' is invoked as before. - Crypto operations are now solely described by the flat 'cryptop' structure. The linked list of descriptors has been removed. A separate enum has been added to describe the type of data buffer in use instead of using CRYPTO_F_* flags to make it easier to add more types in the future if needed (e.g. wired userspace buffers for zero-copy). It will also make it easier to re-introduce separate input and output buffers (in-kernel TLS would benefit from this). Try to make the flags related to IV handling less insane: - CRYPTO_F_IV_SEPARATE means that the IV is stored in the 'crp_iv' member of the operation structure. If this flag is not set, the IV is stored in the data buffer at the 'crp_iv_start' offset. - CRYPTO_F_IV_GENERATE means that a random IV should be generated and stored into the data buffer. This cannot be used with CRYPTO_F_IV_SEPARATE. If a consumer wants to deal with explicit vs implicit IVs, etc. it can always generate the IV however it needs and store partial IVs in the buffer and the full IV/nonce in crp_iv and set CRYPTO_F_IV_SEPARATE. The layout of the buffer is now described via fields in cryptop. crp_aad_start and crp_aad_length define the boundaries of any AAD. Previously with GCM and CCM you defined an auth crd with this range, but for ETA your auth crd had to span both the AAD and plaintext (and they had to be adjacent). crp_payload_start and crp_payload_length define the boundaries of the plaintext/ciphertext. Modes that only do a single operation (COMPRESS, CIPHER, DIGEST) should only use this region and leave the AAD region empty. If a digest is present (or should be generated), it's starting location is marked by crp_digest_start. Instead of using the CRD_F_ENCRYPT flag to determine the direction of the operation, cryptop now includes an 'op' field defining the operation to perform. For digests I've added a new VERIFY digest mode which assumes a digest is present in the input and fails the request with EBADMSG if it doesn't match the internally-computed digest. GCM and CCM already assumed this, and the new AEAD mode requires this for decryption. The new ETA mode now also requires this for decryption, so IPsec and GELI no longer do their own authentication verification. Simple DIGEST operations can also do this, though there are no in-tree consumers. To eventually support some refcounting to close races, the session cookie is now passed to crypto_getop() and clients should no longer set crp_sesssion directly. - Assymteric crypto operation structures should be allocated via crypto_getkreq() and freed via crypto_freekreq(). This permits the crypto layer to track open asym requests and close races with a driver trying to unregister while asym requests are in flight. - crypto_copyback, crypto_copydata, crypto_apply, and crypto_contiguous_subsegment now accept the 'crp' object as the first parameter instead of individual members. This makes it easier to deal with different buffer types in the future as well as separate input and output buffers. It's also simpler for driver writers to use. - bus_dmamap_load_crp() loads a DMA mapping for a crypto buffer. This understands the various types of buffers so that drivers that use DMA do not have to be aware of different buffer types. - Helper routines now exist to build an auth context for HMAC IPAD and OPAD. This reduces some duplicated work among drivers. - Key buffers are now treated as const throughout the framework and in device drivers. However, session key buffers provided when a session is created are expected to remain alive for the duration of the session. - GCM and CCM sessions now only specify a cipher algorithm and a cipher key. The redundant auth information is not needed or used. - For cryptosoft, split up the code a bit such that the 'process' callback now invokes a function pointer in the session. This function pointer is set based on the mode (in effect) though it simplifies a few edge cases that would otherwise be in the switch in 'process'. It does split up GCM vs CCM which I think is more readable even if there is some duplication. - I changed /dev/crypto to support GMAC requests using CRYPTO_AES_NIST_GMAC as an auth algorithm and updated cryptocheck to work with it. - Combined cipher and auth sessions via /dev/crypto now always use ETA mode. The COP_F_CIPHER_FIRST flag is now a no-op that is ignored. This was actually documented as being true in crypto(4) before, but the code had not implemented this before I added the CIPHER_FIRST flag. - I have not yet updated /dev/crypto to be aware of explicit modes for sessions. I will probably do that at some point in the future as well as teach it about IV/nonce and tag lengths for AEAD so we can support all of the NIST KAT tests for GCM and CCM. - I've split up the exising crypto.9 manpage into several pages of which many are written from scratch. - I have converted all drivers and consumers in the tree and verified that they compile, but I have not tested all of them. I have tested the following drivers: - cryptosoft - aesni (AES only) - blake2 - ccr and the following consumers: - cryptodev - IPsec - ktls_ocf - GELI (lightly) I have not tested the following: - ccp - aesni with sha - hifn - kgssapi_krb5 - ubsec - padlock - safe - armv8_crypto (aarch64) - glxsb (i386) - sec (ppc) - cesa (armv7) - cryptocteon (mips64) - nlmsec (mips64) Discussed with: cem Relnotes: yes Sponsored by: Chelsio Communications Differential Revision: https://reviews.freebsd.org/D23677
2020-03-27 18:25:23 +00:00
error = copyin(cop->iv, crp->crp_iv, cse->ivsize);
if (error) {
2014-12-12 19:56:36 +00:00
SDT_PROBE1(opencrypto, dev, ioctl, error, __LINE__);
goto bail;
}
Refactor driver and consumer interfaces for OCF (in-kernel crypto). - The linked list of cryptoini structures used in session initialization is replaced with a new flat structure: struct crypto_session_params. This session includes a new mode to define how the other fields should be interpreted. Available modes include: - COMPRESS (for compression/decompression) - CIPHER (for simply encryption/decryption) - DIGEST (computing and verifying digests) - AEAD (combined auth and encryption such as AES-GCM and AES-CCM) - ETA (combined auth and encryption using encrypt-then-authenticate) Additional modes could be added in the future (e.g. if we wanted to support TLS MtE for AES-CBC in the kernel we could add a new mode for that. TLS modes might also affect how AAD is interpreted, etc.) The flat structure also includes the key lengths and algorithms as before. However, code doesn't have to walk the linked list and switch on the algorithm to determine which key is the auth key vs encryption key. The 'csp_auth_*' fields are always used for auth keys and settings and 'csp_cipher_*' for cipher. (Compression algorithms are stored in csp_cipher_alg.) - Drivers no longer register a list of supported algorithms. This doesn't quite work when you factor in modes (e.g. a driver might support both AES-CBC and SHA2-256-HMAC separately but not combined for ETA). Instead, a new 'crypto_probesession' method has been added to the kobj interface for symmteric crypto drivers. This method returns a negative value on success (similar to how device_probe works) and the crypto framework uses this value to pick the "best" driver. There are three constants for hardware (e.g. ccr), accelerated software (e.g. aesni), and plain software (cryptosoft) that give preference in that order. One effect of this is that if you request only hardware when creating a new session, you will no longer get a session using accelerated software. Another effect is that the default setting to disallow software crypto via /dev/crypto now disables accelerated software. Once a driver is chosen, 'crypto_newsession' is invoked as before. - Crypto operations are now solely described by the flat 'cryptop' structure. The linked list of descriptors has been removed. A separate enum has been added to describe the type of data buffer in use instead of using CRYPTO_F_* flags to make it easier to add more types in the future if needed (e.g. wired userspace buffers for zero-copy). It will also make it easier to re-introduce separate input and output buffers (in-kernel TLS would benefit from this). Try to make the flags related to IV handling less insane: - CRYPTO_F_IV_SEPARATE means that the IV is stored in the 'crp_iv' member of the operation structure. If this flag is not set, the IV is stored in the data buffer at the 'crp_iv_start' offset. - CRYPTO_F_IV_GENERATE means that a random IV should be generated and stored into the data buffer. This cannot be used with CRYPTO_F_IV_SEPARATE. If a consumer wants to deal with explicit vs implicit IVs, etc. it can always generate the IV however it needs and store partial IVs in the buffer and the full IV/nonce in crp_iv and set CRYPTO_F_IV_SEPARATE. The layout of the buffer is now described via fields in cryptop. crp_aad_start and crp_aad_length define the boundaries of any AAD. Previously with GCM and CCM you defined an auth crd with this range, but for ETA your auth crd had to span both the AAD and plaintext (and they had to be adjacent). crp_payload_start and crp_payload_length define the boundaries of the plaintext/ciphertext. Modes that only do a single operation (COMPRESS, CIPHER, DIGEST) should only use this region and leave the AAD region empty. If a digest is present (or should be generated), it's starting location is marked by crp_digest_start. Instead of using the CRD_F_ENCRYPT flag to determine the direction of the operation, cryptop now includes an 'op' field defining the operation to perform. For digests I've added a new VERIFY digest mode which assumes a digest is present in the input and fails the request with EBADMSG if it doesn't match the internally-computed digest. GCM and CCM already assumed this, and the new AEAD mode requires this for decryption. The new ETA mode now also requires this for decryption, so IPsec and GELI no longer do their own authentication verification. Simple DIGEST operations can also do this, though there are no in-tree consumers. To eventually support some refcounting to close races, the session cookie is now passed to crypto_getop() and clients should no longer set crp_sesssion directly. - Assymteric crypto operation structures should be allocated via crypto_getkreq() and freed via crypto_freekreq(). This permits the crypto layer to track open asym requests and close races with a driver trying to unregister while asym requests are in flight. - crypto_copyback, crypto_copydata, crypto_apply, and crypto_contiguous_subsegment now accept the 'crp' object as the first parameter instead of individual members. This makes it easier to deal with different buffer types in the future as well as separate input and output buffers. It's also simpler for driver writers to use. - bus_dmamap_load_crp() loads a DMA mapping for a crypto buffer. This understands the various types of buffers so that drivers that use DMA do not have to be aware of different buffer types. - Helper routines now exist to build an auth context for HMAC IPAD and OPAD. This reduces some duplicated work among drivers. - Key buffers are now treated as const throughout the framework and in device drivers. However, session key buffers provided when a session is created are expected to remain alive for the duration of the session. - GCM and CCM sessions now only specify a cipher algorithm and a cipher key. The redundant auth information is not needed or used. - For cryptosoft, split up the code a bit such that the 'process' callback now invokes a function pointer in the session. This function pointer is set based on the mode (in effect) though it simplifies a few edge cases that would otherwise be in the switch in 'process'. It does split up GCM vs CCM which I think is more readable even if there is some duplication. - I changed /dev/crypto to support GMAC requests using CRYPTO_AES_NIST_GMAC as an auth algorithm and updated cryptocheck to work with it. - Combined cipher and auth sessions via /dev/crypto now always use ETA mode. The COP_F_CIPHER_FIRST flag is now a no-op that is ignored. This was actually documented as being true in crypto(4) before, but the code had not implemented this before I added the CIPHER_FIRST flag. - I have not yet updated /dev/crypto to be aware of explicit modes for sessions. I will probably do that at some point in the future as well as teach it about IV/nonce and tag lengths for AEAD so we can support all of the NIST KAT tests for GCM and CCM. - I've split up the exising crypto.9 manpage into several pages of which many are written from scratch. - I have converted all drivers and consumers in the tree and verified that they compile, but I have not tested all of them. I have tested the following drivers: - cryptosoft - aesni (AES only) - blake2 - ccr and the following consumers: - cryptodev - IPsec - ktls_ocf - GELI (lightly) I have not tested the following: - ccp - aesni with sha - hifn - kgssapi_krb5 - ubsec - padlock - safe - armv8_crypto (aarch64) - glxsb (i386) - sec (ppc) - cesa (armv7) - cryptocteon (mips64) - nlmsec (mips64) Discussed with: cem Relnotes: yes Sponsored by: Chelsio Communications Differential Revision: https://reviews.freebsd.org/D23677
2020-03-27 18:25:23 +00:00
crp->crp_flags |= CRYPTO_F_IV_SEPARATE;
} else if (cse->ivsize != 0) {
if (crp->crp_payload_length < cse->ivsize) {
SDT_PROBE1(opencrypto, dev, ioctl, error, __LINE__);
error = EINVAL;
goto bail;
}
Refactor driver and consumer interfaces for OCF (in-kernel crypto). - The linked list of cryptoini structures used in session initialization is replaced with a new flat structure: struct crypto_session_params. This session includes a new mode to define how the other fields should be interpreted. Available modes include: - COMPRESS (for compression/decompression) - CIPHER (for simply encryption/decryption) - DIGEST (computing and verifying digests) - AEAD (combined auth and encryption such as AES-GCM and AES-CCM) - ETA (combined auth and encryption using encrypt-then-authenticate) Additional modes could be added in the future (e.g. if we wanted to support TLS MtE for AES-CBC in the kernel we could add a new mode for that. TLS modes might also affect how AAD is interpreted, etc.) The flat structure also includes the key lengths and algorithms as before. However, code doesn't have to walk the linked list and switch on the algorithm to determine which key is the auth key vs encryption key. The 'csp_auth_*' fields are always used for auth keys and settings and 'csp_cipher_*' for cipher. (Compression algorithms are stored in csp_cipher_alg.) - Drivers no longer register a list of supported algorithms. This doesn't quite work when you factor in modes (e.g. a driver might support both AES-CBC and SHA2-256-HMAC separately but not combined for ETA). Instead, a new 'crypto_probesession' method has been added to the kobj interface for symmteric crypto drivers. This method returns a negative value on success (similar to how device_probe works) and the crypto framework uses this value to pick the "best" driver. There are three constants for hardware (e.g. ccr), accelerated software (e.g. aesni), and plain software (cryptosoft) that give preference in that order. One effect of this is that if you request only hardware when creating a new session, you will no longer get a session using accelerated software. Another effect is that the default setting to disallow software crypto via /dev/crypto now disables accelerated software. Once a driver is chosen, 'crypto_newsession' is invoked as before. - Crypto operations are now solely described by the flat 'cryptop' structure. The linked list of descriptors has been removed. A separate enum has been added to describe the type of data buffer in use instead of using CRYPTO_F_* flags to make it easier to add more types in the future if needed (e.g. wired userspace buffers for zero-copy). It will also make it easier to re-introduce separate input and output buffers (in-kernel TLS would benefit from this). Try to make the flags related to IV handling less insane: - CRYPTO_F_IV_SEPARATE means that the IV is stored in the 'crp_iv' member of the operation structure. If this flag is not set, the IV is stored in the data buffer at the 'crp_iv_start' offset. - CRYPTO_F_IV_GENERATE means that a random IV should be generated and stored into the data buffer. This cannot be used with CRYPTO_F_IV_SEPARATE. If a consumer wants to deal with explicit vs implicit IVs, etc. it can always generate the IV however it needs and store partial IVs in the buffer and the full IV/nonce in crp_iv and set CRYPTO_F_IV_SEPARATE. The layout of the buffer is now described via fields in cryptop. crp_aad_start and crp_aad_length define the boundaries of any AAD. Previously with GCM and CCM you defined an auth crd with this range, but for ETA your auth crd had to span both the AAD and plaintext (and they had to be adjacent). crp_payload_start and crp_payload_length define the boundaries of the plaintext/ciphertext. Modes that only do a single operation (COMPRESS, CIPHER, DIGEST) should only use this region and leave the AAD region empty. If a digest is present (or should be generated), it's starting location is marked by crp_digest_start. Instead of using the CRD_F_ENCRYPT flag to determine the direction of the operation, cryptop now includes an 'op' field defining the operation to perform. For digests I've added a new VERIFY digest mode which assumes a digest is present in the input and fails the request with EBADMSG if it doesn't match the internally-computed digest. GCM and CCM already assumed this, and the new AEAD mode requires this for decryption. The new ETA mode now also requires this for decryption, so IPsec and GELI no longer do their own authentication verification. Simple DIGEST operations can also do this, though there are no in-tree consumers. To eventually support some refcounting to close races, the session cookie is now passed to crypto_getop() and clients should no longer set crp_sesssion directly. - Assymteric crypto operation structures should be allocated via crypto_getkreq() and freed via crypto_freekreq(). This permits the crypto layer to track open asym requests and close races with a driver trying to unregister while asym requests are in flight. - crypto_copyback, crypto_copydata, crypto_apply, and crypto_contiguous_subsegment now accept the 'crp' object as the first parameter instead of individual members. This makes it easier to deal with different buffer types in the future as well as separate input and output buffers. It's also simpler for driver writers to use. - bus_dmamap_load_crp() loads a DMA mapping for a crypto buffer. This understands the various types of buffers so that drivers that use DMA do not have to be aware of different buffer types. - Helper routines now exist to build an auth context for HMAC IPAD and OPAD. This reduces some duplicated work among drivers. - Key buffers are now treated as const throughout the framework and in device drivers. However, session key buffers provided when a session is created are expected to remain alive for the duration of the session. - GCM and CCM sessions now only specify a cipher algorithm and a cipher key. The redundant auth information is not needed or used. - For cryptosoft, split up the code a bit such that the 'process' callback now invokes a function pointer in the session. This function pointer is set based on the mode (in effect) though it simplifies a few edge cases that would otherwise be in the switch in 'process'. It does split up GCM vs CCM which I think is more readable even if there is some duplication. - I changed /dev/crypto to support GMAC requests using CRYPTO_AES_NIST_GMAC as an auth algorithm and updated cryptocheck to work with it. - Combined cipher and auth sessions via /dev/crypto now always use ETA mode. The COP_F_CIPHER_FIRST flag is now a no-op that is ignored. This was actually documented as being true in crypto(4) before, but the code had not implemented this before I added the CIPHER_FIRST flag. - I have not yet updated /dev/crypto to be aware of explicit modes for sessions. I will probably do that at some point in the future as well as teach it about IV/nonce and tag lengths for AEAD so we can support all of the NIST KAT tests for GCM and CCM. - I've split up the exising crypto.9 manpage into several pages of which many are written from scratch. - I have converted all drivers and consumers in the tree and verified that they compile, but I have not tested all of them. I have tested the following drivers: - cryptosoft - aesni (AES only) - blake2 - ccr and the following consumers: - cryptodev - IPsec - ktls_ocf - GELI (lightly) I have not tested the following: - ccp - aesni with sha - hifn - kgssapi_krb5 - ubsec - padlock - safe - armv8_crypto (aarch64) - glxsb (i386) - sec (ppc) - cesa (armv7) - cryptocteon (mips64) - nlmsec (mips64) Discussed with: cem Relnotes: yes Sponsored by: Chelsio Communications Differential Revision: https://reviews.freebsd.org/D23677
2020-03-27 18:25:23 +00:00
crp->crp_iv_start = 0;
crp->crp_payload_length -= cse->ivsize;
if (crp->crp_payload_length != 0)
crp->crp_payload_start = cse->ivsize;
dst += cse->ivsize;
Refactor driver and consumer interfaces for OCF (in-kernel crypto). - The linked list of cryptoini structures used in session initialization is replaced with a new flat structure: struct crypto_session_params. This session includes a new mode to define how the other fields should be interpreted. Available modes include: - COMPRESS (for compression/decompression) - CIPHER (for simply encryption/decryption) - DIGEST (computing and verifying digests) - AEAD (combined auth and encryption such as AES-GCM and AES-CCM) - ETA (combined auth and encryption using encrypt-then-authenticate) Additional modes could be added in the future (e.g. if we wanted to support TLS MtE for AES-CBC in the kernel we could add a new mode for that. TLS modes might also affect how AAD is interpreted, etc.) The flat structure also includes the key lengths and algorithms as before. However, code doesn't have to walk the linked list and switch on the algorithm to determine which key is the auth key vs encryption key. The 'csp_auth_*' fields are always used for auth keys and settings and 'csp_cipher_*' for cipher. (Compression algorithms are stored in csp_cipher_alg.) - Drivers no longer register a list of supported algorithms. This doesn't quite work when you factor in modes (e.g. a driver might support both AES-CBC and SHA2-256-HMAC separately but not combined for ETA). Instead, a new 'crypto_probesession' method has been added to the kobj interface for symmteric crypto drivers. This method returns a negative value on success (similar to how device_probe works) and the crypto framework uses this value to pick the "best" driver. There are three constants for hardware (e.g. ccr), accelerated software (e.g. aesni), and plain software (cryptosoft) that give preference in that order. One effect of this is that if you request only hardware when creating a new session, you will no longer get a session using accelerated software. Another effect is that the default setting to disallow software crypto via /dev/crypto now disables accelerated software. Once a driver is chosen, 'crypto_newsession' is invoked as before. - Crypto operations are now solely described by the flat 'cryptop' structure. The linked list of descriptors has been removed. A separate enum has been added to describe the type of data buffer in use instead of using CRYPTO_F_* flags to make it easier to add more types in the future if needed (e.g. wired userspace buffers for zero-copy). It will also make it easier to re-introduce separate input and output buffers (in-kernel TLS would benefit from this). Try to make the flags related to IV handling less insane: - CRYPTO_F_IV_SEPARATE means that the IV is stored in the 'crp_iv' member of the operation structure. If this flag is not set, the IV is stored in the data buffer at the 'crp_iv_start' offset. - CRYPTO_F_IV_GENERATE means that a random IV should be generated and stored into the data buffer. This cannot be used with CRYPTO_F_IV_SEPARATE. If a consumer wants to deal with explicit vs implicit IVs, etc. it can always generate the IV however it needs and store partial IVs in the buffer and the full IV/nonce in crp_iv and set CRYPTO_F_IV_SEPARATE. The layout of the buffer is now described via fields in cryptop. crp_aad_start and crp_aad_length define the boundaries of any AAD. Previously with GCM and CCM you defined an auth crd with this range, but for ETA your auth crd had to span both the AAD and plaintext (and they had to be adjacent). crp_payload_start and crp_payload_length define the boundaries of the plaintext/ciphertext. Modes that only do a single operation (COMPRESS, CIPHER, DIGEST) should only use this region and leave the AAD region empty. If a digest is present (or should be generated), it's starting location is marked by crp_digest_start. Instead of using the CRD_F_ENCRYPT flag to determine the direction of the operation, cryptop now includes an 'op' field defining the operation to perform. For digests I've added a new VERIFY digest mode which assumes a digest is present in the input and fails the request with EBADMSG if it doesn't match the internally-computed digest. GCM and CCM already assumed this, and the new AEAD mode requires this for decryption. The new ETA mode now also requires this for decryption, so IPsec and GELI no longer do their own authentication verification. Simple DIGEST operations can also do this, though there are no in-tree consumers. To eventually support some refcounting to close races, the session cookie is now passed to crypto_getop() and clients should no longer set crp_sesssion directly. - Assymteric crypto operation structures should be allocated via crypto_getkreq() and freed via crypto_freekreq(). This permits the crypto layer to track open asym requests and close races with a driver trying to unregister while asym requests are in flight. - crypto_copyback, crypto_copydata, crypto_apply, and crypto_contiguous_subsegment now accept the 'crp' object as the first parameter instead of individual members. This makes it easier to deal with different buffer types in the future as well as separate input and output buffers. It's also simpler for driver writers to use. - bus_dmamap_load_crp() loads a DMA mapping for a crypto buffer. This understands the various types of buffers so that drivers that use DMA do not have to be aware of different buffer types. - Helper routines now exist to build an auth context for HMAC IPAD and OPAD. This reduces some duplicated work among drivers. - Key buffers are now treated as const throughout the framework and in device drivers. However, session key buffers provided when a session is created are expected to remain alive for the duration of the session. - GCM and CCM sessions now only specify a cipher algorithm and a cipher key. The redundant auth information is not needed or used. - For cryptosoft, split up the code a bit such that the 'process' callback now invokes a function pointer in the session. This function pointer is set based on the mode (in effect) though it simplifies a few edge cases that would otherwise be in the switch in 'process'. It does split up GCM vs CCM which I think is more readable even if there is some duplication. - I changed /dev/crypto to support GMAC requests using CRYPTO_AES_NIST_GMAC as an auth algorithm and updated cryptocheck to work with it. - Combined cipher and auth sessions via /dev/crypto now always use ETA mode. The COP_F_CIPHER_FIRST flag is now a no-op that is ignored. This was actually documented as being true in crypto(4) before, but the code had not implemented this before I added the CIPHER_FIRST flag. - I have not yet updated /dev/crypto to be aware of explicit modes for sessions. I will probably do that at some point in the future as well as teach it about IV/nonce and tag lengths for AEAD so we can support all of the NIST KAT tests for GCM and CCM. - I've split up the exising crypto.9 manpage into several pages of which many are written from scratch. - I have converted all drivers and consumers in the tree and verified that they compile, but I have not tested all of them. I have tested the following drivers: - cryptosoft - aesni (AES only) - blake2 - ccr and the following consumers: - cryptodev - IPsec - ktls_ocf - GELI (lightly) I have not tested the following: - ccp - aesni with sha - hifn - kgssapi_krb5 - ubsec - padlock - safe - armv8_crypto (aarch64) - glxsb (i386) - sec (ppc) - cesa (armv7) - cryptocteon (mips64) - nlmsec (mips64) Discussed with: cem Relnotes: yes Sponsored by: Chelsio Communications Differential Revision: https://reviews.freebsd.org/D23677
2020-03-27 18:25:23 +00:00
}
if (crp->crp_op & CRYPTO_OP_VERIFY_DIGEST) {
error = copyin(cop->mac, cod->buf + crp->crp_digest_start,
cse->hashsize);
Refactor driver and consumer interfaces for OCF (in-kernel crypto). - The linked list of cryptoini structures used in session initialization is replaced with a new flat structure: struct crypto_session_params. This session includes a new mode to define how the other fields should be interpreted. Available modes include: - COMPRESS (for compression/decompression) - CIPHER (for simply encryption/decryption) - DIGEST (computing and verifying digests) - AEAD (combined auth and encryption such as AES-GCM and AES-CCM) - ETA (combined auth and encryption using encrypt-then-authenticate) Additional modes could be added in the future (e.g. if we wanted to support TLS MtE for AES-CBC in the kernel we could add a new mode for that. TLS modes might also affect how AAD is interpreted, etc.) The flat structure also includes the key lengths and algorithms as before. However, code doesn't have to walk the linked list and switch on the algorithm to determine which key is the auth key vs encryption key. The 'csp_auth_*' fields are always used for auth keys and settings and 'csp_cipher_*' for cipher. (Compression algorithms are stored in csp_cipher_alg.) - Drivers no longer register a list of supported algorithms. This doesn't quite work when you factor in modes (e.g. a driver might support both AES-CBC and SHA2-256-HMAC separately but not combined for ETA). Instead, a new 'crypto_probesession' method has been added to the kobj interface for symmteric crypto drivers. This method returns a negative value on success (similar to how device_probe works) and the crypto framework uses this value to pick the "best" driver. There are three constants for hardware (e.g. ccr), accelerated software (e.g. aesni), and plain software (cryptosoft) that give preference in that order. One effect of this is that if you request only hardware when creating a new session, you will no longer get a session using accelerated software. Another effect is that the default setting to disallow software crypto via /dev/crypto now disables accelerated software. Once a driver is chosen, 'crypto_newsession' is invoked as before. - Crypto operations are now solely described by the flat 'cryptop' structure. The linked list of descriptors has been removed. A separate enum has been added to describe the type of data buffer in use instead of using CRYPTO_F_* flags to make it easier to add more types in the future if needed (e.g. wired userspace buffers for zero-copy). It will also make it easier to re-introduce separate input and output buffers (in-kernel TLS would benefit from this). Try to make the flags related to IV handling less insane: - CRYPTO_F_IV_SEPARATE means that the IV is stored in the 'crp_iv' member of the operation structure. If this flag is not set, the IV is stored in the data buffer at the 'crp_iv_start' offset. - CRYPTO_F_IV_GENERATE means that a random IV should be generated and stored into the data buffer. This cannot be used with CRYPTO_F_IV_SEPARATE. If a consumer wants to deal with explicit vs implicit IVs, etc. it can always generate the IV however it needs and store partial IVs in the buffer and the full IV/nonce in crp_iv and set CRYPTO_F_IV_SEPARATE. The layout of the buffer is now described via fields in cryptop. crp_aad_start and crp_aad_length define the boundaries of any AAD. Previously with GCM and CCM you defined an auth crd with this range, but for ETA your auth crd had to span both the AAD and plaintext (and they had to be adjacent). crp_payload_start and crp_payload_length define the boundaries of the plaintext/ciphertext. Modes that only do a single operation (COMPRESS, CIPHER, DIGEST) should only use this region and leave the AAD region empty. If a digest is present (or should be generated), it's starting location is marked by crp_digest_start. Instead of using the CRD_F_ENCRYPT flag to determine the direction of the operation, cryptop now includes an 'op' field defining the operation to perform. For digests I've added a new VERIFY digest mode which assumes a digest is present in the input and fails the request with EBADMSG if it doesn't match the internally-computed digest. GCM and CCM already assumed this, and the new AEAD mode requires this for decryption. The new ETA mode now also requires this for decryption, so IPsec and GELI no longer do their own authentication verification. Simple DIGEST operations can also do this, though there are no in-tree consumers. To eventually support some refcounting to close races, the session cookie is now passed to crypto_getop() and clients should no longer set crp_sesssion directly. - Assymteric crypto operation structures should be allocated via crypto_getkreq() and freed via crypto_freekreq(). This permits the crypto layer to track open asym requests and close races with a driver trying to unregister while asym requests are in flight. - crypto_copyback, crypto_copydata, crypto_apply, and crypto_contiguous_subsegment now accept the 'crp' object as the first parameter instead of individual members. This makes it easier to deal with different buffer types in the future as well as separate input and output buffers. It's also simpler for driver writers to use. - bus_dmamap_load_crp() loads a DMA mapping for a crypto buffer. This understands the various types of buffers so that drivers that use DMA do not have to be aware of different buffer types. - Helper routines now exist to build an auth context for HMAC IPAD and OPAD. This reduces some duplicated work among drivers. - Key buffers are now treated as const throughout the framework and in device drivers. However, session key buffers provided when a session is created are expected to remain alive for the duration of the session. - GCM and CCM sessions now only specify a cipher algorithm and a cipher key. The redundant auth information is not needed or used. - For cryptosoft, split up the code a bit such that the 'process' callback now invokes a function pointer in the session. This function pointer is set based on the mode (in effect) though it simplifies a few edge cases that would otherwise be in the switch in 'process'. It does split up GCM vs CCM which I think is more readable even if there is some duplication. - I changed /dev/crypto to support GMAC requests using CRYPTO_AES_NIST_GMAC as an auth algorithm and updated cryptocheck to work with it. - Combined cipher and auth sessions via /dev/crypto now always use ETA mode. The COP_F_CIPHER_FIRST flag is now a no-op that is ignored. This was actually documented as being true in crypto(4) before, but the code had not implemented this before I added the CIPHER_FIRST flag. - I have not yet updated /dev/crypto to be aware of explicit modes for sessions. I will probably do that at some point in the future as well as teach it about IV/nonce and tag lengths for AEAD so we can support all of the NIST KAT tests for GCM and CCM. - I've split up the exising crypto.9 manpage into several pages of which many are written from scratch. - I have converted all drivers and consumers in the tree and verified that they compile, but I have not tested all of them. I have tested the following drivers: - cryptosoft - aesni (AES only) - blake2 - ccr and the following consumers: - cryptodev - IPsec - ktls_ocf - GELI (lightly) I have not tested the following: - ccp - aesni with sha - hifn - kgssapi_krb5 - ubsec - padlock - safe - armv8_crypto (aarch64) - glxsb (i386) - sec (ppc) - cesa (armv7) - cryptocteon (mips64) - nlmsec (mips64) Discussed with: cem Relnotes: yes Sponsored by: Chelsio Communications Differential Revision: https://reviews.freebsd.org/D23677
2020-03-27 18:25:23 +00:00
if (error) {
2014-12-12 19:56:36 +00:00
SDT_PROBE1(opencrypto, dev, ioctl, error, __LINE__);
goto bail;
2014-12-12 19:56:36 +00:00
}
}
again:
/*
* Let the dispatch run unlocked, then, interlock against the
* callback before checking if the operation completed and going
* to sleep. This insures drivers don't inherit our lock which
* results in a lock order reversal between crypto_dispatch forced
* entry and the crypto_done callback into us.
*/
error = crypto_dispatch(crp);
2014-12-12 19:56:36 +00:00
if (error != 0) {
SDT_PROBE1(opencrypto, dev, ioctl, error, __LINE__);
goto bail;
2014-12-12 19:56:36 +00:00
}
mtx_lock(&cse->lock);
while (!cod->done)
mtx_sleep(cod, &cse->lock, PWAIT, "crydev", 0);
mtx_unlock(&cse->lock);
if (crp->crp_etype == EAGAIN) {
crp->crp_etype = 0;
crp->crp_flags &= ~CRYPTO_F_DONE;
cod->done = false;
goto again;
}
if (crp->crp_etype != 0) {
2014-12-12 19:56:36 +00:00
SDT_PROBE1(opencrypto, dev, ioctl, error, __LINE__);
error = crp->crp_etype;
goto bail;
}
Refactor driver and consumer interfaces for OCF (in-kernel crypto). - The linked list of cryptoini structures used in session initialization is replaced with a new flat structure: struct crypto_session_params. This session includes a new mode to define how the other fields should be interpreted. Available modes include: - COMPRESS (for compression/decompression) - CIPHER (for simply encryption/decryption) - DIGEST (computing and verifying digests) - AEAD (combined auth and encryption such as AES-GCM and AES-CCM) - ETA (combined auth and encryption using encrypt-then-authenticate) Additional modes could be added in the future (e.g. if we wanted to support TLS MtE for AES-CBC in the kernel we could add a new mode for that. TLS modes might also affect how AAD is interpreted, etc.) The flat structure also includes the key lengths and algorithms as before. However, code doesn't have to walk the linked list and switch on the algorithm to determine which key is the auth key vs encryption key. The 'csp_auth_*' fields are always used for auth keys and settings and 'csp_cipher_*' for cipher. (Compression algorithms are stored in csp_cipher_alg.) - Drivers no longer register a list of supported algorithms. This doesn't quite work when you factor in modes (e.g. a driver might support both AES-CBC and SHA2-256-HMAC separately but not combined for ETA). Instead, a new 'crypto_probesession' method has been added to the kobj interface for symmteric crypto drivers. This method returns a negative value on success (similar to how device_probe works) and the crypto framework uses this value to pick the "best" driver. There are three constants for hardware (e.g. ccr), accelerated software (e.g. aesni), and plain software (cryptosoft) that give preference in that order. One effect of this is that if you request only hardware when creating a new session, you will no longer get a session using accelerated software. Another effect is that the default setting to disallow software crypto via /dev/crypto now disables accelerated software. Once a driver is chosen, 'crypto_newsession' is invoked as before. - Crypto operations are now solely described by the flat 'cryptop' structure. The linked list of descriptors has been removed. A separate enum has been added to describe the type of data buffer in use instead of using CRYPTO_F_* flags to make it easier to add more types in the future if needed (e.g. wired userspace buffers for zero-copy). It will also make it easier to re-introduce separate input and output buffers (in-kernel TLS would benefit from this). Try to make the flags related to IV handling less insane: - CRYPTO_F_IV_SEPARATE means that the IV is stored in the 'crp_iv' member of the operation structure. If this flag is not set, the IV is stored in the data buffer at the 'crp_iv_start' offset. - CRYPTO_F_IV_GENERATE means that a random IV should be generated and stored into the data buffer. This cannot be used with CRYPTO_F_IV_SEPARATE. If a consumer wants to deal with explicit vs implicit IVs, etc. it can always generate the IV however it needs and store partial IVs in the buffer and the full IV/nonce in crp_iv and set CRYPTO_F_IV_SEPARATE. The layout of the buffer is now described via fields in cryptop. crp_aad_start and crp_aad_length define the boundaries of any AAD. Previously with GCM and CCM you defined an auth crd with this range, but for ETA your auth crd had to span both the AAD and plaintext (and they had to be adjacent). crp_payload_start and crp_payload_length define the boundaries of the plaintext/ciphertext. Modes that only do a single operation (COMPRESS, CIPHER, DIGEST) should only use this region and leave the AAD region empty. If a digest is present (or should be generated), it's starting location is marked by crp_digest_start. Instead of using the CRD_F_ENCRYPT flag to determine the direction of the operation, cryptop now includes an 'op' field defining the operation to perform. For digests I've added a new VERIFY digest mode which assumes a digest is present in the input and fails the request with EBADMSG if it doesn't match the internally-computed digest. GCM and CCM already assumed this, and the new AEAD mode requires this for decryption. The new ETA mode now also requires this for decryption, so IPsec and GELI no longer do their own authentication verification. Simple DIGEST operations can also do this, though there are no in-tree consumers. To eventually support some refcounting to close races, the session cookie is now passed to crypto_getop() and clients should no longer set crp_sesssion directly. - Assymteric crypto operation structures should be allocated via crypto_getkreq() and freed via crypto_freekreq(). This permits the crypto layer to track open asym requests and close races with a driver trying to unregister while asym requests are in flight. - crypto_copyback, crypto_copydata, crypto_apply, and crypto_contiguous_subsegment now accept the 'crp' object as the first parameter instead of individual members. This makes it easier to deal with different buffer types in the future as well as separate input and output buffers. It's also simpler for driver writers to use. - bus_dmamap_load_crp() loads a DMA mapping for a crypto buffer. This understands the various types of buffers so that drivers that use DMA do not have to be aware of different buffer types. - Helper routines now exist to build an auth context for HMAC IPAD and OPAD. This reduces some duplicated work among drivers. - Key buffers are now treated as const throughout the framework and in device drivers. However, session key buffers provided when a session is created are expected to remain alive for the duration of the session. - GCM and CCM sessions now only specify a cipher algorithm and a cipher key. The redundant auth information is not needed or used. - For cryptosoft, split up the code a bit such that the 'process' callback now invokes a function pointer in the session. This function pointer is set based on the mode (in effect) though it simplifies a few edge cases that would otherwise be in the switch in 'process'. It does split up GCM vs CCM which I think is more readable even if there is some duplication. - I changed /dev/crypto to support GMAC requests using CRYPTO_AES_NIST_GMAC as an auth algorithm and updated cryptocheck to work with it. - Combined cipher and auth sessions via /dev/crypto now always use ETA mode. The COP_F_CIPHER_FIRST flag is now a no-op that is ignored. This was actually documented as being true in crypto(4) before, but the code had not implemented this before I added the CIPHER_FIRST flag. - I have not yet updated /dev/crypto to be aware of explicit modes for sessions. I will probably do that at some point in the future as well as teach it about IV/nonce and tag lengths for AEAD so we can support all of the NIST KAT tests for GCM and CCM. - I've split up the exising crypto.9 manpage into several pages of which many are written from scratch. - I have converted all drivers and consumers in the tree and verified that they compile, but I have not tested all of them. I have tested the following drivers: - cryptosoft - aesni (AES only) - blake2 - ccr and the following consumers: - cryptodev - IPsec - ktls_ocf - GELI (lightly) I have not tested the following: - ccp - aesni with sha - hifn - kgssapi_krb5 - ubsec - padlock - safe - armv8_crypto (aarch64) - glxsb (i386) - sec (ppc) - cesa (armv7) - cryptocteon (mips64) - nlmsec (mips64) Discussed with: cem Relnotes: yes Sponsored by: Chelsio Communications Differential Revision: https://reviews.freebsd.org/D23677
2020-03-27 18:25:23 +00:00
if (cop->dst != NULL) {
error = copyout(cod->obuf != NULL ? cod->obuf :
cod->buf + crp->crp_payload_start, dst,
crp->crp_payload_length);
Refactor driver and consumer interfaces for OCF (in-kernel crypto). - The linked list of cryptoini structures used in session initialization is replaced with a new flat structure: struct crypto_session_params. This session includes a new mode to define how the other fields should be interpreted. Available modes include: - COMPRESS (for compression/decompression) - CIPHER (for simply encryption/decryption) - DIGEST (computing and verifying digests) - AEAD (combined auth and encryption such as AES-GCM and AES-CCM) - ETA (combined auth and encryption using encrypt-then-authenticate) Additional modes could be added in the future (e.g. if we wanted to support TLS MtE for AES-CBC in the kernel we could add a new mode for that. TLS modes might also affect how AAD is interpreted, etc.) The flat structure also includes the key lengths and algorithms as before. However, code doesn't have to walk the linked list and switch on the algorithm to determine which key is the auth key vs encryption key. The 'csp_auth_*' fields are always used for auth keys and settings and 'csp_cipher_*' for cipher. (Compression algorithms are stored in csp_cipher_alg.) - Drivers no longer register a list of supported algorithms. This doesn't quite work when you factor in modes (e.g. a driver might support both AES-CBC and SHA2-256-HMAC separately but not combined for ETA). Instead, a new 'crypto_probesession' method has been added to the kobj interface for symmteric crypto drivers. This method returns a negative value on success (similar to how device_probe works) and the crypto framework uses this value to pick the "best" driver. There are three constants for hardware (e.g. ccr), accelerated software (e.g. aesni), and plain software (cryptosoft) that give preference in that order. One effect of this is that if you request only hardware when creating a new session, you will no longer get a session using accelerated software. Another effect is that the default setting to disallow software crypto via /dev/crypto now disables accelerated software. Once a driver is chosen, 'crypto_newsession' is invoked as before. - Crypto operations are now solely described by the flat 'cryptop' structure. The linked list of descriptors has been removed. A separate enum has been added to describe the type of data buffer in use instead of using CRYPTO_F_* flags to make it easier to add more types in the future if needed (e.g. wired userspace buffers for zero-copy). It will also make it easier to re-introduce separate input and output buffers (in-kernel TLS would benefit from this). Try to make the flags related to IV handling less insane: - CRYPTO_F_IV_SEPARATE means that the IV is stored in the 'crp_iv' member of the operation structure. If this flag is not set, the IV is stored in the data buffer at the 'crp_iv_start' offset. - CRYPTO_F_IV_GENERATE means that a random IV should be generated and stored into the data buffer. This cannot be used with CRYPTO_F_IV_SEPARATE. If a consumer wants to deal with explicit vs implicit IVs, etc. it can always generate the IV however it needs and store partial IVs in the buffer and the full IV/nonce in crp_iv and set CRYPTO_F_IV_SEPARATE. The layout of the buffer is now described via fields in cryptop. crp_aad_start and crp_aad_length define the boundaries of any AAD. Previously with GCM and CCM you defined an auth crd with this range, but for ETA your auth crd had to span both the AAD and plaintext (and they had to be adjacent). crp_payload_start and crp_payload_length define the boundaries of the plaintext/ciphertext. Modes that only do a single operation (COMPRESS, CIPHER, DIGEST) should only use this region and leave the AAD region empty. If a digest is present (or should be generated), it's starting location is marked by crp_digest_start. Instead of using the CRD_F_ENCRYPT flag to determine the direction of the operation, cryptop now includes an 'op' field defining the operation to perform. For digests I've added a new VERIFY digest mode which assumes a digest is present in the input and fails the request with EBADMSG if it doesn't match the internally-computed digest. GCM and CCM already assumed this, and the new AEAD mode requires this for decryption. The new ETA mode now also requires this for decryption, so IPsec and GELI no longer do their own authentication verification. Simple DIGEST operations can also do this, though there are no in-tree consumers. To eventually support some refcounting to close races, the session cookie is now passed to crypto_getop() and clients should no longer set crp_sesssion directly. - Assymteric crypto operation structures should be allocated via crypto_getkreq() and freed via crypto_freekreq(). This permits the crypto layer to track open asym requests and close races with a driver trying to unregister while asym requests are in flight. - crypto_copyback, crypto_copydata, crypto_apply, and crypto_contiguous_subsegment now accept the 'crp' object as the first parameter instead of individual members. This makes it easier to deal with different buffer types in the future as well as separate input and output buffers. It's also simpler for driver writers to use. - bus_dmamap_load_crp() loads a DMA mapping for a crypto buffer. This understands the various types of buffers so that drivers that use DMA do not have to be aware of different buffer types. - Helper routines now exist to build an auth context for HMAC IPAD and OPAD. This reduces some duplicated work among drivers. - Key buffers are now treated as const throughout the framework and in device drivers. However, session key buffers provided when a session is created are expected to remain alive for the duration of the session. - GCM and CCM sessions now only specify a cipher algorithm and a cipher key. The redundant auth information is not needed or used. - For cryptosoft, split up the code a bit such that the 'process' callback now invokes a function pointer in the session. This function pointer is set based on the mode (in effect) though it simplifies a few edge cases that would otherwise be in the switch in 'process'. It does split up GCM vs CCM which I think is more readable even if there is some duplication. - I changed /dev/crypto to support GMAC requests using CRYPTO_AES_NIST_GMAC as an auth algorithm and updated cryptocheck to work with it. - Combined cipher and auth sessions via /dev/crypto now always use ETA mode. The COP_F_CIPHER_FIRST flag is now a no-op that is ignored. This was actually documented as being true in crypto(4) before, but the code had not implemented this before I added the CIPHER_FIRST flag. - I have not yet updated /dev/crypto to be aware of explicit modes for sessions. I will probably do that at some point in the future as well as teach it about IV/nonce and tag lengths for AEAD so we can support all of the NIST KAT tests for GCM and CCM. - I've split up the exising crypto.9 manpage into several pages of which many are written from scratch. - I have converted all drivers and consumers in the tree and verified that they compile, but I have not tested all of them. I have tested the following drivers: - cryptosoft - aesni (AES only) - blake2 - ccr and the following consumers: - cryptodev - IPsec - ktls_ocf - GELI (lightly) I have not tested the following: - ccp - aesni with sha - hifn - kgssapi_krb5 - ubsec - padlock - safe - armv8_crypto (aarch64) - glxsb (i386) - sec (ppc) - cesa (armv7) - cryptocteon (mips64) - nlmsec (mips64) Discussed with: cem Relnotes: yes Sponsored by: Chelsio Communications Differential Revision: https://reviews.freebsd.org/D23677
2020-03-27 18:25:23 +00:00
if (error) {
SDT_PROBE1(opencrypto, dev, ioctl, error, __LINE__);
goto bail;
}
2014-12-12 19:56:36 +00:00
}
if (cop->mac != NULL && (crp->crp_op & CRYPTO_OP_VERIFY_DIGEST) == 0) {
error = copyout((cod->obuf != NULL ? cod->obuf : cod->buf) +
crp->crp_digest_start, cop->mac, cse->hashsize);
Refactor driver and consumer interfaces for OCF (in-kernel crypto). - The linked list of cryptoini structures used in session initialization is replaced with a new flat structure: struct crypto_session_params. This session includes a new mode to define how the other fields should be interpreted. Available modes include: - COMPRESS (for compression/decompression) - CIPHER (for simply encryption/decryption) - DIGEST (computing and verifying digests) - AEAD (combined auth and encryption such as AES-GCM and AES-CCM) - ETA (combined auth and encryption using encrypt-then-authenticate) Additional modes could be added in the future (e.g. if we wanted to support TLS MtE for AES-CBC in the kernel we could add a new mode for that. TLS modes might also affect how AAD is interpreted, etc.) The flat structure also includes the key lengths and algorithms as before. However, code doesn't have to walk the linked list and switch on the algorithm to determine which key is the auth key vs encryption key. The 'csp_auth_*' fields are always used for auth keys and settings and 'csp_cipher_*' for cipher. (Compression algorithms are stored in csp_cipher_alg.) - Drivers no longer register a list of supported algorithms. This doesn't quite work when you factor in modes (e.g. a driver might support both AES-CBC and SHA2-256-HMAC separately but not combined for ETA). Instead, a new 'crypto_probesession' method has been added to the kobj interface for symmteric crypto drivers. This method returns a negative value on success (similar to how device_probe works) and the crypto framework uses this value to pick the "best" driver. There are three constants for hardware (e.g. ccr), accelerated software (e.g. aesni), and plain software (cryptosoft) that give preference in that order. One effect of this is that if you request only hardware when creating a new session, you will no longer get a session using accelerated software. Another effect is that the default setting to disallow software crypto via /dev/crypto now disables accelerated software. Once a driver is chosen, 'crypto_newsession' is invoked as before. - Crypto operations are now solely described by the flat 'cryptop' structure. The linked list of descriptors has been removed. A separate enum has been added to describe the type of data buffer in use instead of using CRYPTO_F_* flags to make it easier to add more types in the future if needed (e.g. wired userspace buffers for zero-copy). It will also make it easier to re-introduce separate input and output buffers (in-kernel TLS would benefit from this). Try to make the flags related to IV handling less insane: - CRYPTO_F_IV_SEPARATE means that the IV is stored in the 'crp_iv' member of the operation structure. If this flag is not set, the IV is stored in the data buffer at the 'crp_iv_start' offset. - CRYPTO_F_IV_GENERATE means that a random IV should be generated and stored into the data buffer. This cannot be used with CRYPTO_F_IV_SEPARATE. If a consumer wants to deal with explicit vs implicit IVs, etc. it can always generate the IV however it needs and store partial IVs in the buffer and the full IV/nonce in crp_iv and set CRYPTO_F_IV_SEPARATE. The layout of the buffer is now described via fields in cryptop. crp_aad_start and crp_aad_length define the boundaries of any AAD. Previously with GCM and CCM you defined an auth crd with this range, but for ETA your auth crd had to span both the AAD and plaintext (and they had to be adjacent). crp_payload_start and crp_payload_length define the boundaries of the plaintext/ciphertext. Modes that only do a single operation (COMPRESS, CIPHER, DIGEST) should only use this region and leave the AAD region empty. If a digest is present (or should be generated), it's starting location is marked by crp_digest_start. Instead of using the CRD_F_ENCRYPT flag to determine the direction of the operation, cryptop now includes an 'op' field defining the operation to perform. For digests I've added a new VERIFY digest mode which assumes a digest is present in the input and fails the request with EBADMSG if it doesn't match the internally-computed digest. GCM and CCM already assumed this, and the new AEAD mode requires this for decryption. The new ETA mode now also requires this for decryption, so IPsec and GELI no longer do their own authentication verification. Simple DIGEST operations can also do this, though there are no in-tree consumers. To eventually support some refcounting to close races, the session cookie is now passed to crypto_getop() and clients should no longer set crp_sesssion directly. - Assymteric crypto operation structures should be allocated via crypto_getkreq() and freed via crypto_freekreq(). This permits the crypto layer to track open asym requests and close races with a driver trying to unregister while asym requests are in flight. - crypto_copyback, crypto_copydata, crypto_apply, and crypto_contiguous_subsegment now accept the 'crp' object as the first parameter instead of individual members. This makes it easier to deal with different buffer types in the future as well as separate input and output buffers. It's also simpler for driver writers to use. - bus_dmamap_load_crp() loads a DMA mapping for a crypto buffer. This understands the various types of buffers so that drivers that use DMA do not have to be aware of different buffer types. - Helper routines now exist to build an auth context for HMAC IPAD and OPAD. This reduces some duplicated work among drivers. - Key buffers are now treated as const throughout the framework and in device drivers. However, session key buffers provided when a session is created are expected to remain alive for the duration of the session. - GCM and CCM sessions now only specify a cipher algorithm and a cipher key. The redundant auth information is not needed or used. - For cryptosoft, split up the code a bit such that the 'process' callback now invokes a function pointer in the session. This function pointer is set based on the mode (in effect) though it simplifies a few edge cases that would otherwise be in the switch in 'process'. It does split up GCM vs CCM which I think is more readable even if there is some duplication. - I changed /dev/crypto to support GMAC requests using CRYPTO_AES_NIST_GMAC as an auth algorithm and updated cryptocheck to work with it. - Combined cipher and auth sessions via /dev/crypto now always use ETA mode. The COP_F_CIPHER_FIRST flag is now a no-op that is ignored. This was actually documented as being true in crypto(4) before, but the code had not implemented this before I added the CIPHER_FIRST flag. - I have not yet updated /dev/crypto to be aware of explicit modes for sessions. I will probably do that at some point in the future as well as teach it about IV/nonce and tag lengths for AEAD so we can support all of the NIST KAT tests for GCM and CCM. - I've split up the exising crypto.9 manpage into several pages of which many are written from scratch. - I have converted all drivers and consumers in the tree and verified that they compile, but I have not tested all of them. I have tested the following drivers: - cryptosoft - aesni (AES only) - blake2 - ccr and the following consumers: - cryptodev - IPsec - ktls_ocf - GELI (lightly) I have not tested the following: - ccp - aesni with sha - hifn - kgssapi_krb5 - ubsec - padlock - safe - armv8_crypto (aarch64) - glxsb (i386) - sec (ppc) - cesa (armv7) - cryptocteon (mips64) - nlmsec (mips64) Discussed with: cem Relnotes: yes Sponsored by: Chelsio Communications Differential Revision: https://reviews.freebsd.org/D23677
2020-03-27 18:25:23 +00:00
if (error) {
SDT_PROBE1(opencrypto, dev, ioctl, error, __LINE__);
goto bail;
}
2014-12-12 19:56:36 +00:00
}
bail:
crypto_freereq(crp);
cod_free(cod);
return (error);
}
2014-12-12 19:56:36 +00:00
static int
cryptodev_aead(struct csession *cse, struct crypt_aead *caead)
2014-12-12 19:56:36 +00:00
{
const struct crypto_session_params *csp;
struct cryptop_data *cod = NULL;
2014-12-12 19:56:36 +00:00
struct cryptop *crp = NULL;
char *dst;
2014-12-12 19:56:36 +00:00
int error;
if (caead->len > 256*1024-4 || caead->aadlen > 256*1024-4) {
SDT_PROBE1(opencrypto, dev, ioctl, error, __LINE__);
2014-12-12 19:56:36 +00:00
return (E2BIG);
}
2014-12-12 19:56:36 +00:00
if ((caead->len % cse->blocksize) != 0) {
SDT_PROBE1(opencrypto, dev, ioctl, error, __LINE__);
return (EINVAL);
}
if (cse->hashsize == 0 || caead->tag == NULL) {
SDT_PROBE1(opencrypto, dev, ioctl, error, __LINE__);
2014-12-12 19:56:36 +00:00
return (EINVAL);
}
2014-12-12 19:56:36 +00:00
Refactor driver and consumer interfaces for OCF (in-kernel crypto). - The linked list of cryptoini structures used in session initialization is replaced with a new flat structure: struct crypto_session_params. This session includes a new mode to define how the other fields should be interpreted. Available modes include: - COMPRESS (for compression/decompression) - CIPHER (for simply encryption/decryption) - DIGEST (computing and verifying digests) - AEAD (combined auth and encryption such as AES-GCM and AES-CCM) - ETA (combined auth and encryption using encrypt-then-authenticate) Additional modes could be added in the future (e.g. if we wanted to support TLS MtE for AES-CBC in the kernel we could add a new mode for that. TLS modes might also affect how AAD is interpreted, etc.) The flat structure also includes the key lengths and algorithms as before. However, code doesn't have to walk the linked list and switch on the algorithm to determine which key is the auth key vs encryption key. The 'csp_auth_*' fields are always used for auth keys and settings and 'csp_cipher_*' for cipher. (Compression algorithms are stored in csp_cipher_alg.) - Drivers no longer register a list of supported algorithms. This doesn't quite work when you factor in modes (e.g. a driver might support both AES-CBC and SHA2-256-HMAC separately but not combined for ETA). Instead, a new 'crypto_probesession' method has been added to the kobj interface for symmteric crypto drivers. This method returns a negative value on success (similar to how device_probe works) and the crypto framework uses this value to pick the "best" driver. There are three constants for hardware (e.g. ccr), accelerated software (e.g. aesni), and plain software (cryptosoft) that give preference in that order. One effect of this is that if you request only hardware when creating a new session, you will no longer get a session using accelerated software. Another effect is that the default setting to disallow software crypto via /dev/crypto now disables accelerated software. Once a driver is chosen, 'crypto_newsession' is invoked as before. - Crypto operations are now solely described by the flat 'cryptop' structure. The linked list of descriptors has been removed. A separate enum has been added to describe the type of data buffer in use instead of using CRYPTO_F_* flags to make it easier to add more types in the future if needed (e.g. wired userspace buffers for zero-copy). It will also make it easier to re-introduce separate input and output buffers (in-kernel TLS would benefit from this). Try to make the flags related to IV handling less insane: - CRYPTO_F_IV_SEPARATE means that the IV is stored in the 'crp_iv' member of the operation structure. If this flag is not set, the IV is stored in the data buffer at the 'crp_iv_start' offset. - CRYPTO_F_IV_GENERATE means that a random IV should be generated and stored into the data buffer. This cannot be used with CRYPTO_F_IV_SEPARATE. If a consumer wants to deal with explicit vs implicit IVs, etc. it can always generate the IV however it needs and store partial IVs in the buffer and the full IV/nonce in crp_iv and set CRYPTO_F_IV_SEPARATE. The layout of the buffer is now described via fields in cryptop. crp_aad_start and crp_aad_length define the boundaries of any AAD. Previously with GCM and CCM you defined an auth crd with this range, but for ETA your auth crd had to span both the AAD and plaintext (and they had to be adjacent). crp_payload_start and crp_payload_length define the boundaries of the plaintext/ciphertext. Modes that only do a single operation (COMPRESS, CIPHER, DIGEST) should only use this region and leave the AAD region empty. If a digest is present (or should be generated), it's starting location is marked by crp_digest_start. Instead of using the CRD_F_ENCRYPT flag to determine the direction of the operation, cryptop now includes an 'op' field defining the operation to perform. For digests I've added a new VERIFY digest mode which assumes a digest is present in the input and fails the request with EBADMSG if it doesn't match the internally-computed digest. GCM and CCM already assumed this, and the new AEAD mode requires this for decryption. The new ETA mode now also requires this for decryption, so IPsec and GELI no longer do their own authentication verification. Simple DIGEST operations can also do this, though there are no in-tree consumers. To eventually support some refcounting to close races, the session cookie is now passed to crypto_getop() and clients should no longer set crp_sesssion directly. - Assymteric crypto operation structures should be allocated via crypto_getkreq() and freed via crypto_freekreq(). This permits the crypto layer to track open asym requests and close races with a driver trying to unregister while asym requests are in flight. - crypto_copyback, crypto_copydata, crypto_apply, and crypto_contiguous_subsegment now accept the 'crp' object as the first parameter instead of individual members. This makes it easier to deal with different buffer types in the future as well as separate input and output buffers. It's also simpler for driver writers to use. - bus_dmamap_load_crp() loads a DMA mapping for a crypto buffer. This understands the various types of buffers so that drivers that use DMA do not have to be aware of different buffer types. - Helper routines now exist to build an auth context for HMAC IPAD and OPAD. This reduces some duplicated work among drivers. - Key buffers are now treated as const throughout the framework and in device drivers. However, session key buffers provided when a session is created are expected to remain alive for the duration of the session. - GCM and CCM sessions now only specify a cipher algorithm and a cipher key. The redundant auth information is not needed or used. - For cryptosoft, split up the code a bit such that the 'process' callback now invokes a function pointer in the session. This function pointer is set based on the mode (in effect) though it simplifies a few edge cases that would otherwise be in the switch in 'process'. It does split up GCM vs CCM which I think is more readable even if there is some duplication. - I changed /dev/crypto to support GMAC requests using CRYPTO_AES_NIST_GMAC as an auth algorithm and updated cryptocheck to work with it. - Combined cipher and auth sessions via /dev/crypto now always use ETA mode. The COP_F_CIPHER_FIRST flag is now a no-op that is ignored. This was actually documented as being true in crypto(4) before, but the code had not implemented this before I added the CIPHER_FIRST flag. - I have not yet updated /dev/crypto to be aware of explicit modes for sessions. I will probably do that at some point in the future as well as teach it about IV/nonce and tag lengths for AEAD so we can support all of the NIST KAT tests for GCM and CCM. - I've split up the exising crypto.9 manpage into several pages of which many are written from scratch. - I have converted all drivers and consumers in the tree and verified that they compile, but I have not tested all of them. I have tested the following drivers: - cryptosoft - aesni (AES only) - blake2 - ccr and the following consumers: - cryptodev - IPsec - ktls_ocf - GELI (lightly) I have not tested the following: - ccp - aesni with sha - hifn - kgssapi_krb5 - ubsec - padlock - safe - armv8_crypto (aarch64) - glxsb (i386) - sec (ppc) - cesa (armv7) - cryptocteon (mips64) - nlmsec (mips64) Discussed with: cem Relnotes: yes Sponsored by: Chelsio Communications Differential Revision: https://reviews.freebsd.org/D23677
2020-03-27 18:25:23 +00:00
/*
* The COP_F_CIPHER_FIRST flag predates explicit session
* modes, but the only way it was used was for EtA so allow it
* as long as it is consistent with EtA.
*/
if (caead->flags & COP_F_CIPHER_FIRST) {
if (caead->op != COP_ENCRYPT) {
SDT_PROBE1(opencrypto, dev, ioctl, error, __LINE__);
return (EINVAL);
}
}
cod = cod_alloc(cse, caead->aadlen, caead->len + cse->hashsize);
dst = caead->dst;
2014-12-12 19:56:36 +00:00
Refactor driver and consumer interfaces for OCF (in-kernel crypto). - The linked list of cryptoini structures used in session initialization is replaced with a new flat structure: struct crypto_session_params. This session includes a new mode to define how the other fields should be interpreted. Available modes include: - COMPRESS (for compression/decompression) - CIPHER (for simply encryption/decryption) - DIGEST (computing and verifying digests) - AEAD (combined auth and encryption such as AES-GCM and AES-CCM) - ETA (combined auth and encryption using encrypt-then-authenticate) Additional modes could be added in the future (e.g. if we wanted to support TLS MtE for AES-CBC in the kernel we could add a new mode for that. TLS modes might also affect how AAD is interpreted, etc.) The flat structure also includes the key lengths and algorithms as before. However, code doesn't have to walk the linked list and switch on the algorithm to determine which key is the auth key vs encryption key. The 'csp_auth_*' fields are always used for auth keys and settings and 'csp_cipher_*' for cipher. (Compression algorithms are stored in csp_cipher_alg.) - Drivers no longer register a list of supported algorithms. This doesn't quite work when you factor in modes (e.g. a driver might support both AES-CBC and SHA2-256-HMAC separately but not combined for ETA). Instead, a new 'crypto_probesession' method has been added to the kobj interface for symmteric crypto drivers. This method returns a negative value on success (similar to how device_probe works) and the crypto framework uses this value to pick the "best" driver. There are three constants for hardware (e.g. ccr), accelerated software (e.g. aesni), and plain software (cryptosoft) that give preference in that order. One effect of this is that if you request only hardware when creating a new session, you will no longer get a session using accelerated software. Another effect is that the default setting to disallow software crypto via /dev/crypto now disables accelerated software. Once a driver is chosen, 'crypto_newsession' is invoked as before. - Crypto operations are now solely described by the flat 'cryptop' structure. The linked list of descriptors has been removed. A separate enum has been added to describe the type of data buffer in use instead of using CRYPTO_F_* flags to make it easier to add more types in the future if needed (e.g. wired userspace buffers for zero-copy). It will also make it easier to re-introduce separate input and output buffers (in-kernel TLS would benefit from this). Try to make the flags related to IV handling less insane: - CRYPTO_F_IV_SEPARATE means that the IV is stored in the 'crp_iv' member of the operation structure. If this flag is not set, the IV is stored in the data buffer at the 'crp_iv_start' offset. - CRYPTO_F_IV_GENERATE means that a random IV should be generated and stored into the data buffer. This cannot be used with CRYPTO_F_IV_SEPARATE. If a consumer wants to deal with explicit vs implicit IVs, etc. it can always generate the IV however it needs and store partial IVs in the buffer and the full IV/nonce in crp_iv and set CRYPTO_F_IV_SEPARATE. The layout of the buffer is now described via fields in cryptop. crp_aad_start and crp_aad_length define the boundaries of any AAD. Previously with GCM and CCM you defined an auth crd with this range, but for ETA your auth crd had to span both the AAD and plaintext (and they had to be adjacent). crp_payload_start and crp_payload_length define the boundaries of the plaintext/ciphertext. Modes that only do a single operation (COMPRESS, CIPHER, DIGEST) should only use this region and leave the AAD region empty. If a digest is present (or should be generated), it's starting location is marked by crp_digest_start. Instead of using the CRD_F_ENCRYPT flag to determine the direction of the operation, cryptop now includes an 'op' field defining the operation to perform. For digests I've added a new VERIFY digest mode which assumes a digest is present in the input and fails the request with EBADMSG if it doesn't match the internally-computed digest. GCM and CCM already assumed this, and the new AEAD mode requires this for decryption. The new ETA mode now also requires this for decryption, so IPsec and GELI no longer do their own authentication verification. Simple DIGEST operations can also do this, though there are no in-tree consumers. To eventually support some refcounting to close races, the session cookie is now passed to crypto_getop() and clients should no longer set crp_sesssion directly. - Assymteric crypto operation structures should be allocated via crypto_getkreq() and freed via crypto_freekreq(). This permits the crypto layer to track open asym requests and close races with a driver trying to unregister while asym requests are in flight. - crypto_copyback, crypto_copydata, crypto_apply, and crypto_contiguous_subsegment now accept the 'crp' object as the first parameter instead of individual members. This makes it easier to deal with different buffer types in the future as well as separate input and output buffers. It's also simpler for driver writers to use. - bus_dmamap_load_crp() loads a DMA mapping for a crypto buffer. This understands the various types of buffers so that drivers that use DMA do not have to be aware of different buffer types. - Helper routines now exist to build an auth context for HMAC IPAD and OPAD. This reduces some duplicated work among drivers. - Key buffers are now treated as const throughout the framework and in device drivers. However, session key buffers provided when a session is created are expected to remain alive for the duration of the session. - GCM and CCM sessions now only specify a cipher algorithm and a cipher key. The redundant auth information is not needed or used. - For cryptosoft, split up the code a bit such that the 'process' callback now invokes a function pointer in the session. This function pointer is set based on the mode (in effect) though it simplifies a few edge cases that would otherwise be in the switch in 'process'. It does split up GCM vs CCM which I think is more readable even if there is some duplication. - I changed /dev/crypto to support GMAC requests using CRYPTO_AES_NIST_GMAC as an auth algorithm and updated cryptocheck to work with it. - Combined cipher and auth sessions via /dev/crypto now always use ETA mode. The COP_F_CIPHER_FIRST flag is now a no-op that is ignored. This was actually documented as being true in crypto(4) before, but the code had not implemented this before I added the CIPHER_FIRST flag. - I have not yet updated /dev/crypto to be aware of explicit modes for sessions. I will probably do that at some point in the future as well as teach it about IV/nonce and tag lengths for AEAD so we can support all of the NIST KAT tests for GCM and CCM. - I've split up the exising crypto.9 manpage into several pages of which many are written from scratch. - I have converted all drivers and consumers in the tree and verified that they compile, but I have not tested all of them. I have tested the following drivers: - cryptosoft - aesni (AES only) - blake2 - ccr and the following consumers: - cryptodev - IPsec - ktls_ocf - GELI (lightly) I have not tested the following: - ccp - aesni with sha - hifn - kgssapi_krb5 - ubsec - padlock - safe - armv8_crypto (aarch64) - glxsb (i386) - sec (ppc) - cesa (armv7) - cryptocteon (mips64) - nlmsec (mips64) Discussed with: cem Relnotes: yes Sponsored by: Chelsio Communications Differential Revision: https://reviews.freebsd.org/D23677
2020-03-27 18:25:23 +00:00
crp = crypto_getreq(cse->cses, M_WAITOK);
if (cod->aad != NULL)
error = copyin(caead->aad, cod->aad, caead->aadlen);
else
error = copyin(caead->aad, cod->buf, caead->aadlen);
Refactor driver and consumer interfaces for OCF (in-kernel crypto). - The linked list of cryptoini structures used in session initialization is replaced with a new flat structure: struct crypto_session_params. This session includes a new mode to define how the other fields should be interpreted. Available modes include: - COMPRESS (for compression/decompression) - CIPHER (for simply encryption/decryption) - DIGEST (computing and verifying digests) - AEAD (combined auth and encryption such as AES-GCM and AES-CCM) - ETA (combined auth and encryption using encrypt-then-authenticate) Additional modes could be added in the future (e.g. if we wanted to support TLS MtE for AES-CBC in the kernel we could add a new mode for that. TLS modes might also affect how AAD is interpreted, etc.) The flat structure also includes the key lengths and algorithms as before. However, code doesn't have to walk the linked list and switch on the algorithm to determine which key is the auth key vs encryption key. The 'csp_auth_*' fields are always used for auth keys and settings and 'csp_cipher_*' for cipher. (Compression algorithms are stored in csp_cipher_alg.) - Drivers no longer register a list of supported algorithms. This doesn't quite work when you factor in modes (e.g. a driver might support both AES-CBC and SHA2-256-HMAC separately but not combined for ETA). Instead, a new 'crypto_probesession' method has been added to the kobj interface for symmteric crypto drivers. This method returns a negative value on success (similar to how device_probe works) and the crypto framework uses this value to pick the "best" driver. There are three constants for hardware (e.g. ccr), accelerated software (e.g. aesni), and plain software (cryptosoft) that give preference in that order. One effect of this is that if you request only hardware when creating a new session, you will no longer get a session using accelerated software. Another effect is that the default setting to disallow software crypto via /dev/crypto now disables accelerated software. Once a driver is chosen, 'crypto_newsession' is invoked as before. - Crypto operations are now solely described by the flat 'cryptop' structure. The linked list of descriptors has been removed. A separate enum has been added to describe the type of data buffer in use instead of using CRYPTO_F_* flags to make it easier to add more types in the future if needed (e.g. wired userspace buffers for zero-copy). It will also make it easier to re-introduce separate input and output buffers (in-kernel TLS would benefit from this). Try to make the flags related to IV handling less insane: - CRYPTO_F_IV_SEPARATE means that the IV is stored in the 'crp_iv' member of the operation structure. If this flag is not set, the IV is stored in the data buffer at the 'crp_iv_start' offset. - CRYPTO_F_IV_GENERATE means that a random IV should be generated and stored into the data buffer. This cannot be used with CRYPTO_F_IV_SEPARATE. If a consumer wants to deal with explicit vs implicit IVs, etc. it can always generate the IV however it needs and store partial IVs in the buffer and the full IV/nonce in crp_iv and set CRYPTO_F_IV_SEPARATE. The layout of the buffer is now described via fields in cryptop. crp_aad_start and crp_aad_length define the boundaries of any AAD. Previously with GCM and CCM you defined an auth crd with this range, but for ETA your auth crd had to span both the AAD and plaintext (and they had to be adjacent). crp_payload_start and crp_payload_length define the boundaries of the plaintext/ciphertext. Modes that only do a single operation (COMPRESS, CIPHER, DIGEST) should only use this region and leave the AAD region empty. If a digest is present (or should be generated), it's starting location is marked by crp_digest_start. Instead of using the CRD_F_ENCRYPT flag to determine the direction of the operation, cryptop now includes an 'op' field defining the operation to perform. For digests I've added a new VERIFY digest mode which assumes a digest is present in the input and fails the request with EBADMSG if it doesn't match the internally-computed digest. GCM and CCM already assumed this, and the new AEAD mode requires this for decryption. The new ETA mode now also requires this for decryption, so IPsec and GELI no longer do their own authentication verification. Simple DIGEST operations can also do this, though there are no in-tree consumers. To eventually support some refcounting to close races, the session cookie is now passed to crypto_getop() and clients should no longer set crp_sesssion directly. - Assymteric crypto operation structures should be allocated via crypto_getkreq() and freed via crypto_freekreq(). This permits the crypto layer to track open asym requests and close races with a driver trying to unregister while asym requests are in flight. - crypto_copyback, crypto_copydata, crypto_apply, and crypto_contiguous_subsegment now accept the 'crp' object as the first parameter instead of individual members. This makes it easier to deal with different buffer types in the future as well as separate input and output buffers. It's also simpler for driver writers to use. - bus_dmamap_load_crp() loads a DMA mapping for a crypto buffer. This understands the various types of buffers so that drivers that use DMA do not have to be aware of different buffer types. - Helper routines now exist to build an auth context for HMAC IPAD and OPAD. This reduces some duplicated work among drivers. - Key buffers are now treated as const throughout the framework and in device drivers. However, session key buffers provided when a session is created are expected to remain alive for the duration of the session. - GCM and CCM sessions now only specify a cipher algorithm and a cipher key. The redundant auth information is not needed or used. - For cryptosoft, split up the code a bit such that the 'process' callback now invokes a function pointer in the session. This function pointer is set based on the mode (in effect) though it simplifies a few edge cases that would otherwise be in the switch in 'process'. It does split up GCM vs CCM which I think is more readable even if there is some duplication. - I changed /dev/crypto to support GMAC requests using CRYPTO_AES_NIST_GMAC as an auth algorithm and updated cryptocheck to work with it. - Combined cipher and auth sessions via /dev/crypto now always use ETA mode. The COP_F_CIPHER_FIRST flag is now a no-op that is ignored. This was actually documented as being true in crypto(4) before, but the code had not implemented this before I added the CIPHER_FIRST flag. - I have not yet updated /dev/crypto to be aware of explicit modes for sessions. I will probably do that at some point in the future as well as teach it about IV/nonce and tag lengths for AEAD so we can support all of the NIST KAT tests for GCM and CCM. - I've split up the exising crypto.9 manpage into several pages of which many are written from scratch. - I have converted all drivers and consumers in the tree and verified that they compile, but I have not tested all of them. I have tested the following drivers: - cryptosoft - aesni (AES only) - blake2 - ccr and the following consumers: - cryptodev - IPsec - ktls_ocf - GELI (lightly) I have not tested the following: - ccp - aesni with sha - hifn - kgssapi_krb5 - ubsec - padlock - safe - armv8_crypto (aarch64) - glxsb (i386) - sec (ppc) - cesa (armv7) - cryptocteon (mips64) - nlmsec (mips64) Discussed with: cem Relnotes: yes Sponsored by: Chelsio Communications Differential Revision: https://reviews.freebsd.org/D23677
2020-03-27 18:25:23 +00:00
if (error) {
SDT_PROBE1(opencrypto, dev, ioctl, error, __LINE__);
2014-12-12 19:56:36 +00:00
goto bail;
}
crp->crp_aad = cod->aad;
Refactor driver and consumer interfaces for OCF (in-kernel crypto). - The linked list of cryptoini structures used in session initialization is replaced with a new flat structure: struct crypto_session_params. This session includes a new mode to define how the other fields should be interpreted. Available modes include: - COMPRESS (for compression/decompression) - CIPHER (for simply encryption/decryption) - DIGEST (computing and verifying digests) - AEAD (combined auth and encryption such as AES-GCM and AES-CCM) - ETA (combined auth and encryption using encrypt-then-authenticate) Additional modes could be added in the future (e.g. if we wanted to support TLS MtE for AES-CBC in the kernel we could add a new mode for that. TLS modes might also affect how AAD is interpreted, etc.) The flat structure also includes the key lengths and algorithms as before. However, code doesn't have to walk the linked list and switch on the algorithm to determine which key is the auth key vs encryption key. The 'csp_auth_*' fields are always used for auth keys and settings and 'csp_cipher_*' for cipher. (Compression algorithms are stored in csp_cipher_alg.) - Drivers no longer register a list of supported algorithms. This doesn't quite work when you factor in modes (e.g. a driver might support both AES-CBC and SHA2-256-HMAC separately but not combined for ETA). Instead, a new 'crypto_probesession' method has been added to the kobj interface for symmteric crypto drivers. This method returns a negative value on success (similar to how device_probe works) and the crypto framework uses this value to pick the "best" driver. There are three constants for hardware (e.g. ccr), accelerated software (e.g. aesni), and plain software (cryptosoft) that give preference in that order. One effect of this is that if you request only hardware when creating a new session, you will no longer get a session using accelerated software. Another effect is that the default setting to disallow software crypto via /dev/crypto now disables accelerated software. Once a driver is chosen, 'crypto_newsession' is invoked as before. - Crypto operations are now solely described by the flat 'cryptop' structure. The linked list of descriptors has been removed. A separate enum has been added to describe the type of data buffer in use instead of using CRYPTO_F_* flags to make it easier to add more types in the future if needed (e.g. wired userspace buffers for zero-copy). It will also make it easier to re-introduce separate input and output buffers (in-kernel TLS would benefit from this). Try to make the flags related to IV handling less insane: - CRYPTO_F_IV_SEPARATE means that the IV is stored in the 'crp_iv' member of the operation structure. If this flag is not set, the IV is stored in the data buffer at the 'crp_iv_start' offset. - CRYPTO_F_IV_GENERATE means that a random IV should be generated and stored into the data buffer. This cannot be used with CRYPTO_F_IV_SEPARATE. If a consumer wants to deal with explicit vs implicit IVs, etc. it can always generate the IV however it needs and store partial IVs in the buffer and the full IV/nonce in crp_iv and set CRYPTO_F_IV_SEPARATE. The layout of the buffer is now described via fields in cryptop. crp_aad_start and crp_aad_length define the boundaries of any AAD. Previously with GCM and CCM you defined an auth crd with this range, but for ETA your auth crd had to span both the AAD and plaintext (and they had to be adjacent). crp_payload_start and crp_payload_length define the boundaries of the plaintext/ciphertext. Modes that only do a single operation (COMPRESS, CIPHER, DIGEST) should only use this region and leave the AAD region empty. If a digest is present (or should be generated), it's starting location is marked by crp_digest_start. Instead of using the CRD_F_ENCRYPT flag to determine the direction of the operation, cryptop now includes an 'op' field defining the operation to perform. For digests I've added a new VERIFY digest mode which assumes a digest is present in the input and fails the request with EBADMSG if it doesn't match the internally-computed digest. GCM and CCM already assumed this, and the new AEAD mode requires this for decryption. The new ETA mode now also requires this for decryption, so IPsec and GELI no longer do their own authentication verification. Simple DIGEST operations can also do this, though there are no in-tree consumers. To eventually support some refcounting to close races, the session cookie is now passed to crypto_getop() and clients should no longer set crp_sesssion directly. - Assymteric crypto operation structures should be allocated via crypto_getkreq() and freed via crypto_freekreq(). This permits the crypto layer to track open asym requests and close races with a driver trying to unregister while asym requests are in flight. - crypto_copyback, crypto_copydata, crypto_apply, and crypto_contiguous_subsegment now accept the 'crp' object as the first parameter instead of individual members. This makes it easier to deal with different buffer types in the future as well as separate input and output buffers. It's also simpler for driver writers to use. - bus_dmamap_load_crp() loads a DMA mapping for a crypto buffer. This understands the various types of buffers so that drivers that use DMA do not have to be aware of different buffer types. - Helper routines now exist to build an auth context for HMAC IPAD and OPAD. This reduces some duplicated work among drivers. - Key buffers are now treated as const throughout the framework and in device drivers. However, session key buffers provided when a session is created are expected to remain alive for the duration of the session. - GCM and CCM sessions now only specify a cipher algorithm and a cipher key. The redundant auth information is not needed or used. - For cryptosoft, split up the code a bit such that the 'process' callback now invokes a function pointer in the session. This function pointer is set based on the mode (in effect) though it simplifies a few edge cases that would otherwise be in the switch in 'process'. It does split up GCM vs CCM which I think is more readable even if there is some duplication. - I changed /dev/crypto to support GMAC requests using CRYPTO_AES_NIST_GMAC as an auth algorithm and updated cryptocheck to work with it. - Combined cipher and auth sessions via /dev/crypto now always use ETA mode. The COP_F_CIPHER_FIRST flag is now a no-op that is ignored. This was actually documented as being true in crypto(4) before, but the code had not implemented this before I added the CIPHER_FIRST flag. - I have not yet updated /dev/crypto to be aware of explicit modes for sessions. I will probably do that at some point in the future as well as teach it about IV/nonce and tag lengths for AEAD so we can support all of the NIST KAT tests for GCM and CCM. - I've split up the exising crypto.9 manpage into several pages of which many are written from scratch. - I have converted all drivers and consumers in the tree and verified that they compile, but I have not tested all of them. I have tested the following drivers: - cryptosoft - aesni (AES only) - blake2 - ccr and the following consumers: - cryptodev - IPsec - ktls_ocf - GELI (lightly) I have not tested the following: - ccp - aesni with sha - hifn - kgssapi_krb5 - ubsec - padlock - safe - armv8_crypto (aarch64) - glxsb (i386) - sec (ppc) - cesa (armv7) - cryptocteon (mips64) - nlmsec (mips64) Discussed with: cem Relnotes: yes Sponsored by: Chelsio Communications Differential Revision: https://reviews.freebsd.org/D23677
2020-03-27 18:25:23 +00:00
crp->crp_aad_start = 0;
crp->crp_aad_length = caead->aadlen;
2014-12-12 19:56:36 +00:00
if (cod->aad != NULL)
crp->crp_payload_start = 0;
else
crp->crp_payload_start = caead->aadlen;
error = copyin(caead->src, cod->buf + crp->crp_payload_start,
caead->len);
Refactor driver and consumer interfaces for OCF (in-kernel crypto). - The linked list of cryptoini structures used in session initialization is replaced with a new flat structure: struct crypto_session_params. This session includes a new mode to define how the other fields should be interpreted. Available modes include: - COMPRESS (for compression/decompression) - CIPHER (for simply encryption/decryption) - DIGEST (computing and verifying digests) - AEAD (combined auth and encryption such as AES-GCM and AES-CCM) - ETA (combined auth and encryption using encrypt-then-authenticate) Additional modes could be added in the future (e.g. if we wanted to support TLS MtE for AES-CBC in the kernel we could add a new mode for that. TLS modes might also affect how AAD is interpreted, etc.) The flat structure also includes the key lengths and algorithms as before. However, code doesn't have to walk the linked list and switch on the algorithm to determine which key is the auth key vs encryption key. The 'csp_auth_*' fields are always used for auth keys and settings and 'csp_cipher_*' for cipher. (Compression algorithms are stored in csp_cipher_alg.) - Drivers no longer register a list of supported algorithms. This doesn't quite work when you factor in modes (e.g. a driver might support both AES-CBC and SHA2-256-HMAC separately but not combined for ETA). Instead, a new 'crypto_probesession' method has been added to the kobj interface for symmteric crypto drivers. This method returns a negative value on success (similar to how device_probe works) and the crypto framework uses this value to pick the "best" driver. There are three constants for hardware (e.g. ccr), accelerated software (e.g. aesni), and plain software (cryptosoft) that give preference in that order. One effect of this is that if you request only hardware when creating a new session, you will no longer get a session using accelerated software. Another effect is that the default setting to disallow software crypto via /dev/crypto now disables accelerated software. Once a driver is chosen, 'crypto_newsession' is invoked as before. - Crypto operations are now solely described by the flat 'cryptop' structure. The linked list of descriptors has been removed. A separate enum has been added to describe the type of data buffer in use instead of using CRYPTO_F_* flags to make it easier to add more types in the future if needed (e.g. wired userspace buffers for zero-copy). It will also make it easier to re-introduce separate input and output buffers (in-kernel TLS would benefit from this). Try to make the flags related to IV handling less insane: - CRYPTO_F_IV_SEPARATE means that the IV is stored in the 'crp_iv' member of the operation structure. If this flag is not set, the IV is stored in the data buffer at the 'crp_iv_start' offset. - CRYPTO_F_IV_GENERATE means that a random IV should be generated and stored into the data buffer. This cannot be used with CRYPTO_F_IV_SEPARATE. If a consumer wants to deal with explicit vs implicit IVs, etc. it can always generate the IV however it needs and store partial IVs in the buffer and the full IV/nonce in crp_iv and set CRYPTO_F_IV_SEPARATE. The layout of the buffer is now described via fields in cryptop. crp_aad_start and crp_aad_length define the boundaries of any AAD. Previously with GCM and CCM you defined an auth crd with this range, but for ETA your auth crd had to span both the AAD and plaintext (and they had to be adjacent). crp_payload_start and crp_payload_length define the boundaries of the plaintext/ciphertext. Modes that only do a single operation (COMPRESS, CIPHER, DIGEST) should only use this region and leave the AAD region empty. If a digest is present (or should be generated), it's starting location is marked by crp_digest_start. Instead of using the CRD_F_ENCRYPT flag to determine the direction of the operation, cryptop now includes an 'op' field defining the operation to perform. For digests I've added a new VERIFY digest mode which assumes a digest is present in the input and fails the request with EBADMSG if it doesn't match the internally-computed digest. GCM and CCM already assumed this, and the new AEAD mode requires this for decryption. The new ETA mode now also requires this for decryption, so IPsec and GELI no longer do their own authentication verification. Simple DIGEST operations can also do this, though there are no in-tree consumers. To eventually support some refcounting to close races, the session cookie is now passed to crypto_getop() and clients should no longer set crp_sesssion directly. - Assymteric crypto operation structures should be allocated via crypto_getkreq() and freed via crypto_freekreq(). This permits the crypto layer to track open asym requests and close races with a driver trying to unregister while asym requests are in flight. - crypto_copyback, crypto_copydata, crypto_apply, and crypto_contiguous_subsegment now accept the 'crp' object as the first parameter instead of individual members. This makes it easier to deal with different buffer types in the future as well as separate input and output buffers. It's also simpler for driver writers to use. - bus_dmamap_load_crp() loads a DMA mapping for a crypto buffer. This understands the various types of buffers so that drivers that use DMA do not have to be aware of different buffer types. - Helper routines now exist to build an auth context for HMAC IPAD and OPAD. This reduces some duplicated work among drivers. - Key buffers are now treated as const throughout the framework and in device drivers. However, session key buffers provided when a session is created are expected to remain alive for the duration of the session. - GCM and CCM sessions now only specify a cipher algorithm and a cipher key. The redundant auth information is not needed or used. - For cryptosoft, split up the code a bit such that the 'process' callback now invokes a function pointer in the session. This function pointer is set based on the mode (in effect) though it simplifies a few edge cases that would otherwise be in the switch in 'process'. It does split up GCM vs CCM which I think is more readable even if there is some duplication. - I changed /dev/crypto to support GMAC requests using CRYPTO_AES_NIST_GMAC as an auth algorithm and updated cryptocheck to work with it. - Combined cipher and auth sessions via /dev/crypto now always use ETA mode. The COP_F_CIPHER_FIRST flag is now a no-op that is ignored. This was actually documented as being true in crypto(4) before, but the code had not implemented this before I added the CIPHER_FIRST flag. - I have not yet updated /dev/crypto to be aware of explicit modes for sessions. I will probably do that at some point in the future as well as teach it about IV/nonce and tag lengths for AEAD so we can support all of the NIST KAT tests for GCM and CCM. - I've split up the exising crypto.9 manpage into several pages of which many are written from scratch. - I have converted all drivers and consumers in the tree and verified that they compile, but I have not tested all of them. I have tested the following drivers: - cryptosoft - aesni (AES only) - blake2 - ccr and the following consumers: - cryptodev - IPsec - ktls_ocf - GELI (lightly) I have not tested the following: - ccp - aesni with sha - hifn - kgssapi_krb5 - ubsec - padlock - safe - armv8_crypto (aarch64) - glxsb (i386) - sec (ppc) - cesa (armv7) - cryptocteon (mips64) - nlmsec (mips64) Discussed with: cem Relnotes: yes Sponsored by: Chelsio Communications Differential Revision: https://reviews.freebsd.org/D23677
2020-03-27 18:25:23 +00:00
if (error) {
SDT_PROBE1(opencrypto, dev, ioctl, error, __LINE__);
2014-12-12 19:56:36 +00:00
goto bail;
}
Refactor driver and consumer interfaces for OCF (in-kernel crypto). - The linked list of cryptoini structures used in session initialization is replaced with a new flat structure: struct crypto_session_params. This session includes a new mode to define how the other fields should be interpreted. Available modes include: - COMPRESS (for compression/decompression) - CIPHER (for simply encryption/decryption) - DIGEST (computing and verifying digests) - AEAD (combined auth and encryption such as AES-GCM and AES-CCM) - ETA (combined auth and encryption using encrypt-then-authenticate) Additional modes could be added in the future (e.g. if we wanted to support TLS MtE for AES-CBC in the kernel we could add a new mode for that. TLS modes might also affect how AAD is interpreted, etc.) The flat structure also includes the key lengths and algorithms as before. However, code doesn't have to walk the linked list and switch on the algorithm to determine which key is the auth key vs encryption key. The 'csp_auth_*' fields are always used for auth keys and settings and 'csp_cipher_*' for cipher. (Compression algorithms are stored in csp_cipher_alg.) - Drivers no longer register a list of supported algorithms. This doesn't quite work when you factor in modes (e.g. a driver might support both AES-CBC and SHA2-256-HMAC separately but not combined for ETA). Instead, a new 'crypto_probesession' method has been added to the kobj interface for symmteric crypto drivers. This method returns a negative value on success (similar to how device_probe works) and the crypto framework uses this value to pick the "best" driver. There are three constants for hardware (e.g. ccr), accelerated software (e.g. aesni), and plain software (cryptosoft) that give preference in that order. One effect of this is that if you request only hardware when creating a new session, you will no longer get a session using accelerated software. Another effect is that the default setting to disallow software crypto via /dev/crypto now disables accelerated software. Once a driver is chosen, 'crypto_newsession' is invoked as before. - Crypto operations are now solely described by the flat 'cryptop' structure. The linked list of descriptors has been removed. A separate enum has been added to describe the type of data buffer in use instead of using CRYPTO_F_* flags to make it easier to add more types in the future if needed (e.g. wired userspace buffers for zero-copy). It will also make it easier to re-introduce separate input and output buffers (in-kernel TLS would benefit from this). Try to make the flags related to IV handling less insane: - CRYPTO_F_IV_SEPARATE means that the IV is stored in the 'crp_iv' member of the operation structure. If this flag is not set, the IV is stored in the data buffer at the 'crp_iv_start' offset. - CRYPTO_F_IV_GENERATE means that a random IV should be generated and stored into the data buffer. This cannot be used with CRYPTO_F_IV_SEPARATE. If a consumer wants to deal with explicit vs implicit IVs, etc. it can always generate the IV however it needs and store partial IVs in the buffer and the full IV/nonce in crp_iv and set CRYPTO_F_IV_SEPARATE. The layout of the buffer is now described via fields in cryptop. crp_aad_start and crp_aad_length define the boundaries of any AAD. Previously with GCM and CCM you defined an auth crd with this range, but for ETA your auth crd had to span both the AAD and plaintext (and they had to be adjacent). crp_payload_start and crp_payload_length define the boundaries of the plaintext/ciphertext. Modes that only do a single operation (COMPRESS, CIPHER, DIGEST) should only use this region and leave the AAD region empty. If a digest is present (or should be generated), it's starting location is marked by crp_digest_start. Instead of using the CRD_F_ENCRYPT flag to determine the direction of the operation, cryptop now includes an 'op' field defining the operation to perform. For digests I've added a new VERIFY digest mode which assumes a digest is present in the input and fails the request with EBADMSG if it doesn't match the internally-computed digest. GCM and CCM already assumed this, and the new AEAD mode requires this for decryption. The new ETA mode now also requires this for decryption, so IPsec and GELI no longer do their own authentication verification. Simple DIGEST operations can also do this, though there are no in-tree consumers. To eventually support some refcounting to close races, the session cookie is now passed to crypto_getop() and clients should no longer set crp_sesssion directly. - Assymteric crypto operation structures should be allocated via crypto_getkreq() and freed via crypto_freekreq(). This permits the crypto layer to track open asym requests and close races with a driver trying to unregister while asym requests are in flight. - crypto_copyback, crypto_copydata, crypto_apply, and crypto_contiguous_subsegment now accept the 'crp' object as the first parameter instead of individual members. This makes it easier to deal with different buffer types in the future as well as separate input and output buffers. It's also simpler for driver writers to use. - bus_dmamap_load_crp() loads a DMA mapping for a crypto buffer. This understands the various types of buffers so that drivers that use DMA do not have to be aware of different buffer types. - Helper routines now exist to build an auth context for HMAC IPAD and OPAD. This reduces some duplicated work among drivers. - Key buffers are now treated as const throughout the framework and in device drivers. However, session key buffers provided when a session is created are expected to remain alive for the duration of the session. - GCM and CCM sessions now only specify a cipher algorithm and a cipher key. The redundant auth information is not needed or used. - For cryptosoft, split up the code a bit such that the 'process' callback now invokes a function pointer in the session. This function pointer is set based on the mode (in effect) though it simplifies a few edge cases that would otherwise be in the switch in 'process'. It does split up GCM vs CCM which I think is more readable even if there is some duplication. - I changed /dev/crypto to support GMAC requests using CRYPTO_AES_NIST_GMAC as an auth algorithm and updated cryptocheck to work with it. - Combined cipher and auth sessions via /dev/crypto now always use ETA mode. The COP_F_CIPHER_FIRST flag is now a no-op that is ignored. This was actually documented as being true in crypto(4) before, but the code had not implemented this before I added the CIPHER_FIRST flag. - I have not yet updated /dev/crypto to be aware of explicit modes for sessions. I will probably do that at some point in the future as well as teach it about IV/nonce and tag lengths for AEAD so we can support all of the NIST KAT tests for GCM and CCM. - I've split up the exising crypto.9 manpage into several pages of which many are written from scratch. - I have converted all drivers and consumers in the tree and verified that they compile, but I have not tested all of them. I have tested the following drivers: - cryptosoft - aesni (AES only) - blake2 - ccr and the following consumers: - cryptodev - IPsec - ktls_ocf - GELI (lightly) I have not tested the following: - ccp - aesni with sha - hifn - kgssapi_krb5 - ubsec - padlock - safe - armv8_crypto (aarch64) - glxsb (i386) - sec (ppc) - cesa (armv7) - cryptocteon (mips64) - nlmsec (mips64) Discussed with: cem Relnotes: yes Sponsored by: Chelsio Communications Differential Revision: https://reviews.freebsd.org/D23677
2020-03-27 18:25:23 +00:00
crp->crp_payload_length = caead->len;
if (caead->op == COP_ENCRYPT && cod->obuf != NULL)
crp->crp_digest_start = crp->crp_payload_output_start +
caead->len;
else
crp->crp_digest_start = crp->crp_payload_start + caead->len;
Refactor driver and consumer interfaces for OCF (in-kernel crypto). - The linked list of cryptoini structures used in session initialization is replaced with a new flat structure: struct crypto_session_params. This session includes a new mode to define how the other fields should be interpreted. Available modes include: - COMPRESS (for compression/decompression) - CIPHER (for simply encryption/decryption) - DIGEST (computing and verifying digests) - AEAD (combined auth and encryption such as AES-GCM and AES-CCM) - ETA (combined auth and encryption using encrypt-then-authenticate) Additional modes could be added in the future (e.g. if we wanted to support TLS MtE for AES-CBC in the kernel we could add a new mode for that. TLS modes might also affect how AAD is interpreted, etc.) The flat structure also includes the key lengths and algorithms as before. However, code doesn't have to walk the linked list and switch on the algorithm to determine which key is the auth key vs encryption key. The 'csp_auth_*' fields are always used for auth keys and settings and 'csp_cipher_*' for cipher. (Compression algorithms are stored in csp_cipher_alg.) - Drivers no longer register a list of supported algorithms. This doesn't quite work when you factor in modes (e.g. a driver might support both AES-CBC and SHA2-256-HMAC separately but not combined for ETA). Instead, a new 'crypto_probesession' method has been added to the kobj interface for symmteric crypto drivers. This method returns a negative value on success (similar to how device_probe works) and the crypto framework uses this value to pick the "best" driver. There are three constants for hardware (e.g. ccr), accelerated software (e.g. aesni), and plain software (cryptosoft) that give preference in that order. One effect of this is that if you request only hardware when creating a new session, you will no longer get a session using accelerated software. Another effect is that the default setting to disallow software crypto via /dev/crypto now disables accelerated software. Once a driver is chosen, 'crypto_newsession' is invoked as before. - Crypto operations are now solely described by the flat 'cryptop' structure. The linked list of descriptors has been removed. A separate enum has been added to describe the type of data buffer in use instead of using CRYPTO_F_* flags to make it easier to add more types in the future if needed (e.g. wired userspace buffers for zero-copy). It will also make it easier to re-introduce separate input and output buffers (in-kernel TLS would benefit from this). Try to make the flags related to IV handling less insane: - CRYPTO_F_IV_SEPARATE means that the IV is stored in the 'crp_iv' member of the operation structure. If this flag is not set, the IV is stored in the data buffer at the 'crp_iv_start' offset. - CRYPTO_F_IV_GENERATE means that a random IV should be generated and stored into the data buffer. This cannot be used with CRYPTO_F_IV_SEPARATE. If a consumer wants to deal with explicit vs implicit IVs, etc. it can always generate the IV however it needs and store partial IVs in the buffer and the full IV/nonce in crp_iv and set CRYPTO_F_IV_SEPARATE. The layout of the buffer is now described via fields in cryptop. crp_aad_start and crp_aad_length define the boundaries of any AAD. Previously with GCM and CCM you defined an auth crd with this range, but for ETA your auth crd had to span both the AAD and plaintext (and they had to be adjacent). crp_payload_start and crp_payload_length define the boundaries of the plaintext/ciphertext. Modes that only do a single operation (COMPRESS, CIPHER, DIGEST) should only use this region and leave the AAD region empty. If a digest is present (or should be generated), it's starting location is marked by crp_digest_start. Instead of using the CRD_F_ENCRYPT flag to determine the direction of the operation, cryptop now includes an 'op' field defining the operation to perform. For digests I've added a new VERIFY digest mode which assumes a digest is present in the input and fails the request with EBADMSG if it doesn't match the internally-computed digest. GCM and CCM already assumed this, and the new AEAD mode requires this for decryption. The new ETA mode now also requires this for decryption, so IPsec and GELI no longer do their own authentication verification. Simple DIGEST operations can also do this, though there are no in-tree consumers. To eventually support some refcounting to close races, the session cookie is now passed to crypto_getop() and clients should no longer set crp_sesssion directly. - Assymteric crypto operation structures should be allocated via crypto_getkreq() and freed via crypto_freekreq(). This permits the crypto layer to track open asym requests and close races with a driver trying to unregister while asym requests are in flight. - crypto_copyback, crypto_copydata, crypto_apply, and crypto_contiguous_subsegment now accept the 'crp' object as the first parameter instead of individual members. This makes it easier to deal with different buffer types in the future as well as separate input and output buffers. It's also simpler for driver writers to use. - bus_dmamap_load_crp() loads a DMA mapping for a crypto buffer. This understands the various types of buffers so that drivers that use DMA do not have to be aware of different buffer types. - Helper routines now exist to build an auth context for HMAC IPAD and OPAD. This reduces some duplicated work among drivers. - Key buffers are now treated as const throughout the framework and in device drivers. However, session key buffers provided when a session is created are expected to remain alive for the duration of the session. - GCM and CCM sessions now only specify a cipher algorithm and a cipher key. The redundant auth information is not needed or used. - For cryptosoft, split up the code a bit such that the 'process' callback now invokes a function pointer in the session. This function pointer is set based on the mode (in effect) though it simplifies a few edge cases that would otherwise be in the switch in 'process'. It does split up GCM vs CCM which I think is more readable even if there is some duplication. - I changed /dev/crypto to support GMAC requests using CRYPTO_AES_NIST_GMAC as an auth algorithm and updated cryptocheck to work with it. - Combined cipher and auth sessions via /dev/crypto now always use ETA mode. The COP_F_CIPHER_FIRST flag is now a no-op that is ignored. This was actually documented as being true in crypto(4) before, but the code had not implemented this before I added the CIPHER_FIRST flag. - I have not yet updated /dev/crypto to be aware of explicit modes for sessions. I will probably do that at some point in the future as well as teach it about IV/nonce and tag lengths for AEAD so we can support all of the NIST KAT tests for GCM and CCM. - I've split up the exising crypto.9 manpage into several pages of which many are written from scratch. - I have converted all drivers and consumers in the tree and verified that they compile, but I have not tested all of them. I have tested the following drivers: - cryptosoft - aesni (AES only) - blake2 - ccr and the following consumers: - cryptodev - IPsec - ktls_ocf - GELI (lightly) I have not tested the following: - ccp - aesni with sha - hifn - kgssapi_krb5 - ubsec - padlock - safe - armv8_crypto (aarch64) - glxsb (i386) - sec (ppc) - cesa (armv7) - cryptocteon (mips64) - nlmsec (mips64) Discussed with: cem Relnotes: yes Sponsored by: Chelsio Communications Differential Revision: https://reviews.freebsd.org/D23677
2020-03-27 18:25:23 +00:00
csp = crypto_get_params(cse->cses);
switch (csp->csp_mode) {
Refactor driver and consumer interfaces for OCF (in-kernel crypto). - The linked list of cryptoini structures used in session initialization is replaced with a new flat structure: struct crypto_session_params. This session includes a new mode to define how the other fields should be interpreted. Available modes include: - COMPRESS (for compression/decompression) - CIPHER (for simply encryption/decryption) - DIGEST (computing and verifying digests) - AEAD (combined auth and encryption such as AES-GCM and AES-CCM) - ETA (combined auth and encryption using encrypt-then-authenticate) Additional modes could be added in the future (e.g. if we wanted to support TLS MtE for AES-CBC in the kernel we could add a new mode for that. TLS modes might also affect how AAD is interpreted, etc.) The flat structure also includes the key lengths and algorithms as before. However, code doesn't have to walk the linked list and switch on the algorithm to determine which key is the auth key vs encryption key. The 'csp_auth_*' fields are always used for auth keys and settings and 'csp_cipher_*' for cipher. (Compression algorithms are stored in csp_cipher_alg.) - Drivers no longer register a list of supported algorithms. This doesn't quite work when you factor in modes (e.g. a driver might support both AES-CBC and SHA2-256-HMAC separately but not combined for ETA). Instead, a new 'crypto_probesession' method has been added to the kobj interface for symmteric crypto drivers. This method returns a negative value on success (similar to how device_probe works) and the crypto framework uses this value to pick the "best" driver. There are three constants for hardware (e.g. ccr), accelerated software (e.g. aesni), and plain software (cryptosoft) that give preference in that order. One effect of this is that if you request only hardware when creating a new session, you will no longer get a session using accelerated software. Another effect is that the default setting to disallow software crypto via /dev/crypto now disables accelerated software. Once a driver is chosen, 'crypto_newsession' is invoked as before. - Crypto operations are now solely described by the flat 'cryptop' structure. The linked list of descriptors has been removed. A separate enum has been added to describe the type of data buffer in use instead of using CRYPTO_F_* flags to make it easier to add more types in the future if needed (e.g. wired userspace buffers for zero-copy). It will also make it easier to re-introduce separate input and output buffers (in-kernel TLS would benefit from this). Try to make the flags related to IV handling less insane: - CRYPTO_F_IV_SEPARATE means that the IV is stored in the 'crp_iv' member of the operation structure. If this flag is not set, the IV is stored in the data buffer at the 'crp_iv_start' offset. - CRYPTO_F_IV_GENERATE means that a random IV should be generated and stored into the data buffer. This cannot be used with CRYPTO_F_IV_SEPARATE. If a consumer wants to deal with explicit vs implicit IVs, etc. it can always generate the IV however it needs and store partial IVs in the buffer and the full IV/nonce in crp_iv and set CRYPTO_F_IV_SEPARATE. The layout of the buffer is now described via fields in cryptop. crp_aad_start and crp_aad_length define the boundaries of any AAD. Previously with GCM and CCM you defined an auth crd with this range, but for ETA your auth crd had to span both the AAD and plaintext (and they had to be adjacent). crp_payload_start and crp_payload_length define the boundaries of the plaintext/ciphertext. Modes that only do a single operation (COMPRESS, CIPHER, DIGEST) should only use this region and leave the AAD region empty. If a digest is present (or should be generated), it's starting location is marked by crp_digest_start. Instead of using the CRD_F_ENCRYPT flag to determine the direction of the operation, cryptop now includes an 'op' field defining the operation to perform. For digests I've added a new VERIFY digest mode which assumes a digest is present in the input and fails the request with EBADMSG if it doesn't match the internally-computed digest. GCM and CCM already assumed this, and the new AEAD mode requires this for decryption. The new ETA mode now also requires this for decryption, so IPsec and GELI no longer do their own authentication verification. Simple DIGEST operations can also do this, though there are no in-tree consumers. To eventually support some refcounting to close races, the session cookie is now passed to crypto_getop() and clients should no longer set crp_sesssion directly. - Assymteric crypto operation structures should be allocated via crypto_getkreq() and freed via crypto_freekreq(). This permits the crypto layer to track open asym requests and close races with a driver trying to unregister while asym requests are in flight. - crypto_copyback, crypto_copydata, crypto_apply, and crypto_contiguous_subsegment now accept the 'crp' object as the first parameter instead of individual members. This makes it easier to deal with different buffer types in the future as well as separate input and output buffers. It's also simpler for driver writers to use. - bus_dmamap_load_crp() loads a DMA mapping for a crypto buffer. This understands the various types of buffers so that drivers that use DMA do not have to be aware of different buffer types. - Helper routines now exist to build an auth context for HMAC IPAD and OPAD. This reduces some duplicated work among drivers. - Key buffers are now treated as const throughout the framework and in device drivers. However, session key buffers provided when a session is created are expected to remain alive for the duration of the session. - GCM and CCM sessions now only specify a cipher algorithm and a cipher key. The redundant auth information is not needed or used. - For cryptosoft, split up the code a bit such that the 'process' callback now invokes a function pointer in the session. This function pointer is set based on the mode (in effect) though it simplifies a few edge cases that would otherwise be in the switch in 'process'. It does split up GCM vs CCM which I think is more readable even if there is some duplication. - I changed /dev/crypto to support GMAC requests using CRYPTO_AES_NIST_GMAC as an auth algorithm and updated cryptocheck to work with it. - Combined cipher and auth sessions via /dev/crypto now always use ETA mode. The COP_F_CIPHER_FIRST flag is now a no-op that is ignored. This was actually documented as being true in crypto(4) before, but the code had not implemented this before I added the CIPHER_FIRST flag. - I have not yet updated /dev/crypto to be aware of explicit modes for sessions. I will probably do that at some point in the future as well as teach it about IV/nonce and tag lengths for AEAD so we can support all of the NIST KAT tests for GCM and CCM. - I've split up the exising crypto.9 manpage into several pages of which many are written from scratch. - I have converted all drivers and consumers in the tree and verified that they compile, but I have not tested all of them. I have tested the following drivers: - cryptosoft - aesni (AES only) - blake2 - ccr and the following consumers: - cryptodev - IPsec - ktls_ocf - GELI (lightly) I have not tested the following: - ccp - aesni with sha - hifn - kgssapi_krb5 - ubsec - padlock - safe - armv8_crypto (aarch64) - glxsb (i386) - sec (ppc) - cesa (armv7) - cryptocteon (mips64) - nlmsec (mips64) Discussed with: cem Relnotes: yes Sponsored by: Chelsio Communications Differential Revision: https://reviews.freebsd.org/D23677
2020-03-27 18:25:23 +00:00
case CSP_MODE_AEAD:
case CSP_MODE_ETA:
switch (caead->op) {
case COP_ENCRYPT:
crp->crp_op = CRYPTO_OP_ENCRYPT |
CRYPTO_OP_COMPUTE_DIGEST;
break;
case COP_DECRYPT:
crp->crp_op = CRYPTO_OP_DECRYPT |
CRYPTO_OP_VERIFY_DIGEST;
break;
default:
SDT_PROBE1(opencrypto, dev, ioctl, error, __LINE__);
error = EINVAL;
goto bail;
}
break;
default:
SDT_PROBE1(opencrypto, dev, ioctl, error, __LINE__);
Refactor driver and consumer interfaces for OCF (in-kernel crypto). - The linked list of cryptoini structures used in session initialization is replaced with a new flat structure: struct crypto_session_params. This session includes a new mode to define how the other fields should be interpreted. Available modes include: - COMPRESS (for compression/decompression) - CIPHER (for simply encryption/decryption) - DIGEST (computing and verifying digests) - AEAD (combined auth and encryption such as AES-GCM and AES-CCM) - ETA (combined auth and encryption using encrypt-then-authenticate) Additional modes could be added in the future (e.g. if we wanted to support TLS MtE for AES-CBC in the kernel we could add a new mode for that. TLS modes might also affect how AAD is interpreted, etc.) The flat structure also includes the key lengths and algorithms as before. However, code doesn't have to walk the linked list and switch on the algorithm to determine which key is the auth key vs encryption key. The 'csp_auth_*' fields are always used for auth keys and settings and 'csp_cipher_*' for cipher. (Compression algorithms are stored in csp_cipher_alg.) - Drivers no longer register a list of supported algorithms. This doesn't quite work when you factor in modes (e.g. a driver might support both AES-CBC and SHA2-256-HMAC separately but not combined for ETA). Instead, a new 'crypto_probesession' method has been added to the kobj interface for symmteric crypto drivers. This method returns a negative value on success (similar to how device_probe works) and the crypto framework uses this value to pick the "best" driver. There are three constants for hardware (e.g. ccr), accelerated software (e.g. aesni), and plain software (cryptosoft) that give preference in that order. One effect of this is that if you request only hardware when creating a new session, you will no longer get a session using accelerated software. Another effect is that the default setting to disallow software crypto via /dev/crypto now disables accelerated software. Once a driver is chosen, 'crypto_newsession' is invoked as before. - Crypto operations are now solely described by the flat 'cryptop' structure. The linked list of descriptors has been removed. A separate enum has been added to describe the type of data buffer in use instead of using CRYPTO_F_* flags to make it easier to add more types in the future if needed (e.g. wired userspace buffers for zero-copy). It will also make it easier to re-introduce separate input and output buffers (in-kernel TLS would benefit from this). Try to make the flags related to IV handling less insane: - CRYPTO_F_IV_SEPARATE means that the IV is stored in the 'crp_iv' member of the operation structure. If this flag is not set, the IV is stored in the data buffer at the 'crp_iv_start' offset. - CRYPTO_F_IV_GENERATE means that a random IV should be generated and stored into the data buffer. This cannot be used with CRYPTO_F_IV_SEPARATE. If a consumer wants to deal with explicit vs implicit IVs, etc. it can always generate the IV however it needs and store partial IVs in the buffer and the full IV/nonce in crp_iv and set CRYPTO_F_IV_SEPARATE. The layout of the buffer is now described via fields in cryptop. crp_aad_start and crp_aad_length define the boundaries of any AAD. Previously with GCM and CCM you defined an auth crd with this range, but for ETA your auth crd had to span both the AAD and plaintext (and they had to be adjacent). crp_payload_start and crp_payload_length define the boundaries of the plaintext/ciphertext. Modes that only do a single operation (COMPRESS, CIPHER, DIGEST) should only use this region and leave the AAD region empty. If a digest is present (or should be generated), it's starting location is marked by crp_digest_start. Instead of using the CRD_F_ENCRYPT flag to determine the direction of the operation, cryptop now includes an 'op' field defining the operation to perform. For digests I've added a new VERIFY digest mode which assumes a digest is present in the input and fails the request with EBADMSG if it doesn't match the internally-computed digest. GCM and CCM already assumed this, and the new AEAD mode requires this for decryption. The new ETA mode now also requires this for decryption, so IPsec and GELI no longer do their own authentication verification. Simple DIGEST operations can also do this, though there are no in-tree consumers. To eventually support some refcounting to close races, the session cookie is now passed to crypto_getop() and clients should no longer set crp_sesssion directly. - Assymteric crypto operation structures should be allocated via crypto_getkreq() and freed via crypto_freekreq(). This permits the crypto layer to track open asym requests and close races with a driver trying to unregister while asym requests are in flight. - crypto_copyback, crypto_copydata, crypto_apply, and crypto_contiguous_subsegment now accept the 'crp' object as the first parameter instead of individual members. This makes it easier to deal with different buffer types in the future as well as separate input and output buffers. It's also simpler for driver writers to use. - bus_dmamap_load_crp() loads a DMA mapping for a crypto buffer. This understands the various types of buffers so that drivers that use DMA do not have to be aware of different buffer types. - Helper routines now exist to build an auth context for HMAC IPAD and OPAD. This reduces some duplicated work among drivers. - Key buffers are now treated as const throughout the framework and in device drivers. However, session key buffers provided when a session is created are expected to remain alive for the duration of the session. - GCM and CCM sessions now only specify a cipher algorithm and a cipher key. The redundant auth information is not needed or used. - For cryptosoft, split up the code a bit such that the 'process' callback now invokes a function pointer in the session. This function pointer is set based on the mode (in effect) though it simplifies a few edge cases that would otherwise be in the switch in 'process'. It does split up GCM vs CCM which I think is more readable even if there is some duplication. - I changed /dev/crypto to support GMAC requests using CRYPTO_AES_NIST_GMAC as an auth algorithm and updated cryptocheck to work with it. - Combined cipher and auth sessions via /dev/crypto now always use ETA mode. The COP_F_CIPHER_FIRST flag is now a no-op that is ignored. This was actually documented as being true in crypto(4) before, but the code had not implemented this before I added the CIPHER_FIRST flag. - I have not yet updated /dev/crypto to be aware of explicit modes for sessions. I will probably do that at some point in the future as well as teach it about IV/nonce and tag lengths for AEAD so we can support all of the NIST KAT tests for GCM and CCM. - I've split up the exising crypto.9 manpage into several pages of which many are written from scratch. - I have converted all drivers and consumers in the tree and verified that they compile, but I have not tested all of them. I have tested the following drivers: - cryptosoft - aesni (AES only) - blake2 - ccr and the following consumers: - cryptodev - IPsec - ktls_ocf - GELI (lightly) I have not tested the following: - ccp - aesni with sha - hifn - kgssapi_krb5 - ubsec - padlock - safe - armv8_crypto (aarch64) - glxsb (i386) - sec (ppc) - cesa (armv7) - cryptocteon (mips64) - nlmsec (mips64) Discussed with: cem Relnotes: yes Sponsored by: Chelsio Communications Differential Revision: https://reviews.freebsd.org/D23677
2020-03-27 18:25:23 +00:00
error = EINVAL;
2014-12-12 19:56:36 +00:00
goto bail;
}
2014-12-12 19:56:36 +00:00
Refactor driver and consumer interfaces for OCF (in-kernel crypto). - The linked list of cryptoini structures used in session initialization is replaced with a new flat structure: struct crypto_session_params. This session includes a new mode to define how the other fields should be interpreted. Available modes include: - COMPRESS (for compression/decompression) - CIPHER (for simply encryption/decryption) - DIGEST (computing and verifying digests) - AEAD (combined auth and encryption such as AES-GCM and AES-CCM) - ETA (combined auth and encryption using encrypt-then-authenticate) Additional modes could be added in the future (e.g. if we wanted to support TLS MtE for AES-CBC in the kernel we could add a new mode for that. TLS modes might also affect how AAD is interpreted, etc.) The flat structure also includes the key lengths and algorithms as before. However, code doesn't have to walk the linked list and switch on the algorithm to determine which key is the auth key vs encryption key. The 'csp_auth_*' fields are always used for auth keys and settings and 'csp_cipher_*' for cipher. (Compression algorithms are stored in csp_cipher_alg.) - Drivers no longer register a list of supported algorithms. This doesn't quite work when you factor in modes (e.g. a driver might support both AES-CBC and SHA2-256-HMAC separately but not combined for ETA). Instead, a new 'crypto_probesession' method has been added to the kobj interface for symmteric crypto drivers. This method returns a negative value on success (similar to how device_probe works) and the crypto framework uses this value to pick the "best" driver. There are three constants for hardware (e.g. ccr), accelerated software (e.g. aesni), and plain software (cryptosoft) that give preference in that order. One effect of this is that if you request only hardware when creating a new session, you will no longer get a session using accelerated software. Another effect is that the default setting to disallow software crypto via /dev/crypto now disables accelerated software. Once a driver is chosen, 'crypto_newsession' is invoked as before. - Crypto operations are now solely described by the flat 'cryptop' structure. The linked list of descriptors has been removed. A separate enum has been added to describe the type of data buffer in use instead of using CRYPTO_F_* flags to make it easier to add more types in the future if needed (e.g. wired userspace buffers for zero-copy). It will also make it easier to re-introduce separate input and output buffers (in-kernel TLS would benefit from this). Try to make the flags related to IV handling less insane: - CRYPTO_F_IV_SEPARATE means that the IV is stored in the 'crp_iv' member of the operation structure. If this flag is not set, the IV is stored in the data buffer at the 'crp_iv_start' offset. - CRYPTO_F_IV_GENERATE means that a random IV should be generated and stored into the data buffer. This cannot be used with CRYPTO_F_IV_SEPARATE. If a consumer wants to deal with explicit vs implicit IVs, etc. it can always generate the IV however it needs and store partial IVs in the buffer and the full IV/nonce in crp_iv and set CRYPTO_F_IV_SEPARATE. The layout of the buffer is now described via fields in cryptop. crp_aad_start and crp_aad_length define the boundaries of any AAD. Previously with GCM and CCM you defined an auth crd with this range, but for ETA your auth crd had to span both the AAD and plaintext (and they had to be adjacent). crp_payload_start and crp_payload_length define the boundaries of the plaintext/ciphertext. Modes that only do a single operation (COMPRESS, CIPHER, DIGEST) should only use this region and leave the AAD region empty. If a digest is present (or should be generated), it's starting location is marked by crp_digest_start. Instead of using the CRD_F_ENCRYPT flag to determine the direction of the operation, cryptop now includes an 'op' field defining the operation to perform. For digests I've added a new VERIFY digest mode which assumes a digest is present in the input and fails the request with EBADMSG if it doesn't match the internally-computed digest. GCM and CCM already assumed this, and the new AEAD mode requires this for decryption. The new ETA mode now also requires this for decryption, so IPsec and GELI no longer do their own authentication verification. Simple DIGEST operations can also do this, though there are no in-tree consumers. To eventually support some refcounting to close races, the session cookie is now passed to crypto_getop() and clients should no longer set crp_sesssion directly. - Assymteric crypto operation structures should be allocated via crypto_getkreq() and freed via crypto_freekreq(). This permits the crypto layer to track open asym requests and close races with a driver trying to unregister while asym requests are in flight. - crypto_copyback, crypto_copydata, crypto_apply, and crypto_contiguous_subsegment now accept the 'crp' object as the first parameter instead of individual members. This makes it easier to deal with different buffer types in the future as well as separate input and output buffers. It's also simpler for driver writers to use. - bus_dmamap_load_crp() loads a DMA mapping for a crypto buffer. This understands the various types of buffers so that drivers that use DMA do not have to be aware of different buffer types. - Helper routines now exist to build an auth context for HMAC IPAD and OPAD. This reduces some duplicated work among drivers. - Key buffers are now treated as const throughout the framework and in device drivers. However, session key buffers provided when a session is created are expected to remain alive for the duration of the session. - GCM and CCM sessions now only specify a cipher algorithm and a cipher key. The redundant auth information is not needed or used. - For cryptosoft, split up the code a bit such that the 'process' callback now invokes a function pointer in the session. This function pointer is set based on the mode (in effect) though it simplifies a few edge cases that would otherwise be in the switch in 'process'. It does split up GCM vs CCM which I think is more readable even if there is some duplication. - I changed /dev/crypto to support GMAC requests using CRYPTO_AES_NIST_GMAC as an auth algorithm and updated cryptocheck to work with it. - Combined cipher and auth sessions via /dev/crypto now always use ETA mode. The COP_F_CIPHER_FIRST flag is now a no-op that is ignored. This was actually documented as being true in crypto(4) before, but the code had not implemented this before I added the CIPHER_FIRST flag. - I have not yet updated /dev/crypto to be aware of explicit modes for sessions. I will probably do that at some point in the future as well as teach it about IV/nonce and tag lengths for AEAD so we can support all of the NIST KAT tests for GCM and CCM. - I've split up the exising crypto.9 manpage into several pages of which many are written from scratch. - I have converted all drivers and consumers in the tree and verified that they compile, but I have not tested all of them. I have tested the following drivers: - cryptosoft - aesni (AES only) - blake2 - ccr and the following consumers: - cryptodev - IPsec - ktls_ocf - GELI (lightly) I have not tested the following: - ccp - aesni with sha - hifn - kgssapi_krb5 - ubsec - padlock - safe - armv8_crypto (aarch64) - glxsb (i386) - sec (ppc) - cesa (armv7) - cryptocteon (mips64) - nlmsec (mips64) Discussed with: cem Relnotes: yes Sponsored by: Chelsio Communications Differential Revision: https://reviews.freebsd.org/D23677
2020-03-27 18:25:23 +00:00
crp->crp_flags = CRYPTO_F_CBIMM | (caead->flags & COP_F_BATCH);
crypto_use_buf(crp, cod->buf, crp->crp_payload_start + caead->len +
Add support for optional separate output buffers to in-kernel crypto. Some crypto consumers such as GELI and KTLS for file-backed sendfile need to store their output in a separate buffer from the input. Currently these consumers copy the contents of the input buffer into the output buffer and queue an in-place crypto operation on the output buffer. Using a separate output buffer avoids this copy. - Create a new 'struct crypto_buffer' describing a crypto buffer containing a type and type-specific fields. crp_ilen is gone, instead buffers that use a flat kernel buffer have a cb_buf_len field for their length. The length of other buffer types is inferred from the backing store (e.g. uio_resid for a uio). Requests now have two such structures: crp_buf for the input buffer, and crp_obuf for the output buffer. - Consumers now use helper functions (crypto_use_*, e.g. crypto_use_mbuf()) to configure the input buffer. If an output buffer is not configured, the request still modifies the input buffer in-place. A consumer uses a second set of helper functions (crypto_use_output_*) to configure an output buffer. - Consumers must request support for separate output buffers when creating a crypto session via the CSP_F_SEPARATE_OUTPUT flag and are only permitted to queue a request with a separate output buffer on sessions with this flag set. Existing drivers already reject sessions with unknown flags, so this permits drivers to be modified to support this extension without requiring all drivers to change. - Several data-related functions now have matching versions that operate on an explicit buffer (e.g. crypto_apply_buf, crypto_contiguous_subsegment_buf, bus_dma_load_crp_buf). - Most of the existing data-related functions operate on the input buffer. However crypto_copyback always writes to the output buffer if a request uses a separate output buffer. - For the regions in input/output buffers, the following conventions are followed: - AAD and IV are always present in input only and their fields are offsets into the input buffer. - payload is always present in both buffers. If a request uses a separate output buffer, it must set a new crp_payload_start_output field to the offset of the payload in the output buffer. - digest is in the input buffer for verify operations, and in the output buffer for compute operations. crp_digest_start is relative to the appropriate buffer. - Add a crypto buffer cursor abstraction. This is a more general form of some bits in the cryptosoft driver that tried to always use uio's. However, compared to the original code, this avoids rewalking the uio iovec array for requests with multiple vectors. It also avoids allocate an iovec array for mbufs and populating it by instead walking the mbuf chain directly. - Update the cryptosoft(4) driver to support separate output buffers making use of the cursor abstraction. Sponsored by: Netflix Differential Revision: https://reviews.freebsd.org/D24545
2020-05-25 22:12:04 +00:00
cse->hashsize);
if (cod->obuf != NULL)
crypto_use_output_buf(crp, cod->obuf, caead->len +
cse->hashsize);
crp->crp_callback = cryptodev_cb;
crp->crp_opaque = cod;
2014-12-12 19:56:36 +00:00
if (caead->iv) {
Refactor driver and consumer interfaces for OCF (in-kernel crypto). - The linked list of cryptoini structures used in session initialization is replaced with a new flat structure: struct crypto_session_params. This session includes a new mode to define how the other fields should be interpreted. Available modes include: - COMPRESS (for compression/decompression) - CIPHER (for simply encryption/decryption) - DIGEST (computing and verifying digests) - AEAD (combined auth and encryption such as AES-GCM and AES-CCM) - ETA (combined auth and encryption using encrypt-then-authenticate) Additional modes could be added in the future (e.g. if we wanted to support TLS MtE for AES-CBC in the kernel we could add a new mode for that. TLS modes might also affect how AAD is interpreted, etc.) The flat structure also includes the key lengths and algorithms as before. However, code doesn't have to walk the linked list and switch on the algorithm to determine which key is the auth key vs encryption key. The 'csp_auth_*' fields are always used for auth keys and settings and 'csp_cipher_*' for cipher. (Compression algorithms are stored in csp_cipher_alg.) - Drivers no longer register a list of supported algorithms. This doesn't quite work when you factor in modes (e.g. a driver might support both AES-CBC and SHA2-256-HMAC separately but not combined for ETA). Instead, a new 'crypto_probesession' method has been added to the kobj interface for symmteric crypto drivers. This method returns a negative value on success (similar to how device_probe works) and the crypto framework uses this value to pick the "best" driver. There are three constants for hardware (e.g. ccr), accelerated software (e.g. aesni), and plain software (cryptosoft) that give preference in that order. One effect of this is that if you request only hardware when creating a new session, you will no longer get a session using accelerated software. Another effect is that the default setting to disallow software crypto via /dev/crypto now disables accelerated software. Once a driver is chosen, 'crypto_newsession' is invoked as before. - Crypto operations are now solely described by the flat 'cryptop' structure. The linked list of descriptors has been removed. A separate enum has been added to describe the type of data buffer in use instead of using CRYPTO_F_* flags to make it easier to add more types in the future if needed (e.g. wired userspace buffers for zero-copy). It will also make it easier to re-introduce separate input and output buffers (in-kernel TLS would benefit from this). Try to make the flags related to IV handling less insane: - CRYPTO_F_IV_SEPARATE means that the IV is stored in the 'crp_iv' member of the operation structure. If this flag is not set, the IV is stored in the data buffer at the 'crp_iv_start' offset. - CRYPTO_F_IV_GENERATE means that a random IV should be generated and stored into the data buffer. This cannot be used with CRYPTO_F_IV_SEPARATE. If a consumer wants to deal with explicit vs implicit IVs, etc. it can always generate the IV however it needs and store partial IVs in the buffer and the full IV/nonce in crp_iv and set CRYPTO_F_IV_SEPARATE. The layout of the buffer is now described via fields in cryptop. crp_aad_start and crp_aad_length define the boundaries of any AAD. Previously with GCM and CCM you defined an auth crd with this range, but for ETA your auth crd had to span both the AAD and plaintext (and they had to be adjacent). crp_payload_start and crp_payload_length define the boundaries of the plaintext/ciphertext. Modes that only do a single operation (COMPRESS, CIPHER, DIGEST) should only use this region and leave the AAD region empty. If a digest is present (or should be generated), it's starting location is marked by crp_digest_start. Instead of using the CRD_F_ENCRYPT flag to determine the direction of the operation, cryptop now includes an 'op' field defining the operation to perform. For digests I've added a new VERIFY digest mode which assumes a digest is present in the input and fails the request with EBADMSG if it doesn't match the internally-computed digest. GCM and CCM already assumed this, and the new AEAD mode requires this for decryption. The new ETA mode now also requires this for decryption, so IPsec and GELI no longer do their own authentication verification. Simple DIGEST operations can also do this, though there are no in-tree consumers. To eventually support some refcounting to close races, the session cookie is now passed to crypto_getop() and clients should no longer set crp_sesssion directly. - Assymteric crypto operation structures should be allocated via crypto_getkreq() and freed via crypto_freekreq(). This permits the crypto layer to track open asym requests and close races with a driver trying to unregister while asym requests are in flight. - crypto_copyback, crypto_copydata, crypto_apply, and crypto_contiguous_subsegment now accept the 'crp' object as the first parameter instead of individual members. This makes it easier to deal with different buffer types in the future as well as separate input and output buffers. It's also simpler for driver writers to use. - bus_dmamap_load_crp() loads a DMA mapping for a crypto buffer. This understands the various types of buffers so that drivers that use DMA do not have to be aware of different buffer types. - Helper routines now exist to build an auth context for HMAC IPAD and OPAD. This reduces some duplicated work among drivers. - Key buffers are now treated as const throughout the framework and in device drivers. However, session key buffers provided when a session is created are expected to remain alive for the duration of the session. - GCM and CCM sessions now only specify a cipher algorithm and a cipher key. The redundant auth information is not needed or used. - For cryptosoft, split up the code a bit such that the 'process' callback now invokes a function pointer in the session. This function pointer is set based on the mode (in effect) though it simplifies a few edge cases that would otherwise be in the switch in 'process'. It does split up GCM vs CCM which I think is more readable even if there is some duplication. - I changed /dev/crypto to support GMAC requests using CRYPTO_AES_NIST_GMAC as an auth algorithm and updated cryptocheck to work with it. - Combined cipher and auth sessions via /dev/crypto now always use ETA mode. The COP_F_CIPHER_FIRST flag is now a no-op that is ignored. This was actually documented as being true in crypto(4) before, but the code had not implemented this before I added the CIPHER_FIRST flag. - I have not yet updated /dev/crypto to be aware of explicit modes for sessions. I will probably do that at some point in the future as well as teach it about IV/nonce and tag lengths for AEAD so we can support all of the NIST KAT tests for GCM and CCM. - I've split up the exising crypto.9 manpage into several pages of which many are written from scratch. - I have converted all drivers and consumers in the tree and verified that they compile, but I have not tested all of them. I have tested the following drivers: - cryptosoft - aesni (AES only) - blake2 - ccr and the following consumers: - cryptodev - IPsec - ktls_ocf - GELI (lightly) I have not tested the following: - ccp - aesni with sha - hifn - kgssapi_krb5 - ubsec - padlock - safe - armv8_crypto (aarch64) - glxsb (i386) - sec (ppc) - cesa (armv7) - cryptocteon (mips64) - nlmsec (mips64) Discussed with: cem Relnotes: yes Sponsored by: Chelsio Communications Differential Revision: https://reviews.freebsd.org/D23677
2020-03-27 18:25:23 +00:00
/*
* Permit a 16-byte IV for AES-XTS, but only use the
* first 8 bytes as a block number.
*/
if (csp->csp_mode == CSP_MODE_ETA &&
csp->csp_cipher_alg == CRYPTO_AES_XTS &&
caead->ivlen == AES_BLOCK_LEN)
Refactor driver and consumer interfaces for OCF (in-kernel crypto). - The linked list of cryptoini structures used in session initialization is replaced with a new flat structure: struct crypto_session_params. This session includes a new mode to define how the other fields should be interpreted. Available modes include: - COMPRESS (for compression/decompression) - CIPHER (for simply encryption/decryption) - DIGEST (computing and verifying digests) - AEAD (combined auth and encryption such as AES-GCM and AES-CCM) - ETA (combined auth and encryption using encrypt-then-authenticate) Additional modes could be added in the future (e.g. if we wanted to support TLS MtE for AES-CBC in the kernel we could add a new mode for that. TLS modes might also affect how AAD is interpreted, etc.) The flat structure also includes the key lengths and algorithms as before. However, code doesn't have to walk the linked list and switch on the algorithm to determine which key is the auth key vs encryption key. The 'csp_auth_*' fields are always used for auth keys and settings and 'csp_cipher_*' for cipher. (Compression algorithms are stored in csp_cipher_alg.) - Drivers no longer register a list of supported algorithms. This doesn't quite work when you factor in modes (e.g. a driver might support both AES-CBC and SHA2-256-HMAC separately but not combined for ETA). Instead, a new 'crypto_probesession' method has been added to the kobj interface for symmteric crypto drivers. This method returns a negative value on success (similar to how device_probe works) and the crypto framework uses this value to pick the "best" driver. There are three constants for hardware (e.g. ccr), accelerated software (e.g. aesni), and plain software (cryptosoft) that give preference in that order. One effect of this is that if you request only hardware when creating a new session, you will no longer get a session using accelerated software. Another effect is that the default setting to disallow software crypto via /dev/crypto now disables accelerated software. Once a driver is chosen, 'crypto_newsession' is invoked as before. - Crypto operations are now solely described by the flat 'cryptop' structure. The linked list of descriptors has been removed. A separate enum has been added to describe the type of data buffer in use instead of using CRYPTO_F_* flags to make it easier to add more types in the future if needed (e.g. wired userspace buffers for zero-copy). It will also make it easier to re-introduce separate input and output buffers (in-kernel TLS would benefit from this). Try to make the flags related to IV handling less insane: - CRYPTO_F_IV_SEPARATE means that the IV is stored in the 'crp_iv' member of the operation structure. If this flag is not set, the IV is stored in the data buffer at the 'crp_iv_start' offset. - CRYPTO_F_IV_GENERATE means that a random IV should be generated and stored into the data buffer. This cannot be used with CRYPTO_F_IV_SEPARATE. If a consumer wants to deal with explicit vs implicit IVs, etc. it can always generate the IV however it needs and store partial IVs in the buffer and the full IV/nonce in crp_iv and set CRYPTO_F_IV_SEPARATE. The layout of the buffer is now described via fields in cryptop. crp_aad_start and crp_aad_length define the boundaries of any AAD. Previously with GCM and CCM you defined an auth crd with this range, but for ETA your auth crd had to span both the AAD and plaintext (and they had to be adjacent). crp_payload_start and crp_payload_length define the boundaries of the plaintext/ciphertext. Modes that only do a single operation (COMPRESS, CIPHER, DIGEST) should only use this region and leave the AAD region empty. If a digest is present (or should be generated), it's starting location is marked by crp_digest_start. Instead of using the CRD_F_ENCRYPT flag to determine the direction of the operation, cryptop now includes an 'op' field defining the operation to perform. For digests I've added a new VERIFY digest mode which assumes a digest is present in the input and fails the request with EBADMSG if it doesn't match the internally-computed digest. GCM and CCM already assumed this, and the new AEAD mode requires this for decryption. The new ETA mode now also requires this for decryption, so IPsec and GELI no longer do their own authentication verification. Simple DIGEST operations can also do this, though there are no in-tree consumers. To eventually support some refcounting to close races, the session cookie is now passed to crypto_getop() and clients should no longer set crp_sesssion directly. - Assymteric crypto operation structures should be allocated via crypto_getkreq() and freed via crypto_freekreq(). This permits the crypto layer to track open asym requests and close races with a driver trying to unregister while asym requests are in flight. - crypto_copyback, crypto_copydata, crypto_apply, and crypto_contiguous_subsegment now accept the 'crp' object as the first parameter instead of individual members. This makes it easier to deal with different buffer types in the future as well as separate input and output buffers. It's also simpler for driver writers to use. - bus_dmamap_load_crp() loads a DMA mapping for a crypto buffer. This understands the various types of buffers so that drivers that use DMA do not have to be aware of different buffer types. - Helper routines now exist to build an auth context for HMAC IPAD and OPAD. This reduces some duplicated work among drivers. - Key buffers are now treated as const throughout the framework and in device drivers. However, session key buffers provided when a session is created are expected to remain alive for the duration of the session. - GCM and CCM sessions now only specify a cipher algorithm and a cipher key. The redundant auth information is not needed or used. - For cryptosoft, split up the code a bit such that the 'process' callback now invokes a function pointer in the session. This function pointer is set based on the mode (in effect) though it simplifies a few edge cases that would otherwise be in the switch in 'process'. It does split up GCM vs CCM which I think is more readable even if there is some duplication. - I changed /dev/crypto to support GMAC requests using CRYPTO_AES_NIST_GMAC as an auth algorithm and updated cryptocheck to work with it. - Combined cipher and auth sessions via /dev/crypto now always use ETA mode. The COP_F_CIPHER_FIRST flag is now a no-op that is ignored. This was actually documented as being true in crypto(4) before, but the code had not implemented this before I added the CIPHER_FIRST flag. - I have not yet updated /dev/crypto to be aware of explicit modes for sessions. I will probably do that at some point in the future as well as teach it about IV/nonce and tag lengths for AEAD so we can support all of the NIST KAT tests for GCM and CCM. - I've split up the exising crypto.9 manpage into several pages of which many are written from scratch. - I have converted all drivers and consumers in the tree and verified that they compile, but I have not tested all of them. I have tested the following drivers: - cryptosoft - aesni (AES only) - blake2 - ccr and the following consumers: - cryptodev - IPsec - ktls_ocf - GELI (lightly) I have not tested the following: - ccp - aesni with sha - hifn - kgssapi_krb5 - ubsec - padlock - safe - armv8_crypto (aarch64) - glxsb (i386) - sec (ppc) - cesa (armv7) - cryptocteon (mips64) - nlmsec (mips64) Discussed with: cem Relnotes: yes Sponsored by: Chelsio Communications Differential Revision: https://reviews.freebsd.org/D23677
2020-03-27 18:25:23 +00:00
caead->ivlen = AES_XTS_IV_LEN;
if (cse->ivsize == 0) {
SDT_PROBE1(opencrypto, dev, ioctl, error, __LINE__);
error = EINVAL;
goto bail;
}
Refactor driver and consumer interfaces for OCF (in-kernel crypto). - The linked list of cryptoini structures used in session initialization is replaced with a new flat structure: struct crypto_session_params. This session includes a new mode to define how the other fields should be interpreted. Available modes include: - COMPRESS (for compression/decompression) - CIPHER (for simply encryption/decryption) - DIGEST (computing and verifying digests) - AEAD (combined auth and encryption such as AES-GCM and AES-CCM) - ETA (combined auth and encryption using encrypt-then-authenticate) Additional modes could be added in the future (e.g. if we wanted to support TLS MtE for AES-CBC in the kernel we could add a new mode for that. TLS modes might also affect how AAD is interpreted, etc.) The flat structure also includes the key lengths and algorithms as before. However, code doesn't have to walk the linked list and switch on the algorithm to determine which key is the auth key vs encryption key. The 'csp_auth_*' fields are always used for auth keys and settings and 'csp_cipher_*' for cipher. (Compression algorithms are stored in csp_cipher_alg.) - Drivers no longer register a list of supported algorithms. This doesn't quite work when you factor in modes (e.g. a driver might support both AES-CBC and SHA2-256-HMAC separately but not combined for ETA). Instead, a new 'crypto_probesession' method has been added to the kobj interface for symmteric crypto drivers. This method returns a negative value on success (similar to how device_probe works) and the crypto framework uses this value to pick the "best" driver. There are three constants for hardware (e.g. ccr), accelerated software (e.g. aesni), and plain software (cryptosoft) that give preference in that order. One effect of this is that if you request only hardware when creating a new session, you will no longer get a session using accelerated software. Another effect is that the default setting to disallow software crypto via /dev/crypto now disables accelerated software. Once a driver is chosen, 'crypto_newsession' is invoked as before. - Crypto operations are now solely described by the flat 'cryptop' structure. The linked list of descriptors has been removed. A separate enum has been added to describe the type of data buffer in use instead of using CRYPTO_F_* flags to make it easier to add more types in the future if needed (e.g. wired userspace buffers for zero-copy). It will also make it easier to re-introduce separate input and output buffers (in-kernel TLS would benefit from this). Try to make the flags related to IV handling less insane: - CRYPTO_F_IV_SEPARATE means that the IV is stored in the 'crp_iv' member of the operation structure. If this flag is not set, the IV is stored in the data buffer at the 'crp_iv_start' offset. - CRYPTO_F_IV_GENERATE means that a random IV should be generated and stored into the data buffer. This cannot be used with CRYPTO_F_IV_SEPARATE. If a consumer wants to deal with explicit vs implicit IVs, etc. it can always generate the IV however it needs and store partial IVs in the buffer and the full IV/nonce in crp_iv and set CRYPTO_F_IV_SEPARATE. The layout of the buffer is now described via fields in cryptop. crp_aad_start and crp_aad_length define the boundaries of any AAD. Previously with GCM and CCM you defined an auth crd with this range, but for ETA your auth crd had to span both the AAD and plaintext (and they had to be adjacent). crp_payload_start and crp_payload_length define the boundaries of the plaintext/ciphertext. Modes that only do a single operation (COMPRESS, CIPHER, DIGEST) should only use this region and leave the AAD region empty. If a digest is present (or should be generated), it's starting location is marked by crp_digest_start. Instead of using the CRD_F_ENCRYPT flag to determine the direction of the operation, cryptop now includes an 'op' field defining the operation to perform. For digests I've added a new VERIFY digest mode which assumes a digest is present in the input and fails the request with EBADMSG if it doesn't match the internally-computed digest. GCM and CCM already assumed this, and the new AEAD mode requires this for decryption. The new ETA mode now also requires this for decryption, so IPsec and GELI no longer do their own authentication verification. Simple DIGEST operations can also do this, though there are no in-tree consumers. To eventually support some refcounting to close races, the session cookie is now passed to crypto_getop() and clients should no longer set crp_sesssion directly. - Assymteric crypto operation structures should be allocated via crypto_getkreq() and freed via crypto_freekreq(). This permits the crypto layer to track open asym requests and close races with a driver trying to unregister while asym requests are in flight. - crypto_copyback, crypto_copydata, crypto_apply, and crypto_contiguous_subsegment now accept the 'crp' object as the first parameter instead of individual members. This makes it easier to deal with different buffer types in the future as well as separate input and output buffers. It's also simpler for driver writers to use. - bus_dmamap_load_crp() loads a DMA mapping for a crypto buffer. This understands the various types of buffers so that drivers that use DMA do not have to be aware of different buffer types. - Helper routines now exist to build an auth context for HMAC IPAD and OPAD. This reduces some duplicated work among drivers. - Key buffers are now treated as const throughout the framework and in device drivers. However, session key buffers provided when a session is created are expected to remain alive for the duration of the session. - GCM and CCM sessions now only specify a cipher algorithm and a cipher key. The redundant auth information is not needed or used. - For cryptosoft, split up the code a bit such that the 'process' callback now invokes a function pointer in the session. This function pointer is set based on the mode (in effect) though it simplifies a few edge cases that would otherwise be in the switch in 'process'. It does split up GCM vs CCM which I think is more readable even if there is some duplication. - I changed /dev/crypto to support GMAC requests using CRYPTO_AES_NIST_GMAC as an auth algorithm and updated cryptocheck to work with it. - Combined cipher and auth sessions via /dev/crypto now always use ETA mode. The COP_F_CIPHER_FIRST flag is now a no-op that is ignored. This was actually documented as being true in crypto(4) before, but the code had not implemented this before I added the CIPHER_FIRST flag. - I have not yet updated /dev/crypto to be aware of explicit modes for sessions. I will probably do that at some point in the future as well as teach it about IV/nonce and tag lengths for AEAD so we can support all of the NIST KAT tests for GCM and CCM. - I've split up the exising crypto.9 manpage into several pages of which many are written from scratch. - I have converted all drivers and consumers in the tree and verified that they compile, but I have not tested all of them. I have tested the following drivers: - cryptosoft - aesni (AES only) - blake2 - ccr and the following consumers: - cryptodev - IPsec - ktls_ocf - GELI (lightly) I have not tested the following: - ccp - aesni with sha - hifn - kgssapi_krb5 - ubsec - padlock - safe - armv8_crypto (aarch64) - glxsb (i386) - sec (ppc) - cesa (armv7) - cryptocteon (mips64) - nlmsec (mips64) Discussed with: cem Relnotes: yes Sponsored by: Chelsio Communications Differential Revision: https://reviews.freebsd.org/D23677
2020-03-27 18:25:23 +00:00
if (caead->ivlen != cse->ivsize) {
2014-12-12 19:56:36 +00:00
error = EINVAL;
SDT_PROBE1(opencrypto, dev, ioctl, error, __LINE__);
2014-12-12 19:56:36 +00:00
goto bail;
}
Refactor driver and consumer interfaces for OCF (in-kernel crypto). - The linked list of cryptoini structures used in session initialization is replaced with a new flat structure: struct crypto_session_params. This session includes a new mode to define how the other fields should be interpreted. Available modes include: - COMPRESS (for compression/decompression) - CIPHER (for simply encryption/decryption) - DIGEST (computing and verifying digests) - AEAD (combined auth and encryption such as AES-GCM and AES-CCM) - ETA (combined auth and encryption using encrypt-then-authenticate) Additional modes could be added in the future (e.g. if we wanted to support TLS MtE for AES-CBC in the kernel we could add a new mode for that. TLS modes might also affect how AAD is interpreted, etc.) The flat structure also includes the key lengths and algorithms as before. However, code doesn't have to walk the linked list and switch on the algorithm to determine which key is the auth key vs encryption key. The 'csp_auth_*' fields are always used for auth keys and settings and 'csp_cipher_*' for cipher. (Compression algorithms are stored in csp_cipher_alg.) - Drivers no longer register a list of supported algorithms. This doesn't quite work when you factor in modes (e.g. a driver might support both AES-CBC and SHA2-256-HMAC separately but not combined for ETA). Instead, a new 'crypto_probesession' method has been added to the kobj interface for symmteric crypto drivers. This method returns a negative value on success (similar to how device_probe works) and the crypto framework uses this value to pick the "best" driver. There are three constants for hardware (e.g. ccr), accelerated software (e.g. aesni), and plain software (cryptosoft) that give preference in that order. One effect of this is that if you request only hardware when creating a new session, you will no longer get a session using accelerated software. Another effect is that the default setting to disallow software crypto via /dev/crypto now disables accelerated software. Once a driver is chosen, 'crypto_newsession' is invoked as before. - Crypto operations are now solely described by the flat 'cryptop' structure. The linked list of descriptors has been removed. A separate enum has been added to describe the type of data buffer in use instead of using CRYPTO_F_* flags to make it easier to add more types in the future if needed (e.g. wired userspace buffers for zero-copy). It will also make it easier to re-introduce separate input and output buffers (in-kernel TLS would benefit from this). Try to make the flags related to IV handling less insane: - CRYPTO_F_IV_SEPARATE means that the IV is stored in the 'crp_iv' member of the operation structure. If this flag is not set, the IV is stored in the data buffer at the 'crp_iv_start' offset. - CRYPTO_F_IV_GENERATE means that a random IV should be generated and stored into the data buffer. This cannot be used with CRYPTO_F_IV_SEPARATE. If a consumer wants to deal with explicit vs implicit IVs, etc. it can always generate the IV however it needs and store partial IVs in the buffer and the full IV/nonce in crp_iv and set CRYPTO_F_IV_SEPARATE. The layout of the buffer is now described via fields in cryptop. crp_aad_start and crp_aad_length define the boundaries of any AAD. Previously with GCM and CCM you defined an auth crd with this range, but for ETA your auth crd had to span both the AAD and plaintext (and they had to be adjacent). crp_payload_start and crp_payload_length define the boundaries of the plaintext/ciphertext. Modes that only do a single operation (COMPRESS, CIPHER, DIGEST) should only use this region and leave the AAD region empty. If a digest is present (or should be generated), it's starting location is marked by crp_digest_start. Instead of using the CRD_F_ENCRYPT flag to determine the direction of the operation, cryptop now includes an 'op' field defining the operation to perform. For digests I've added a new VERIFY digest mode which assumes a digest is present in the input and fails the request with EBADMSG if it doesn't match the internally-computed digest. GCM and CCM already assumed this, and the new AEAD mode requires this for decryption. The new ETA mode now also requires this for decryption, so IPsec and GELI no longer do their own authentication verification. Simple DIGEST operations can also do this, though there are no in-tree consumers. To eventually support some refcounting to close races, the session cookie is now passed to crypto_getop() and clients should no longer set crp_sesssion directly. - Assymteric crypto operation structures should be allocated via crypto_getkreq() and freed via crypto_freekreq(). This permits the crypto layer to track open asym requests and close races with a driver trying to unregister while asym requests are in flight. - crypto_copyback, crypto_copydata, crypto_apply, and crypto_contiguous_subsegment now accept the 'crp' object as the first parameter instead of individual members. This makes it easier to deal with different buffer types in the future as well as separate input and output buffers. It's also simpler for driver writers to use. - bus_dmamap_load_crp() loads a DMA mapping for a crypto buffer. This understands the various types of buffers so that drivers that use DMA do not have to be aware of different buffer types. - Helper routines now exist to build an auth context for HMAC IPAD and OPAD. This reduces some duplicated work among drivers. - Key buffers are now treated as const throughout the framework and in device drivers. However, session key buffers provided when a session is created are expected to remain alive for the duration of the session. - GCM and CCM sessions now only specify a cipher algorithm and a cipher key. The redundant auth information is not needed or used. - For cryptosoft, split up the code a bit such that the 'process' callback now invokes a function pointer in the session. This function pointer is set based on the mode (in effect) though it simplifies a few edge cases that would otherwise be in the switch in 'process'. It does split up GCM vs CCM which I think is more readable even if there is some duplication. - I changed /dev/crypto to support GMAC requests using CRYPTO_AES_NIST_GMAC as an auth algorithm and updated cryptocheck to work with it. - Combined cipher and auth sessions via /dev/crypto now always use ETA mode. The COP_F_CIPHER_FIRST flag is now a no-op that is ignored. This was actually documented as being true in crypto(4) before, but the code had not implemented this before I added the CIPHER_FIRST flag. - I have not yet updated /dev/crypto to be aware of explicit modes for sessions. I will probably do that at some point in the future as well as teach it about IV/nonce and tag lengths for AEAD so we can support all of the NIST KAT tests for GCM and CCM. - I've split up the exising crypto.9 manpage into several pages of which many are written from scratch. - I have converted all drivers and consumers in the tree and verified that they compile, but I have not tested all of them. I have tested the following drivers: - cryptosoft - aesni (AES only) - blake2 - ccr and the following consumers: - cryptodev - IPsec - ktls_ocf - GELI (lightly) I have not tested the following: - ccp - aesni with sha - hifn - kgssapi_krb5 - ubsec - padlock - safe - armv8_crypto (aarch64) - glxsb (i386) - sec (ppc) - cesa (armv7) - cryptocteon (mips64) - nlmsec (mips64) Discussed with: cem Relnotes: yes Sponsored by: Chelsio Communications Differential Revision: https://reviews.freebsd.org/D23677
2020-03-27 18:25:23 +00:00
error = copyin(caead->iv, crp->crp_iv, cse->ivsize);
if (error) {
SDT_PROBE1(opencrypto, dev, ioctl, error, __LINE__);
2014-12-12 19:56:36 +00:00
goto bail;
}
Refactor driver and consumer interfaces for OCF (in-kernel crypto). - The linked list of cryptoini structures used in session initialization is replaced with a new flat structure: struct crypto_session_params. This session includes a new mode to define how the other fields should be interpreted. Available modes include: - COMPRESS (for compression/decompression) - CIPHER (for simply encryption/decryption) - DIGEST (computing and verifying digests) - AEAD (combined auth and encryption such as AES-GCM and AES-CCM) - ETA (combined auth and encryption using encrypt-then-authenticate) Additional modes could be added in the future (e.g. if we wanted to support TLS MtE for AES-CBC in the kernel we could add a new mode for that. TLS modes might also affect how AAD is interpreted, etc.) The flat structure also includes the key lengths and algorithms as before. However, code doesn't have to walk the linked list and switch on the algorithm to determine which key is the auth key vs encryption key. The 'csp_auth_*' fields are always used for auth keys and settings and 'csp_cipher_*' for cipher. (Compression algorithms are stored in csp_cipher_alg.) - Drivers no longer register a list of supported algorithms. This doesn't quite work when you factor in modes (e.g. a driver might support both AES-CBC and SHA2-256-HMAC separately but not combined for ETA). Instead, a new 'crypto_probesession' method has been added to the kobj interface for symmteric crypto drivers. This method returns a negative value on success (similar to how device_probe works) and the crypto framework uses this value to pick the "best" driver. There are three constants for hardware (e.g. ccr), accelerated software (e.g. aesni), and plain software (cryptosoft) that give preference in that order. One effect of this is that if you request only hardware when creating a new session, you will no longer get a session using accelerated software. Another effect is that the default setting to disallow software crypto via /dev/crypto now disables accelerated software. Once a driver is chosen, 'crypto_newsession' is invoked as before. - Crypto operations are now solely described by the flat 'cryptop' structure. The linked list of descriptors has been removed. A separate enum has been added to describe the type of data buffer in use instead of using CRYPTO_F_* flags to make it easier to add more types in the future if needed (e.g. wired userspace buffers for zero-copy). It will also make it easier to re-introduce separate input and output buffers (in-kernel TLS would benefit from this). Try to make the flags related to IV handling less insane: - CRYPTO_F_IV_SEPARATE means that the IV is stored in the 'crp_iv' member of the operation structure. If this flag is not set, the IV is stored in the data buffer at the 'crp_iv_start' offset. - CRYPTO_F_IV_GENERATE means that a random IV should be generated and stored into the data buffer. This cannot be used with CRYPTO_F_IV_SEPARATE. If a consumer wants to deal with explicit vs implicit IVs, etc. it can always generate the IV however it needs and store partial IVs in the buffer and the full IV/nonce in crp_iv and set CRYPTO_F_IV_SEPARATE. The layout of the buffer is now described via fields in cryptop. crp_aad_start and crp_aad_length define the boundaries of any AAD. Previously with GCM and CCM you defined an auth crd with this range, but for ETA your auth crd had to span both the AAD and plaintext (and they had to be adjacent). crp_payload_start and crp_payload_length define the boundaries of the plaintext/ciphertext. Modes that only do a single operation (COMPRESS, CIPHER, DIGEST) should only use this region and leave the AAD region empty. If a digest is present (or should be generated), it's starting location is marked by crp_digest_start. Instead of using the CRD_F_ENCRYPT flag to determine the direction of the operation, cryptop now includes an 'op' field defining the operation to perform. For digests I've added a new VERIFY digest mode which assumes a digest is present in the input and fails the request with EBADMSG if it doesn't match the internally-computed digest. GCM and CCM already assumed this, and the new AEAD mode requires this for decryption. The new ETA mode now also requires this for decryption, so IPsec and GELI no longer do their own authentication verification. Simple DIGEST operations can also do this, though there are no in-tree consumers. To eventually support some refcounting to close races, the session cookie is now passed to crypto_getop() and clients should no longer set crp_sesssion directly. - Assymteric crypto operation structures should be allocated via crypto_getkreq() and freed via crypto_freekreq(). This permits the crypto layer to track open asym requests and close races with a driver trying to unregister while asym requests are in flight. - crypto_copyback, crypto_copydata, crypto_apply, and crypto_contiguous_subsegment now accept the 'crp' object as the first parameter instead of individual members. This makes it easier to deal with different buffer types in the future as well as separate input and output buffers. It's also simpler for driver writers to use. - bus_dmamap_load_crp() loads a DMA mapping for a crypto buffer. This understands the various types of buffers so that drivers that use DMA do not have to be aware of different buffer types. - Helper routines now exist to build an auth context for HMAC IPAD and OPAD. This reduces some duplicated work among drivers. - Key buffers are now treated as const throughout the framework and in device drivers. However, session key buffers provided when a session is created are expected to remain alive for the duration of the session. - GCM and CCM sessions now only specify a cipher algorithm and a cipher key. The redundant auth information is not needed or used. - For cryptosoft, split up the code a bit such that the 'process' callback now invokes a function pointer in the session. This function pointer is set based on the mode (in effect) though it simplifies a few edge cases that would otherwise be in the switch in 'process'. It does split up GCM vs CCM which I think is more readable even if there is some duplication. - I changed /dev/crypto to support GMAC requests using CRYPTO_AES_NIST_GMAC as an auth algorithm and updated cryptocheck to work with it. - Combined cipher and auth sessions via /dev/crypto now always use ETA mode. The COP_F_CIPHER_FIRST flag is now a no-op that is ignored. This was actually documented as being true in crypto(4) before, but the code had not implemented this before I added the CIPHER_FIRST flag. - I have not yet updated /dev/crypto to be aware of explicit modes for sessions. I will probably do that at some point in the future as well as teach it about IV/nonce and tag lengths for AEAD so we can support all of the NIST KAT tests for GCM and CCM. - I've split up the exising crypto.9 manpage into several pages of which many are written from scratch. - I have converted all drivers and consumers in the tree and verified that they compile, but I have not tested all of them. I have tested the following drivers: - cryptosoft - aesni (AES only) - blake2 - ccr and the following consumers: - cryptodev - IPsec - ktls_ocf - GELI (lightly) I have not tested the following: - ccp - aesni with sha - hifn - kgssapi_krb5 - ubsec - padlock - safe - armv8_crypto (aarch64) - glxsb (i386) - sec (ppc) - cesa (armv7) - cryptocteon (mips64) - nlmsec (mips64) Discussed with: cem Relnotes: yes Sponsored by: Chelsio Communications Differential Revision: https://reviews.freebsd.org/D23677
2020-03-27 18:25:23 +00:00
crp->crp_flags |= CRYPTO_F_IV_SEPARATE;
2014-12-12 19:56:36 +00:00
} else {
error = EINVAL;
SDT_PROBE1(opencrypto, dev, ioctl, error, __LINE__);
goto bail;
2014-12-12 19:56:36 +00:00
}
if (crp->crp_op & CRYPTO_OP_VERIFY_DIGEST) {
error = copyin(caead->tag, cod->buf + crp->crp_digest_start,
cse->hashsize);
if (error) {
SDT_PROBE1(opencrypto, dev, ioctl, error, __LINE__);
goto bail;
}
}
2014-12-12 19:56:36 +00:00
again:
/*
* Let the dispatch run unlocked, then, interlock against the
* callback before checking if the operation completed and going
* to sleep. This insures drivers don't inherit our lock which
* results in a lock order reversal between crypto_dispatch forced
* entry and the crypto_done callback into us.
*/
error = crypto_dispatch(crp);
if (error != 0) {
SDT_PROBE1(opencrypto, dev, ioctl, error, __LINE__);
2014-12-12 19:56:36 +00:00
goto bail;
}
2014-12-12 19:56:36 +00:00
mtx_lock(&cse->lock);
while (!cod->done)
mtx_sleep(cod, &cse->lock, PWAIT, "crydev", 0);
mtx_unlock(&cse->lock);
2014-12-12 19:56:36 +00:00
if (crp->crp_etype == EAGAIN) {
crp->crp_etype = 0;
crp->crp_flags &= ~CRYPTO_F_DONE;
cod->done = false;
2014-12-12 19:56:36 +00:00
goto again;
}
if (crp->crp_etype != 0) {
error = crp->crp_etype;
SDT_PROBE1(opencrypto, dev, ioctl, error, __LINE__);
2014-12-12 19:56:36 +00:00
goto bail;
}
Refactor driver and consumer interfaces for OCF (in-kernel crypto). - The linked list of cryptoini structures used in session initialization is replaced with a new flat structure: struct crypto_session_params. This session includes a new mode to define how the other fields should be interpreted. Available modes include: - COMPRESS (for compression/decompression) - CIPHER (for simply encryption/decryption) - DIGEST (computing and verifying digests) - AEAD (combined auth and encryption such as AES-GCM and AES-CCM) - ETA (combined auth and encryption using encrypt-then-authenticate) Additional modes could be added in the future (e.g. if we wanted to support TLS MtE for AES-CBC in the kernel we could add a new mode for that. TLS modes might also affect how AAD is interpreted, etc.) The flat structure also includes the key lengths and algorithms as before. However, code doesn't have to walk the linked list and switch on the algorithm to determine which key is the auth key vs encryption key. The 'csp_auth_*' fields are always used for auth keys and settings and 'csp_cipher_*' for cipher. (Compression algorithms are stored in csp_cipher_alg.) - Drivers no longer register a list of supported algorithms. This doesn't quite work when you factor in modes (e.g. a driver might support both AES-CBC and SHA2-256-HMAC separately but not combined for ETA). Instead, a new 'crypto_probesession' method has been added to the kobj interface for symmteric crypto drivers. This method returns a negative value on success (similar to how device_probe works) and the crypto framework uses this value to pick the "best" driver. There are three constants for hardware (e.g. ccr), accelerated software (e.g. aesni), and plain software (cryptosoft) that give preference in that order. One effect of this is that if you request only hardware when creating a new session, you will no longer get a session using accelerated software. Another effect is that the default setting to disallow software crypto via /dev/crypto now disables accelerated software. Once a driver is chosen, 'crypto_newsession' is invoked as before. - Crypto operations are now solely described by the flat 'cryptop' structure. The linked list of descriptors has been removed. A separate enum has been added to describe the type of data buffer in use instead of using CRYPTO_F_* flags to make it easier to add more types in the future if needed (e.g. wired userspace buffers for zero-copy). It will also make it easier to re-introduce separate input and output buffers (in-kernel TLS would benefit from this). Try to make the flags related to IV handling less insane: - CRYPTO_F_IV_SEPARATE means that the IV is stored in the 'crp_iv' member of the operation structure. If this flag is not set, the IV is stored in the data buffer at the 'crp_iv_start' offset. - CRYPTO_F_IV_GENERATE means that a random IV should be generated and stored into the data buffer. This cannot be used with CRYPTO_F_IV_SEPARATE. If a consumer wants to deal with explicit vs implicit IVs, etc. it can always generate the IV however it needs and store partial IVs in the buffer and the full IV/nonce in crp_iv and set CRYPTO_F_IV_SEPARATE. The layout of the buffer is now described via fields in cryptop. crp_aad_start and crp_aad_length define the boundaries of any AAD. Previously with GCM and CCM you defined an auth crd with this range, but for ETA your auth crd had to span both the AAD and plaintext (and they had to be adjacent). crp_payload_start and crp_payload_length define the boundaries of the plaintext/ciphertext. Modes that only do a single operation (COMPRESS, CIPHER, DIGEST) should only use this region and leave the AAD region empty. If a digest is present (or should be generated), it's starting location is marked by crp_digest_start. Instead of using the CRD_F_ENCRYPT flag to determine the direction of the operation, cryptop now includes an 'op' field defining the operation to perform. For digests I've added a new VERIFY digest mode which assumes a digest is present in the input and fails the request with EBADMSG if it doesn't match the internally-computed digest. GCM and CCM already assumed this, and the new AEAD mode requires this for decryption. The new ETA mode now also requires this for decryption, so IPsec and GELI no longer do their own authentication verification. Simple DIGEST operations can also do this, though there are no in-tree consumers. To eventually support some refcounting to close races, the session cookie is now passed to crypto_getop() and clients should no longer set crp_sesssion directly. - Assymteric crypto operation structures should be allocated via crypto_getkreq() and freed via crypto_freekreq(). This permits the crypto layer to track open asym requests and close races with a driver trying to unregister while asym requests are in flight. - crypto_copyback, crypto_copydata, crypto_apply, and crypto_contiguous_subsegment now accept the 'crp' object as the first parameter instead of individual members. This makes it easier to deal with different buffer types in the future as well as separate input and output buffers. It's also simpler for driver writers to use. - bus_dmamap_load_crp() loads a DMA mapping for a crypto buffer. This understands the various types of buffers so that drivers that use DMA do not have to be aware of different buffer types. - Helper routines now exist to build an auth context for HMAC IPAD and OPAD. This reduces some duplicated work among drivers. - Key buffers are now treated as const throughout the framework and in device drivers. However, session key buffers provided when a session is created are expected to remain alive for the duration of the session. - GCM and CCM sessions now only specify a cipher algorithm and a cipher key. The redundant auth information is not needed or used. - For cryptosoft, split up the code a bit such that the 'process' callback now invokes a function pointer in the session. This function pointer is set based on the mode (in effect) though it simplifies a few edge cases that would otherwise be in the switch in 'process'. It does split up GCM vs CCM which I think is more readable even if there is some duplication. - I changed /dev/crypto to support GMAC requests using CRYPTO_AES_NIST_GMAC as an auth algorithm and updated cryptocheck to work with it. - Combined cipher and auth sessions via /dev/crypto now always use ETA mode. The COP_F_CIPHER_FIRST flag is now a no-op that is ignored. This was actually documented as being true in crypto(4) before, but the code had not implemented this before I added the CIPHER_FIRST flag. - I have not yet updated /dev/crypto to be aware of explicit modes for sessions. I will probably do that at some point in the future as well as teach it about IV/nonce and tag lengths for AEAD so we can support all of the NIST KAT tests for GCM and CCM. - I've split up the exising crypto.9 manpage into several pages of which many are written from scratch. - I have converted all drivers and consumers in the tree and verified that they compile, but I have not tested all of them. I have tested the following drivers: - cryptosoft - aesni (AES only) - blake2 - ccr and the following consumers: - cryptodev - IPsec - ktls_ocf - GELI (lightly) I have not tested the following: - ccp - aesni with sha - hifn - kgssapi_krb5 - ubsec - padlock - safe - armv8_crypto (aarch64) - glxsb (i386) - sec (ppc) - cesa (armv7) - cryptocteon (mips64) - nlmsec (mips64) Discussed with: cem Relnotes: yes Sponsored by: Chelsio Communications Differential Revision: https://reviews.freebsd.org/D23677
2020-03-27 18:25:23 +00:00
if (caead->dst != NULL) {
error = copyout(cod->obuf != NULL ? cod->obuf :
cod->buf + crp->crp_payload_start, dst,
crp->crp_payload_length);
Refactor driver and consumer interfaces for OCF (in-kernel crypto). - The linked list of cryptoini structures used in session initialization is replaced with a new flat structure: struct crypto_session_params. This session includes a new mode to define how the other fields should be interpreted. Available modes include: - COMPRESS (for compression/decompression) - CIPHER (for simply encryption/decryption) - DIGEST (computing and verifying digests) - AEAD (combined auth and encryption such as AES-GCM and AES-CCM) - ETA (combined auth and encryption using encrypt-then-authenticate) Additional modes could be added in the future (e.g. if we wanted to support TLS MtE for AES-CBC in the kernel we could add a new mode for that. TLS modes might also affect how AAD is interpreted, etc.) The flat structure also includes the key lengths and algorithms as before. However, code doesn't have to walk the linked list and switch on the algorithm to determine which key is the auth key vs encryption key. The 'csp_auth_*' fields are always used for auth keys and settings and 'csp_cipher_*' for cipher. (Compression algorithms are stored in csp_cipher_alg.) - Drivers no longer register a list of supported algorithms. This doesn't quite work when you factor in modes (e.g. a driver might support both AES-CBC and SHA2-256-HMAC separately but not combined for ETA). Instead, a new 'crypto_probesession' method has been added to the kobj interface for symmteric crypto drivers. This method returns a negative value on success (similar to how device_probe works) and the crypto framework uses this value to pick the "best" driver. There are three constants for hardware (e.g. ccr), accelerated software (e.g. aesni), and plain software (cryptosoft) that give preference in that order. One effect of this is that if you request only hardware when creating a new session, you will no longer get a session using accelerated software. Another effect is that the default setting to disallow software crypto via /dev/crypto now disables accelerated software. Once a driver is chosen, 'crypto_newsession' is invoked as before. - Crypto operations are now solely described by the flat 'cryptop' structure. The linked list of descriptors has been removed. A separate enum has been added to describe the type of data buffer in use instead of using CRYPTO_F_* flags to make it easier to add more types in the future if needed (e.g. wired userspace buffers for zero-copy). It will also make it easier to re-introduce separate input and output buffers (in-kernel TLS would benefit from this). Try to make the flags related to IV handling less insane: - CRYPTO_F_IV_SEPARATE means that the IV is stored in the 'crp_iv' member of the operation structure. If this flag is not set, the IV is stored in the data buffer at the 'crp_iv_start' offset. - CRYPTO_F_IV_GENERATE means that a random IV should be generated and stored into the data buffer. This cannot be used with CRYPTO_F_IV_SEPARATE. If a consumer wants to deal with explicit vs implicit IVs, etc. it can always generate the IV however it needs and store partial IVs in the buffer and the full IV/nonce in crp_iv and set CRYPTO_F_IV_SEPARATE. The layout of the buffer is now described via fields in cryptop. crp_aad_start and crp_aad_length define the boundaries of any AAD. Previously with GCM and CCM you defined an auth crd with this range, but for ETA your auth crd had to span both the AAD and plaintext (and they had to be adjacent). crp_payload_start and crp_payload_length define the boundaries of the plaintext/ciphertext. Modes that only do a single operation (COMPRESS, CIPHER, DIGEST) should only use this region and leave the AAD region empty. If a digest is present (or should be generated), it's starting location is marked by crp_digest_start. Instead of using the CRD_F_ENCRYPT flag to determine the direction of the operation, cryptop now includes an 'op' field defining the operation to perform. For digests I've added a new VERIFY digest mode which assumes a digest is present in the input and fails the request with EBADMSG if it doesn't match the internally-computed digest. GCM and CCM already assumed this, and the new AEAD mode requires this for decryption. The new ETA mode now also requires this for decryption, so IPsec and GELI no longer do their own authentication verification. Simple DIGEST operations can also do this, though there are no in-tree consumers. To eventually support some refcounting to close races, the session cookie is now passed to crypto_getop() and clients should no longer set crp_sesssion directly. - Assymteric crypto operation structures should be allocated via crypto_getkreq() and freed via crypto_freekreq(). This permits the crypto layer to track open asym requests and close races with a driver trying to unregister while asym requests are in flight. - crypto_copyback, crypto_copydata, crypto_apply, and crypto_contiguous_subsegment now accept the 'crp' object as the first parameter instead of individual members. This makes it easier to deal with different buffer types in the future as well as separate input and output buffers. It's also simpler for driver writers to use. - bus_dmamap_load_crp() loads a DMA mapping for a crypto buffer. This understands the various types of buffers so that drivers that use DMA do not have to be aware of different buffer types. - Helper routines now exist to build an auth context for HMAC IPAD and OPAD. This reduces some duplicated work among drivers. - Key buffers are now treated as const throughout the framework and in device drivers. However, session key buffers provided when a session is created are expected to remain alive for the duration of the session. - GCM and CCM sessions now only specify a cipher algorithm and a cipher key. The redundant auth information is not needed or used. - For cryptosoft, split up the code a bit such that the 'process' callback now invokes a function pointer in the session. This function pointer is set based on the mode (in effect) though it simplifies a few edge cases that would otherwise be in the switch in 'process'. It does split up GCM vs CCM which I think is more readable even if there is some duplication. - I changed /dev/crypto to support GMAC requests using CRYPTO_AES_NIST_GMAC as an auth algorithm and updated cryptocheck to work with it. - Combined cipher and auth sessions via /dev/crypto now always use ETA mode. The COP_F_CIPHER_FIRST flag is now a no-op that is ignored. This was actually documented as being true in crypto(4) before, but the code had not implemented this before I added the CIPHER_FIRST flag. - I have not yet updated /dev/crypto to be aware of explicit modes for sessions. I will probably do that at some point in the future as well as teach it about IV/nonce and tag lengths for AEAD so we can support all of the NIST KAT tests for GCM and CCM. - I've split up the exising crypto.9 manpage into several pages of which many are written from scratch. - I have converted all drivers and consumers in the tree and verified that they compile, but I have not tested all of them. I have tested the following drivers: - cryptosoft - aesni (AES only) - blake2 - ccr and the following consumers: - cryptodev - IPsec - ktls_ocf - GELI (lightly) I have not tested the following: - ccp - aesni with sha - hifn - kgssapi_krb5 - ubsec - padlock - safe - armv8_crypto (aarch64) - glxsb (i386) - sec (ppc) - cesa (armv7) - cryptocteon (mips64) - nlmsec (mips64) Discussed with: cem Relnotes: yes Sponsored by: Chelsio Communications Differential Revision: https://reviews.freebsd.org/D23677
2020-03-27 18:25:23 +00:00
if (error) {
SDT_PROBE1(opencrypto, dev, ioctl, error, __LINE__);
goto bail;
}
}
2014-12-12 19:56:36 +00:00
if ((crp->crp_op & CRYPTO_OP_VERIFY_DIGEST) == 0) {
error = copyout((cod->obuf != NULL ? cod->obuf : cod->buf) +
crp->crp_digest_start, caead->tag, cse->hashsize);
if (error) {
SDT_PROBE1(opencrypto, dev, ioctl, error, __LINE__);
goto bail;
}
}
2014-12-12 19:56:36 +00:00
bail:
crypto_freereq(crp);
cod_free(cod);
2014-12-12 19:56:36 +00:00
return (error);
}
static int
cryptodev_find(struct crypt_find_op *find)
{
device_t dev;
2014-12-12 19:56:36 +00:00
size_t fnlen = sizeof find->name;
if (find->crid != -1) {
dev = crypto_find_device_byhid(find->crid);
if (dev == NULL)
return (ENOENT);
2014-12-12 19:56:36 +00:00
strncpy(find->name, device_get_nameunit(dev), fnlen);
find->name[fnlen - 1] = '\x0';
} else {
2014-12-12 19:56:36 +00:00
find->name[fnlen - 1] = '\x0';
find->crid = crypto_find_driver(find->name);
if (find->crid == -1)
return (ENOENT);
}
return (0);
}
static void
fcrypt_dtor(void *data)
{
struct fcrypt *fcr = data;
struct csession *cse;
while ((cse = TAILQ_FIRST(&fcr->csessions))) {
TAILQ_REMOVE(&fcr->csessions, cse, next);
KASSERT(refcount_load(&cse->refs) == 1,
("%s: crypto session %p with %d refs", __func__, cse,
refcount_load(&cse->refs)));
cse_free(cse);
}
mtx_destroy(&fcr->lock);
free(fcr, M_CRYPTODEV);
}
static int
crypto_open(struct cdev *dev, int oflags, int devtype, struct thread *td)
{
struct fcrypt *fcr;
int error;
fcr = malloc(sizeof(struct fcrypt), M_CRYPTODEV, M_WAITOK | M_ZERO);
TAILQ_INIT(&fcr->csessions);
mtx_init(&fcr->lock, "fcrypt", NULL, MTX_DEF);
error = devfs_set_cdevpriv(fcr, fcrypt_dtor);
if (error)
fcrypt_dtor(fcr);
return (error);
}
static int
crypto_ioctl(struct cdev *dev, u_long cmd, caddr_t data, int flag,
struct thread *td)
{
struct fcrypt *fcr;
struct csession *cse;
struct session2_op *sop;
struct crypt_op *cop;
struct crypt_aead *caead;
uint32_t ses;
int error = 0;
union {
struct session2_op sopc;
#ifdef COMPAT_FREEBSD32
struct crypt_op copc;
struct crypt_aead aeadc;
#endif
} thunk;
#ifdef COMPAT_FREEBSD32
u_long cmd32;
void *data32;
cmd32 = 0;
data32 = NULL;
switch (cmd) {
case CIOCGSESSION32:
cmd32 = cmd;
data32 = data;
cmd = CIOCGSESSION;
data = (void *)&thunk.sopc;
session_op_from_32((struct session_op32 *)data32, &thunk.sopc);
break;
case CIOCGSESSION232:
cmd32 = cmd;
data32 = data;
cmd = CIOCGSESSION2;
data = (void *)&thunk.sopc;
session2_op_from_32((struct session2_op32 *)data32,
&thunk.sopc);
break;
case CIOCCRYPT32:
cmd32 = cmd;
data32 = data;
cmd = CIOCCRYPT;
data = (void *)&thunk.copc;
crypt_op_from_32((struct crypt_op32 *)data32, &thunk.copc);
break;
case CIOCCRYPTAEAD32:
cmd32 = cmd;
data32 = data;
cmd = CIOCCRYPTAEAD;
data = (void *)&thunk.aeadc;
crypt_aead_from_32((struct crypt_aead32 *)data32, &thunk.aeadc);
break;
}
#endif
devfs_get_cdevpriv((void **)&fcr);
switch (cmd) {
#ifdef COMPAT_FREEBSD12
case CRIOGET:
/*
* NB: This may fail in cases that the old
* implementation did not if the current process has
* restricted filesystem access (e.g. running in a
* jail that does not expose /dev/crypto or in
* capability mode).
*/
error = kern_openat(td, AT_FDCWD, "/dev/crypto", UIO_SYSSPACE,
O_RDWR, 0);
if (error == 0)
*(uint32_t *)data = td->td_retval[0];
break;
#endif
case CIOCGSESSION:
case CIOCGSESSION2:
if (cmd == CIOCGSESSION) {
session2_op_from_op((void *)data, &thunk.sopc);
sop = &thunk.sopc;
} else
sop = (struct session2_op *)data;
error = cse_create(fcr, sop);
if (cmd == CIOCGSESSION && error == 0)
session2_op_to_op(sop, (void *)data);
break;
case CIOCFSESSION:
ses = *(uint32_t *)data;
if (!cse_delete(fcr, ses)) {
SDT_PROBE1(opencrypto, dev, ioctl, error, __LINE__);
return (EINVAL);
}
break;
case CIOCCRYPT:
cop = (struct crypt_op *)data;
cse = cse_find(fcr, cop->ses);
if (cse == NULL) {
SDT_PROBE1(opencrypto, dev, ioctl, error, __LINE__);
return (EINVAL);
}
error = cryptodev_op(cse, cop);
cse_free(cse);
break;
case CIOCFINDDEV:
error = cryptodev_find((struct crypt_find_op *)data);
break;
case CIOCCRYPTAEAD:
caead = (struct crypt_aead *)data;
cse = cse_find(fcr, caead->ses);
if (cse == NULL) {
SDT_PROBE1(opencrypto, dev, ioctl, error, __LINE__);
return (EINVAL);
}
error = cryptodev_aead(cse, caead);
cse_free(cse);
break;
default:
error = EINVAL;
SDT_PROBE1(opencrypto, dev, ioctl, error, __LINE__);
break;
}
#ifdef COMPAT_FREEBSD32
switch (cmd32) {
case CIOCGSESSION32:
if (error == 0)
session_op_to_32((void *)data, data32);
break;
case CIOCGSESSION232:
if (error == 0)
session2_op_to_32((void *)data, data32);
break;
case CIOCCRYPT32:
if (error == 0)
crypt_op_to_32((void *)data, data32);
break;
case CIOCCRYPTAEAD32:
if (error == 0)
crypt_aead_to_32((void *)data, data32);
break;
}
#endif
return (error);
}
static struct cdevsw crypto_cdevsw = {
.d_version = D_VERSION,
.d_open = crypto_open,
.d_ioctl = crypto_ioctl,
.d_name = "crypto",
};
static struct cdev *crypto_dev;
/*
* Initialization code, both for static and dynamic loading.
*/
static int
cryptodev_modevent(module_t mod, int type, void *unused)
{
switch (type) {
case MOD_LOAD:
if (bootverbose)
printf("crypto: <crypto device>\n");
crypto_dev = make_dev(&crypto_cdevsw, 0,
UID_ROOT, GID_WHEEL, 0666,
"crypto");
return 0;
case MOD_UNLOAD:
/*XXX disallow if active sessions */
destroy_dev(crypto_dev);
return 0;
}
return EINVAL;
}
static moduledata_t cryptodev_mod = {
"cryptodev",
cryptodev_modevent,
0
};
MODULE_VERSION(cryptodev, 1);
DECLARE_MODULE(cryptodev, cryptodev_mod, SI_SUB_PSEUDO, SI_ORDER_ANY);
MODULE_DEPEND(cryptodev, crypto, 1, 1, 1);
MODULE_DEPEND(cryptodev, zlib, 1, 1, 1);