d/freebsd-dev
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity
1838dec318
freebsd-dev/sys/crypto/ccp/ccp.c

792 lines
17 KiB
C
Raw Normal View History

Add ccp(4): experimental driver for AMD Crypto Co-Processor * Registers TRNG source for random(4) * Finds available queues, LSBs; allocates static objects * Allocates a shared MSI-X for all queues. The hardware does not have separate interrupts per queue. Working interrupt mode driver. * Computes SHA hashes, HMAC. Passes cryptotest.py, cryptocheck tests. * Does AES-CBC, CTR mode, and XTS. cryptotest.py and cryptocheck pass. * Support for "authenc" (AES + HMAC). (SHA1 seems to result in "unaligned" cleartext inputs from cryptocheck -- which the engine cannot handle. SHA2 seems to work fine.) * GCM passes for block-multiple AAD, input lengths Largely based on ccr(4), part of cxgbe(4). Rough performance averages on AMD Ryzen 1950X (4kB buffer): aesni: SHA1: ~8300 Mb/s SHA256: ~8000 Mb/s ccp: ~630 Mb/s SHA256: ~660 Mb/s SHA512: ~700 Mb/s cryptosoft: ~1800 Mb/s SHA256: ~1800 Mb/s SHA512: ~2700 Mb/s As you can see, performance is poor in comparison to aesni(4) and even cryptosoft (due to high setup cost). At a larger buffer size (128kB), throughput is a little better (but still worse than aesni(4)): aesni: SHA1:~10400 Mb/s SHA256: ~9950 Mb/s ccp: ~2200 Mb/s SHA256: ~2600 Mb/s SHA512: ~3800 Mb/s cryptosoft: ~1750 Mb/s SHA256: ~1800 Mb/s SHA512: ~2700 Mb/s AES performance has a similar story: aesni: 4kB: ~11250 Mb/s 128kB: ~11250 Mb/s ccp: ~350 Mb/s 128kB: ~4600 Mb/s cryptosoft: ~1750 Mb/s 128kB: ~1700 Mb/s This driver is EXPERIMENTAL. You should verify cryptographic results on typical and corner case inputs from your application against a known- good implementation. Sponsored by: Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D12723
2018-01-18 22:01:30 +00:00
/*-
* SPDX-License-Identifier: BSD-2-Clause-FreeBSD
*
* Copyright (c) 2017 Chelsio Communications, Inc.
* Copyright (c) 2017 Conrad Meyer <cem@FreeBSD.org>
* All rights reserved.
* Largely borrowed from ccr(4), Written by: John Baldwin <jhb@FreeBSD.org>
*
* 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.
*
* THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``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 OR CONTRIBUTORS 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.
*/
#include <sys/cdefs.h>
__FBSDID("$FreeBSD$");
#include "opt_ddb.h"
Extract eventfilter declarations to sys/_eventfilter.h This allows replacing "sys/eventfilter.h" includes with "sys/_eventfilter.h" in other header files (e.g., sys/{bus,conf,cpu}.h) and reduces header pollution substantially. EVENTHANDLER_DECLARE and EVENTHANDLER_LIST_DECLAREs were moved out of .c files into appropriate headers (e.g., sys/proc.h, powernv/opal.h). As a side effect of reduced header pollution, many .c files and headers no longer contain needed definitions. The remainder of the patch addresses adding appropriate includes to fix those files. LOCK_DEBUG and LOCK_FILE_LINE_ARG are moved to sys/_lock.h, as required by sys/mutex.h since r326106 (but silently protected by header pollution prior to this change). No functional change (intended). Of course, any out of tree modules that relied on header pollution for sys/eventhandler.h, sys/lock.h, or sys/mutex.h inclusion need to be fixed. __FreeBSD_version has been bumped.
2019-05-20 00:38:23 +00:00
#include <sys/param.h>
Add ccp(4): experimental driver for AMD Crypto Co-Processor * Registers TRNG source for random(4) * Finds available queues, LSBs; allocates static objects * Allocates a shared MSI-X for all queues. The hardware does not have separate interrupts per queue. Working interrupt mode driver. * Computes SHA hashes, HMAC. Passes cryptotest.py, cryptocheck tests. * Does AES-CBC, CTR mode, and XTS. cryptotest.py and cryptocheck pass. * Support for "authenc" (AES + HMAC). (SHA1 seems to result in "unaligned" cleartext inputs from cryptocheck -- which the engine cannot handle. SHA2 seems to work fine.) * GCM passes for block-multiple AAD, input lengths Largely based on ccr(4), part of cxgbe(4). Rough performance averages on AMD Ryzen 1950X (4kB buffer): aesni: SHA1: ~8300 Mb/s SHA256: ~8000 Mb/s ccp: ~630 Mb/s SHA256: ~660 Mb/s SHA512: ~700 Mb/s cryptosoft: ~1800 Mb/s SHA256: ~1800 Mb/s SHA512: ~2700 Mb/s As you can see, performance is poor in comparison to aesni(4) and even cryptosoft (due to high setup cost). At a larger buffer size (128kB), throughput is a little better (but still worse than aesni(4)): aesni: SHA1:~10400 Mb/s SHA256: ~9950 Mb/s ccp: ~2200 Mb/s SHA256: ~2600 Mb/s SHA512: ~3800 Mb/s cryptosoft: ~1750 Mb/s SHA256: ~1800 Mb/s SHA512: ~2700 Mb/s AES performance has a similar story: aesni: 4kB: ~11250 Mb/s 128kB: ~11250 Mb/s ccp: ~350 Mb/s 128kB: ~4600 Mb/s cryptosoft: ~1750 Mb/s 128kB: ~1700 Mb/s This driver is EXPERIMENTAL. You should verify cryptographic results on typical and corner case inputs from your application against a known- good implementation. Sponsored by: Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D12723
2018-01-18 22:01:30 +00:00
#include <sys/bus.h>
#include <sys/lock.h>
#include <sys/kernel.h>
#include <sys/malloc.h>
#include <sys/mutex.h>
#include <sys/module.h>
#include <sys/random.h>
#include <sys/sglist.h>
#include <sys/sysctl.h>
#ifdef DDB
#include <ddb/ddb.h>
#endif
#include <dev/pci/pcivar.h>
#include <dev/random/randomdev.h>
#include <opencrypto/cryptodev.h>
#include <opencrypto/xform.h>
#include "cryptodev_if.h"
#include "ccp.h"
#include "ccp_hardware.h"
MALLOC_DEFINE(M_CCP, "ccp", "AMD CCP crypto");
/*
* Need a global softc available for garbage random_source API, which lacks any
* context pointer. It's also handy for debugging.
*/
struct ccp_softc *g_ccp_softc;
bool g_debug_print = false;
SYSCTL_BOOL(_hw_ccp, OID_AUTO, debug, CTLFLAG_RWTUN, &g_debug_print, 0,
"Set to enable debugging log messages");
static struct pciid {
uint32_t devid;
const char *desc;
} ccp_ids[] = {
{ 0x14561022, "AMD CCP-5a" },
{ 0x14681022, "AMD CCP-5b" },
ccp: add a new hardware ID, found on AMD Ryzen 3 3200U pciconf reports the device as: Family 17h (Models 10h-1fh) Platform Security Processor dmesg: ccp0: <AMD CCP-5a> mem 0xfe500000-0xfe5fffff,0xfe68c000-0xfe68dfff at device 0.2 on pci4 crypto: assign ccp0 driver id 1, flags 0x1000000 MFC after: 1 week
2020-04-24 10:20:54 +00:00
{ 0x15df1022, "AMD CCP-5a" },
Add ccp(4): experimental driver for AMD Crypto Co-Processor * Registers TRNG source for random(4) * Finds available queues, LSBs; allocates static objects * Allocates a shared MSI-X for all queues. The hardware does not have separate interrupts per queue. Working interrupt mode driver. * Computes SHA hashes, HMAC. Passes cryptotest.py, cryptocheck tests. * Does AES-CBC, CTR mode, and XTS. cryptotest.py and cryptocheck pass. * Support for "authenc" (AES + HMAC). (SHA1 seems to result in "unaligned" cleartext inputs from cryptocheck -- which the engine cannot handle. SHA2 seems to work fine.) * GCM passes for block-multiple AAD, input lengths Largely based on ccr(4), part of cxgbe(4). Rough performance averages on AMD Ryzen 1950X (4kB buffer): aesni: SHA1: ~8300 Mb/s SHA256: ~8000 Mb/s ccp: ~630 Mb/s SHA256: ~660 Mb/s SHA512: ~700 Mb/s cryptosoft: ~1800 Mb/s SHA256: ~1800 Mb/s SHA512: ~2700 Mb/s As you can see, performance is poor in comparison to aesni(4) and even cryptosoft (due to high setup cost). At a larger buffer size (128kB), throughput is a little better (but still worse than aesni(4)): aesni: SHA1:~10400 Mb/s SHA256: ~9950 Mb/s ccp: ~2200 Mb/s SHA256: ~2600 Mb/s SHA512: ~3800 Mb/s cryptosoft: ~1750 Mb/s SHA256: ~1800 Mb/s SHA512: ~2700 Mb/s AES performance has a similar story: aesni: 4kB: ~11250 Mb/s 128kB: ~11250 Mb/s ccp: ~350 Mb/s 128kB: ~4600 Mb/s cryptosoft: ~1750 Mb/s 128kB: ~1700 Mb/s This driver is EXPERIMENTAL. You should verify cryptographic results on typical and corner case inputs from your application against a known- good implementation. Sponsored by: Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D12723
2018-01-18 22:01:30 +00:00
};
static struct random_source random_ccp = {
.rs_ident = "AMD CCP TRNG",
.rs_source = RANDOM_PURE_CCP,
.rs_read = random_ccp_read,
};
/*
* ccp_populate_sglist() generates a scatter/gather list that covers the entire
* crypto operation buffer.
*/
static int
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
ccp_populate_sglist(struct sglist *sg, struct crypto_buffer *cb)
Add ccp(4): experimental driver for AMD Crypto Co-Processor * Registers TRNG source for random(4) * Finds available queues, LSBs; allocates static objects * Allocates a shared MSI-X for all queues. The hardware does not have separate interrupts per queue. Working interrupt mode driver. * Computes SHA hashes, HMAC. Passes cryptotest.py, cryptocheck tests. * Does AES-CBC, CTR mode, and XTS. cryptotest.py and cryptocheck pass. * Support for "authenc" (AES + HMAC). (SHA1 seems to result in "unaligned" cleartext inputs from cryptocheck -- which the engine cannot handle. SHA2 seems to work fine.) * GCM passes for block-multiple AAD, input lengths Largely based on ccr(4), part of cxgbe(4). Rough performance averages on AMD Ryzen 1950X (4kB buffer): aesni: SHA1: ~8300 Mb/s SHA256: ~8000 Mb/s ccp: ~630 Mb/s SHA256: ~660 Mb/s SHA512: ~700 Mb/s cryptosoft: ~1800 Mb/s SHA256: ~1800 Mb/s SHA512: ~2700 Mb/s As you can see, performance is poor in comparison to aesni(4) and even cryptosoft (due to high setup cost). At a larger buffer size (128kB), throughput is a little better (but still worse than aesni(4)): aesni: SHA1:~10400 Mb/s SHA256: ~9950 Mb/s ccp: ~2200 Mb/s SHA256: ~2600 Mb/s SHA512: ~3800 Mb/s cryptosoft: ~1750 Mb/s SHA256: ~1800 Mb/s SHA512: ~2700 Mb/s AES performance has a similar story: aesni: 4kB: ~11250 Mb/s 128kB: ~11250 Mb/s ccp: ~350 Mb/s 128kB: ~4600 Mb/s cryptosoft: ~1750 Mb/s 128kB: ~1700 Mb/s This driver is EXPERIMENTAL. You should verify cryptographic results on typical and corner case inputs from your application against a known- good implementation. Sponsored by: Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D12723
2018-01-18 22:01:30 +00:00
{
int error;
sglist_reset(sg);
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
switch (cb->cb_type) {
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 CRYPTO_BUF_MBUF:
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
error = sglist_append_mbuf(sg, cb->cb_mbuf);
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;
crypto: Add a new type of crypto buffer for a single mbuf. This is intended for use in KTLS transmit where each TLS record is described by a single mbuf that is itself queued in the socket buffer. Using the existing CRYPTO_BUF_MBUF would result in bus_dmamap_load_crp() walking additional mbufs in the socket buffer that are not relevant, but generating a S/G list that potentially exceeds the limit of the tag (while also wasting CPU cycles). Reviewed by: markj Sponsored by: Netflix Differential Revision: https://reviews.freebsd.org/D30136
2021-05-25 23:59:18 +00:00
case CRYPTO_BUF_SINGLE_MBUF:
error = sglist_append_single_mbuf(sg, cb->cb_mbuf);
break;
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 CRYPTO_BUF_UIO:
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
error = sglist_append_uio(sg, cb->cb_uio);
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;
case CRYPTO_BUF_CONTIG:
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
error = sglist_append(sg, cb->cb_buf, cb->cb_buf_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
break;
crypto(9): add CRYPTO_BUF_VMPAGE crypto(9) functions can now be used on buffers composed of an array of vm_page_t structures, such as those stored in an unmapped struct bio. It requires the running to kernel to support the direct memory map, so not all architectures can use it. Reviewed by: markj, kib, jhb, mjg, mat, bcr (manpages) MFC after: 1 week Sponsored by: Axcient Differential Revision: https://reviews.freebsd.org/D25671
2020-08-26 02:37:42 +00:00
case CRYPTO_BUF_VMPAGE:
error = sglist_append_vmpages(sg, cb->cb_vm_page,
cb->cb_vm_page_len, cb->cb_vm_page_offset);
break;
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
default:
error = EINVAL;
}
Add ccp(4): experimental driver for AMD Crypto Co-Processor * Registers TRNG source for random(4) * Finds available queues, LSBs; allocates static objects * Allocates a shared MSI-X for all queues. The hardware does not have separate interrupts per queue. Working interrupt mode driver. * Computes SHA hashes, HMAC. Passes cryptotest.py, cryptocheck tests. * Does AES-CBC, CTR mode, and XTS. cryptotest.py and cryptocheck pass. * Support for "authenc" (AES + HMAC). (SHA1 seems to result in "unaligned" cleartext inputs from cryptocheck -- which the engine cannot handle. SHA2 seems to work fine.) * GCM passes for block-multiple AAD, input lengths Largely based on ccr(4), part of cxgbe(4). Rough performance averages on AMD Ryzen 1950X (4kB buffer): aesni: SHA1: ~8300 Mb/s SHA256: ~8000 Mb/s ccp: ~630 Mb/s SHA256: ~660 Mb/s SHA512: ~700 Mb/s cryptosoft: ~1800 Mb/s SHA256: ~1800 Mb/s SHA512: ~2700 Mb/s As you can see, performance is poor in comparison to aesni(4) and even cryptosoft (due to high setup cost). At a larger buffer size (128kB), throughput is a little better (but still worse than aesni(4)): aesni: SHA1:~10400 Mb/s SHA256: ~9950 Mb/s ccp: ~2200 Mb/s SHA256: ~2600 Mb/s SHA512: ~3800 Mb/s cryptosoft: ~1750 Mb/s SHA256: ~1800 Mb/s SHA512: ~2700 Mb/s AES performance has a similar story: aesni: 4kB: ~11250 Mb/s 128kB: ~11250 Mb/s ccp: ~350 Mb/s 128kB: ~4600 Mb/s cryptosoft: ~1750 Mb/s 128kB: ~1700 Mb/s This driver is EXPERIMENTAL. You should verify cryptographic results on typical and corner case inputs from your application against a known- good implementation. Sponsored by: Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D12723
2018-01-18 22:01:30 +00:00
return (error);
}
static int
ccp_probe(device_t dev)
{
struct pciid *ip;
uint32_t id;
id = pci_get_devid(dev);
for (ip = ccp_ids; ip < &ccp_ids[nitems(ccp_ids)]; ip++) {
if (id == ip->devid) {
device_set_desc(dev, ip->desc);
return (0);
}
}
return (ENXIO);
}
static void
ccp_initialize_queues(struct ccp_softc *sc)
{
struct ccp_queue *qp;
size_t i;
for (i = 0; i < nitems(sc->queues); i++) {
qp = &sc->queues[i];
qp->cq_softc = sc;
qp->cq_qindex = i;
mtx_init(&qp->cq_lock, "ccp queue", NULL, MTX_DEF);
/* XXX - arbitrarily chosen sizes */
qp->cq_sg_crp = sglist_alloc(32, M_WAITOK);
/* Two more SGEs than sg_crp to accommodate ipad. */
qp->cq_sg_ulptx = sglist_alloc(34, M_WAITOK);
qp->cq_sg_dst = sglist_alloc(2, M_WAITOK);
}
}
static void
ccp_free_queues(struct ccp_softc *sc)
{
struct ccp_queue *qp;
size_t i;
for (i = 0; i < nitems(sc->queues); i++) {
qp = &sc->queues[i];
mtx_destroy(&qp->cq_lock);
sglist_free(qp->cq_sg_crp);
sglist_free(qp->cq_sg_ulptx);
sglist_free(qp->cq_sg_dst);
}
}
static int
ccp_attach(device_t dev)
{
struct ccp_softc *sc;
int error;
sc = device_get_softc(dev);
sc->dev = dev;
OpenCrypto: Convert sessions to opaque handles instead of integers Track session objects in the framework, and pass handles between the framework (OCF), consumers, and drivers. Avoid redundancy and complexity in individual drivers by allocating session memory in the framework and providing it to drivers in ::newsession(). Session handles are no longer integers with information encoded in various high bits. Use of the CRYPTO_SESID2FOO() macros should be replaced with the appropriate crypto_ses2foo() function on the opaque session handle. Convert OCF drivers (in particular, cryptosoft, as well as myriad others) to the opaque handle interface. Discard existing session tracking as much as possible (quick pass). There may be additional code ripe for deletion. Convert OCF consumers (ipsec, geom_eli, krb5, cryptodev) to handle-style interface. The conversion is largely mechnical. The change is documented in crypto.9. Inspired by https://lists.freebsd.org/pipermail/freebsd-arch/2018-January/018835.html . No objection from: ae (ipsec portion) Reported by: jhb
2018-07-18 00:56:25 +00:00
sc->cid = crypto_get_driverid(dev, sizeof(struct ccp_session),
CRYPTOCAP_F_HARDWARE);
Add ccp(4): experimental driver for AMD Crypto Co-Processor * Registers TRNG source for random(4) * Finds available queues, LSBs; allocates static objects * Allocates a shared MSI-X for all queues. The hardware does not have separate interrupts per queue. Working interrupt mode driver. * Computes SHA hashes, HMAC. Passes cryptotest.py, cryptocheck tests. * Does AES-CBC, CTR mode, and XTS. cryptotest.py and cryptocheck pass. * Support for "authenc" (AES + HMAC). (SHA1 seems to result in "unaligned" cleartext inputs from cryptocheck -- which the engine cannot handle. SHA2 seems to work fine.) * GCM passes for block-multiple AAD, input lengths Largely based on ccr(4), part of cxgbe(4). Rough performance averages on AMD Ryzen 1950X (4kB buffer): aesni: SHA1: ~8300 Mb/s SHA256: ~8000 Mb/s ccp: ~630 Mb/s SHA256: ~660 Mb/s SHA512: ~700 Mb/s cryptosoft: ~1800 Mb/s SHA256: ~1800 Mb/s SHA512: ~2700 Mb/s As you can see, performance is poor in comparison to aesni(4) and even cryptosoft (due to high setup cost). At a larger buffer size (128kB), throughput is a little better (but still worse than aesni(4)): aesni: SHA1:~10400 Mb/s SHA256: ~9950 Mb/s ccp: ~2200 Mb/s SHA256: ~2600 Mb/s SHA512: ~3800 Mb/s cryptosoft: ~1750 Mb/s SHA256: ~1800 Mb/s SHA512: ~2700 Mb/s AES performance has a similar story: aesni: 4kB: ~11250 Mb/s 128kB: ~11250 Mb/s ccp: ~350 Mb/s 128kB: ~4600 Mb/s cryptosoft: ~1750 Mb/s 128kB: ~1700 Mb/s This driver is EXPERIMENTAL. You should verify cryptographic results on typical and corner case inputs from your application against a known- good implementation. Sponsored by: Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D12723
2018-01-18 22:01:30 +00:00
if (sc->cid < 0) {
device_printf(dev, "could not get crypto driver id\n");
return (ENXIO);
}
error = ccp_hw_attach(dev);
if (error != 0)
return (error);
mtx_init(&sc->lock, "ccp", NULL, MTX_DEF);
ccp_initialize_queues(sc);
if (g_ccp_softc == NULL) {
g_ccp_softc = sc;
if ((sc->hw_features & VERSION_CAP_TRNG) != 0)
random_source_register(&random_ccp);
}
return (0);
}
static int
ccp_detach(device_t dev)
{
struct ccp_softc *sc;
sc = device_get_softc(dev);
mtx_lock(&sc->lock);
sc->detaching = true;
mtx_unlock(&sc->lock);
crypto_unregister_all(sc->cid);
if (g_ccp_softc == sc && (sc->hw_features & VERSION_CAP_TRNG) != 0)
random_source_deregister(&random_ccp);
ccp_hw_detach(dev);
ccp_free_queues(sc);
if (g_ccp_softc == sc)
g_ccp_softc = NULL;
mtx_destroy(&sc->lock);
return (0);
}
static void
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
ccp_init_hmac_digest(struct ccp_session *s, const char *key, int klen)
Add ccp(4): experimental driver for AMD Crypto Co-Processor * Registers TRNG source for random(4) * Finds available queues, LSBs; allocates static objects * Allocates a shared MSI-X for all queues. The hardware does not have separate interrupts per queue. Working interrupt mode driver. * Computes SHA hashes, HMAC. Passes cryptotest.py, cryptocheck tests. * Does AES-CBC, CTR mode, and XTS. cryptotest.py and cryptocheck pass. * Support for "authenc" (AES + HMAC). (SHA1 seems to result in "unaligned" cleartext inputs from cryptocheck -- which the engine cannot handle. SHA2 seems to work fine.) * GCM passes for block-multiple AAD, input lengths Largely based on ccr(4), part of cxgbe(4). Rough performance averages on AMD Ryzen 1950X (4kB buffer): aesni: SHA1: ~8300 Mb/s SHA256: ~8000 Mb/s ccp: ~630 Mb/s SHA256: ~660 Mb/s SHA512: ~700 Mb/s cryptosoft: ~1800 Mb/s SHA256: ~1800 Mb/s SHA512: ~2700 Mb/s As you can see, performance is poor in comparison to aesni(4) and even cryptosoft (due to high setup cost). At a larger buffer size (128kB), throughput is a little better (but still worse than aesni(4)): aesni: SHA1:~10400 Mb/s SHA256: ~9950 Mb/s ccp: ~2200 Mb/s SHA256: ~2600 Mb/s SHA512: ~3800 Mb/s cryptosoft: ~1750 Mb/s SHA256: ~1800 Mb/s SHA512: ~2700 Mb/s AES performance has a similar story: aesni: 4kB: ~11250 Mb/s 128kB: ~11250 Mb/s ccp: ~350 Mb/s 128kB: ~4600 Mb/s cryptosoft: ~1750 Mb/s 128kB: ~1700 Mb/s This driver is EXPERIMENTAL. You should verify cryptographic results on typical and corner case inputs from your application against a known- good implementation. Sponsored by: Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D12723
2018-01-18 22:01:30 +00:00
{
union authctx auth_ctx;
crypto: Constify all transform descriptors No functional change intended. Reviewed by: ae, jhb MFC after: 1 week Sponsored by: The FreeBSD Foundation Differential Revision: https://reviews.freebsd.org/D31196
2021-07-26 20:41:05 +00:00
const struct auth_hash *axf;
Add ccp(4): experimental driver for AMD Crypto Co-Processor * Registers TRNG source for random(4) * Finds available queues, LSBs; allocates static objects * Allocates a shared MSI-X for all queues. The hardware does not have separate interrupts per queue. Working interrupt mode driver. * Computes SHA hashes, HMAC. Passes cryptotest.py, cryptocheck tests. * Does AES-CBC, CTR mode, and XTS. cryptotest.py and cryptocheck pass. * Support for "authenc" (AES + HMAC). (SHA1 seems to result in "unaligned" cleartext inputs from cryptocheck -- which the engine cannot handle. SHA2 seems to work fine.) * GCM passes for block-multiple AAD, input lengths Largely based on ccr(4), part of cxgbe(4). Rough performance averages on AMD Ryzen 1950X (4kB buffer): aesni: SHA1: ~8300 Mb/s SHA256: ~8000 Mb/s ccp: ~630 Mb/s SHA256: ~660 Mb/s SHA512: ~700 Mb/s cryptosoft: ~1800 Mb/s SHA256: ~1800 Mb/s SHA512: ~2700 Mb/s As you can see, performance is poor in comparison to aesni(4) and even cryptosoft (due to high setup cost). At a larger buffer size (128kB), throughput is a little better (but still worse than aesni(4)): aesni: SHA1:~10400 Mb/s SHA256: ~9950 Mb/s ccp: ~2200 Mb/s SHA256: ~2600 Mb/s SHA512: ~3800 Mb/s cryptosoft: ~1750 Mb/s SHA256: ~1800 Mb/s SHA512: ~2700 Mb/s AES performance has a similar story: aesni: 4kB: ~11250 Mb/s 128kB: ~11250 Mb/s ccp: ~350 Mb/s 128kB: ~4600 Mb/s cryptosoft: ~1750 Mb/s 128kB: ~1700 Mb/s This driver is EXPERIMENTAL. You should verify cryptographic results on typical and corner case inputs from your application against a known- good implementation. Sponsored by: Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D12723
2018-01-18 22:01:30 +00:00
u_int i;
/*
* If the key is larger than the block size, use the digest of
* the key as the key instead.
*/
axf = s->hmac.auth_hash;
if (klen > axf->blocksize) {
axf->Init(&auth_ctx);
axf->Update(&auth_ctx, key, klen);
axf->Final(s->hmac.ipad, &auth_ctx);
explicit_bzero(&auth_ctx, sizeof(auth_ctx));
klen = axf->hashsize;
} else
memcpy(s->hmac.ipad, key, klen);
memset(s->hmac.ipad + klen, 0, axf->blocksize - klen);
memcpy(s->hmac.opad, s->hmac.ipad, axf->blocksize);
for (i = 0; i < axf->blocksize; i++) {
s->hmac.ipad[i] ^= HMAC_IPAD_VAL;
s->hmac.opad[i] ^= HMAC_OPAD_VAL;
}
}
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
static bool
Add ccp(4): experimental driver for AMD Crypto Co-Processor * Registers TRNG source for random(4) * Finds available queues, LSBs; allocates static objects * Allocates a shared MSI-X for all queues. The hardware does not have separate interrupts per queue. Working interrupt mode driver. * Computes SHA hashes, HMAC. Passes cryptotest.py, cryptocheck tests. * Does AES-CBC, CTR mode, and XTS. cryptotest.py and cryptocheck pass. * Support for "authenc" (AES + HMAC). (SHA1 seems to result in "unaligned" cleartext inputs from cryptocheck -- which the engine cannot handle. SHA2 seems to work fine.) * GCM passes for block-multiple AAD, input lengths Largely based on ccr(4), part of cxgbe(4). Rough performance averages on AMD Ryzen 1950X (4kB buffer): aesni: SHA1: ~8300 Mb/s SHA256: ~8000 Mb/s ccp: ~630 Mb/s SHA256: ~660 Mb/s SHA512: ~700 Mb/s cryptosoft: ~1800 Mb/s SHA256: ~1800 Mb/s SHA512: ~2700 Mb/s As you can see, performance is poor in comparison to aesni(4) and even cryptosoft (due to high setup cost). At a larger buffer size (128kB), throughput is a little better (but still worse than aesni(4)): aesni: SHA1:~10400 Mb/s SHA256: ~9950 Mb/s ccp: ~2200 Mb/s SHA256: ~2600 Mb/s SHA512: ~3800 Mb/s cryptosoft: ~1750 Mb/s SHA256: ~1800 Mb/s SHA512: ~2700 Mb/s AES performance has a similar story: aesni: 4kB: ~11250 Mb/s 128kB: ~11250 Mb/s ccp: ~350 Mb/s 128kB: ~4600 Mb/s cryptosoft: ~1750 Mb/s 128kB: ~1700 Mb/s This driver is EXPERIMENTAL. You should verify cryptographic results on typical and corner case inputs from your application against a known- good implementation. Sponsored by: Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D12723
2018-01-18 22:01:30 +00:00
ccp_aes_check_keylen(int alg, int klen)
{
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 (klen * 8) {
Add ccp(4): experimental driver for AMD Crypto Co-Processor * Registers TRNG source for random(4) * Finds available queues, LSBs; allocates static objects * Allocates a shared MSI-X for all queues. The hardware does not have separate interrupts per queue. Working interrupt mode driver. * Computes SHA hashes, HMAC. Passes cryptotest.py, cryptocheck tests. * Does AES-CBC, CTR mode, and XTS. cryptotest.py and cryptocheck pass. * Support for "authenc" (AES + HMAC). (SHA1 seems to result in "unaligned" cleartext inputs from cryptocheck -- which the engine cannot handle. SHA2 seems to work fine.) * GCM passes for block-multiple AAD, input lengths Largely based on ccr(4), part of cxgbe(4). Rough performance averages on AMD Ryzen 1950X (4kB buffer): aesni: SHA1: ~8300 Mb/s SHA256: ~8000 Mb/s ccp: ~630 Mb/s SHA256: ~660 Mb/s SHA512: ~700 Mb/s cryptosoft: ~1800 Mb/s SHA256: ~1800 Mb/s SHA512: ~2700 Mb/s As you can see, performance is poor in comparison to aesni(4) and even cryptosoft (due to high setup cost). At a larger buffer size (128kB), throughput is a little better (but still worse than aesni(4)): aesni: SHA1:~10400 Mb/s SHA256: ~9950 Mb/s ccp: ~2200 Mb/s SHA256: ~2600 Mb/s SHA512: ~3800 Mb/s cryptosoft: ~1750 Mb/s SHA256: ~1800 Mb/s SHA512: ~2700 Mb/s AES performance has a similar story: aesni: 4kB: ~11250 Mb/s 128kB: ~11250 Mb/s ccp: ~350 Mb/s 128kB: ~4600 Mb/s cryptosoft: ~1750 Mb/s 128kB: ~1700 Mb/s This driver is EXPERIMENTAL. You should verify cryptographic results on typical and corner case inputs from your application against a known- good implementation. Sponsored by: Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D12723
2018-01-18 22:01:30 +00:00
case 128:
case 192:
if (alg == CRYPTO_AES_XTS)
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
return (false);
Add ccp(4): experimental driver for AMD Crypto Co-Processor * Registers TRNG source for random(4) * Finds available queues, LSBs; allocates static objects * Allocates a shared MSI-X for all queues. The hardware does not have separate interrupts per queue. Working interrupt mode driver. * Computes SHA hashes, HMAC. Passes cryptotest.py, cryptocheck tests. * Does AES-CBC, CTR mode, and XTS. cryptotest.py and cryptocheck pass. * Support for "authenc" (AES + HMAC). (SHA1 seems to result in "unaligned" cleartext inputs from cryptocheck -- which the engine cannot handle. SHA2 seems to work fine.) * GCM passes for block-multiple AAD, input lengths Largely based on ccr(4), part of cxgbe(4). Rough performance averages on AMD Ryzen 1950X (4kB buffer): aesni: SHA1: ~8300 Mb/s SHA256: ~8000 Mb/s ccp: ~630 Mb/s SHA256: ~660 Mb/s SHA512: ~700 Mb/s cryptosoft: ~1800 Mb/s SHA256: ~1800 Mb/s SHA512: ~2700 Mb/s As you can see, performance is poor in comparison to aesni(4) and even cryptosoft (due to high setup cost). At a larger buffer size (128kB), throughput is a little better (but still worse than aesni(4)): aesni: SHA1:~10400 Mb/s SHA256: ~9950 Mb/s ccp: ~2200 Mb/s SHA256: ~2600 Mb/s SHA512: ~3800 Mb/s cryptosoft: ~1750 Mb/s SHA256: ~1800 Mb/s SHA512: ~2700 Mb/s AES performance has a similar story: aesni: 4kB: ~11250 Mb/s 128kB: ~11250 Mb/s ccp: ~350 Mb/s 128kB: ~4600 Mb/s cryptosoft: ~1750 Mb/s 128kB: ~1700 Mb/s This driver is EXPERIMENTAL. You should verify cryptographic results on typical and corner case inputs from your application against a known- good implementation. Sponsored by: Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D12723
2018-01-18 22:01:30 +00:00
break;
case 256:
break;
case 512:
if (alg != CRYPTO_AES_XTS)
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
return (false);
Add ccp(4): experimental driver for AMD Crypto Co-Processor * Registers TRNG source for random(4) * Finds available queues, LSBs; allocates static objects * Allocates a shared MSI-X for all queues. The hardware does not have separate interrupts per queue. Working interrupt mode driver. * Computes SHA hashes, HMAC. Passes cryptotest.py, cryptocheck tests. * Does AES-CBC, CTR mode, and XTS. cryptotest.py and cryptocheck pass. * Support for "authenc" (AES + HMAC). (SHA1 seems to result in "unaligned" cleartext inputs from cryptocheck -- which the engine cannot handle. SHA2 seems to work fine.) * GCM passes for block-multiple AAD, input lengths Largely based on ccr(4), part of cxgbe(4). Rough performance averages on AMD Ryzen 1950X (4kB buffer): aesni: SHA1: ~8300 Mb/s SHA256: ~8000 Mb/s ccp: ~630 Mb/s SHA256: ~660 Mb/s SHA512: ~700 Mb/s cryptosoft: ~1800 Mb/s SHA256: ~1800 Mb/s SHA512: ~2700 Mb/s As you can see, performance is poor in comparison to aesni(4) and even cryptosoft (due to high setup cost). At a larger buffer size (128kB), throughput is a little better (but still worse than aesni(4)): aesni: SHA1:~10400 Mb/s SHA256: ~9950 Mb/s ccp: ~2200 Mb/s SHA256: ~2600 Mb/s SHA512: ~3800 Mb/s cryptosoft: ~1750 Mb/s SHA256: ~1800 Mb/s SHA512: ~2700 Mb/s AES performance has a similar story: aesni: 4kB: ~11250 Mb/s 128kB: ~11250 Mb/s ccp: ~350 Mb/s 128kB: ~4600 Mb/s cryptosoft: ~1750 Mb/s 128kB: ~1700 Mb/s This driver is EXPERIMENTAL. You should verify cryptographic results on typical and corner case inputs from your application against a known- good implementation. Sponsored by: Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D12723
2018-01-18 22:01:30 +00:00
break;
default:
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
return (false);
Add ccp(4): experimental driver for AMD Crypto Co-Processor * Registers TRNG source for random(4) * Finds available queues, LSBs; allocates static objects * Allocates a shared MSI-X for all queues. The hardware does not have separate interrupts per queue. Working interrupt mode driver. * Computes SHA hashes, HMAC. Passes cryptotest.py, cryptocheck tests. * Does AES-CBC, CTR mode, and XTS. cryptotest.py and cryptocheck pass. * Support for "authenc" (AES + HMAC). (SHA1 seems to result in "unaligned" cleartext inputs from cryptocheck -- which the engine cannot handle. SHA2 seems to work fine.) * GCM passes for block-multiple AAD, input lengths Largely based on ccr(4), part of cxgbe(4). Rough performance averages on AMD Ryzen 1950X (4kB buffer): aesni: SHA1: ~8300 Mb/s SHA256: ~8000 Mb/s ccp: ~630 Mb/s SHA256: ~660 Mb/s SHA512: ~700 Mb/s cryptosoft: ~1800 Mb/s SHA256: ~1800 Mb/s SHA512: ~2700 Mb/s As you can see, performance is poor in comparison to aesni(4) and even cryptosoft (due to high setup cost). At a larger buffer size (128kB), throughput is a little better (but still worse than aesni(4)): aesni: SHA1:~10400 Mb/s SHA256: ~9950 Mb/s ccp: ~2200 Mb/s SHA256: ~2600 Mb/s SHA512: ~3800 Mb/s cryptosoft: ~1750 Mb/s SHA256: ~1800 Mb/s SHA512: ~2700 Mb/s AES performance has a similar story: aesni: 4kB: ~11250 Mb/s 128kB: ~11250 Mb/s ccp: ~350 Mb/s 128kB: ~4600 Mb/s cryptosoft: ~1750 Mb/s 128kB: ~1700 Mb/s This driver is EXPERIMENTAL. You should verify cryptographic results on typical and corner case inputs from your application against a known- good implementation. Sponsored by: Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D12723
2018-01-18 22:01:30 +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
return (true);
Add ccp(4): experimental driver for AMD Crypto Co-Processor * Registers TRNG source for random(4) * Finds available queues, LSBs; allocates static objects * Allocates a shared MSI-X for all queues. The hardware does not have separate interrupts per queue. Working interrupt mode driver. * Computes SHA hashes, HMAC. Passes cryptotest.py, cryptocheck tests. * Does AES-CBC, CTR mode, and XTS. cryptotest.py and cryptocheck pass. * Support for "authenc" (AES + HMAC). (SHA1 seems to result in "unaligned" cleartext inputs from cryptocheck -- which the engine cannot handle. SHA2 seems to work fine.) * GCM passes for block-multiple AAD, input lengths Largely based on ccr(4), part of cxgbe(4). Rough performance averages on AMD Ryzen 1950X (4kB buffer): aesni: SHA1: ~8300 Mb/s SHA256: ~8000 Mb/s ccp: ~630 Mb/s SHA256: ~660 Mb/s SHA512: ~700 Mb/s cryptosoft: ~1800 Mb/s SHA256: ~1800 Mb/s SHA512: ~2700 Mb/s As you can see, performance is poor in comparison to aesni(4) and even cryptosoft (due to high setup cost). At a larger buffer size (128kB), throughput is a little better (but still worse than aesni(4)): aesni: SHA1:~10400 Mb/s SHA256: ~9950 Mb/s ccp: ~2200 Mb/s SHA256: ~2600 Mb/s SHA512: ~3800 Mb/s cryptosoft: ~1750 Mb/s SHA256: ~1800 Mb/s SHA512: ~2700 Mb/s AES performance has a similar story: aesni: 4kB: ~11250 Mb/s 128kB: ~11250 Mb/s ccp: ~350 Mb/s 128kB: ~4600 Mb/s cryptosoft: ~1750 Mb/s 128kB: ~1700 Mb/s This driver is EXPERIMENTAL. You should verify cryptographic results on typical and corner case inputs from your application against a known- good implementation. Sponsored by: Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D12723
2018-01-18 22:01:30 +00:00
}
static void
ccp_aes_setkey(struct ccp_session *s, int alg, const void *key, int klen)
{
unsigned kbits;
if (alg == CRYPTO_AES_XTS)
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
kbits = (klen / 2) * 8;
Add ccp(4): experimental driver for AMD Crypto Co-Processor * Registers TRNG source for random(4) * Finds available queues, LSBs; allocates static objects * Allocates a shared MSI-X for all queues. The hardware does not have separate interrupts per queue. Working interrupt mode driver. * Computes SHA hashes, HMAC. Passes cryptotest.py, cryptocheck tests. * Does AES-CBC, CTR mode, and XTS. cryptotest.py and cryptocheck pass. * Support for "authenc" (AES + HMAC). (SHA1 seems to result in "unaligned" cleartext inputs from cryptocheck -- which the engine cannot handle. SHA2 seems to work fine.) * GCM passes for block-multiple AAD, input lengths Largely based on ccr(4), part of cxgbe(4). Rough performance averages on AMD Ryzen 1950X (4kB buffer): aesni: SHA1: ~8300 Mb/s SHA256: ~8000 Mb/s ccp: ~630 Mb/s SHA256: ~660 Mb/s SHA512: ~700 Mb/s cryptosoft: ~1800 Mb/s SHA256: ~1800 Mb/s SHA512: ~2700 Mb/s As you can see, performance is poor in comparison to aesni(4) and even cryptosoft (due to high setup cost). At a larger buffer size (128kB), throughput is a little better (but still worse than aesni(4)): aesni: SHA1:~10400 Mb/s SHA256: ~9950 Mb/s ccp: ~2200 Mb/s SHA256: ~2600 Mb/s SHA512: ~3800 Mb/s cryptosoft: ~1750 Mb/s SHA256: ~1800 Mb/s SHA512: ~2700 Mb/s AES performance has a similar story: aesni: 4kB: ~11250 Mb/s 128kB: ~11250 Mb/s ccp: ~350 Mb/s 128kB: ~4600 Mb/s cryptosoft: ~1750 Mb/s 128kB: ~1700 Mb/s This driver is EXPERIMENTAL. You should verify cryptographic results on typical and corner case inputs from your application against a known- good implementation. Sponsored by: Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D12723
2018-01-18 22:01:30 +00:00
else
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
kbits = klen * 8;
Add ccp(4): experimental driver for AMD Crypto Co-Processor * Registers TRNG source for random(4) * Finds available queues, LSBs; allocates static objects * Allocates a shared MSI-X for all queues. The hardware does not have separate interrupts per queue. Working interrupt mode driver. * Computes SHA hashes, HMAC. Passes cryptotest.py, cryptocheck tests. * Does AES-CBC, CTR mode, and XTS. cryptotest.py and cryptocheck pass. * Support for "authenc" (AES + HMAC). (SHA1 seems to result in "unaligned" cleartext inputs from cryptocheck -- which the engine cannot handle. SHA2 seems to work fine.) * GCM passes for block-multiple AAD, input lengths Largely based on ccr(4), part of cxgbe(4). Rough performance averages on AMD Ryzen 1950X (4kB buffer): aesni: SHA1: ~8300 Mb/s SHA256: ~8000 Mb/s ccp: ~630 Mb/s SHA256: ~660 Mb/s SHA512: ~700 Mb/s cryptosoft: ~1800 Mb/s SHA256: ~1800 Mb/s SHA512: ~2700 Mb/s As you can see, performance is poor in comparison to aesni(4) and even cryptosoft (due to high setup cost). At a larger buffer size (128kB), throughput is a little better (but still worse than aesni(4)): aesni: SHA1:~10400 Mb/s SHA256: ~9950 Mb/s ccp: ~2200 Mb/s SHA256: ~2600 Mb/s SHA512: ~3800 Mb/s cryptosoft: ~1750 Mb/s SHA256: ~1800 Mb/s SHA512: ~2700 Mb/s AES performance has a similar story: aesni: 4kB: ~11250 Mb/s 128kB: ~11250 Mb/s ccp: ~350 Mb/s 128kB: ~4600 Mb/s cryptosoft: ~1750 Mb/s 128kB: ~1700 Mb/s This driver is EXPERIMENTAL. You should verify cryptographic results on typical and corner case inputs from your application against a known- good implementation. Sponsored by: Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D12723
2018-01-18 22:01:30 +00:00
switch (kbits) {
case 128:
s->blkcipher.cipher_type = CCP_AES_TYPE_128;
break;
case 192:
s->blkcipher.cipher_type = CCP_AES_TYPE_192;
break;
case 256:
s->blkcipher.cipher_type = CCP_AES_TYPE_256;
break;
default:
panic("should not get here");
}
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
s->blkcipher.key_len = klen;
Add ccp(4): experimental driver for AMD Crypto Co-Processor * Registers TRNG source for random(4) * Finds available queues, LSBs; allocates static objects * Allocates a shared MSI-X for all queues. The hardware does not have separate interrupts per queue. Working interrupt mode driver. * Computes SHA hashes, HMAC. Passes cryptotest.py, cryptocheck tests. * Does AES-CBC, CTR mode, and XTS. cryptotest.py and cryptocheck pass. * Support for "authenc" (AES + HMAC). (SHA1 seems to result in "unaligned" cleartext inputs from cryptocheck -- which the engine cannot handle. SHA2 seems to work fine.) * GCM passes for block-multiple AAD, input lengths Largely based on ccr(4), part of cxgbe(4). Rough performance averages on AMD Ryzen 1950X (4kB buffer): aesni: SHA1: ~8300 Mb/s SHA256: ~8000 Mb/s ccp: ~630 Mb/s SHA256: ~660 Mb/s SHA512: ~700 Mb/s cryptosoft: ~1800 Mb/s SHA256: ~1800 Mb/s SHA512: ~2700 Mb/s As you can see, performance is poor in comparison to aesni(4) and even cryptosoft (due to high setup cost). At a larger buffer size (128kB), throughput is a little better (but still worse than aesni(4)): aesni: SHA1:~10400 Mb/s SHA256: ~9950 Mb/s ccp: ~2200 Mb/s SHA256: ~2600 Mb/s SHA512: ~3800 Mb/s cryptosoft: ~1750 Mb/s SHA256: ~1800 Mb/s SHA512: ~2700 Mb/s AES performance has a similar story: aesni: 4kB: ~11250 Mb/s 128kB: ~11250 Mb/s ccp: ~350 Mb/s 128kB: ~4600 Mb/s cryptosoft: ~1750 Mb/s 128kB: ~1700 Mb/s This driver is EXPERIMENTAL. You should verify cryptographic results on typical and corner case inputs from your application against a known- good implementation. Sponsored by: Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D12723
2018-01-18 22:01:30 +00:00
memcpy(s->blkcipher.enckey, key, s->blkcipher.key_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
static bool
ccp_auth_supported(struct ccp_softc *sc,
const struct crypto_session_params *csp)
{
if ((sc->hw_features & VERSION_CAP_SHA) == 0)
return (false);
switch (csp->csp_auth_alg) {
case CRYPTO_SHA1_HMAC:
case CRYPTO_SHA2_256_HMAC:
case CRYPTO_SHA2_384_HMAC:
case CRYPTO_SHA2_512_HMAC:
if (csp->csp_auth_key == NULL)
return (false);
break;
default:
return (false);
}
return (true);
}
static bool
ccp_cipher_supported(struct ccp_softc *sc,
const struct crypto_session_params *csp)
{
if ((sc->hw_features & VERSION_CAP_AES) == 0)
return (false);
switch (csp->csp_cipher_alg) {
case CRYPTO_AES_CBC:
if (csp->csp_ivlen != AES_BLOCK_LEN)
return (false);
break;
case CRYPTO_AES_ICM:
if (csp->csp_ivlen != AES_BLOCK_LEN)
return (false);
break;
case CRYPTO_AES_XTS:
if (csp->csp_ivlen != AES_XTS_IV_LEN)
return (false);
break;
default:
return (false);
}
return (ccp_aes_check_keylen(csp->csp_cipher_alg,
csp->csp_cipher_klen));
}
Add ccp(4): experimental driver for AMD Crypto Co-Processor * Registers TRNG source for random(4) * Finds available queues, LSBs; allocates static objects * Allocates a shared MSI-X for all queues. The hardware does not have separate interrupts per queue. Working interrupt mode driver. * Computes SHA hashes, HMAC. Passes cryptotest.py, cryptocheck tests. * Does AES-CBC, CTR mode, and XTS. cryptotest.py and cryptocheck pass. * Support for "authenc" (AES + HMAC). (SHA1 seems to result in "unaligned" cleartext inputs from cryptocheck -- which the engine cannot handle. SHA2 seems to work fine.) * GCM passes for block-multiple AAD, input lengths Largely based on ccr(4), part of cxgbe(4). Rough performance averages on AMD Ryzen 1950X (4kB buffer): aesni: SHA1: ~8300 Mb/s SHA256: ~8000 Mb/s ccp: ~630 Mb/s SHA256: ~660 Mb/s SHA512: ~700 Mb/s cryptosoft: ~1800 Mb/s SHA256: ~1800 Mb/s SHA512: ~2700 Mb/s As you can see, performance is poor in comparison to aesni(4) and even cryptosoft (due to high setup cost). At a larger buffer size (128kB), throughput is a little better (but still worse than aesni(4)): aesni: SHA1:~10400 Mb/s SHA256: ~9950 Mb/s ccp: ~2200 Mb/s SHA256: ~2600 Mb/s SHA512: ~3800 Mb/s cryptosoft: ~1750 Mb/s SHA256: ~1800 Mb/s SHA512: ~2700 Mb/s AES performance has a similar story: aesni: 4kB: ~11250 Mb/s 128kB: ~11250 Mb/s ccp: ~350 Mb/s 128kB: ~4600 Mb/s cryptosoft: ~1750 Mb/s 128kB: ~1700 Mb/s This driver is EXPERIMENTAL. You should verify cryptographic results on typical and corner case inputs from your application against a known- good implementation. Sponsored by: Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D12723
2018-01-18 22:01:30 +00:00
static int
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
ccp_probesession(device_t dev, const struct crypto_session_params *csp)
Add ccp(4): experimental driver for AMD Crypto Co-Processor * Registers TRNG source for random(4) * Finds available queues, LSBs; allocates static objects * Allocates a shared MSI-X for all queues. The hardware does not have separate interrupts per queue. Working interrupt mode driver. * Computes SHA hashes, HMAC. Passes cryptotest.py, cryptocheck tests. * Does AES-CBC, CTR mode, and XTS. cryptotest.py and cryptocheck pass. * Support for "authenc" (AES + HMAC). (SHA1 seems to result in "unaligned" cleartext inputs from cryptocheck -- which the engine cannot handle. SHA2 seems to work fine.) * GCM passes for block-multiple AAD, input lengths Largely based on ccr(4), part of cxgbe(4). Rough performance averages on AMD Ryzen 1950X (4kB buffer): aesni: SHA1: ~8300 Mb/s SHA256: ~8000 Mb/s ccp: ~630 Mb/s SHA256: ~660 Mb/s SHA512: ~700 Mb/s cryptosoft: ~1800 Mb/s SHA256: ~1800 Mb/s SHA512: ~2700 Mb/s As you can see, performance is poor in comparison to aesni(4) and even cryptosoft (due to high setup cost). At a larger buffer size (128kB), throughput is a little better (but still worse than aesni(4)): aesni: SHA1:~10400 Mb/s SHA256: ~9950 Mb/s ccp: ~2200 Mb/s SHA256: ~2600 Mb/s SHA512: ~3800 Mb/s cryptosoft: ~1750 Mb/s SHA256: ~1800 Mb/s SHA512: ~2700 Mb/s AES performance has a similar story: aesni: 4kB: ~11250 Mb/s 128kB: ~11250 Mb/s ccp: ~350 Mb/s 128kB: ~4600 Mb/s cryptosoft: ~1750 Mb/s 128kB: ~1700 Mb/s This driver is EXPERIMENTAL. You should verify cryptographic results on typical and corner case inputs from your application against a known- good implementation. Sponsored by: Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D12723
2018-01-18 22:01:30 +00:00
{
struct ccp_softc *sc;
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 (csp->csp_flags != 0)
Add ccp(4): experimental driver for AMD Crypto Co-Processor * Registers TRNG source for random(4) * Finds available queues, LSBs; allocates static objects * Allocates a shared MSI-X for all queues. The hardware does not have separate interrupts per queue. Working interrupt mode driver. * Computes SHA hashes, HMAC. Passes cryptotest.py, cryptocheck tests. * Does AES-CBC, CTR mode, and XTS. cryptotest.py and cryptocheck pass. * Support for "authenc" (AES + HMAC). (SHA1 seems to result in "unaligned" cleartext inputs from cryptocheck -- which the engine cannot handle. SHA2 seems to work fine.) * GCM passes for block-multiple AAD, input lengths Largely based on ccr(4), part of cxgbe(4). Rough performance averages on AMD Ryzen 1950X (4kB buffer): aesni: SHA1: ~8300 Mb/s SHA256: ~8000 Mb/s ccp: ~630 Mb/s SHA256: ~660 Mb/s SHA512: ~700 Mb/s cryptosoft: ~1800 Mb/s SHA256: ~1800 Mb/s SHA512: ~2700 Mb/s As you can see, performance is poor in comparison to aesni(4) and even cryptosoft (due to high setup cost). At a larger buffer size (128kB), throughput is a little better (but still worse than aesni(4)): aesni: SHA1:~10400 Mb/s SHA256: ~9950 Mb/s ccp: ~2200 Mb/s SHA256: ~2600 Mb/s SHA512: ~3800 Mb/s cryptosoft: ~1750 Mb/s SHA256: ~1800 Mb/s SHA512: ~2700 Mb/s AES performance has a similar story: aesni: 4kB: ~11250 Mb/s 128kB: ~11250 Mb/s ccp: ~350 Mb/s 128kB: ~4600 Mb/s cryptosoft: ~1750 Mb/s 128kB: ~1700 Mb/s This driver is EXPERIMENTAL. You should verify cryptographic results on typical and corner case inputs from your application against a known- good implementation. Sponsored by: Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D12723
2018-01-18 22:01:30 +00:00
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
sc = device_get_softc(dev);
switch (csp->csp_mode) {
case CSP_MODE_DIGEST:
if (!ccp_auth_supported(sc, csp))
return (EINVAL);
break;
case CSP_MODE_CIPHER:
if (!ccp_cipher_supported(sc, csp))
return (EINVAL);
break;
case CSP_MODE_AEAD:
switch (csp->csp_cipher_alg) {
Add ccp(4): experimental driver for AMD Crypto Co-Processor * Registers TRNG source for random(4) * Finds available queues, LSBs; allocates static objects * Allocates a shared MSI-X for all queues. The hardware does not have separate interrupts per queue. Working interrupt mode driver. * Computes SHA hashes, HMAC. Passes cryptotest.py, cryptocheck tests. * Does AES-CBC, CTR mode, and XTS. cryptotest.py and cryptocheck pass. * Support for "authenc" (AES + HMAC). (SHA1 seems to result in "unaligned" cleartext inputs from cryptocheck -- which the engine cannot handle. SHA2 seems to work fine.) * GCM passes for block-multiple AAD, input lengths Largely based on ccr(4), part of cxgbe(4). Rough performance averages on AMD Ryzen 1950X (4kB buffer): aesni: SHA1: ~8300 Mb/s SHA256: ~8000 Mb/s ccp: ~630 Mb/s SHA256: ~660 Mb/s SHA512: ~700 Mb/s cryptosoft: ~1800 Mb/s SHA256: ~1800 Mb/s SHA512: ~2700 Mb/s As you can see, performance is poor in comparison to aesni(4) and even cryptosoft (due to high setup cost). At a larger buffer size (128kB), throughput is a little better (but still worse than aesni(4)): aesni: SHA1:~10400 Mb/s SHA256: ~9950 Mb/s ccp: ~2200 Mb/s SHA256: ~2600 Mb/s SHA512: ~3800 Mb/s cryptosoft: ~1750 Mb/s SHA256: ~1800 Mb/s SHA512: ~2700 Mb/s AES performance has a similar story: aesni: 4kB: ~11250 Mb/s 128kB: ~11250 Mb/s ccp: ~350 Mb/s 128kB: ~4600 Mb/s cryptosoft: ~1750 Mb/s 128kB: ~1700 Mb/s This driver is EXPERIMENTAL. You should verify cryptographic results on typical and corner case inputs from your application against a known- good implementation. Sponsored by: Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D12723
2018-01-18 22:01:30 +00:00
case CRYPTO_AES_NIST_GCM_16:
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 ((sc->hw_features & VERSION_CAP_AES) == 0)
Add ccp(4): experimental driver for AMD Crypto Co-Processor * Registers TRNG source for random(4) * Finds available queues, LSBs; allocates static objects * Allocates a shared MSI-X for all queues. The hardware does not have separate interrupts per queue. Working interrupt mode driver. * Computes SHA hashes, HMAC. Passes cryptotest.py, cryptocheck tests. * Does AES-CBC, CTR mode, and XTS. cryptotest.py and cryptocheck pass. * Support for "authenc" (AES + HMAC). (SHA1 seems to result in "unaligned" cleartext inputs from cryptocheck -- which the engine cannot handle. SHA2 seems to work fine.) * GCM passes for block-multiple AAD, input lengths Largely based on ccr(4), part of cxgbe(4). Rough performance averages on AMD Ryzen 1950X (4kB buffer): aesni: SHA1: ~8300 Mb/s SHA256: ~8000 Mb/s ccp: ~630 Mb/s SHA256: ~660 Mb/s SHA512: ~700 Mb/s cryptosoft: ~1800 Mb/s SHA256: ~1800 Mb/s SHA512: ~2700 Mb/s As you can see, performance is poor in comparison to aesni(4) and even cryptosoft (due to high setup cost). At a larger buffer size (128kB), throughput is a little better (but still worse than aesni(4)): aesni: SHA1:~10400 Mb/s SHA256: ~9950 Mb/s ccp: ~2200 Mb/s SHA256: ~2600 Mb/s SHA512: ~3800 Mb/s cryptosoft: ~1750 Mb/s SHA256: ~1800 Mb/s SHA512: ~2700 Mb/s AES performance has a similar story: aesni: 4kB: ~11250 Mb/s 128kB: ~11250 Mb/s ccp: ~350 Mb/s 128kB: ~4600 Mb/s cryptosoft: ~1750 Mb/s 128kB: ~1700 Mb/s This driver is EXPERIMENTAL. You should verify cryptographic results on typical and corner case inputs from your application against a known- good implementation. Sponsored by: Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D12723
2018-01-18 22:01:30 +00:00
return (EINVAL);
break;
default:
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
break;
case CSP_MODE_ETA:
if (!ccp_auth_supported(sc, csp) ||
!ccp_cipher_supported(sc, csp))
return (EINVAL);
break;
default:
Add ccp(4): experimental driver for AMD Crypto Co-Processor * Registers TRNG source for random(4) * Finds available queues, LSBs; allocates static objects * Allocates a shared MSI-X for all queues. The hardware does not have separate interrupts per queue. Working interrupt mode driver. * Computes SHA hashes, HMAC. Passes cryptotest.py, cryptocheck tests. * Does AES-CBC, CTR mode, and XTS. cryptotest.py and cryptocheck pass. * Support for "authenc" (AES + HMAC). (SHA1 seems to result in "unaligned" cleartext inputs from cryptocheck -- which the engine cannot handle. SHA2 seems to work fine.) * GCM passes for block-multiple AAD, input lengths Largely based on ccr(4), part of cxgbe(4). Rough performance averages on AMD Ryzen 1950X (4kB buffer): aesni: SHA1: ~8300 Mb/s SHA256: ~8000 Mb/s ccp: ~630 Mb/s SHA256: ~660 Mb/s SHA512: ~700 Mb/s cryptosoft: ~1800 Mb/s SHA256: ~1800 Mb/s SHA512: ~2700 Mb/s As you can see, performance is poor in comparison to aesni(4) and even cryptosoft (due to high setup cost). At a larger buffer size (128kB), throughput is a little better (but still worse than aesni(4)): aesni: SHA1:~10400 Mb/s SHA256: ~9950 Mb/s ccp: ~2200 Mb/s SHA256: ~2600 Mb/s SHA512: ~3800 Mb/s cryptosoft: ~1750 Mb/s SHA256: ~1800 Mb/s SHA512: ~2700 Mb/s AES performance has a similar story: aesni: 4kB: ~11250 Mb/s 128kB: ~11250 Mb/s ccp: ~350 Mb/s 128kB: ~4600 Mb/s cryptosoft: ~1750 Mb/s 128kB: ~1700 Mb/s This driver is EXPERIMENTAL. You should verify cryptographic results on typical and corner case inputs from your application against a known- good implementation. Sponsored by: Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D12723
2018-01-18 22:01:30 +00:00
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
}
return (CRYPTODEV_PROBE_HARDWARE);
}
static int
ccp_newsession(device_t dev, crypto_session_t cses,
const struct crypto_session_params *csp)
{
struct ccp_softc *sc;
struct ccp_session *s;
crypto: Constify all transform descriptors No functional change intended. Reviewed by: ae, jhb MFC after: 1 week Sponsored by: The FreeBSD Foundation Differential Revision: https://reviews.freebsd.org/D31196
2021-07-26 20:41:05 +00:00
const struct auth_hash *auth_hash;
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
enum ccp_aes_mode cipher_mode;
unsigned auth_mode;
unsigned q;
/* XXX reconcile auth_mode with use by ccp_sha */
switch (csp->csp_auth_alg) {
case CRYPTO_SHA1_HMAC:
auth_hash = &auth_hash_hmac_sha1;
auth_mode = SHA1;
break;
case CRYPTO_SHA2_256_HMAC:
auth_hash = &auth_hash_hmac_sha2_256;
auth_mode = SHA2_256;
break;
case CRYPTO_SHA2_384_HMAC:
auth_hash = &auth_hash_hmac_sha2_384;
auth_mode = SHA2_384;
break;
case CRYPTO_SHA2_512_HMAC:
auth_hash = &auth_hash_hmac_sha2_512;
auth_mode = SHA2_512;
break;
default:
auth_hash = NULL;
auth_mode = 0;
break;
}
switch (csp->csp_cipher_alg) {
case CRYPTO_AES_CBC:
cipher_mode = CCP_AES_MODE_CBC;
break;
case CRYPTO_AES_ICM:
cipher_mode = CCP_AES_MODE_CTR;
break;
case CRYPTO_AES_NIST_GCM_16:
cipher_mode = CCP_AES_MODE_GCTR;
break;
case CRYPTO_AES_XTS:
cipher_mode = CCP_AES_MODE_XTS;
break;
default:
cipher_mode = CCP_AES_MODE_ECB;
break;
}
Add ccp(4): experimental driver for AMD Crypto Co-Processor * Registers TRNG source for random(4) * Finds available queues, LSBs; allocates static objects * Allocates a shared MSI-X for all queues. The hardware does not have separate interrupts per queue. Working interrupt mode driver. * Computes SHA hashes, HMAC. Passes cryptotest.py, cryptocheck tests. * Does AES-CBC, CTR mode, and XTS. cryptotest.py and cryptocheck pass. * Support for "authenc" (AES + HMAC). (SHA1 seems to result in "unaligned" cleartext inputs from cryptocheck -- which the engine cannot handle. SHA2 seems to work fine.) * GCM passes for block-multiple AAD, input lengths Largely based on ccr(4), part of cxgbe(4). Rough performance averages on AMD Ryzen 1950X (4kB buffer): aesni: SHA1: ~8300 Mb/s SHA256: ~8000 Mb/s ccp: ~630 Mb/s SHA256: ~660 Mb/s SHA512: ~700 Mb/s cryptosoft: ~1800 Mb/s SHA256: ~1800 Mb/s SHA512: ~2700 Mb/s As you can see, performance is poor in comparison to aesni(4) and even cryptosoft (due to high setup cost). At a larger buffer size (128kB), throughput is a little better (but still worse than aesni(4)): aesni: SHA1:~10400 Mb/s SHA256: ~9950 Mb/s ccp: ~2200 Mb/s SHA256: ~2600 Mb/s SHA512: ~3800 Mb/s cryptosoft: ~1750 Mb/s SHA256: ~1800 Mb/s SHA512: ~2700 Mb/s AES performance has a similar story: aesni: 4kB: ~11250 Mb/s 128kB: ~11250 Mb/s ccp: ~350 Mb/s 128kB: ~4600 Mb/s cryptosoft: ~1750 Mb/s 128kB: ~1700 Mb/s This driver is EXPERIMENTAL. You should verify cryptographic results on typical and corner case inputs from your application against a known- good implementation. Sponsored by: Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D12723
2018-01-18 22:01:30 +00:00
sc = device_get_softc(dev);
mtx_lock(&sc->lock);
if (sc->detaching) {
mtx_unlock(&sc->lock);
return (ENXIO);
}
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
s = crypto_get_driver_session(cses);
Add ccp(4): experimental driver for AMD Crypto Co-Processor * Registers TRNG source for random(4) * Finds available queues, LSBs; allocates static objects * Allocates a shared MSI-X for all queues. The hardware does not have separate interrupts per queue. Working interrupt mode driver. * Computes SHA hashes, HMAC. Passes cryptotest.py, cryptocheck tests. * Does AES-CBC, CTR mode, and XTS. cryptotest.py and cryptocheck pass. * Support for "authenc" (AES + HMAC). (SHA1 seems to result in "unaligned" cleartext inputs from cryptocheck -- which the engine cannot handle. SHA2 seems to work fine.) * GCM passes for block-multiple AAD, input lengths Largely based on ccr(4), part of cxgbe(4). Rough performance averages on AMD Ryzen 1950X (4kB buffer): aesni: SHA1: ~8300 Mb/s SHA256: ~8000 Mb/s ccp: ~630 Mb/s SHA256: ~660 Mb/s SHA512: ~700 Mb/s cryptosoft: ~1800 Mb/s SHA256: ~1800 Mb/s SHA512: ~2700 Mb/s As you can see, performance is poor in comparison to aesni(4) and even cryptosoft (due to high setup cost). At a larger buffer size (128kB), throughput is a little better (but still worse than aesni(4)): aesni: SHA1:~10400 Mb/s SHA256: ~9950 Mb/s ccp: ~2200 Mb/s SHA256: ~2600 Mb/s SHA512: ~3800 Mb/s cryptosoft: ~1750 Mb/s SHA256: ~1800 Mb/s SHA512: ~2700 Mb/s AES performance has a similar story: aesni: 4kB: ~11250 Mb/s 128kB: ~11250 Mb/s ccp: ~350 Mb/s 128kB: ~4600 Mb/s cryptosoft: ~1750 Mb/s 128kB: ~1700 Mb/s This driver is EXPERIMENTAL. You should verify cryptographic results on typical and corner case inputs from your application against a known- good implementation. Sponsored by: Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D12723
2018-01-18 22:01:30 +00:00
/* Just grab the first usable queue for now. */
for (q = 0; q < nitems(sc->queues); q++)
if ((sc->valid_queues & (1 << q)) != 0)
break;
if (q == nitems(sc->queues)) {
mtx_unlock(&sc->lock);
return (ENXIO);
}
s->queue = q;
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 (csp->csp_mode) {
case CSP_MODE_AEAD:
Add ccp(4): experimental driver for AMD Crypto Co-Processor * Registers TRNG source for random(4) * Finds available queues, LSBs; allocates static objects * Allocates a shared MSI-X for all queues. The hardware does not have separate interrupts per queue. Working interrupt mode driver. * Computes SHA hashes, HMAC. Passes cryptotest.py, cryptocheck tests. * Does AES-CBC, CTR mode, and XTS. cryptotest.py and cryptocheck pass. * Support for "authenc" (AES + HMAC). (SHA1 seems to result in "unaligned" cleartext inputs from cryptocheck -- which the engine cannot handle. SHA2 seems to work fine.) * GCM passes for block-multiple AAD, input lengths Largely based on ccr(4), part of cxgbe(4). Rough performance averages on AMD Ryzen 1950X (4kB buffer): aesni: SHA1: ~8300 Mb/s SHA256: ~8000 Mb/s ccp: ~630 Mb/s SHA256: ~660 Mb/s SHA512: ~700 Mb/s cryptosoft: ~1800 Mb/s SHA256: ~1800 Mb/s SHA512: ~2700 Mb/s As you can see, performance is poor in comparison to aesni(4) and even cryptosoft (due to high setup cost). At a larger buffer size (128kB), throughput is a little better (but still worse than aesni(4)): aesni: SHA1:~10400 Mb/s SHA256: ~9950 Mb/s ccp: ~2200 Mb/s SHA256: ~2600 Mb/s SHA512: ~3800 Mb/s cryptosoft: ~1750 Mb/s SHA256: ~1800 Mb/s SHA512: ~2700 Mb/s AES performance has a similar story: aesni: 4kB: ~11250 Mb/s 128kB: ~11250 Mb/s ccp: ~350 Mb/s 128kB: ~4600 Mb/s cryptosoft: ~1750 Mb/s 128kB: ~1700 Mb/s This driver is EXPERIMENTAL. You should verify cryptographic results on typical and corner case inputs from your application against a known- good implementation. Sponsored by: Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D12723
2018-01-18 22:01:30 +00:00
s->mode = GCM;
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;
case CSP_MODE_ETA:
Add ccp(4): experimental driver for AMD Crypto Co-Processor * Registers TRNG source for random(4) * Finds available queues, LSBs; allocates static objects * Allocates a shared MSI-X for all queues. The hardware does not have separate interrupts per queue. Working interrupt mode driver. * Computes SHA hashes, HMAC. Passes cryptotest.py, cryptocheck tests. * Does AES-CBC, CTR mode, and XTS. cryptotest.py and cryptocheck pass. * Support for "authenc" (AES + HMAC). (SHA1 seems to result in "unaligned" cleartext inputs from cryptocheck -- which the engine cannot handle. SHA2 seems to work fine.) * GCM passes for block-multiple AAD, input lengths Largely based on ccr(4), part of cxgbe(4). Rough performance averages on AMD Ryzen 1950X (4kB buffer): aesni: SHA1: ~8300 Mb/s SHA256: ~8000 Mb/s ccp: ~630 Mb/s SHA256: ~660 Mb/s SHA512: ~700 Mb/s cryptosoft: ~1800 Mb/s SHA256: ~1800 Mb/s SHA512: ~2700 Mb/s As you can see, performance is poor in comparison to aesni(4) and even cryptosoft (due to high setup cost). At a larger buffer size (128kB), throughput is a little better (but still worse than aesni(4)): aesni: SHA1:~10400 Mb/s SHA256: ~9950 Mb/s ccp: ~2200 Mb/s SHA256: ~2600 Mb/s SHA512: ~3800 Mb/s cryptosoft: ~1750 Mb/s SHA256: ~1800 Mb/s SHA512: ~2700 Mb/s AES performance has a similar story: aesni: 4kB: ~11250 Mb/s 128kB: ~11250 Mb/s ccp: ~350 Mb/s 128kB: ~4600 Mb/s cryptosoft: ~1750 Mb/s 128kB: ~1700 Mb/s This driver is EXPERIMENTAL. You should verify cryptographic results on typical and corner case inputs from your application against a known- good implementation. Sponsored by: Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D12723
2018-01-18 22:01:30 +00:00
s->mode = AUTHENC;
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;
case CSP_MODE_DIGEST:
Add ccp(4): experimental driver for AMD Crypto Co-Processor * Registers TRNG source for random(4) * Finds available queues, LSBs; allocates static objects * Allocates a shared MSI-X for all queues. The hardware does not have separate interrupts per queue. Working interrupt mode driver. * Computes SHA hashes, HMAC. Passes cryptotest.py, cryptocheck tests. * Does AES-CBC, CTR mode, and XTS. cryptotest.py and cryptocheck pass. * Support for "authenc" (AES + HMAC). (SHA1 seems to result in "unaligned" cleartext inputs from cryptocheck -- which the engine cannot handle. SHA2 seems to work fine.) * GCM passes for block-multiple AAD, input lengths Largely based on ccr(4), part of cxgbe(4). Rough performance averages on AMD Ryzen 1950X (4kB buffer): aesni: SHA1: ~8300 Mb/s SHA256: ~8000 Mb/s ccp: ~630 Mb/s SHA256: ~660 Mb/s SHA512: ~700 Mb/s cryptosoft: ~1800 Mb/s SHA256: ~1800 Mb/s SHA512: ~2700 Mb/s As you can see, performance is poor in comparison to aesni(4) and even cryptosoft (due to high setup cost). At a larger buffer size (128kB), throughput is a little better (but still worse than aesni(4)): aesni: SHA1:~10400 Mb/s SHA256: ~9950 Mb/s ccp: ~2200 Mb/s SHA256: ~2600 Mb/s SHA512: ~3800 Mb/s cryptosoft: ~1750 Mb/s SHA256: ~1800 Mb/s SHA512: ~2700 Mb/s AES performance has a similar story: aesni: 4kB: ~11250 Mb/s 128kB: ~11250 Mb/s ccp: ~350 Mb/s 128kB: ~4600 Mb/s cryptosoft: ~1750 Mb/s 128kB: ~1700 Mb/s This driver is EXPERIMENTAL. You should verify cryptographic results on typical and corner case inputs from your application against a known- good implementation. Sponsored by: Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D12723
2018-01-18 22:01:30 +00:00
s->mode = HMAC;
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;
case CSP_MODE_CIPHER:
Add ccp(4): experimental driver for AMD Crypto Co-Processor * Registers TRNG source for random(4) * Finds available queues, LSBs; allocates static objects * Allocates a shared MSI-X for all queues. The hardware does not have separate interrupts per queue. Working interrupt mode driver. * Computes SHA hashes, HMAC. Passes cryptotest.py, cryptocheck tests. * Does AES-CBC, CTR mode, and XTS. cryptotest.py and cryptocheck pass. * Support for "authenc" (AES + HMAC). (SHA1 seems to result in "unaligned" cleartext inputs from cryptocheck -- which the engine cannot handle. SHA2 seems to work fine.) * GCM passes for block-multiple AAD, input lengths Largely based on ccr(4), part of cxgbe(4). Rough performance averages on AMD Ryzen 1950X (4kB buffer): aesni: SHA1: ~8300 Mb/s SHA256: ~8000 Mb/s ccp: ~630 Mb/s SHA256: ~660 Mb/s SHA512: ~700 Mb/s cryptosoft: ~1800 Mb/s SHA256: ~1800 Mb/s SHA512: ~2700 Mb/s As you can see, performance is poor in comparison to aesni(4) and even cryptosoft (due to high setup cost). At a larger buffer size (128kB), throughput is a little better (but still worse than aesni(4)): aesni: SHA1:~10400 Mb/s SHA256: ~9950 Mb/s ccp: ~2200 Mb/s SHA256: ~2600 Mb/s SHA512: ~3800 Mb/s cryptosoft: ~1750 Mb/s SHA256: ~1800 Mb/s SHA512: ~2700 Mb/s AES performance has a similar story: aesni: 4kB: ~11250 Mb/s 128kB: ~11250 Mb/s ccp: ~350 Mb/s 128kB: ~4600 Mb/s cryptosoft: ~1750 Mb/s 128kB: ~1700 Mb/s This driver is EXPERIMENTAL. You should verify cryptographic results on typical and corner case inputs from your application against a known- good implementation. Sponsored by: Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D12723
2018-01-18 22:01:30 +00:00
s->mode = BLKCIPHER;
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;
Add ccp(4): experimental driver for AMD Crypto Co-Processor * Registers TRNG source for random(4) * Finds available queues, LSBs; allocates static objects * Allocates a shared MSI-X for all queues. The hardware does not have separate interrupts per queue. Working interrupt mode driver. * Computes SHA hashes, HMAC. Passes cryptotest.py, cryptocheck tests. * Does AES-CBC, CTR mode, and XTS. cryptotest.py and cryptocheck pass. * Support for "authenc" (AES + HMAC). (SHA1 seems to result in "unaligned" cleartext inputs from cryptocheck -- which the engine cannot handle. SHA2 seems to work fine.) * GCM passes for block-multiple AAD, input lengths Largely based on ccr(4), part of cxgbe(4). Rough performance averages on AMD Ryzen 1950X (4kB buffer): aesni: SHA1: ~8300 Mb/s SHA256: ~8000 Mb/s ccp: ~630 Mb/s SHA256: ~660 Mb/s SHA512: ~700 Mb/s cryptosoft: ~1800 Mb/s SHA256: ~1800 Mb/s SHA512: ~2700 Mb/s As you can see, performance is poor in comparison to aesni(4) and even cryptosoft (due to high setup cost). At a larger buffer size (128kB), throughput is a little better (but still worse than aesni(4)): aesni: SHA1:~10400 Mb/s SHA256: ~9950 Mb/s ccp: ~2200 Mb/s SHA256: ~2600 Mb/s SHA512: ~3800 Mb/s cryptosoft: ~1750 Mb/s SHA256: ~1800 Mb/s SHA512: ~2700 Mb/s AES performance has a similar story: aesni: 4kB: ~11250 Mb/s 128kB: ~11250 Mb/s ccp: ~350 Mb/s 128kB: ~4600 Mb/s cryptosoft: ~1750 Mb/s 128kB: ~1700 Mb/s This driver is EXPERIMENTAL. You should verify cryptographic results on typical and corner case inputs from your application against a known- good implementation. Sponsored by: Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D12723
2018-01-18 22:01:30 +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
if (s->mode == GCM) {
if (csp->csp_auth_mlen == 0)
Add ccp(4): experimental driver for AMD Crypto Co-Processor * Registers TRNG source for random(4) * Finds available queues, LSBs; allocates static objects * Allocates a shared MSI-X for all queues. The hardware does not have separate interrupts per queue. Working interrupt mode driver. * Computes SHA hashes, HMAC. Passes cryptotest.py, cryptocheck tests. * Does AES-CBC, CTR mode, and XTS. cryptotest.py and cryptocheck pass. * Support for "authenc" (AES + HMAC). (SHA1 seems to result in "unaligned" cleartext inputs from cryptocheck -- which the engine cannot handle. SHA2 seems to work fine.) * GCM passes for block-multiple AAD, input lengths Largely based on ccr(4), part of cxgbe(4). Rough performance averages on AMD Ryzen 1950X (4kB buffer): aesni: SHA1: ~8300 Mb/s SHA256: ~8000 Mb/s ccp: ~630 Mb/s SHA256: ~660 Mb/s SHA512: ~700 Mb/s cryptosoft: ~1800 Mb/s SHA256: ~1800 Mb/s SHA512: ~2700 Mb/s As you can see, performance is poor in comparison to aesni(4) and even cryptosoft (due to high setup cost). At a larger buffer size (128kB), throughput is a little better (but still worse than aesni(4)): aesni: SHA1:~10400 Mb/s SHA256: ~9950 Mb/s ccp: ~2200 Mb/s SHA256: ~2600 Mb/s SHA512: ~3800 Mb/s cryptosoft: ~1750 Mb/s SHA256: ~1800 Mb/s SHA512: ~2700 Mb/s AES performance has a similar story: aesni: 4kB: ~11250 Mb/s 128kB: ~11250 Mb/s ccp: ~350 Mb/s 128kB: ~4600 Mb/s cryptosoft: ~1750 Mb/s 128kB: ~1700 Mb/s This driver is EXPERIMENTAL. You should verify cryptographic results on typical and corner case inputs from your application against a known- good implementation. Sponsored by: Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D12723
2018-01-18 22:01:30 +00:00
s->gmac.hash_len = AES_GMAC_HASH_LEN;
else
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
s->gmac.hash_len = csp->csp_auth_mlen;
} else if (auth_hash != NULL) {
Add ccp(4): experimental driver for AMD Crypto Co-Processor * Registers TRNG source for random(4) * Finds available queues, LSBs; allocates static objects * Allocates a shared MSI-X for all queues. The hardware does not have separate interrupts per queue. Working interrupt mode driver. * Computes SHA hashes, HMAC. Passes cryptotest.py, cryptocheck tests. * Does AES-CBC, CTR mode, and XTS. cryptotest.py and cryptocheck pass. * Support for "authenc" (AES + HMAC). (SHA1 seems to result in "unaligned" cleartext inputs from cryptocheck -- which the engine cannot handle. SHA2 seems to work fine.) * GCM passes for block-multiple AAD, input lengths Largely based on ccr(4), part of cxgbe(4). Rough performance averages on AMD Ryzen 1950X (4kB buffer): aesni: SHA1: ~8300 Mb/s SHA256: ~8000 Mb/s ccp: ~630 Mb/s SHA256: ~660 Mb/s SHA512: ~700 Mb/s cryptosoft: ~1800 Mb/s SHA256: ~1800 Mb/s SHA512: ~2700 Mb/s As you can see, performance is poor in comparison to aesni(4) and even cryptosoft (due to high setup cost). At a larger buffer size (128kB), throughput is a little better (but still worse than aesni(4)): aesni: SHA1:~10400 Mb/s SHA256: ~9950 Mb/s ccp: ~2200 Mb/s SHA256: ~2600 Mb/s SHA512: ~3800 Mb/s cryptosoft: ~1750 Mb/s SHA256: ~1800 Mb/s SHA512: ~2700 Mb/s AES performance has a similar story: aesni: 4kB: ~11250 Mb/s 128kB: ~11250 Mb/s ccp: ~350 Mb/s 128kB: ~4600 Mb/s cryptosoft: ~1750 Mb/s 128kB: ~1700 Mb/s This driver is EXPERIMENTAL. You should verify cryptographic results on typical and corner case inputs from your application against a known- good implementation. Sponsored by: Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D12723
2018-01-18 22:01:30 +00:00
s->hmac.auth_hash = auth_hash;
s->hmac.auth_mode = auth_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
if (csp->csp_auth_mlen == 0)
Add ccp(4): experimental driver for AMD Crypto Co-Processor * Registers TRNG source for random(4) * Finds available queues, LSBs; allocates static objects * Allocates a shared MSI-X for all queues. The hardware does not have separate interrupts per queue. Working interrupt mode driver. * Computes SHA hashes, HMAC. Passes cryptotest.py, cryptocheck tests. * Does AES-CBC, CTR mode, and XTS. cryptotest.py and cryptocheck pass. * Support for "authenc" (AES + HMAC). (SHA1 seems to result in "unaligned" cleartext inputs from cryptocheck -- which the engine cannot handle. SHA2 seems to work fine.) * GCM passes for block-multiple AAD, input lengths Largely based on ccr(4), part of cxgbe(4). Rough performance averages on AMD Ryzen 1950X (4kB buffer): aesni: SHA1: ~8300 Mb/s SHA256: ~8000 Mb/s ccp: ~630 Mb/s SHA256: ~660 Mb/s SHA512: ~700 Mb/s cryptosoft: ~1800 Mb/s SHA256: ~1800 Mb/s SHA512: ~2700 Mb/s As you can see, performance is poor in comparison to aesni(4) and even cryptosoft (due to high setup cost). At a larger buffer size (128kB), throughput is a little better (but still worse than aesni(4)): aesni: SHA1:~10400 Mb/s SHA256: ~9950 Mb/s ccp: ~2200 Mb/s SHA256: ~2600 Mb/s SHA512: ~3800 Mb/s cryptosoft: ~1750 Mb/s SHA256: ~1800 Mb/s SHA512: ~2700 Mb/s AES performance has a similar story: aesni: 4kB: ~11250 Mb/s 128kB: ~11250 Mb/s ccp: ~350 Mb/s 128kB: ~4600 Mb/s cryptosoft: ~1750 Mb/s 128kB: ~1700 Mb/s This driver is EXPERIMENTAL. You should verify cryptographic results on typical and corner case inputs from your application against a known- good implementation. Sponsored by: Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D12723
2018-01-18 22:01:30 +00:00
s->hmac.hash_len = auth_hash->hashsize;
else
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
s->hmac.hash_len = csp->csp_auth_mlen;
ccp_init_hmac_digest(s, csp->csp_auth_key, csp->csp_auth_klen);
Add ccp(4): experimental driver for AMD Crypto Co-Processor * Registers TRNG source for random(4) * Finds available queues, LSBs; allocates static objects * Allocates a shared MSI-X for all queues. The hardware does not have separate interrupts per queue. Working interrupt mode driver. * Computes SHA hashes, HMAC. Passes cryptotest.py, cryptocheck tests. * Does AES-CBC, CTR mode, and XTS. cryptotest.py and cryptocheck pass. * Support for "authenc" (AES + HMAC). (SHA1 seems to result in "unaligned" cleartext inputs from cryptocheck -- which the engine cannot handle. SHA2 seems to work fine.) * GCM passes for block-multiple AAD, input lengths Largely based on ccr(4), part of cxgbe(4). Rough performance averages on AMD Ryzen 1950X (4kB buffer): aesni: SHA1: ~8300 Mb/s SHA256: ~8000 Mb/s ccp: ~630 Mb/s SHA256: ~660 Mb/s SHA512: ~700 Mb/s cryptosoft: ~1800 Mb/s SHA256: ~1800 Mb/s SHA512: ~2700 Mb/s As you can see, performance is poor in comparison to aesni(4) and even cryptosoft (due to high setup cost). At a larger buffer size (128kB), throughput is a little better (but still worse than aesni(4)): aesni: SHA1:~10400 Mb/s SHA256: ~9950 Mb/s ccp: ~2200 Mb/s SHA256: ~2600 Mb/s SHA512: ~3800 Mb/s cryptosoft: ~1750 Mb/s SHA256: ~1800 Mb/s SHA512: ~2700 Mb/s AES performance has a similar story: aesni: 4kB: ~11250 Mb/s 128kB: ~11250 Mb/s ccp: ~350 Mb/s 128kB: ~4600 Mb/s cryptosoft: ~1750 Mb/s 128kB: ~1700 Mb/s This driver is EXPERIMENTAL. You should verify cryptographic results on typical and corner case inputs from your application against a known- good implementation. Sponsored by: Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D12723
2018-01-18 22:01:30 +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
if (cipher_mode != CCP_AES_MODE_ECB) {
Add ccp(4): experimental driver for AMD Crypto Co-Processor * Registers TRNG source for random(4) * Finds available queues, LSBs; allocates static objects * Allocates a shared MSI-X for all queues. The hardware does not have separate interrupts per queue. Working interrupt mode driver. * Computes SHA hashes, HMAC. Passes cryptotest.py, cryptocheck tests. * Does AES-CBC, CTR mode, and XTS. cryptotest.py and cryptocheck pass. * Support for "authenc" (AES + HMAC). (SHA1 seems to result in "unaligned" cleartext inputs from cryptocheck -- which the engine cannot handle. SHA2 seems to work fine.) * GCM passes for block-multiple AAD, input lengths Largely based on ccr(4), part of cxgbe(4). Rough performance averages on AMD Ryzen 1950X (4kB buffer): aesni: SHA1: ~8300 Mb/s SHA256: ~8000 Mb/s ccp: ~630 Mb/s SHA256: ~660 Mb/s SHA512: ~700 Mb/s cryptosoft: ~1800 Mb/s SHA256: ~1800 Mb/s SHA512: ~2700 Mb/s As you can see, performance is poor in comparison to aesni(4) and even cryptosoft (due to high setup cost). At a larger buffer size (128kB), throughput is a little better (but still worse than aesni(4)): aesni: SHA1:~10400 Mb/s SHA256: ~9950 Mb/s ccp: ~2200 Mb/s SHA256: ~2600 Mb/s SHA512: ~3800 Mb/s cryptosoft: ~1750 Mb/s SHA256: ~1800 Mb/s SHA512: ~2700 Mb/s AES performance has a similar story: aesni: 4kB: ~11250 Mb/s 128kB: ~11250 Mb/s ccp: ~350 Mb/s 128kB: ~4600 Mb/s cryptosoft: ~1750 Mb/s 128kB: ~1700 Mb/s This driver is EXPERIMENTAL. You should verify cryptographic results on typical and corner case inputs from your application against a known- good implementation. Sponsored by: Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D12723
2018-01-18 22:01:30 +00:00
s->blkcipher.cipher_mode = cipher_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
if (csp->csp_cipher_key != NULL)
ccp_aes_setkey(s, csp->csp_cipher_alg,
csp->csp_cipher_key, csp->csp_cipher_klen);
Add ccp(4): experimental driver for AMD Crypto Co-Processor * Registers TRNG source for random(4) * Finds available queues, LSBs; allocates static objects * Allocates a shared MSI-X for all queues. The hardware does not have separate interrupts per queue. Working interrupt mode driver. * Computes SHA hashes, HMAC. Passes cryptotest.py, cryptocheck tests. * Does AES-CBC, CTR mode, and XTS. cryptotest.py and cryptocheck pass. * Support for "authenc" (AES + HMAC). (SHA1 seems to result in "unaligned" cleartext inputs from cryptocheck -- which the engine cannot handle. SHA2 seems to work fine.) * GCM passes for block-multiple AAD, input lengths Largely based on ccr(4), part of cxgbe(4). Rough performance averages on AMD Ryzen 1950X (4kB buffer): aesni: SHA1: ~8300 Mb/s SHA256: ~8000 Mb/s ccp: ~630 Mb/s SHA256: ~660 Mb/s SHA512: ~700 Mb/s cryptosoft: ~1800 Mb/s SHA256: ~1800 Mb/s SHA512: ~2700 Mb/s As you can see, performance is poor in comparison to aesni(4) and even cryptosoft (due to high setup cost). At a larger buffer size (128kB), throughput is a little better (but still worse than aesni(4)): aesni: SHA1:~10400 Mb/s SHA256: ~9950 Mb/s ccp: ~2200 Mb/s SHA256: ~2600 Mb/s SHA512: ~3800 Mb/s cryptosoft: ~1750 Mb/s SHA256: ~1800 Mb/s SHA512: ~2700 Mb/s AES performance has a similar story: aesni: 4kB: ~11250 Mb/s 128kB: ~11250 Mb/s ccp: ~350 Mb/s 128kB: ~4600 Mb/s cryptosoft: ~1750 Mb/s 128kB: ~1700 Mb/s This driver is EXPERIMENTAL. You should verify cryptographic results on typical and corner case inputs from your application against a known- good implementation. Sponsored by: Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D12723
2018-01-18 22:01:30 +00:00
}
s->active = true;
mtx_unlock(&sc->lock);
return (0);
}
OpenCrypto: Convert sessions to opaque handles instead of integers Track session objects in the framework, and pass handles between the framework (OCF), consumers, and drivers. Avoid redundancy and complexity in individual drivers by allocating session memory in the framework and providing it to drivers in ::newsession(). Session handles are no longer integers with information encoded in various high bits. Use of the CRYPTO_SESID2FOO() macros should be replaced with the appropriate crypto_ses2foo() function on the opaque session handle. Convert OCF drivers (in particular, cryptosoft, as well as myriad others) to the opaque handle interface. Discard existing session tracking as much as possible (quick pass). There may be additional code ripe for deletion. Convert OCF consumers (ipsec, geom_eli, krb5, cryptodev) to handle-style interface. The conversion is largely mechnical. The change is documented in crypto.9. Inspired by https://lists.freebsd.org/pipermail/freebsd-arch/2018-January/018835.html . No objection from: ae (ipsec portion) Reported by: jhb
2018-07-18 00:56:25 +00:00
static void
ccp_freesession(device_t dev, crypto_session_t cses)
Add ccp(4): experimental driver for AMD Crypto Co-Processor * Registers TRNG source for random(4) * Finds available queues, LSBs; allocates static objects * Allocates a shared MSI-X for all queues. The hardware does not have separate interrupts per queue. Working interrupt mode driver. * Computes SHA hashes, HMAC. Passes cryptotest.py, cryptocheck tests. * Does AES-CBC, CTR mode, and XTS. cryptotest.py and cryptocheck pass. * Support for "authenc" (AES + HMAC). (SHA1 seems to result in "unaligned" cleartext inputs from cryptocheck -- which the engine cannot handle. SHA2 seems to work fine.) * GCM passes for block-multiple AAD, input lengths Largely based on ccr(4), part of cxgbe(4). Rough performance averages on AMD Ryzen 1950X (4kB buffer): aesni: SHA1: ~8300 Mb/s SHA256: ~8000 Mb/s ccp: ~630 Mb/s SHA256: ~660 Mb/s SHA512: ~700 Mb/s cryptosoft: ~1800 Mb/s SHA256: ~1800 Mb/s SHA512: ~2700 Mb/s As you can see, performance is poor in comparison to aesni(4) and even cryptosoft (due to high setup cost). At a larger buffer size (128kB), throughput is a little better (but still worse than aesni(4)): aesni: SHA1:~10400 Mb/s SHA256: ~9950 Mb/s ccp: ~2200 Mb/s SHA256: ~2600 Mb/s SHA512: ~3800 Mb/s cryptosoft: ~1750 Mb/s SHA256: ~1800 Mb/s SHA512: ~2700 Mb/s AES performance has a similar story: aesni: 4kB: ~11250 Mb/s 128kB: ~11250 Mb/s ccp: ~350 Mb/s 128kB: ~4600 Mb/s cryptosoft: ~1750 Mb/s 128kB: ~1700 Mb/s This driver is EXPERIMENTAL. You should verify cryptographic results on typical and corner case inputs from your application against a known- good implementation. Sponsored by: Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D12723
2018-01-18 22:01:30 +00:00
{
OpenCrypto: Convert sessions to opaque handles instead of integers Track session objects in the framework, and pass handles between the framework (OCF), consumers, and drivers. Avoid redundancy and complexity in individual drivers by allocating session memory in the framework and providing it to drivers in ::newsession(). Session handles are no longer integers with information encoded in various high bits. Use of the CRYPTO_SESID2FOO() macros should be replaced with the appropriate crypto_ses2foo() function on the opaque session handle. Convert OCF drivers (in particular, cryptosoft, as well as myriad others) to the opaque handle interface. Discard existing session tracking as much as possible (quick pass). There may be additional code ripe for deletion. Convert OCF consumers (ipsec, geom_eli, krb5, cryptodev) to handle-style interface. The conversion is largely mechnical. The change is documented in crypto.9. Inspired by https://lists.freebsd.org/pipermail/freebsd-arch/2018-January/018835.html . No objection from: ae (ipsec portion) Reported by: jhb
2018-07-18 00:56:25 +00:00
struct ccp_session *s;
Add ccp(4): experimental driver for AMD Crypto Co-Processor * Registers TRNG source for random(4) * Finds available queues, LSBs; allocates static objects * Allocates a shared MSI-X for all queues. The hardware does not have separate interrupts per queue. Working interrupt mode driver. * Computes SHA hashes, HMAC. Passes cryptotest.py, cryptocheck tests. * Does AES-CBC, CTR mode, and XTS. cryptotest.py and cryptocheck pass. * Support for "authenc" (AES + HMAC). (SHA1 seems to result in "unaligned" cleartext inputs from cryptocheck -- which the engine cannot handle. SHA2 seems to work fine.) * GCM passes for block-multiple AAD, input lengths Largely based on ccr(4), part of cxgbe(4). Rough performance averages on AMD Ryzen 1950X (4kB buffer): aesni: SHA1: ~8300 Mb/s SHA256: ~8000 Mb/s ccp: ~630 Mb/s SHA256: ~660 Mb/s SHA512: ~700 Mb/s cryptosoft: ~1800 Mb/s SHA256: ~1800 Mb/s SHA512: ~2700 Mb/s As you can see, performance is poor in comparison to aesni(4) and even cryptosoft (due to high setup cost). At a larger buffer size (128kB), throughput is a little better (but still worse than aesni(4)): aesni: SHA1:~10400 Mb/s SHA256: ~9950 Mb/s ccp: ~2200 Mb/s SHA256: ~2600 Mb/s SHA512: ~3800 Mb/s cryptosoft: ~1750 Mb/s SHA256: ~1800 Mb/s SHA512: ~2700 Mb/s AES performance has a similar story: aesni: 4kB: ~11250 Mb/s 128kB: ~11250 Mb/s ccp: ~350 Mb/s 128kB: ~4600 Mb/s cryptosoft: ~1750 Mb/s 128kB: ~1700 Mb/s This driver is EXPERIMENTAL. You should verify cryptographic results on typical and corner case inputs from your application against a known- good implementation. Sponsored by: Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D12723
2018-01-18 22:01:30 +00:00
OpenCrypto: Convert sessions to opaque handles instead of integers Track session objects in the framework, and pass handles between the framework (OCF), consumers, and drivers. Avoid redundancy and complexity in individual drivers by allocating session memory in the framework and providing it to drivers in ::newsession(). Session handles are no longer integers with information encoded in various high bits. Use of the CRYPTO_SESID2FOO() macros should be replaced with the appropriate crypto_ses2foo() function on the opaque session handle. Convert OCF drivers (in particular, cryptosoft, as well as myriad others) to the opaque handle interface. Discard existing session tracking as much as possible (quick pass). There may be additional code ripe for deletion. Convert OCF consumers (ipsec, geom_eli, krb5, cryptodev) to handle-style interface. The conversion is largely mechnical. The change is documented in crypto.9. Inspired by https://lists.freebsd.org/pipermail/freebsd-arch/2018-January/018835.html . No objection from: ae (ipsec portion) Reported by: jhb
2018-07-18 00:56:25 +00:00
s = crypto_get_driver_session(cses);
if (s->pending != 0)
device_printf(dev,
"session %p freed with %d pending requests\n", s,
s->pending);
s->active = false;
Add ccp(4): experimental driver for AMD Crypto Co-Processor * Registers TRNG source for random(4) * Finds available queues, LSBs; allocates static objects * Allocates a shared MSI-X for all queues. The hardware does not have separate interrupts per queue. Working interrupt mode driver. * Computes SHA hashes, HMAC. Passes cryptotest.py, cryptocheck tests. * Does AES-CBC, CTR mode, and XTS. cryptotest.py and cryptocheck pass. * Support for "authenc" (AES + HMAC). (SHA1 seems to result in "unaligned" cleartext inputs from cryptocheck -- which the engine cannot handle. SHA2 seems to work fine.) * GCM passes for block-multiple AAD, input lengths Largely based on ccr(4), part of cxgbe(4). Rough performance averages on AMD Ryzen 1950X (4kB buffer): aesni: SHA1: ~8300 Mb/s SHA256: ~8000 Mb/s ccp: ~630 Mb/s SHA256: ~660 Mb/s SHA512: ~700 Mb/s cryptosoft: ~1800 Mb/s SHA256: ~1800 Mb/s SHA512: ~2700 Mb/s As you can see, performance is poor in comparison to aesni(4) and even cryptosoft (due to high setup cost). At a larger buffer size (128kB), throughput is a little better (but still worse than aesni(4)): aesni: SHA1:~10400 Mb/s SHA256: ~9950 Mb/s ccp: ~2200 Mb/s SHA256: ~2600 Mb/s SHA512: ~3800 Mb/s cryptosoft: ~1750 Mb/s SHA256: ~1800 Mb/s SHA512: ~2700 Mb/s AES performance has a similar story: aesni: 4kB: ~11250 Mb/s 128kB: ~11250 Mb/s ccp: ~350 Mb/s 128kB: ~4600 Mb/s cryptosoft: ~1750 Mb/s 128kB: ~1700 Mb/s This driver is EXPERIMENTAL. You should verify cryptographic results on typical and corner case inputs from your application against a known- good implementation. Sponsored by: Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D12723
2018-01-18 22:01:30 +00:00
}
static int
ccp_process(device_t dev, struct cryptop *crp, int hint)
{
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
const struct crypto_session_params *csp;
Add ccp(4): experimental driver for AMD Crypto Co-Processor * Registers TRNG source for random(4) * Finds available queues, LSBs; allocates static objects * Allocates a shared MSI-X for all queues. The hardware does not have separate interrupts per queue. Working interrupt mode driver. * Computes SHA hashes, HMAC. Passes cryptotest.py, cryptocheck tests. * Does AES-CBC, CTR mode, and XTS. cryptotest.py and cryptocheck pass. * Support for "authenc" (AES + HMAC). (SHA1 seems to result in "unaligned" cleartext inputs from cryptocheck -- which the engine cannot handle. SHA2 seems to work fine.) * GCM passes for block-multiple AAD, input lengths Largely based on ccr(4), part of cxgbe(4). Rough performance averages on AMD Ryzen 1950X (4kB buffer): aesni: SHA1: ~8300 Mb/s SHA256: ~8000 Mb/s ccp: ~630 Mb/s SHA256: ~660 Mb/s SHA512: ~700 Mb/s cryptosoft: ~1800 Mb/s SHA256: ~1800 Mb/s SHA512: ~2700 Mb/s As you can see, performance is poor in comparison to aesni(4) and even cryptosoft (due to high setup cost). At a larger buffer size (128kB), throughput is a little better (but still worse than aesni(4)): aesni: SHA1:~10400 Mb/s SHA256: ~9950 Mb/s ccp: ~2200 Mb/s SHA256: ~2600 Mb/s SHA512: ~3800 Mb/s cryptosoft: ~1750 Mb/s SHA256: ~1800 Mb/s SHA512: ~2700 Mb/s AES performance has a similar story: aesni: 4kB: ~11250 Mb/s 128kB: ~11250 Mb/s ccp: ~350 Mb/s 128kB: ~4600 Mb/s cryptosoft: ~1750 Mb/s 128kB: ~1700 Mb/s This driver is EXPERIMENTAL. You should verify cryptographic results on typical and corner case inputs from your application against a known- good implementation. Sponsored by: Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D12723
2018-01-18 22:01:30 +00:00
struct ccp_softc *sc;
struct ccp_queue *qp;
struct ccp_session *s;
int error;
bool qpheld;
qpheld = false;
qp = NULL;
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(crp->crp_session);
OpenCrypto: Convert sessions to opaque handles instead of integers Track session objects in the framework, and pass handles between the framework (OCF), consumers, and drivers. Avoid redundancy and complexity in individual drivers by allocating session memory in the framework and providing it to drivers in ::newsession(). Session handles are no longer integers with information encoded in various high bits. Use of the CRYPTO_SESID2FOO() macros should be replaced with the appropriate crypto_ses2foo() function on the opaque session handle. Convert OCF drivers (in particular, cryptosoft, as well as myriad others) to the opaque handle interface. Discard existing session tracking as much as possible (quick pass). There may be additional code ripe for deletion. Convert OCF consumers (ipsec, geom_eli, krb5, cryptodev) to handle-style interface. The conversion is largely mechnical. The change is documented in crypto.9. Inspired by https://lists.freebsd.org/pipermail/freebsd-arch/2018-January/018835.html . No objection from: ae (ipsec portion) Reported by: jhb
2018-07-18 00:56:25 +00:00
s = crypto_get_driver_session(crp->crp_session);
Add ccp(4): experimental driver for AMD Crypto Co-Processor * Registers TRNG source for random(4) * Finds available queues, LSBs; allocates static objects * Allocates a shared MSI-X for all queues. The hardware does not have separate interrupts per queue. Working interrupt mode driver. * Computes SHA hashes, HMAC. Passes cryptotest.py, cryptocheck tests. * Does AES-CBC, CTR mode, and XTS. cryptotest.py and cryptocheck pass. * Support for "authenc" (AES + HMAC). (SHA1 seems to result in "unaligned" cleartext inputs from cryptocheck -- which the engine cannot handle. SHA2 seems to work fine.) * GCM passes for block-multiple AAD, input lengths Largely based on ccr(4), part of cxgbe(4). Rough performance averages on AMD Ryzen 1950X (4kB buffer): aesni: SHA1: ~8300 Mb/s SHA256: ~8000 Mb/s ccp: ~630 Mb/s SHA256: ~660 Mb/s SHA512: ~700 Mb/s cryptosoft: ~1800 Mb/s SHA256: ~1800 Mb/s SHA512: ~2700 Mb/s As you can see, performance is poor in comparison to aesni(4) and even cryptosoft (due to high setup cost). At a larger buffer size (128kB), throughput is a little better (but still worse than aesni(4)): aesni: SHA1:~10400 Mb/s SHA256: ~9950 Mb/s ccp: ~2200 Mb/s SHA256: ~2600 Mb/s SHA512: ~3800 Mb/s cryptosoft: ~1750 Mb/s SHA256: ~1800 Mb/s SHA512: ~2700 Mb/s AES performance has a similar story: aesni: 4kB: ~11250 Mb/s 128kB: ~11250 Mb/s ccp: ~350 Mb/s 128kB: ~4600 Mb/s cryptosoft: ~1750 Mb/s 128kB: ~1700 Mb/s This driver is EXPERIMENTAL. You should verify cryptographic results on typical and corner case inputs from your application against a known- good implementation. Sponsored by: Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D12723
2018-01-18 22:01:30 +00:00
sc = device_get_softc(dev);
mtx_lock(&sc->lock);
qp = &sc->queues[s->queue];
mtx_unlock(&sc->lock);
error = ccp_queue_acquire_reserve(qp, 1 /* placeholder */, M_NOWAIT);
if (error != 0)
goto out;
qpheld = true;
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
error = ccp_populate_sglist(qp->cq_sg_crp, &crp->crp_buf);
Add ccp(4): experimental driver for AMD Crypto Co-Processor * Registers TRNG source for random(4) * Finds available queues, LSBs; allocates static objects * Allocates a shared MSI-X for all queues. The hardware does not have separate interrupts per queue. Working interrupt mode driver. * Computes SHA hashes, HMAC. Passes cryptotest.py, cryptocheck tests. * Does AES-CBC, CTR mode, and XTS. cryptotest.py and cryptocheck pass. * Support for "authenc" (AES + HMAC). (SHA1 seems to result in "unaligned" cleartext inputs from cryptocheck -- which the engine cannot handle. SHA2 seems to work fine.) * GCM passes for block-multiple AAD, input lengths Largely based on ccr(4), part of cxgbe(4). Rough performance averages on AMD Ryzen 1950X (4kB buffer): aesni: SHA1: ~8300 Mb/s SHA256: ~8000 Mb/s ccp: ~630 Mb/s SHA256: ~660 Mb/s SHA512: ~700 Mb/s cryptosoft: ~1800 Mb/s SHA256: ~1800 Mb/s SHA512: ~2700 Mb/s As you can see, performance is poor in comparison to aesni(4) and even cryptosoft (due to high setup cost). At a larger buffer size (128kB), throughput is a little better (but still worse than aesni(4)): aesni: SHA1:~10400 Mb/s SHA256: ~9950 Mb/s ccp: ~2200 Mb/s SHA256: ~2600 Mb/s SHA512: ~3800 Mb/s cryptosoft: ~1750 Mb/s SHA256: ~1800 Mb/s SHA512: ~2700 Mb/s AES performance has a similar story: aesni: 4kB: ~11250 Mb/s 128kB: ~11250 Mb/s ccp: ~350 Mb/s 128kB: ~4600 Mb/s cryptosoft: ~1750 Mb/s 128kB: ~1700 Mb/s This driver is EXPERIMENTAL. You should verify cryptographic results on typical and corner case inputs from your application against a known- good implementation. Sponsored by: Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D12723
2018-01-18 22:01:30 +00:00
if (error != 0)
goto out;
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_auth_key != NULL) {
KASSERT(s->hmac.auth_hash != NULL, ("auth key without HMAC"));
ccp_init_hmac_digest(s, crp->crp_auth_key, csp->csp_auth_klen);
}
if (crp->crp_cipher_key != NULL)
ccp_aes_setkey(s, csp->csp_cipher_alg, crp->crp_cipher_key,
csp->csp_cipher_klen);
Add ccp(4): experimental driver for AMD Crypto Co-Processor * Registers TRNG source for random(4) * Finds available queues, LSBs; allocates static objects * Allocates a shared MSI-X for all queues. The hardware does not have separate interrupts per queue. Working interrupt mode driver. * Computes SHA hashes, HMAC. Passes cryptotest.py, cryptocheck tests. * Does AES-CBC, CTR mode, and XTS. cryptotest.py and cryptocheck pass. * Support for "authenc" (AES + HMAC). (SHA1 seems to result in "unaligned" cleartext inputs from cryptocheck -- which the engine cannot handle. SHA2 seems to work fine.) * GCM passes for block-multiple AAD, input lengths Largely based on ccr(4), part of cxgbe(4). Rough performance averages on AMD Ryzen 1950X (4kB buffer): aesni: SHA1: ~8300 Mb/s SHA256: ~8000 Mb/s ccp: ~630 Mb/s SHA256: ~660 Mb/s SHA512: ~700 Mb/s cryptosoft: ~1800 Mb/s SHA256: ~1800 Mb/s SHA512: ~2700 Mb/s As you can see, performance is poor in comparison to aesni(4) and even cryptosoft (due to high setup cost). At a larger buffer size (128kB), throughput is a little better (but still worse than aesni(4)): aesni: SHA1:~10400 Mb/s SHA256: ~9950 Mb/s ccp: ~2200 Mb/s SHA256: ~2600 Mb/s SHA512: ~3800 Mb/s cryptosoft: ~1750 Mb/s SHA256: ~1800 Mb/s SHA512: ~2700 Mb/s AES performance has a similar story: aesni: 4kB: ~11250 Mb/s 128kB: ~11250 Mb/s ccp: ~350 Mb/s 128kB: ~4600 Mb/s cryptosoft: ~1750 Mb/s 128kB: ~1700 Mb/s This driver is EXPERIMENTAL. You should verify cryptographic results on typical and corner case inputs from your application against a known- good implementation. Sponsored by: Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D12723
2018-01-18 22:01:30 +00:00
switch (s->mode) {
case HMAC:
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 (s->pending != 0) {
error = EAGAIN;
break;
}
Add ccp(4): experimental driver for AMD Crypto Co-Processor * Registers TRNG source for random(4) * Finds available queues, LSBs; allocates static objects * Allocates a shared MSI-X for all queues. The hardware does not have separate interrupts per queue. Working interrupt mode driver. * Computes SHA hashes, HMAC. Passes cryptotest.py, cryptocheck tests. * Does AES-CBC, CTR mode, and XTS. cryptotest.py and cryptocheck pass. * Support for "authenc" (AES + HMAC). (SHA1 seems to result in "unaligned" cleartext inputs from cryptocheck -- which the engine cannot handle. SHA2 seems to work fine.) * GCM passes for block-multiple AAD, input lengths Largely based on ccr(4), part of cxgbe(4). Rough performance averages on AMD Ryzen 1950X (4kB buffer): aesni: SHA1: ~8300 Mb/s SHA256: ~8000 Mb/s ccp: ~630 Mb/s SHA256: ~660 Mb/s SHA512: ~700 Mb/s cryptosoft: ~1800 Mb/s SHA256: ~1800 Mb/s SHA512: ~2700 Mb/s As you can see, performance is poor in comparison to aesni(4) and even cryptosoft (due to high setup cost). At a larger buffer size (128kB), throughput is a little better (but still worse than aesni(4)): aesni: SHA1:~10400 Mb/s SHA256: ~9950 Mb/s ccp: ~2200 Mb/s SHA256: ~2600 Mb/s SHA512: ~3800 Mb/s cryptosoft: ~1750 Mb/s SHA256: ~1800 Mb/s SHA512: ~2700 Mb/s AES performance has a similar story: aesni: 4kB: ~11250 Mb/s 128kB: ~11250 Mb/s ccp: ~350 Mb/s 128kB: ~4600 Mb/s cryptosoft: ~1750 Mb/s 128kB: ~1700 Mb/s This driver is EXPERIMENTAL. You should verify cryptographic results on typical and corner case inputs from your application against a known- good implementation. Sponsored by: Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D12723
2018-01-18 22:01:30 +00:00
error = ccp_hmac(qp, s, crp);
break;
case BLKCIPHER:
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 (s->pending != 0) {
error = EAGAIN;
break;
Add ccp(4): experimental driver for AMD Crypto Co-Processor * Registers TRNG source for random(4) * Finds available queues, LSBs; allocates static objects * Allocates a shared MSI-X for all queues. The hardware does not have separate interrupts per queue. Working interrupt mode driver. * Computes SHA hashes, HMAC. Passes cryptotest.py, cryptocheck tests. * Does AES-CBC, CTR mode, and XTS. cryptotest.py and cryptocheck pass. * Support for "authenc" (AES + HMAC). (SHA1 seems to result in "unaligned" cleartext inputs from cryptocheck -- which the engine cannot handle. SHA2 seems to work fine.) * GCM passes for block-multiple AAD, input lengths Largely based on ccr(4), part of cxgbe(4). Rough performance averages on AMD Ryzen 1950X (4kB buffer): aesni: SHA1: ~8300 Mb/s SHA256: ~8000 Mb/s ccp: ~630 Mb/s SHA256: ~660 Mb/s SHA512: ~700 Mb/s cryptosoft: ~1800 Mb/s SHA256: ~1800 Mb/s SHA512: ~2700 Mb/s As you can see, performance is poor in comparison to aesni(4) and even cryptosoft (due to high setup cost). At a larger buffer size (128kB), throughput is a little better (but still worse than aesni(4)): aesni: SHA1:~10400 Mb/s SHA256: ~9950 Mb/s ccp: ~2200 Mb/s SHA256: ~2600 Mb/s SHA512: ~3800 Mb/s cryptosoft: ~1750 Mb/s SHA256: ~1800 Mb/s SHA512: ~2700 Mb/s AES performance has a similar story: aesni: 4kB: ~11250 Mb/s 128kB: ~11250 Mb/s ccp: ~350 Mb/s 128kB: ~4600 Mb/s cryptosoft: ~1750 Mb/s 128kB: ~1700 Mb/s This driver is EXPERIMENTAL. You should verify cryptographic results on typical and corner case inputs from your application against a known- good implementation. Sponsored by: Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D12723
2018-01-18 22:01:30 +00:00
}
error = ccp_blkcipher(qp, s, crp);
break;
case AUTHENC:
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 (s->pending != 0) {
error = EAGAIN;
Add ccp(4): experimental driver for AMD Crypto Co-Processor * Registers TRNG source for random(4) * Finds available queues, LSBs; allocates static objects * Allocates a shared MSI-X for all queues. The hardware does not have separate interrupts per queue. Working interrupt mode driver. * Computes SHA hashes, HMAC. Passes cryptotest.py, cryptocheck tests. * Does AES-CBC, CTR mode, and XTS. cryptotest.py and cryptocheck pass. * Support for "authenc" (AES + HMAC). (SHA1 seems to result in "unaligned" cleartext inputs from cryptocheck -- which the engine cannot handle. SHA2 seems to work fine.) * GCM passes for block-multiple AAD, input lengths Largely based on ccr(4), part of cxgbe(4). Rough performance averages on AMD Ryzen 1950X (4kB buffer): aesni: SHA1: ~8300 Mb/s SHA256: ~8000 Mb/s ccp: ~630 Mb/s SHA256: ~660 Mb/s SHA512: ~700 Mb/s cryptosoft: ~1800 Mb/s SHA256: ~1800 Mb/s SHA512: ~2700 Mb/s As you can see, performance is poor in comparison to aesni(4) and even cryptosoft (due to high setup cost). At a larger buffer size (128kB), throughput is a little better (but still worse than aesni(4)): aesni: SHA1:~10400 Mb/s SHA256: ~9950 Mb/s ccp: ~2200 Mb/s SHA256: ~2600 Mb/s SHA512: ~3800 Mb/s cryptosoft: ~1750 Mb/s SHA256: ~1800 Mb/s SHA512: ~2700 Mb/s AES performance has a similar story: aesni: 4kB: ~11250 Mb/s 128kB: ~11250 Mb/s ccp: ~350 Mb/s 128kB: ~4600 Mb/s cryptosoft: ~1750 Mb/s 128kB: ~1700 Mb/s This driver is EXPERIMENTAL. You should verify cryptographic results on typical and corner case inputs from your application against a known- good implementation. Sponsored by: Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D12723
2018-01-18 22:01:30 +00:00
break;
}
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 = ccp_authenc(qp, s, crp);
Add ccp(4): experimental driver for AMD Crypto Co-Processor * Registers TRNG source for random(4) * Finds available queues, LSBs; allocates static objects * Allocates a shared MSI-X for all queues. The hardware does not have separate interrupts per queue. Working interrupt mode driver. * Computes SHA hashes, HMAC. Passes cryptotest.py, cryptocheck tests. * Does AES-CBC, CTR mode, and XTS. cryptotest.py and cryptocheck pass. * Support for "authenc" (AES + HMAC). (SHA1 seems to result in "unaligned" cleartext inputs from cryptocheck -- which the engine cannot handle. SHA2 seems to work fine.) * GCM passes for block-multiple AAD, input lengths Largely based on ccr(4), part of cxgbe(4). Rough performance averages on AMD Ryzen 1950X (4kB buffer): aesni: SHA1: ~8300 Mb/s SHA256: ~8000 Mb/s ccp: ~630 Mb/s SHA256: ~660 Mb/s SHA512: ~700 Mb/s cryptosoft: ~1800 Mb/s SHA256: ~1800 Mb/s SHA512: ~2700 Mb/s As you can see, performance is poor in comparison to aesni(4) and even cryptosoft (due to high setup cost). At a larger buffer size (128kB), throughput is a little better (but still worse than aesni(4)): aesni: SHA1:~10400 Mb/s SHA256: ~9950 Mb/s ccp: ~2200 Mb/s SHA256: ~2600 Mb/s SHA512: ~3800 Mb/s cryptosoft: ~1750 Mb/s SHA256: ~1800 Mb/s SHA512: ~2700 Mb/s AES performance has a similar story: aesni: 4kB: ~11250 Mb/s 128kB: ~11250 Mb/s ccp: ~350 Mb/s 128kB: ~4600 Mb/s cryptosoft: ~1750 Mb/s 128kB: ~1700 Mb/s This driver is EXPERIMENTAL. You should verify cryptographic results on typical and corner case inputs from your application against a known- good implementation. Sponsored by: Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D12723
2018-01-18 22:01:30 +00:00
break;
case GCM:
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 (s->pending != 0) {
error = EAGAIN;
break;
}
error = ccp_gcm(qp, s, crp);
Add ccp(4): experimental driver for AMD Crypto Co-Processor * Registers TRNG source for random(4) * Finds available queues, LSBs; allocates static objects * Allocates a shared MSI-X for all queues. The hardware does not have separate interrupts per queue. Working interrupt mode driver. * Computes SHA hashes, HMAC. Passes cryptotest.py, cryptocheck tests. * Does AES-CBC, CTR mode, and XTS. cryptotest.py and cryptocheck pass. * Support for "authenc" (AES + HMAC). (SHA1 seems to result in "unaligned" cleartext inputs from cryptocheck -- which the engine cannot handle. SHA2 seems to work fine.) * GCM passes for block-multiple AAD, input lengths Largely based on ccr(4), part of cxgbe(4). Rough performance averages on AMD Ryzen 1950X (4kB buffer): aesni: SHA1: ~8300 Mb/s SHA256: ~8000 Mb/s ccp: ~630 Mb/s SHA256: ~660 Mb/s SHA512: ~700 Mb/s cryptosoft: ~1800 Mb/s SHA256: ~1800 Mb/s SHA512: ~2700 Mb/s As you can see, performance is poor in comparison to aesni(4) and even cryptosoft (due to high setup cost). At a larger buffer size (128kB), throughput is a little better (but still worse than aesni(4)): aesni: SHA1:~10400 Mb/s SHA256: ~9950 Mb/s ccp: ~2200 Mb/s SHA256: ~2600 Mb/s SHA512: ~3800 Mb/s cryptosoft: ~1750 Mb/s SHA256: ~1800 Mb/s SHA512: ~2700 Mb/s AES performance has a similar story: aesni: 4kB: ~11250 Mb/s 128kB: ~11250 Mb/s ccp: ~350 Mb/s 128kB: ~4600 Mb/s cryptosoft: ~1750 Mb/s 128kB: ~1700 Mb/s This driver is EXPERIMENTAL. You should verify cryptographic results on typical and corner case inputs from your application against a known- good implementation. Sponsored by: Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D12723
2018-01-18 22:01:30 +00:00
break;
}
if (error == 0)
s->pending++;
out:
if (qpheld) {
if (error != 0) {
/*
* Squash EAGAIN so callers don't uselessly and
* expensively retry if the ring was full.
*/
if (error == EAGAIN)
error = ENOMEM;
ccp_queue_abort(qp);
} else
ccp_queue_release(qp);
}
if (error != 0) {
DPRINTF(dev, "%s: early error:%d\n", __func__, error);
crp->crp_etype = error;
crypto_done(crp);
}
return (0);
}
static device_method_t ccp_methods[] = {
DEVMETHOD(device_probe, ccp_probe),
DEVMETHOD(device_attach, ccp_attach),
DEVMETHOD(device_detach, ccp_detach),
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
DEVMETHOD(cryptodev_probesession, ccp_probesession),
Add ccp(4): experimental driver for AMD Crypto Co-Processor * Registers TRNG source for random(4) * Finds available queues, LSBs; allocates static objects * Allocates a shared MSI-X for all queues. The hardware does not have separate interrupts per queue. Working interrupt mode driver. * Computes SHA hashes, HMAC. Passes cryptotest.py, cryptocheck tests. * Does AES-CBC, CTR mode, and XTS. cryptotest.py and cryptocheck pass. * Support for "authenc" (AES + HMAC). (SHA1 seems to result in "unaligned" cleartext inputs from cryptocheck -- which the engine cannot handle. SHA2 seems to work fine.) * GCM passes for block-multiple AAD, input lengths Largely based on ccr(4), part of cxgbe(4). Rough performance averages on AMD Ryzen 1950X (4kB buffer): aesni: SHA1: ~8300 Mb/s SHA256: ~8000 Mb/s ccp: ~630 Mb/s SHA256: ~660 Mb/s SHA512: ~700 Mb/s cryptosoft: ~1800 Mb/s SHA256: ~1800 Mb/s SHA512: ~2700 Mb/s As you can see, performance is poor in comparison to aesni(4) and even cryptosoft (due to high setup cost). At a larger buffer size (128kB), throughput is a little better (but still worse than aesni(4)): aesni: SHA1:~10400 Mb/s SHA256: ~9950 Mb/s ccp: ~2200 Mb/s SHA256: ~2600 Mb/s SHA512: ~3800 Mb/s cryptosoft: ~1750 Mb/s SHA256: ~1800 Mb/s SHA512: ~2700 Mb/s AES performance has a similar story: aesni: 4kB: ~11250 Mb/s 128kB: ~11250 Mb/s ccp: ~350 Mb/s 128kB: ~4600 Mb/s cryptosoft: ~1750 Mb/s 128kB: ~1700 Mb/s This driver is EXPERIMENTAL. You should verify cryptographic results on typical and corner case inputs from your application against a known- good implementation. Sponsored by: Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D12723
2018-01-18 22:01:30 +00:00
DEVMETHOD(cryptodev_newsession, ccp_newsession),
DEVMETHOD(cryptodev_freesession, ccp_freesession),
DEVMETHOD(cryptodev_process, ccp_process),
DEVMETHOD_END
};
static driver_t ccp_driver = {
"ccp",
ccp_methods,
sizeof(struct ccp_softc)
};
crypto: Remove unused devclass arguments to DRIVER_MODULE.
2022-05-06 22:46:58 +00:00
DRIVER_MODULE(ccp, pci, ccp_driver, NULL, NULL);
Add ccp(4): experimental driver for AMD Crypto Co-Processor * Registers TRNG source for random(4) * Finds available queues, LSBs; allocates static objects * Allocates a shared MSI-X for all queues. The hardware does not have separate interrupts per queue. Working interrupt mode driver. * Computes SHA hashes, HMAC. Passes cryptotest.py, cryptocheck tests. * Does AES-CBC, CTR mode, and XTS. cryptotest.py and cryptocheck pass. * Support for "authenc" (AES + HMAC). (SHA1 seems to result in "unaligned" cleartext inputs from cryptocheck -- which the engine cannot handle. SHA2 seems to work fine.) * GCM passes for block-multiple AAD, input lengths Largely based on ccr(4), part of cxgbe(4). Rough performance averages on AMD Ryzen 1950X (4kB buffer): aesni: SHA1: ~8300 Mb/s SHA256: ~8000 Mb/s ccp: ~630 Mb/s SHA256: ~660 Mb/s SHA512: ~700 Mb/s cryptosoft: ~1800 Mb/s SHA256: ~1800 Mb/s SHA512: ~2700 Mb/s As you can see, performance is poor in comparison to aesni(4) and even cryptosoft (due to high setup cost). At a larger buffer size (128kB), throughput is a little better (but still worse than aesni(4)): aesni: SHA1:~10400 Mb/s SHA256: ~9950 Mb/s ccp: ~2200 Mb/s SHA256: ~2600 Mb/s SHA512: ~3800 Mb/s cryptosoft: ~1750 Mb/s SHA256: ~1800 Mb/s SHA512: ~2700 Mb/s AES performance has a similar story: aesni: 4kB: ~11250 Mb/s 128kB: ~11250 Mb/s ccp: ~350 Mb/s 128kB: ~4600 Mb/s cryptosoft: ~1750 Mb/s 128kB: ~1700 Mb/s This driver is EXPERIMENTAL. You should verify cryptographic results on typical and corner case inputs from your application against a known- good implementation. Sponsored by: Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D12723
2018-01-18 22:01:30 +00:00
MODULE_VERSION(ccp, 1);
MODULE_DEPEND(ccp, crypto, 1, 1, 1);
MODULE_DEPEND(ccp, random_device, 1, 1, 1);
Don't load ccp automatically with devmatch Remove the PNP info for the moment from the driver. It's an experimental driver (as noted in r328150). It's performance is about 1/10th that of aesni. It will often panic when used with GELI (PR 2279820). It's not in our best interest to have such a driver be autoloaded by default. Approved by: re@ (rgrimes) Reviewed By: cem@ Differential Review: https://reviews.freebsd.org/D16959
2018-08-31 01:01:16 +00:00
#if 0 /* There are enough known issues that we shouldn't load automatically */
Reapply, with minor tweaks, r338025, from the original commit: Remove unused and easy to misuse PNP macro parameter Inspired by r338025, just remove the element size parameter to the MODULE_PNP_INFO macro entirely. The 'table' parameter is now required to have correct pointer (or array) type. Since all invocations of the macro already had this property and the emitted PNP data continues to include the element size, there is no functional change. Mostly done with the coccinelle 'spatch' tool: $ cat modpnpsize0.cocci @normaltables@ identifier b,c; expression a,d,e; declarer MODULE_PNP_INFO; @@ MODULE_PNP_INFO(a,b,c,d, -sizeof(d[0]), e); @singletons@ identifier b,c,d; expression a; declarer MODULE_PNP_INFO; @@ MODULE_PNP_INFO(a,b,c,&d, -sizeof(d), 1); $ rg -l MODULE_PNP_INFO -- sys | \ xargs spatch --in-place --sp-file modpnpsize0.cocci (Note that coccinelle invokes diff(1) via a PATH search and expects diff to tolerate the -B flag, which BSD diff does not. So I had to link gdiff into PATH as diff to use spatch.) Tinderbox'd (-DMAKE_JUST_KERNELS). Approved by: re (glen)
2018-09-26 17:12:14 +00:00
MODULE_PNP_INFO("W32:vendor/device", pci, ccp, ccp_ids,
The PNP info has to follow the module definition. Move it from just after the array to its proper location. Otherwise, the linker.hints file has things out of order and we associated it with whatever was the previous module.
2018-03-25 16:56:49 +00:00
nitems(ccp_ids));
Don't load ccp automatically with devmatch Remove the PNP info for the moment from the driver. It's an experimental driver (as noted in r328150). It's performance is about 1/10th that of aesni. It will often panic when used with GELI (PR 2279820). It's not in our best interest to have such a driver be autoloaded by default. Approved by: re@ (rgrimes) Reviewed By: cem@ Differential Review: https://reviews.freebsd.org/D16959
2018-08-31 01:01:16 +00:00
#endif
Add ccp(4): experimental driver for AMD Crypto Co-Processor * Registers TRNG source for random(4) * Finds available queues, LSBs; allocates static objects * Allocates a shared MSI-X for all queues. The hardware does not have separate interrupts per queue. Working interrupt mode driver. * Computes SHA hashes, HMAC. Passes cryptotest.py, cryptocheck tests. * Does AES-CBC, CTR mode, and XTS. cryptotest.py and cryptocheck pass. * Support for "authenc" (AES + HMAC). (SHA1 seems to result in "unaligned" cleartext inputs from cryptocheck -- which the engine cannot handle. SHA2 seems to work fine.) * GCM passes for block-multiple AAD, input lengths Largely based on ccr(4), part of cxgbe(4). Rough performance averages on AMD Ryzen 1950X (4kB buffer): aesni: SHA1: ~8300 Mb/s SHA256: ~8000 Mb/s ccp: ~630 Mb/s SHA256: ~660 Mb/s SHA512: ~700 Mb/s cryptosoft: ~1800 Mb/s SHA256: ~1800 Mb/s SHA512: ~2700 Mb/s As you can see, performance is poor in comparison to aesni(4) and even cryptosoft (due to high setup cost). At a larger buffer size (128kB), throughput is a little better (but still worse than aesni(4)): aesni: SHA1:~10400 Mb/s SHA256: ~9950 Mb/s ccp: ~2200 Mb/s SHA256: ~2600 Mb/s SHA512: ~3800 Mb/s cryptosoft: ~1750 Mb/s SHA256: ~1800 Mb/s SHA512: ~2700 Mb/s AES performance has a similar story: aesni: 4kB: ~11250 Mb/s 128kB: ~11250 Mb/s ccp: ~350 Mb/s 128kB: ~4600 Mb/s cryptosoft: ~1750 Mb/s 128kB: ~1700 Mb/s This driver is EXPERIMENTAL. You should verify cryptographic results on typical and corner case inputs from your application against a known- good implementation. Sponsored by: Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D12723
2018-01-18 22:01:30 +00:00
static int
ccp_queue_reserve_space(struct ccp_queue *qp, unsigned n, int mflags)
{
struct ccp_softc *sc;
mtx_assert(&qp->cq_lock, MA_OWNED);
sc = qp->cq_softc;
if (n < 1 || n >= (1 << sc->ring_size_order))
return (EINVAL);
while (true) {
if (ccp_queue_get_ring_space(qp) >= n)
return (0);
if ((mflags & M_WAITOK) == 0)
return (EAGAIN);
qp->cq_waiting = true;
msleep(&qp->cq_tail, &qp->cq_lock, 0, "ccpqfull", 0);
}
}
int
ccp_queue_acquire_reserve(struct ccp_queue *qp, unsigned n, int mflags)
{
int error;
mtx_lock(&qp->cq_lock);
qp->cq_acq_tail = qp->cq_tail;
error = ccp_queue_reserve_space(qp, n, mflags);
if (error != 0)
mtx_unlock(&qp->cq_lock);
return (error);
}
void
ccp_queue_release(struct ccp_queue *qp)
{
mtx_assert(&qp->cq_lock, MA_OWNED);
if (qp->cq_tail != qp->cq_acq_tail) {
wmb();
ccp_queue_write_tail(qp);
}
mtx_unlock(&qp->cq_lock);
}
void
ccp_queue_abort(struct ccp_queue *qp)
{
unsigned i;
mtx_assert(&qp->cq_lock, MA_OWNED);
/* Wipe out any descriptors associated with this aborted txn. */
for (i = qp->cq_acq_tail; i != qp->cq_tail;
i = (i + 1) % (1 << qp->cq_softc->ring_size_order)) {
memset(&qp->desc_ring[i], 0, sizeof(qp->desc_ring[i]));
}
qp->cq_tail = qp->cq_acq_tail;
mtx_unlock(&qp->cq_lock);
}
#ifdef DDB
#define _db_show_lock(lo) LOCK_CLASS(lo)->lc_ddb_show(lo)
#define db_show_lock(lk) _db_show_lock(&(lk)->lock_object)
static void
db_show_ccp_sc(struct ccp_softc *sc)
{
db_printf("ccp softc at %p\n", sc);
db_printf(" cid: %d\n", (int)sc->cid);
db_printf(" lock: ");
db_show_lock(&sc->lock);
db_printf(" detaching: %d\n", (int)sc->detaching);
db_printf(" ring_size_order: %u\n", sc->ring_size_order);
db_printf(" hw_version: %d\n", (int)sc->hw_version);
db_printf(" hw_features: %b\n", (int)sc->hw_features,
"\20\24ELFC\23TRNG\22Zip_Compress\16Zip_Decompress\13ECC\12RSA"
"\11SHA\0103DES\07AES");
db_printf(" hw status:\n");
db_ccp_show_hw(sc);
}
static void
db_show_ccp_qp(struct ccp_queue *qp)
{
db_printf(" lock: ");
db_show_lock(&qp->cq_lock);
db_printf(" cq_qindex: %u\n", qp->cq_qindex);
db_printf(" cq_softc: %p\n", qp->cq_softc);
db_printf(" head: %u\n", qp->cq_head);
db_printf(" tail: %u\n", qp->cq_tail);
db_printf(" acq_tail: %u\n", qp->cq_acq_tail);
db_printf(" desc_ring: %p\n", qp->desc_ring);
db_printf(" completions_ring: %p\n", qp->completions_ring);
db_printf(" descriptors (phys): 0x%jx\n",
(uintmax_t)qp->desc_ring_bus_addr);
db_printf(" hw status:\n");
db_ccp_show_queue_hw(qp);
}
DB_SHOW_COMMAND(ccp, db_show_ccp)
{
struct ccp_softc *sc;
unsigned unit, qindex;
if (!have_addr)
goto usage;
unit = (unsigned)addr;
ccp: Use devclass_find to lookup devclass in db_show_ccp. Reviewed by: imp Differential Revision: https://reviews.freebsd.org/D35003
2022-04-21 17:29:14 +00:00
sc = devclass_get_softc(devclass_find("ccp"), unit);
Add ccp(4): experimental driver for AMD Crypto Co-Processor * Registers TRNG source for random(4) * Finds available queues, LSBs; allocates static objects * Allocates a shared MSI-X for all queues. The hardware does not have separate interrupts per queue. Working interrupt mode driver. * Computes SHA hashes, HMAC. Passes cryptotest.py, cryptocheck tests. * Does AES-CBC, CTR mode, and XTS. cryptotest.py and cryptocheck pass. * Support for "authenc" (AES + HMAC). (SHA1 seems to result in "unaligned" cleartext inputs from cryptocheck -- which the engine cannot handle. SHA2 seems to work fine.) * GCM passes for block-multiple AAD, input lengths Largely based on ccr(4), part of cxgbe(4). Rough performance averages on AMD Ryzen 1950X (4kB buffer): aesni: SHA1: ~8300 Mb/s SHA256: ~8000 Mb/s ccp: ~630 Mb/s SHA256: ~660 Mb/s SHA512: ~700 Mb/s cryptosoft: ~1800 Mb/s SHA256: ~1800 Mb/s SHA512: ~2700 Mb/s As you can see, performance is poor in comparison to aesni(4) and even cryptosoft (due to high setup cost). At a larger buffer size (128kB), throughput is a little better (but still worse than aesni(4)): aesni: SHA1:~10400 Mb/s SHA256: ~9950 Mb/s ccp: ~2200 Mb/s SHA256: ~2600 Mb/s SHA512: ~3800 Mb/s cryptosoft: ~1750 Mb/s SHA256: ~1800 Mb/s SHA512: ~2700 Mb/s AES performance has a similar story: aesni: 4kB: ~11250 Mb/s 128kB: ~11250 Mb/s ccp: ~350 Mb/s 128kB: ~4600 Mb/s cryptosoft: ~1750 Mb/s 128kB: ~1700 Mb/s This driver is EXPERIMENTAL. You should verify cryptographic results on typical and corner case inputs from your application against a known- good implementation. Sponsored by: Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D12723
2018-01-18 22:01:30 +00:00
if (sc == NULL) {
db_printf("No such device ccp%u\n", unit);
goto usage;
}
if (count == -1) {
db_show_ccp_sc(sc);
return;
}
qindex = (unsigned)count;
if (qindex >= nitems(sc->queues)) {
db_printf("No such queue %u\n", qindex);
goto usage;
}
db_show_ccp_qp(&sc->queues[qindex]);
return;
usage:
db_printf("usage: show ccp <unit>[,<qindex>]\n");
return;
}
#endif /* DDB */
Reference in New Issue Copy Permalink