OpenE2K
/
linux
6
0
Fork 0
Code Issues Pull Requests Releases Wiki Activity
linux/arch/x86/crypto/sha1_ssse3_glue.c

383 lines
9.1 KiB
C
Raw Normal View History

crypto: sha1 - SSSE3 based SHA1 implementation for x86-64 This is an assembler implementation of the SHA1 algorithm using the Supplemental SSE3 (SSSE3) instructions or, when available, the Advanced Vector Extensions (AVX). Testing with the tcrypt module shows the raw hash performance is up to 2.3 times faster than the C implementation, using 8k data blocks on a Core 2 Duo T5500. For the smalest data set (16 byte) it is still 25% faster. Since this implementation uses SSE/YMM registers it cannot safely be used in every situation, e.g. while an IRQ interrupts a kernel thread. The implementation falls back to the generic SHA1 variant, if using the SSE/YMM registers is not possible. With this algorithm I was able to increase the throughput of a single IPsec link from 344 Mbit/s to 464 Mbit/s on a Core 2 Quad CPU using the SSSE3 variant -- a speedup of +34.8%. Saving and restoring SSE/YMM state might make the actual throughput fluctuate when there are FPU intensive userland applications running. For example, meassuring the performance using iperf2 directly on the machine under test gives wobbling numbers because iperf2 uses the FPU for each packet to check if the reporting interval has expired (in the above test I got min/max/avg: 402/484/464 MBit/s). Using this algorithm on a IPsec gateway gives much more reasonable and stable numbers, albeit not as high as in the directly connected case. Here is the result from an RFC 2544 test run with a EXFO Packet Blazer FTB-8510: frame size sha1-generic sha1-ssse3 delta 64 byte 37.5 MBit/s 37.5 MBit/s 0.0% 128 byte 56.3 MBit/s 62.5 MBit/s +11.0% 256 byte 87.5 MBit/s 100.0 MBit/s +14.3% 512 byte 131.3 MBit/s 150.0 MBit/s +14.2% 1024 byte 162.5 MBit/s 193.8 MBit/s +19.3% 1280 byte 175.0 MBit/s 212.5 MBit/s +21.4% 1420 byte 175.0 MBit/s 218.7 MBit/s +25.0% 1518 byte 150.0 MBit/s 181.2 MBit/s +20.8% The throughput for the largest frame size is lower than for the previous size because the IP packets need to be fragmented in this case to make there way through the IPsec tunnel. Signed-off-by: Mathias Krause <minipli@googlemail.com> Cc: Maxim Locktyukhin <maxim.locktyukhin@intel.com> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2011-08-04 20:19:25 +02:00
/*
* Cryptographic API.
*
* Glue code for the SHA1 Secure Hash Algorithm assembler implementation using
* Supplemental SSE3 instructions.
*
* This file is based on sha1_generic.c
*
* Copyright (c) Alan Smithee.
* Copyright (c) Andrew McDonald <andrew@mcdonald.org.uk>
* Copyright (c) Jean-Francois Dive <jef@linuxbe.org>
* Copyright (c) Mathias Krause <minipli@googlemail.com>
crypto: sha - SHA1 transform x86_64 AVX2 This git patch adds x86_64 AVX2 optimization of SHA1 transform to crypto support. The patch has been tested with 3.14.0-rc1 kernel. On a Haswell desktop, with turbo disabled and all cpus running at maximum frequency, tcrypt shows AVX2 performance improvement from 3% for 256 bytes update to 16% for 1024 bytes update over AVX implementation. This patch adds sha1_avx2_transform(), the glue, build and configuration changes needed for AVX2 optimization of SHA1 transform to crypto support. sha1-ssse3 is one module which adds the necessary optimization support (SSSE3/AVX/AVX2) for the low-level SHA1 transform function. With better optimization support, transform function is overridden as the case may be. In the case of AVX2, due to performance reasons across datablock sizes, the AVX or AVX2 transform function is used at run-time as it suits best. The Makefile change therefore appends the necessary objects to the linkage. Due to this, the patch merely appends AVX2 transform to the existing build mix and Kconfig support and leaves the configuration build support as is. Signed-off-by: Chandramouli Narayanan <mouli@linux.intel.com> Reviewed-by: Marek Vasut <marex@denx.de> Acked-by: H. Peter Anvin <hpa@linux.intel.com> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2014-03-20 23:14:00 +01:00
* Copyright (c) Chandramouli Narayanan <mouli@linux.intel.com>
crypto: sha1 - SSSE3 based SHA1 implementation for x86-64 This is an assembler implementation of the SHA1 algorithm using the Supplemental SSE3 (SSSE3) instructions or, when available, the Advanced Vector Extensions (AVX). Testing with the tcrypt module shows the raw hash performance is up to 2.3 times faster than the C implementation, using 8k data blocks on a Core 2 Duo T5500. For the smalest data set (16 byte) it is still 25% faster. Since this implementation uses SSE/YMM registers it cannot safely be used in every situation, e.g. while an IRQ interrupts a kernel thread. The implementation falls back to the generic SHA1 variant, if using the SSE/YMM registers is not possible. With this algorithm I was able to increase the throughput of a single IPsec link from 344 Mbit/s to 464 Mbit/s on a Core 2 Quad CPU using the SSSE3 variant -- a speedup of +34.8%. Saving and restoring SSE/YMM state might make the actual throughput fluctuate when there are FPU intensive userland applications running. For example, meassuring the performance using iperf2 directly on the machine under test gives wobbling numbers because iperf2 uses the FPU for each packet to check if the reporting interval has expired (in the above test I got min/max/avg: 402/484/464 MBit/s). Using this algorithm on a IPsec gateway gives much more reasonable and stable numbers, albeit not as high as in the directly connected case. Here is the result from an RFC 2544 test run with a EXFO Packet Blazer FTB-8510: frame size sha1-generic sha1-ssse3 delta 64 byte 37.5 MBit/s 37.5 MBit/s 0.0% 128 byte 56.3 MBit/s 62.5 MBit/s +11.0% 256 byte 87.5 MBit/s 100.0 MBit/s +14.3% 512 byte 131.3 MBit/s 150.0 MBit/s +14.2% 1024 byte 162.5 MBit/s 193.8 MBit/s +19.3% 1280 byte 175.0 MBit/s 212.5 MBit/s +21.4% 1420 byte 175.0 MBit/s 218.7 MBit/s +25.0% 1518 byte 150.0 MBit/s 181.2 MBit/s +20.8% The throughput for the largest frame size is lower than for the previous size because the IP packets need to be fragmented in this case to make there way through the IPsec tunnel. Signed-off-by: Mathias Krause <minipli@googlemail.com> Cc: Maxim Locktyukhin <maxim.locktyukhin@intel.com> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2011-08-04 20:19:25 +02:00
*
* This program is free software; you can redistribute it and/or modify it
* under the terms of the GNU General Public License as published by the Free
* Software Foundation; either version 2 of the License, or (at your option)
* any later version.
*
*/
#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
#include <crypto/internal/hash.h>
#include <linux/init.h>
#include <linux/module.h>
#include <linux/mm.h>
#include <linux/cryptohash.h>
#include <linux/types.h>
#include <crypto/sha.h>
crypto: x86/sha1_ssse3 - move SHA-1 SSSE3 implementation to base layer This removes all the boilerplate from the existing implementation, and replaces it with calls into the base layer. Signed-off-by: Ard Biesheuvel <ard.biesheuvel@linaro.org> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2015-04-09 12:55:46 +02:00
#include <crypto/sha1_base.h>
x86/fpu: Rename i387.h to fpu/api.h We already have fpu/types.h, move i387.h to fpu/api.h. The file name has become a misnomer anyway: it offers generic FPU APIs, but is not limited to i387 functionality. Reviewed-by: Borislav Petkov <bp@alien8.de> Cc: Andy Lutomirski <luto@amacapital.net> Cc: Dave Hansen <dave.hansen@linux.intel.com> Cc: Fenghua Yu <fenghua.yu@intel.com> Cc: H. Peter Anvin <hpa@zytor.com> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Oleg Nesterov <oleg@redhat.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Thomas Gleixner <tglx@linutronix.de> Signed-off-by: Ingo Molnar <mingo@kernel.org>
2015-04-24 02:46:00 +02:00
#include <asm/fpu/api.h>
crypto: sha1 - SSSE3 based SHA1 implementation for x86-64 This is an assembler implementation of the SHA1 algorithm using the Supplemental SSE3 (SSSE3) instructions or, when available, the Advanced Vector Extensions (AVX). Testing with the tcrypt module shows the raw hash performance is up to 2.3 times faster than the C implementation, using 8k data blocks on a Core 2 Duo T5500. For the smalest data set (16 byte) it is still 25% faster. Since this implementation uses SSE/YMM registers it cannot safely be used in every situation, e.g. while an IRQ interrupts a kernel thread. The implementation falls back to the generic SHA1 variant, if using the SSE/YMM registers is not possible. With this algorithm I was able to increase the throughput of a single IPsec link from 344 Mbit/s to 464 Mbit/s on a Core 2 Quad CPU using the SSSE3 variant -- a speedup of +34.8%. Saving and restoring SSE/YMM state might make the actual throughput fluctuate when there are FPU intensive userland applications running. For example, meassuring the performance using iperf2 directly on the machine under test gives wobbling numbers because iperf2 uses the FPU for each packet to check if the reporting interval has expired (in the above test I got min/max/avg: 402/484/464 MBit/s). Using this algorithm on a IPsec gateway gives much more reasonable and stable numbers, albeit not as high as in the directly connected case. Here is the result from an RFC 2544 test run with a EXFO Packet Blazer FTB-8510: frame size sha1-generic sha1-ssse3 delta 64 byte 37.5 MBit/s 37.5 MBit/s 0.0% 128 byte 56.3 MBit/s 62.5 MBit/s +11.0% 256 byte 87.5 MBit/s 100.0 MBit/s +14.3% 512 byte 131.3 MBit/s 150.0 MBit/s +14.2% 1024 byte 162.5 MBit/s 193.8 MBit/s +19.3% 1280 byte 175.0 MBit/s 212.5 MBit/s +21.4% 1420 byte 175.0 MBit/s 218.7 MBit/s +25.0% 1518 byte 150.0 MBit/s 181.2 MBit/s +20.8% The throughput for the largest frame size is lower than for the previous size because the IP packets need to be fragmented in this case to make there way through the IPsec tunnel. Signed-off-by: Mathias Krause <minipli@googlemail.com> Cc: Maxim Locktyukhin <maxim.locktyukhin@intel.com> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2011-08-04 20:19:25 +02:00
crypto: x86/sha - Restructure x86 sha1 glue code to expose all the available sha1 transforms Restructure the x86 sha1 glue code so we will expose sha1 transforms based on SSSE3, AVX, AVX2 or SHA-NI extension as separate individual drivers when cpu provides such support. This will make it easy for alternative algorithms to be used if desired and makes the code cleaner and easier to maintain. Signed-off-by: Tim Chen <tim.c.chen@linux.intel.com> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2015-09-17 01:34:53 +02:00
typedef void (sha1_transform_fn)(u32 *digest, const char *data,
unsigned int rounds);
crypto: sha1 - SSSE3 based SHA1 implementation for x86-64 This is an assembler implementation of the SHA1 algorithm using the Supplemental SSE3 (SSSE3) instructions or, when available, the Advanced Vector Extensions (AVX). Testing with the tcrypt module shows the raw hash performance is up to 2.3 times faster than the C implementation, using 8k data blocks on a Core 2 Duo T5500. For the smalest data set (16 byte) it is still 25% faster. Since this implementation uses SSE/YMM registers it cannot safely be used in every situation, e.g. while an IRQ interrupts a kernel thread. The implementation falls back to the generic SHA1 variant, if using the SSE/YMM registers is not possible. With this algorithm I was able to increase the throughput of a single IPsec link from 344 Mbit/s to 464 Mbit/s on a Core 2 Quad CPU using the SSSE3 variant -- a speedup of +34.8%. Saving and restoring SSE/YMM state might make the actual throughput fluctuate when there are FPU intensive userland applications running. For example, meassuring the performance using iperf2 directly on the machine under test gives wobbling numbers because iperf2 uses the FPU for each packet to check if the reporting interval has expired (in the above test I got min/max/avg: 402/484/464 MBit/s). Using this algorithm on a IPsec gateway gives much more reasonable and stable numbers, albeit not as high as in the directly connected case. Here is the result from an RFC 2544 test run with a EXFO Packet Blazer FTB-8510: frame size sha1-generic sha1-ssse3 delta 64 byte 37.5 MBit/s 37.5 MBit/s 0.0% 128 byte 56.3 MBit/s 62.5 MBit/s +11.0% 256 byte 87.5 MBit/s 100.0 MBit/s +14.3% 512 byte 131.3 MBit/s 150.0 MBit/s +14.2% 1024 byte 162.5 MBit/s 193.8 MBit/s +19.3% 1280 byte 175.0 MBit/s 212.5 MBit/s +21.4% 1420 byte 175.0 MBit/s 218.7 MBit/s +25.0% 1518 byte 150.0 MBit/s 181.2 MBit/s +20.8% The throughput for the largest frame size is lower than for the previous size because the IP packets need to be fragmented in this case to make there way through the IPsec tunnel. Signed-off-by: Mathias Krause <minipli@googlemail.com> Cc: Maxim Locktyukhin <maxim.locktyukhin@intel.com> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2011-08-04 20:19:25 +02:00
crypto: x86/sha - Restructure x86 sha1 glue code to expose all the available sha1 transforms Restructure the x86 sha1 glue code so we will expose sha1 transforms based on SSSE3, AVX, AVX2 or SHA-NI extension as separate individual drivers when cpu provides such support. This will make it easy for alternative algorithms to be used if desired and makes the code cleaner and easier to maintain. Signed-off-by: Tim Chen <tim.c.chen@linux.intel.com> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2015-09-17 01:34:53 +02:00
static int sha1_update(struct shash_desc *desc, const u8 *data,
unsigned int len, sha1_transform_fn *sha1_xform)
crypto: sha1 - SSSE3 based SHA1 implementation for x86-64 This is an assembler implementation of the SHA1 algorithm using the Supplemental SSE3 (SSSE3) instructions or, when available, the Advanced Vector Extensions (AVX). Testing with the tcrypt module shows the raw hash performance is up to 2.3 times faster than the C implementation, using 8k data blocks on a Core 2 Duo T5500. For the smalest data set (16 byte) it is still 25% faster. Since this implementation uses SSE/YMM registers it cannot safely be used in every situation, e.g. while an IRQ interrupts a kernel thread. The implementation falls back to the generic SHA1 variant, if using the SSE/YMM registers is not possible. With this algorithm I was able to increase the throughput of a single IPsec link from 344 Mbit/s to 464 Mbit/s on a Core 2 Quad CPU using the SSSE3 variant -- a speedup of +34.8%. Saving and restoring SSE/YMM state might make the actual throughput fluctuate when there are FPU intensive userland applications running. For example, meassuring the performance using iperf2 directly on the machine under test gives wobbling numbers because iperf2 uses the FPU for each packet to check if the reporting interval has expired (in the above test I got min/max/avg: 402/484/464 MBit/s). Using this algorithm on a IPsec gateway gives much more reasonable and stable numbers, albeit not as high as in the directly connected case. Here is the result from an RFC 2544 test run with a EXFO Packet Blazer FTB-8510: frame size sha1-generic sha1-ssse3 delta 64 byte 37.5 MBit/s 37.5 MBit/s 0.0% 128 byte 56.3 MBit/s 62.5 MBit/s +11.0% 256 byte 87.5 MBit/s 100.0 MBit/s +14.3% 512 byte 131.3 MBit/s 150.0 MBit/s +14.2% 1024 byte 162.5 MBit/s 193.8 MBit/s +19.3% 1280 byte 175.0 MBit/s 212.5 MBit/s +21.4% 1420 byte 175.0 MBit/s 218.7 MBit/s +25.0% 1518 byte 150.0 MBit/s 181.2 MBit/s +20.8% The throughput for the largest frame size is lower than for the previous size because the IP packets need to be fragmented in this case to make there way through the IPsec tunnel. Signed-off-by: Mathias Krause <minipli@googlemail.com> Cc: Maxim Locktyukhin <maxim.locktyukhin@intel.com> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2011-08-04 20:19:25 +02:00
{
struct sha1_state *sctx = shash_desc_ctx(desc);
crypto: x86/sha1_ssse3 - move SHA-1 SSSE3 implementation to base layer This removes all the boilerplate from the existing implementation, and replaces it with calls into the base layer. Signed-off-by: Ard Biesheuvel <ard.biesheuvel@linaro.org> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2015-04-09 12:55:46 +02:00
if (!irq_fpu_usable() ||
(sctx->count % SHA1_BLOCK_SIZE) + len < SHA1_BLOCK_SIZE)
return crypto_sha1_update(desc, data, len);
crypto: sha1 - SSSE3 based SHA1 implementation for x86-64 This is an assembler implementation of the SHA1 algorithm using the Supplemental SSE3 (SSSE3) instructions or, when available, the Advanced Vector Extensions (AVX). Testing with the tcrypt module shows the raw hash performance is up to 2.3 times faster than the C implementation, using 8k data blocks on a Core 2 Duo T5500. For the smalest data set (16 byte) it is still 25% faster. Since this implementation uses SSE/YMM registers it cannot safely be used in every situation, e.g. while an IRQ interrupts a kernel thread. The implementation falls back to the generic SHA1 variant, if using the SSE/YMM registers is not possible. With this algorithm I was able to increase the throughput of a single IPsec link from 344 Mbit/s to 464 Mbit/s on a Core 2 Quad CPU using the SSSE3 variant -- a speedup of +34.8%. Saving and restoring SSE/YMM state might make the actual throughput fluctuate when there are FPU intensive userland applications running. For example, meassuring the performance using iperf2 directly on the machine under test gives wobbling numbers because iperf2 uses the FPU for each packet to check if the reporting interval has expired (in the above test I got min/max/avg: 402/484/464 MBit/s). Using this algorithm on a IPsec gateway gives much more reasonable and stable numbers, albeit not as high as in the directly connected case. Here is the result from an RFC 2544 test run with a EXFO Packet Blazer FTB-8510: frame size sha1-generic sha1-ssse3 delta 64 byte 37.5 MBit/s 37.5 MBit/s 0.0% 128 byte 56.3 MBit/s 62.5 MBit/s +11.0% 256 byte 87.5 MBit/s 100.0 MBit/s +14.3% 512 byte 131.3 MBit/s 150.0 MBit/s +14.2% 1024 byte 162.5 MBit/s 193.8 MBit/s +19.3% 1280 byte 175.0 MBit/s 212.5 MBit/s +21.4% 1420 byte 175.0 MBit/s 218.7 MBit/s +25.0% 1518 byte 150.0 MBit/s 181.2 MBit/s +20.8% The throughput for the largest frame size is lower than for the previous size because the IP packets need to be fragmented in this case to make there way through the IPsec tunnel. Signed-off-by: Mathias Krause <minipli@googlemail.com> Cc: Maxim Locktyukhin <maxim.locktyukhin@intel.com> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2011-08-04 20:19:25 +02:00
crypto: x86/sha1_ssse3 - move SHA-1 SSSE3 implementation to base layer This removes all the boilerplate from the existing implementation, and replaces it with calls into the base layer. Signed-off-by: Ard Biesheuvel <ard.biesheuvel@linaro.org> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2015-04-09 12:55:46 +02:00
/* make sure casting to sha1_block_fn() is safe */
BUILD_BUG_ON(offsetof(struct sha1_state, state) != 0);
crypto: sha1 - SSSE3 based SHA1 implementation for x86-64 This is an assembler implementation of the SHA1 algorithm using the Supplemental SSE3 (SSSE3) instructions or, when available, the Advanced Vector Extensions (AVX). Testing with the tcrypt module shows the raw hash performance is up to 2.3 times faster than the C implementation, using 8k data blocks on a Core 2 Duo T5500. For the smalest data set (16 byte) it is still 25% faster. Since this implementation uses SSE/YMM registers it cannot safely be used in every situation, e.g. while an IRQ interrupts a kernel thread. The implementation falls back to the generic SHA1 variant, if using the SSE/YMM registers is not possible. With this algorithm I was able to increase the throughput of a single IPsec link from 344 Mbit/s to 464 Mbit/s on a Core 2 Quad CPU using the SSSE3 variant -- a speedup of +34.8%. Saving and restoring SSE/YMM state might make the actual throughput fluctuate when there are FPU intensive userland applications running. For example, meassuring the performance using iperf2 directly on the machine under test gives wobbling numbers because iperf2 uses the FPU for each packet to check if the reporting interval has expired (in the above test I got min/max/avg: 402/484/464 MBit/s). Using this algorithm on a IPsec gateway gives much more reasonable and stable numbers, albeit not as high as in the directly connected case. Here is the result from an RFC 2544 test run with a EXFO Packet Blazer FTB-8510: frame size sha1-generic sha1-ssse3 delta 64 byte 37.5 MBit/s 37.5 MBit/s 0.0% 128 byte 56.3 MBit/s 62.5 MBit/s +11.0% 256 byte 87.5 MBit/s 100.0 MBit/s +14.3% 512 byte 131.3 MBit/s 150.0 MBit/s +14.2% 1024 byte 162.5 MBit/s 193.8 MBit/s +19.3% 1280 byte 175.0 MBit/s 212.5 MBit/s +21.4% 1420 byte 175.0 MBit/s 218.7 MBit/s +25.0% 1518 byte 150.0 MBit/s 181.2 MBit/s +20.8% The throughput for the largest frame size is lower than for the previous size because the IP packets need to be fragmented in this case to make there way through the IPsec tunnel. Signed-off-by: Mathias Krause <minipli@googlemail.com> Cc: Maxim Locktyukhin <maxim.locktyukhin@intel.com> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2011-08-04 20:19:25 +02:00
crypto: x86/sha1_ssse3 - move SHA-1 SSSE3 implementation to base layer This removes all the boilerplate from the existing implementation, and replaces it with calls into the base layer. Signed-off-by: Ard Biesheuvel <ard.biesheuvel@linaro.org> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2015-04-09 12:55:46 +02:00
kernel_fpu_begin();
sha1_base_do_update(desc, data, len,
crypto: x86/sha - Restructure x86 sha1 glue code to expose all the available sha1 transforms Restructure the x86 sha1 glue code so we will expose sha1 transforms based on SSSE3, AVX, AVX2 or SHA-NI extension as separate individual drivers when cpu provides such support. This will make it easy for alternative algorithms to be used if desired and makes the code cleaner and easier to maintain. Signed-off-by: Tim Chen <tim.c.chen@linux.intel.com> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2015-09-17 01:34:53 +02:00
(sha1_block_fn *)sha1_xform);
crypto: x86/sha1_ssse3 - move SHA-1 SSSE3 implementation to base layer This removes all the boilerplate from the existing implementation, and replaces it with calls into the base layer. Signed-off-by: Ard Biesheuvel <ard.biesheuvel@linaro.org> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2015-04-09 12:55:46 +02:00
kernel_fpu_end();
crypto: sha1 - SSSE3 based SHA1 implementation for x86-64 This is an assembler implementation of the SHA1 algorithm using the Supplemental SSE3 (SSSE3) instructions or, when available, the Advanced Vector Extensions (AVX). Testing with the tcrypt module shows the raw hash performance is up to 2.3 times faster than the C implementation, using 8k data blocks on a Core 2 Duo T5500. For the smalest data set (16 byte) it is still 25% faster. Since this implementation uses SSE/YMM registers it cannot safely be used in every situation, e.g. while an IRQ interrupts a kernel thread. The implementation falls back to the generic SHA1 variant, if using the SSE/YMM registers is not possible. With this algorithm I was able to increase the throughput of a single IPsec link from 344 Mbit/s to 464 Mbit/s on a Core 2 Quad CPU using the SSSE3 variant -- a speedup of +34.8%. Saving and restoring SSE/YMM state might make the actual throughput fluctuate when there are FPU intensive userland applications running. For example, meassuring the performance using iperf2 directly on the machine under test gives wobbling numbers because iperf2 uses the FPU for each packet to check if the reporting interval has expired (in the above test I got min/max/avg: 402/484/464 MBit/s). Using this algorithm on a IPsec gateway gives much more reasonable and stable numbers, albeit not as high as in the directly connected case. Here is the result from an RFC 2544 test run with a EXFO Packet Blazer FTB-8510: frame size sha1-generic sha1-ssse3 delta 64 byte 37.5 MBit/s 37.5 MBit/s 0.0% 128 byte 56.3 MBit/s 62.5 MBit/s +11.0% 256 byte 87.5 MBit/s 100.0 MBit/s +14.3% 512 byte 131.3 MBit/s 150.0 MBit/s +14.2% 1024 byte 162.5 MBit/s 193.8 MBit/s +19.3% 1280 byte 175.0 MBit/s 212.5 MBit/s +21.4% 1420 byte 175.0 MBit/s 218.7 MBit/s +25.0% 1518 byte 150.0 MBit/s 181.2 MBit/s +20.8% The throughput for the largest frame size is lower than for the previous size because the IP packets need to be fragmented in this case to make there way through the IPsec tunnel. Signed-off-by: Mathias Krause <minipli@googlemail.com> Cc: Maxim Locktyukhin <maxim.locktyukhin@intel.com> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2011-08-04 20:19:25 +02:00
return 0;
}
crypto: x86/sha - Restructure x86 sha1 glue code to expose all the available sha1 transforms Restructure the x86 sha1 glue code so we will expose sha1 transforms based on SSSE3, AVX, AVX2 or SHA-NI extension as separate individual drivers when cpu provides such support. This will make it easy for alternative algorithms to be used if desired and makes the code cleaner and easier to maintain. Signed-off-by: Tim Chen <tim.c.chen@linux.intel.com> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2015-09-17 01:34:53 +02:00
static int sha1_finup(struct shash_desc *desc, const u8 *data,
unsigned int len, u8 *out, sha1_transform_fn *sha1_xform)
crypto: sha1 - SSSE3 based SHA1 implementation for x86-64 This is an assembler implementation of the SHA1 algorithm using the Supplemental SSE3 (SSSE3) instructions or, when available, the Advanced Vector Extensions (AVX). Testing with the tcrypt module shows the raw hash performance is up to 2.3 times faster than the C implementation, using 8k data blocks on a Core 2 Duo T5500. For the smalest data set (16 byte) it is still 25% faster. Since this implementation uses SSE/YMM registers it cannot safely be used in every situation, e.g. while an IRQ interrupts a kernel thread. The implementation falls back to the generic SHA1 variant, if using the SSE/YMM registers is not possible. With this algorithm I was able to increase the throughput of a single IPsec link from 344 Mbit/s to 464 Mbit/s on a Core 2 Quad CPU using the SSSE3 variant -- a speedup of +34.8%. Saving and restoring SSE/YMM state might make the actual throughput fluctuate when there are FPU intensive userland applications running. For example, meassuring the performance using iperf2 directly on the machine under test gives wobbling numbers because iperf2 uses the FPU for each packet to check if the reporting interval has expired (in the above test I got min/max/avg: 402/484/464 MBit/s). Using this algorithm on a IPsec gateway gives much more reasonable and stable numbers, albeit not as high as in the directly connected case. Here is the result from an RFC 2544 test run with a EXFO Packet Blazer FTB-8510: frame size sha1-generic sha1-ssse3 delta 64 byte 37.5 MBit/s 37.5 MBit/s 0.0% 128 byte 56.3 MBit/s 62.5 MBit/s +11.0% 256 byte 87.5 MBit/s 100.0 MBit/s +14.3% 512 byte 131.3 MBit/s 150.0 MBit/s +14.2% 1024 byte 162.5 MBit/s 193.8 MBit/s +19.3% 1280 byte 175.0 MBit/s 212.5 MBit/s +21.4% 1420 byte 175.0 MBit/s 218.7 MBit/s +25.0% 1518 byte 150.0 MBit/s 181.2 MBit/s +20.8% The throughput for the largest frame size is lower than for the previous size because the IP packets need to be fragmented in this case to make there way through the IPsec tunnel. Signed-off-by: Mathias Krause <minipli@googlemail.com> Cc: Maxim Locktyukhin <maxim.locktyukhin@intel.com> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2011-08-04 20:19:25 +02:00
{
crypto: x86/sha1_ssse3 - move SHA-1 SSSE3 implementation to base layer This removes all the boilerplate from the existing implementation, and replaces it with calls into the base layer. Signed-off-by: Ard Biesheuvel <ard.biesheuvel@linaro.org> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2015-04-09 12:55:46 +02:00
if (!irq_fpu_usable())
return crypto_sha1_finup(desc, data, len, out);
crypto: sha1 - SSSE3 based SHA1 implementation for x86-64 This is an assembler implementation of the SHA1 algorithm using the Supplemental SSE3 (SSSE3) instructions or, when available, the Advanced Vector Extensions (AVX). Testing with the tcrypt module shows the raw hash performance is up to 2.3 times faster than the C implementation, using 8k data blocks on a Core 2 Duo T5500. For the smalest data set (16 byte) it is still 25% faster. Since this implementation uses SSE/YMM registers it cannot safely be used in every situation, e.g. while an IRQ interrupts a kernel thread. The implementation falls back to the generic SHA1 variant, if using the SSE/YMM registers is not possible. With this algorithm I was able to increase the throughput of a single IPsec link from 344 Mbit/s to 464 Mbit/s on a Core 2 Quad CPU using the SSSE3 variant -- a speedup of +34.8%. Saving and restoring SSE/YMM state might make the actual throughput fluctuate when there are FPU intensive userland applications running. For example, meassuring the performance using iperf2 directly on the machine under test gives wobbling numbers because iperf2 uses the FPU for each packet to check if the reporting interval has expired (in the above test I got min/max/avg: 402/484/464 MBit/s). Using this algorithm on a IPsec gateway gives much more reasonable and stable numbers, albeit not as high as in the directly connected case. Here is the result from an RFC 2544 test run with a EXFO Packet Blazer FTB-8510: frame size sha1-generic sha1-ssse3 delta 64 byte 37.5 MBit/s 37.5 MBit/s 0.0% 128 byte 56.3 MBit/s 62.5 MBit/s +11.0% 256 byte 87.5 MBit/s 100.0 MBit/s +14.3% 512 byte 131.3 MBit/s 150.0 MBit/s +14.2% 1024 byte 162.5 MBit/s 193.8 MBit/s +19.3% 1280 byte 175.0 MBit/s 212.5 MBit/s +21.4% 1420 byte 175.0 MBit/s 218.7 MBit/s +25.0% 1518 byte 150.0 MBit/s 181.2 MBit/s +20.8% The throughput for the largest frame size is lower than for the previous size because the IP packets need to be fragmented in this case to make there way through the IPsec tunnel. Signed-off-by: Mathias Krause <minipli@googlemail.com> Cc: Maxim Locktyukhin <maxim.locktyukhin@intel.com> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2011-08-04 20:19:25 +02:00
crypto: x86/sha1_ssse3 - move SHA-1 SSSE3 implementation to base layer This removes all the boilerplate from the existing implementation, and replaces it with calls into the base layer. Signed-off-by: Ard Biesheuvel <ard.biesheuvel@linaro.org> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2015-04-09 12:55:46 +02:00
kernel_fpu_begin();
if (len)
sha1_base_do_update(desc, data, len,
crypto: x86/sha - Restructure x86 sha1 glue code to expose all the available sha1 transforms Restructure the x86 sha1 glue code so we will expose sha1 transforms based on SSSE3, AVX, AVX2 or SHA-NI extension as separate individual drivers when cpu provides such support. This will make it easy for alternative algorithms to be used if desired and makes the code cleaner and easier to maintain. Signed-off-by: Tim Chen <tim.c.chen@linux.intel.com> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2015-09-17 01:34:53 +02:00
(sha1_block_fn *)sha1_xform);
sha1_base_do_finalize(desc, (sha1_block_fn *)sha1_xform);
crypto: x86/sha1_ssse3 - move SHA-1 SSSE3 implementation to base layer This removes all the boilerplate from the existing implementation, and replaces it with calls into the base layer. Signed-off-by: Ard Biesheuvel <ard.biesheuvel@linaro.org> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2015-04-09 12:55:46 +02:00
kernel_fpu_end();
crypto: sha1 - SSSE3 based SHA1 implementation for x86-64 This is an assembler implementation of the SHA1 algorithm using the Supplemental SSE3 (SSSE3) instructions or, when available, the Advanced Vector Extensions (AVX). Testing with the tcrypt module shows the raw hash performance is up to 2.3 times faster than the C implementation, using 8k data blocks on a Core 2 Duo T5500. For the smalest data set (16 byte) it is still 25% faster. Since this implementation uses SSE/YMM registers it cannot safely be used in every situation, e.g. while an IRQ interrupts a kernel thread. The implementation falls back to the generic SHA1 variant, if using the SSE/YMM registers is not possible. With this algorithm I was able to increase the throughput of a single IPsec link from 344 Mbit/s to 464 Mbit/s on a Core 2 Quad CPU using the SSSE3 variant -- a speedup of +34.8%. Saving and restoring SSE/YMM state might make the actual throughput fluctuate when there are FPU intensive userland applications running. For example, meassuring the performance using iperf2 directly on the machine under test gives wobbling numbers because iperf2 uses the FPU for each packet to check if the reporting interval has expired (in the above test I got min/max/avg: 402/484/464 MBit/s). Using this algorithm on a IPsec gateway gives much more reasonable and stable numbers, albeit not as high as in the directly connected case. Here is the result from an RFC 2544 test run with a EXFO Packet Blazer FTB-8510: frame size sha1-generic sha1-ssse3 delta 64 byte 37.5 MBit/s 37.5 MBit/s 0.0% 128 byte 56.3 MBit/s 62.5 MBit/s +11.0% 256 byte 87.5 MBit/s 100.0 MBit/s +14.3% 512 byte 131.3 MBit/s 150.0 MBit/s +14.2% 1024 byte 162.5 MBit/s 193.8 MBit/s +19.3% 1280 byte 175.0 MBit/s 212.5 MBit/s +21.4% 1420 byte 175.0 MBit/s 218.7 MBit/s +25.0% 1518 byte 150.0 MBit/s 181.2 MBit/s +20.8% The throughput for the largest frame size is lower than for the previous size because the IP packets need to be fragmented in this case to make there way through the IPsec tunnel. Signed-off-by: Mathias Krause <minipli@googlemail.com> Cc: Maxim Locktyukhin <maxim.locktyukhin@intel.com> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2011-08-04 20:19:25 +02:00
crypto: x86/sha1_ssse3 - move SHA-1 SSSE3 implementation to base layer This removes all the boilerplate from the existing implementation, and replaces it with calls into the base layer. Signed-off-by: Ard Biesheuvel <ard.biesheuvel@linaro.org> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2015-04-09 12:55:46 +02:00
return sha1_base_finish(desc, out);
crypto: sha1 - SSSE3 based SHA1 implementation for x86-64 This is an assembler implementation of the SHA1 algorithm using the Supplemental SSE3 (SSSE3) instructions or, when available, the Advanced Vector Extensions (AVX). Testing with the tcrypt module shows the raw hash performance is up to 2.3 times faster than the C implementation, using 8k data blocks on a Core 2 Duo T5500. For the smalest data set (16 byte) it is still 25% faster. Since this implementation uses SSE/YMM registers it cannot safely be used in every situation, e.g. while an IRQ interrupts a kernel thread. The implementation falls back to the generic SHA1 variant, if using the SSE/YMM registers is not possible. With this algorithm I was able to increase the throughput of a single IPsec link from 344 Mbit/s to 464 Mbit/s on a Core 2 Quad CPU using the SSSE3 variant -- a speedup of +34.8%. Saving and restoring SSE/YMM state might make the actual throughput fluctuate when there are FPU intensive userland applications running. For example, meassuring the performance using iperf2 directly on the machine under test gives wobbling numbers because iperf2 uses the FPU for each packet to check if the reporting interval has expired (in the above test I got min/max/avg: 402/484/464 MBit/s). Using this algorithm on a IPsec gateway gives much more reasonable and stable numbers, albeit not as high as in the directly connected case. Here is the result from an RFC 2544 test run with a EXFO Packet Blazer FTB-8510: frame size sha1-generic sha1-ssse3 delta 64 byte 37.5 MBit/s 37.5 MBit/s 0.0% 128 byte 56.3 MBit/s 62.5 MBit/s +11.0% 256 byte 87.5 MBit/s 100.0 MBit/s +14.3% 512 byte 131.3 MBit/s 150.0 MBit/s +14.2% 1024 byte 162.5 MBit/s 193.8 MBit/s +19.3% 1280 byte 175.0 MBit/s 212.5 MBit/s +21.4% 1420 byte 175.0 MBit/s 218.7 MBit/s +25.0% 1518 byte 150.0 MBit/s 181.2 MBit/s +20.8% The throughput for the largest frame size is lower than for the previous size because the IP packets need to be fragmented in this case to make there way through the IPsec tunnel. Signed-off-by: Mathias Krause <minipli@googlemail.com> Cc: Maxim Locktyukhin <maxim.locktyukhin@intel.com> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2011-08-04 20:19:25 +02:00
}
crypto: x86/sha - Restructure x86 sha1 glue code to expose all the available sha1 transforms Restructure the x86 sha1 glue code so we will expose sha1 transforms based on SSSE3, AVX, AVX2 or SHA-NI extension as separate individual drivers when cpu provides such support. This will make it easy for alternative algorithms to be used if desired and makes the code cleaner and easier to maintain. Signed-off-by: Tim Chen <tim.c.chen@linux.intel.com> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2015-09-17 01:34:53 +02:00
asmlinkage void sha1_transform_ssse3(u32 *digest, const char *data,
unsigned int rounds);
static int sha1_ssse3_update(struct shash_desc *desc, const u8 *data,
unsigned int len)
crypto: sha1 - SSSE3 based SHA1 implementation for x86-64 This is an assembler implementation of the SHA1 algorithm using the Supplemental SSE3 (SSSE3) instructions or, when available, the Advanced Vector Extensions (AVX). Testing with the tcrypt module shows the raw hash performance is up to 2.3 times faster than the C implementation, using 8k data blocks on a Core 2 Duo T5500. For the smalest data set (16 byte) it is still 25% faster. Since this implementation uses SSE/YMM registers it cannot safely be used in every situation, e.g. while an IRQ interrupts a kernel thread. The implementation falls back to the generic SHA1 variant, if using the SSE/YMM registers is not possible. With this algorithm I was able to increase the throughput of a single IPsec link from 344 Mbit/s to 464 Mbit/s on a Core 2 Quad CPU using the SSSE3 variant -- a speedup of +34.8%. Saving and restoring SSE/YMM state might make the actual throughput fluctuate when there are FPU intensive userland applications running. For example, meassuring the performance using iperf2 directly on the machine under test gives wobbling numbers because iperf2 uses the FPU for each packet to check if the reporting interval has expired (in the above test I got min/max/avg: 402/484/464 MBit/s). Using this algorithm on a IPsec gateway gives much more reasonable and stable numbers, albeit not as high as in the directly connected case. Here is the result from an RFC 2544 test run with a EXFO Packet Blazer FTB-8510: frame size sha1-generic sha1-ssse3 delta 64 byte 37.5 MBit/s 37.5 MBit/s 0.0% 128 byte 56.3 MBit/s 62.5 MBit/s +11.0% 256 byte 87.5 MBit/s 100.0 MBit/s +14.3% 512 byte 131.3 MBit/s 150.0 MBit/s +14.2% 1024 byte 162.5 MBit/s 193.8 MBit/s +19.3% 1280 byte 175.0 MBit/s 212.5 MBit/s +21.4% 1420 byte 175.0 MBit/s 218.7 MBit/s +25.0% 1518 byte 150.0 MBit/s 181.2 MBit/s +20.8% The throughput for the largest frame size is lower than for the previous size because the IP packets need to be fragmented in this case to make there way through the IPsec tunnel. Signed-off-by: Mathias Krause <minipli@googlemail.com> Cc: Maxim Locktyukhin <maxim.locktyukhin@intel.com> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2011-08-04 20:19:25 +02:00
{
crypto: x86/sha - Restructure x86 sha1 glue code to expose all the available sha1 transforms Restructure the x86 sha1 glue code so we will expose sha1 transforms based on SSSE3, AVX, AVX2 or SHA-NI extension as separate individual drivers when cpu provides such support. This will make it easy for alternative algorithms to be used if desired and makes the code cleaner and easier to maintain. Signed-off-by: Tim Chen <tim.c.chen@linux.intel.com> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2015-09-17 01:34:53 +02:00
return sha1_update(desc, data, len,
(sha1_transform_fn *) sha1_transform_ssse3);
crypto: sha1 - SSSE3 based SHA1 implementation for x86-64 This is an assembler implementation of the SHA1 algorithm using the Supplemental SSE3 (SSSE3) instructions or, when available, the Advanced Vector Extensions (AVX). Testing with the tcrypt module shows the raw hash performance is up to 2.3 times faster than the C implementation, using 8k data blocks on a Core 2 Duo T5500. For the smalest data set (16 byte) it is still 25% faster. Since this implementation uses SSE/YMM registers it cannot safely be used in every situation, e.g. while an IRQ interrupts a kernel thread. The implementation falls back to the generic SHA1 variant, if using the SSE/YMM registers is not possible. With this algorithm I was able to increase the throughput of a single IPsec link from 344 Mbit/s to 464 Mbit/s on a Core 2 Quad CPU using the SSSE3 variant -- a speedup of +34.8%. Saving and restoring SSE/YMM state might make the actual throughput fluctuate when there are FPU intensive userland applications running. For example, meassuring the performance using iperf2 directly on the machine under test gives wobbling numbers because iperf2 uses the FPU for each packet to check if the reporting interval has expired (in the above test I got min/max/avg: 402/484/464 MBit/s). Using this algorithm on a IPsec gateway gives much more reasonable and stable numbers, albeit not as high as in the directly connected case. Here is the result from an RFC 2544 test run with a EXFO Packet Blazer FTB-8510: frame size sha1-generic sha1-ssse3 delta 64 byte 37.5 MBit/s 37.5 MBit/s 0.0% 128 byte 56.3 MBit/s 62.5 MBit/s +11.0% 256 byte 87.5 MBit/s 100.0 MBit/s +14.3% 512 byte 131.3 MBit/s 150.0 MBit/s +14.2% 1024 byte 162.5 MBit/s 193.8 MBit/s +19.3% 1280 byte 175.0 MBit/s 212.5 MBit/s +21.4% 1420 byte 175.0 MBit/s 218.7 MBit/s +25.0% 1518 byte 150.0 MBit/s 181.2 MBit/s +20.8% The throughput for the largest frame size is lower than for the previous size because the IP packets need to be fragmented in this case to make there way through the IPsec tunnel. Signed-off-by: Mathias Krause <minipli@googlemail.com> Cc: Maxim Locktyukhin <maxim.locktyukhin@intel.com> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2011-08-04 20:19:25 +02:00
}
crypto: x86/sha - Restructure x86 sha1 glue code to expose all the available sha1 transforms Restructure the x86 sha1 glue code so we will expose sha1 transforms based on SSSE3, AVX, AVX2 or SHA-NI extension as separate individual drivers when cpu provides such support. This will make it easy for alternative algorithms to be used if desired and makes the code cleaner and easier to maintain. Signed-off-by: Tim Chen <tim.c.chen@linux.intel.com> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2015-09-17 01:34:53 +02:00
static int sha1_ssse3_finup(struct shash_desc *desc, const u8 *data,
unsigned int len, u8 *out)
crypto: sha - SHA1 transform x86_64 AVX2 This git patch adds x86_64 AVX2 optimization of SHA1 transform to crypto support. The patch has been tested with 3.14.0-rc1 kernel. On a Haswell desktop, with turbo disabled and all cpus running at maximum frequency, tcrypt shows AVX2 performance improvement from 3% for 256 bytes update to 16% for 1024 bytes update over AVX implementation. This patch adds sha1_avx2_transform(), the glue, build and configuration changes needed for AVX2 optimization of SHA1 transform to crypto support. sha1-ssse3 is one module which adds the necessary optimization support (SSSE3/AVX/AVX2) for the low-level SHA1 transform function. With better optimization support, transform function is overridden as the case may be. In the case of AVX2, due to performance reasons across datablock sizes, the AVX or AVX2 transform function is used at run-time as it suits best. The Makefile change therefore appends the necessary objects to the linkage. Due to this, the patch merely appends AVX2 transform to the existing build mix and Kconfig support and leaves the configuration build support as is. Signed-off-by: Chandramouli Narayanan <mouli@linux.intel.com> Reviewed-by: Marek Vasut <marex@denx.de> Acked-by: H. Peter Anvin <hpa@linux.intel.com> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2014-03-20 23:14:00 +01:00
{
crypto: x86/sha - Restructure x86 sha1 glue code to expose all the available sha1 transforms Restructure the x86 sha1 glue code so we will expose sha1 transforms based on SSSE3, AVX, AVX2 or SHA-NI extension as separate individual drivers when cpu provides such support. This will make it easy for alternative algorithms to be used if desired and makes the code cleaner and easier to maintain. Signed-off-by: Tim Chen <tim.c.chen@linux.intel.com> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2015-09-17 01:34:53 +02:00
return sha1_finup(desc, data, len, out,
(sha1_transform_fn *) sha1_transform_ssse3);
}
/* Add padding and return the message digest. */
static int sha1_ssse3_final(struct shash_desc *desc, u8 *out)
{
return sha1_ssse3_finup(desc, NULL, 0, out);
crypto: sha - SHA1 transform x86_64 AVX2 This git patch adds x86_64 AVX2 optimization of SHA1 transform to crypto support. The patch has been tested with 3.14.0-rc1 kernel. On a Haswell desktop, with turbo disabled and all cpus running at maximum frequency, tcrypt shows AVX2 performance improvement from 3% for 256 bytes update to 16% for 1024 bytes update over AVX implementation. This patch adds sha1_avx2_transform(), the glue, build and configuration changes needed for AVX2 optimization of SHA1 transform to crypto support. sha1-ssse3 is one module which adds the necessary optimization support (SSSE3/AVX/AVX2) for the low-level SHA1 transform function. With better optimization support, transform function is overridden as the case may be. In the case of AVX2, due to performance reasons across datablock sizes, the AVX or AVX2 transform function is used at run-time as it suits best. The Makefile change therefore appends the necessary objects to the linkage. Due to this, the patch merely appends AVX2 transform to the existing build mix and Kconfig support and leaves the configuration build support as is. Signed-off-by: Chandramouli Narayanan <mouli@linux.intel.com> Reviewed-by: Marek Vasut <marex@denx.de> Acked-by: H. Peter Anvin <hpa@linux.intel.com> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2014-03-20 23:14:00 +01:00
}
crypto: x86/sha - Restructure x86 sha1 glue code to expose all the available sha1 transforms Restructure the x86 sha1 glue code so we will expose sha1 transforms based on SSSE3, AVX, AVX2 or SHA-NI extension as separate individual drivers when cpu provides such support. This will make it easy for alternative algorithms to be used if desired and makes the code cleaner and easier to maintain. Signed-off-by: Tim Chen <tim.c.chen@linux.intel.com> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2015-09-17 01:34:53 +02:00
static struct shash_alg sha1_ssse3_alg = {
crypto: sha1 - SSSE3 based SHA1 implementation for x86-64 This is an assembler implementation of the SHA1 algorithm using the Supplemental SSE3 (SSSE3) instructions or, when available, the Advanced Vector Extensions (AVX). Testing with the tcrypt module shows the raw hash performance is up to 2.3 times faster than the C implementation, using 8k data blocks on a Core 2 Duo T5500. For the smalest data set (16 byte) it is still 25% faster. Since this implementation uses SSE/YMM registers it cannot safely be used in every situation, e.g. while an IRQ interrupts a kernel thread. The implementation falls back to the generic SHA1 variant, if using the SSE/YMM registers is not possible. With this algorithm I was able to increase the throughput of a single IPsec link from 344 Mbit/s to 464 Mbit/s on a Core 2 Quad CPU using the SSSE3 variant -- a speedup of +34.8%. Saving and restoring SSE/YMM state might make the actual throughput fluctuate when there are FPU intensive userland applications running. For example, meassuring the performance using iperf2 directly on the machine under test gives wobbling numbers because iperf2 uses the FPU for each packet to check if the reporting interval has expired (in the above test I got min/max/avg: 402/484/464 MBit/s). Using this algorithm on a IPsec gateway gives much more reasonable and stable numbers, albeit not as high as in the directly connected case. Here is the result from an RFC 2544 test run with a EXFO Packet Blazer FTB-8510: frame size sha1-generic sha1-ssse3 delta 64 byte 37.5 MBit/s 37.5 MBit/s 0.0% 128 byte 56.3 MBit/s 62.5 MBit/s +11.0% 256 byte 87.5 MBit/s 100.0 MBit/s +14.3% 512 byte 131.3 MBit/s 150.0 MBit/s +14.2% 1024 byte 162.5 MBit/s 193.8 MBit/s +19.3% 1280 byte 175.0 MBit/s 212.5 MBit/s +21.4% 1420 byte 175.0 MBit/s 218.7 MBit/s +25.0% 1518 byte 150.0 MBit/s 181.2 MBit/s +20.8% The throughput for the largest frame size is lower than for the previous size because the IP packets need to be fragmented in this case to make there way through the IPsec tunnel. Signed-off-by: Mathias Krause <minipli@googlemail.com> Cc: Maxim Locktyukhin <maxim.locktyukhin@intel.com> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2011-08-04 20:19:25 +02:00
.digestsize = SHA1_DIGEST_SIZE,
crypto: x86/sha1_ssse3 - move SHA-1 SSSE3 implementation to base layer This removes all the boilerplate from the existing implementation, and replaces it with calls into the base layer. Signed-off-by: Ard Biesheuvel <ard.biesheuvel@linaro.org> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2015-04-09 12:55:46 +02:00
.init = sha1_base_init,
crypto: sha1 - SSSE3 based SHA1 implementation for x86-64 This is an assembler implementation of the SHA1 algorithm using the Supplemental SSE3 (SSSE3) instructions or, when available, the Advanced Vector Extensions (AVX). Testing with the tcrypt module shows the raw hash performance is up to 2.3 times faster than the C implementation, using 8k data blocks on a Core 2 Duo T5500. For the smalest data set (16 byte) it is still 25% faster. Since this implementation uses SSE/YMM registers it cannot safely be used in every situation, e.g. while an IRQ interrupts a kernel thread. The implementation falls back to the generic SHA1 variant, if using the SSE/YMM registers is not possible. With this algorithm I was able to increase the throughput of a single IPsec link from 344 Mbit/s to 464 Mbit/s on a Core 2 Quad CPU using the SSSE3 variant -- a speedup of +34.8%. Saving and restoring SSE/YMM state might make the actual throughput fluctuate when there are FPU intensive userland applications running. For example, meassuring the performance using iperf2 directly on the machine under test gives wobbling numbers because iperf2 uses the FPU for each packet to check if the reporting interval has expired (in the above test I got min/max/avg: 402/484/464 MBit/s). Using this algorithm on a IPsec gateway gives much more reasonable and stable numbers, albeit not as high as in the directly connected case. Here is the result from an RFC 2544 test run with a EXFO Packet Blazer FTB-8510: frame size sha1-generic sha1-ssse3 delta 64 byte 37.5 MBit/s 37.5 MBit/s 0.0% 128 byte 56.3 MBit/s 62.5 MBit/s +11.0% 256 byte 87.5 MBit/s 100.0 MBit/s +14.3% 512 byte 131.3 MBit/s 150.0 MBit/s +14.2% 1024 byte 162.5 MBit/s 193.8 MBit/s +19.3% 1280 byte 175.0 MBit/s 212.5 MBit/s +21.4% 1420 byte 175.0 MBit/s 218.7 MBit/s +25.0% 1518 byte 150.0 MBit/s 181.2 MBit/s +20.8% The throughput for the largest frame size is lower than for the previous size because the IP packets need to be fragmented in this case to make there way through the IPsec tunnel. Signed-off-by: Mathias Krause <minipli@googlemail.com> Cc: Maxim Locktyukhin <maxim.locktyukhin@intel.com> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2011-08-04 20:19:25 +02:00
.update = sha1_ssse3_update,
.final = sha1_ssse3_final,
crypto: x86/sha1_ssse3 - move SHA-1 SSSE3 implementation to base layer This removes all the boilerplate from the existing implementation, and replaces it with calls into the base layer. Signed-off-by: Ard Biesheuvel <ard.biesheuvel@linaro.org> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2015-04-09 12:55:46 +02:00
.finup = sha1_ssse3_finup,
crypto: sha1 - SSSE3 based SHA1 implementation for x86-64 This is an assembler implementation of the SHA1 algorithm using the Supplemental SSE3 (SSSE3) instructions or, when available, the Advanced Vector Extensions (AVX). Testing with the tcrypt module shows the raw hash performance is up to 2.3 times faster than the C implementation, using 8k data blocks on a Core 2 Duo T5500. For the smalest data set (16 byte) it is still 25% faster. Since this implementation uses SSE/YMM registers it cannot safely be used in every situation, e.g. while an IRQ interrupts a kernel thread. The implementation falls back to the generic SHA1 variant, if using the SSE/YMM registers is not possible. With this algorithm I was able to increase the throughput of a single IPsec link from 344 Mbit/s to 464 Mbit/s on a Core 2 Quad CPU using the SSSE3 variant -- a speedup of +34.8%. Saving and restoring SSE/YMM state might make the actual throughput fluctuate when there are FPU intensive userland applications running. For example, meassuring the performance using iperf2 directly on the machine under test gives wobbling numbers because iperf2 uses the FPU for each packet to check if the reporting interval has expired (in the above test I got min/max/avg: 402/484/464 MBit/s). Using this algorithm on a IPsec gateway gives much more reasonable and stable numbers, albeit not as high as in the directly connected case. Here is the result from an RFC 2544 test run with a EXFO Packet Blazer FTB-8510: frame size sha1-generic sha1-ssse3 delta 64 byte 37.5 MBit/s 37.5 MBit/s 0.0% 128 byte 56.3 MBit/s 62.5 MBit/s +11.0% 256 byte 87.5 MBit/s 100.0 MBit/s +14.3% 512 byte 131.3 MBit/s 150.0 MBit/s +14.2% 1024 byte 162.5 MBit/s 193.8 MBit/s +19.3% 1280 byte 175.0 MBit/s 212.5 MBit/s +21.4% 1420 byte 175.0 MBit/s 218.7 MBit/s +25.0% 1518 byte 150.0 MBit/s 181.2 MBit/s +20.8% The throughput for the largest frame size is lower than for the previous size because the IP packets need to be fragmented in this case to make there way through the IPsec tunnel. Signed-off-by: Mathias Krause <minipli@googlemail.com> Cc: Maxim Locktyukhin <maxim.locktyukhin@intel.com> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2011-08-04 20:19:25 +02:00
.descsize = sizeof(struct sha1_state),
.base = {
.cra_name = "sha1",
crypto: x86/sha - Restructure x86 sha1 glue code to expose all the available sha1 transforms Restructure the x86 sha1 glue code so we will expose sha1 transforms based on SSSE3, AVX, AVX2 or SHA-NI extension as separate individual drivers when cpu provides such support. This will make it easy for alternative algorithms to be used if desired and makes the code cleaner and easier to maintain. Signed-off-by: Tim Chen <tim.c.chen@linux.intel.com> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2015-09-17 01:34:53 +02:00
.cra_driver_name = "sha1-ssse3",
crypto: sha1 - SSSE3 based SHA1 implementation for x86-64 This is an assembler implementation of the SHA1 algorithm using the Supplemental SSE3 (SSSE3) instructions or, when available, the Advanced Vector Extensions (AVX). Testing with the tcrypt module shows the raw hash performance is up to 2.3 times faster than the C implementation, using 8k data blocks on a Core 2 Duo T5500. For the smalest data set (16 byte) it is still 25% faster. Since this implementation uses SSE/YMM registers it cannot safely be used in every situation, e.g. while an IRQ interrupts a kernel thread. The implementation falls back to the generic SHA1 variant, if using the SSE/YMM registers is not possible. With this algorithm I was able to increase the throughput of a single IPsec link from 344 Mbit/s to 464 Mbit/s on a Core 2 Quad CPU using the SSSE3 variant -- a speedup of +34.8%. Saving and restoring SSE/YMM state might make the actual throughput fluctuate when there are FPU intensive userland applications running. For example, meassuring the performance using iperf2 directly on the machine under test gives wobbling numbers because iperf2 uses the FPU for each packet to check if the reporting interval has expired (in the above test I got min/max/avg: 402/484/464 MBit/s). Using this algorithm on a IPsec gateway gives much more reasonable and stable numbers, albeit not as high as in the directly connected case. Here is the result from an RFC 2544 test run with a EXFO Packet Blazer FTB-8510: frame size sha1-generic sha1-ssse3 delta 64 byte 37.5 MBit/s 37.5 MBit/s 0.0% 128 byte 56.3 MBit/s 62.5 MBit/s +11.0% 256 byte 87.5 MBit/s 100.0 MBit/s +14.3% 512 byte 131.3 MBit/s 150.0 MBit/s +14.2% 1024 byte 162.5 MBit/s 193.8 MBit/s +19.3% 1280 byte 175.0 MBit/s 212.5 MBit/s +21.4% 1420 byte 175.0 MBit/s 218.7 MBit/s +25.0% 1518 byte 150.0 MBit/s 181.2 MBit/s +20.8% The throughput for the largest frame size is lower than for the previous size because the IP packets need to be fragmented in this case to make there way through the IPsec tunnel. Signed-off-by: Mathias Krause <minipli@googlemail.com> Cc: Maxim Locktyukhin <maxim.locktyukhin@intel.com> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2011-08-04 20:19:25 +02:00
.cra_priority = 150,
.cra_flags = CRYPTO_ALG_TYPE_SHASH,
.cra_blocksize = SHA1_BLOCK_SIZE,
.cra_module = THIS_MODULE,
}
};
crypto: x86/sha - Restructure x86 sha1 glue code to expose all the available sha1 transforms Restructure the x86 sha1 glue code so we will expose sha1 transforms based on SSSE3, AVX, AVX2 or SHA-NI extension as separate individual drivers when cpu provides such support. This will make it easy for alternative algorithms to be used if desired and makes the code cleaner and easier to maintain. Signed-off-by: Tim Chen <tim.c.chen@linux.intel.com> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2015-09-17 01:34:53 +02:00
static int register_sha1_ssse3(void)
{
if (boot_cpu_has(X86_FEATURE_SSSE3))
return crypto_register_shash(&sha1_ssse3_alg);
return 0;
}
static void unregister_sha1_ssse3(void)
{
if (boot_cpu_has(X86_FEATURE_SSSE3))
crypto_unregister_shash(&sha1_ssse3_alg);
}
crypto: sha1 - use Kbuild supplied flags for AVX test Commit ea4d26ae ("raid5: add AVX optimized RAID5 checksumming") introduced x86/ arch wide defines for AFLAGS and CFLAGS indicating AVX support in binutils based on the same test we have in x86/crypto/ right now. To minimize duplication drop our implementation in favour to the one in x86/. Signed-off-by: Mathias Krause <minipli@googlemail.com> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2012-05-24 11:13:42 +02:00
#ifdef CONFIG_AS_AVX
crypto: x86/sha - Restructure x86 sha1 glue code to expose all the available sha1 transforms Restructure the x86 sha1 glue code so we will expose sha1 transforms based on SSSE3, AVX, AVX2 or SHA-NI extension as separate individual drivers when cpu provides such support. This will make it easy for alternative algorithms to be used if desired and makes the code cleaner and easier to maintain. Signed-off-by: Tim Chen <tim.c.chen@linux.intel.com> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2015-09-17 01:34:53 +02:00
asmlinkage void sha1_transform_avx(u32 *digest, const char *data,
unsigned int rounds);
static int sha1_avx_update(struct shash_desc *desc, const u8 *data,
unsigned int len)
{
return sha1_update(desc, data, len,
(sha1_transform_fn *) sha1_transform_avx);
}
static int sha1_avx_finup(struct shash_desc *desc, const u8 *data,
unsigned int len, u8 *out)
{
return sha1_finup(desc, data, len, out,
(sha1_transform_fn *) sha1_transform_avx);
}
static int sha1_avx_final(struct shash_desc *desc, u8 *out)
{
return sha1_avx_finup(desc, NULL, 0, out);
}
static struct shash_alg sha1_avx_alg = {
.digestsize = SHA1_DIGEST_SIZE,
.init = sha1_base_init,
.update = sha1_avx_update,
.final = sha1_avx_final,
.finup = sha1_avx_finup,
.descsize = sizeof(struct sha1_state),
.base = {
.cra_name = "sha1",
.cra_driver_name = "sha1-avx",
.cra_priority = 160,
.cra_flags = CRYPTO_ALG_TYPE_SHASH,
.cra_blocksize = SHA1_BLOCK_SIZE,
.cra_module = THIS_MODULE,
}
};
static bool avx_usable(void)
crypto: sha1 - SSSE3 based SHA1 implementation for x86-64 This is an assembler implementation of the SHA1 algorithm using the Supplemental SSE3 (SSSE3) instructions or, when available, the Advanced Vector Extensions (AVX). Testing with the tcrypt module shows the raw hash performance is up to 2.3 times faster than the C implementation, using 8k data blocks on a Core 2 Duo T5500. For the smalest data set (16 byte) it is still 25% faster. Since this implementation uses SSE/YMM registers it cannot safely be used in every situation, e.g. while an IRQ interrupts a kernel thread. The implementation falls back to the generic SHA1 variant, if using the SSE/YMM registers is not possible. With this algorithm I was able to increase the throughput of a single IPsec link from 344 Mbit/s to 464 Mbit/s on a Core 2 Quad CPU using the SSSE3 variant -- a speedup of +34.8%. Saving and restoring SSE/YMM state might make the actual throughput fluctuate when there are FPU intensive userland applications running. For example, meassuring the performance using iperf2 directly on the machine under test gives wobbling numbers because iperf2 uses the FPU for each packet to check if the reporting interval has expired (in the above test I got min/max/avg: 402/484/464 MBit/s). Using this algorithm on a IPsec gateway gives much more reasonable and stable numbers, albeit not as high as in the directly connected case. Here is the result from an RFC 2544 test run with a EXFO Packet Blazer FTB-8510: frame size sha1-generic sha1-ssse3 delta 64 byte 37.5 MBit/s 37.5 MBit/s 0.0% 128 byte 56.3 MBit/s 62.5 MBit/s +11.0% 256 byte 87.5 MBit/s 100.0 MBit/s +14.3% 512 byte 131.3 MBit/s 150.0 MBit/s +14.2% 1024 byte 162.5 MBit/s 193.8 MBit/s +19.3% 1280 byte 175.0 MBit/s 212.5 MBit/s +21.4% 1420 byte 175.0 MBit/s 218.7 MBit/s +25.0% 1518 byte 150.0 MBit/s 181.2 MBit/s +20.8% The throughput for the largest frame size is lower than for the previous size because the IP packets need to be fragmented in this case to make there way through the IPsec tunnel. Signed-off-by: Mathias Krause <minipli@googlemail.com> Cc: Maxim Locktyukhin <maxim.locktyukhin@intel.com> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2011-08-04 20:19:25 +02:00
{
x86/fpu: Rename XSAVE macros There are two concepts that have some confusing naming: 1. Extended State Component numbers (currently called XFEATURE_BIT_*) 2. Extended State Component masks (currently called XSTATE_*) The numbers are (currently) from 0-9. State component 3 is the bounds registers for MPX, for instance. But when we want to enable "state component 3", we go set a bit in XCR0. The bit we set is 1<<3. We can check to see if a state component feature is enabled by looking at its bit. The current 'xfeature_bit's are at best xfeature bit _numbers_. Calling them bits is at best inconsistent with ending the enum list with 'XFEATURES_NR_MAX'. This patch renames the enum to be 'xfeature'. These also happen to be what the Intel documentation calls a "state component". We also want to differentiate these from the "XSTATE_*" macros. The "XSTATE_*" macros are a mask, and we rename them to match. These macros are reasonably widely used so this patch is a wee bit big, but this really is just a rename. The only non-mechanical part of this is the s/XSTATE_EXTEND_MASK/XFEATURE_MASK_EXTEND/ We need a better name for it, but that's another patch. Signed-off-by: Dave Hansen <dave.hansen@linux.intel.com> Cc: Andy Lutomirski <luto@amacapital.net> Cc: Borislav Petkov <bp@alien8.de> Cc: Brian Gerst <brgerst@gmail.com> Cc: Denys Vlasenko <dvlasenk@redhat.com> Cc: Fenghua Yu <fenghua.yu@intel.com> Cc: H. Peter Anvin <hpa@zytor.com> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: Tim Chen <tim.c.chen@linux.intel.com> Cc: dave@sr71.net Cc: linux-kernel@vger.kernel.org Link: http://lkml.kernel.org/r/20150902233126.38653250@viggo.jf.intel.com [ Ported to v4.3-rc1. ] Signed-off-by: Ingo Molnar <mingo@kernel.org>
2015-09-03 01:31:26 +02:00
if (!cpu_has_xfeatures(XFEATURE_MASK_SSE | XFEATURE_MASK_YMM, NULL)) {
x86/cpufeature: Replace cpu_has_avx with boot_cpu_has() usage Signed-off-by: Borislav Petkov <bp@suse.de> Cc: Andy Lutomirski <luto@amacapital.net> Cc: Borislav Petkov <bp@alien8.de> Cc: Brian Gerst <brgerst@gmail.com> Cc: Denys Vlasenko <dvlasenk@redhat.com> Cc: H. Peter Anvin <hpa@zytor.com> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: linux-crypto@vger.kernel.org Link: http://lkml.kernel.org/r/1459801503-15600-4-git-send-email-bp@alien8.de Signed-off-by: Ingo Molnar <mingo@kernel.org>
2016-04-04 22:24:56 +02:00
if (boot_cpu_has(X86_FEATURE_AVX))
x86/fpu, crypto x86/sha1_ssse3: Simplify the sha1_ssse3_mod_init() xfeature checks Use the new 'cpu_has_xfeatures()' function to query AVX CPU support. This has the following advantages to the driver: - Decouples the driver from FPU internals: it's now only using <asm/fpu/api.h>. - Removes detection complexity from the driver, no more raw XGETBV instruction - Shrinks the code a bit. - Standardizes feature name error message printouts across drivers There are also advantages to the x86 FPU code: once all drivers are decoupled from internals we can move them out of common headers and we'll also be able to remove xcr.h. Cc: Andy Lutomirski <luto@amacapital.net> Cc: Borislav Petkov <bp@alien8.de> Cc: Dave Hansen <dave.hansen@linux.intel.com> Cc: Fenghua Yu <fenghua.yu@intel.com> Cc: H. Peter Anvin <hpa@zytor.com> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Oleg Nesterov <oleg@redhat.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Thomas Gleixner <tglx@linutronix.de> Signed-off-by: Ingo Molnar <mingo@kernel.org>
2015-04-28 10:11:24 +02:00
pr_info("AVX detected but unusable.\n");
crypto: sha1 - SSSE3 based SHA1 implementation for x86-64 This is an assembler implementation of the SHA1 algorithm using the Supplemental SSE3 (SSSE3) instructions or, when available, the Advanced Vector Extensions (AVX). Testing with the tcrypt module shows the raw hash performance is up to 2.3 times faster than the C implementation, using 8k data blocks on a Core 2 Duo T5500. For the smalest data set (16 byte) it is still 25% faster. Since this implementation uses SSE/YMM registers it cannot safely be used in every situation, e.g. while an IRQ interrupts a kernel thread. The implementation falls back to the generic SHA1 variant, if using the SSE/YMM registers is not possible. With this algorithm I was able to increase the throughput of a single IPsec link from 344 Mbit/s to 464 Mbit/s on a Core 2 Quad CPU using the SSSE3 variant -- a speedup of +34.8%. Saving and restoring SSE/YMM state might make the actual throughput fluctuate when there are FPU intensive userland applications running. For example, meassuring the performance using iperf2 directly on the machine under test gives wobbling numbers because iperf2 uses the FPU for each packet to check if the reporting interval has expired (in the above test I got min/max/avg: 402/484/464 MBit/s). Using this algorithm on a IPsec gateway gives much more reasonable and stable numbers, albeit not as high as in the directly connected case. Here is the result from an RFC 2544 test run with a EXFO Packet Blazer FTB-8510: frame size sha1-generic sha1-ssse3 delta 64 byte 37.5 MBit/s 37.5 MBit/s 0.0% 128 byte 56.3 MBit/s 62.5 MBit/s +11.0% 256 byte 87.5 MBit/s 100.0 MBit/s +14.3% 512 byte 131.3 MBit/s 150.0 MBit/s +14.2% 1024 byte 162.5 MBit/s 193.8 MBit/s +19.3% 1280 byte 175.0 MBit/s 212.5 MBit/s +21.4% 1420 byte 175.0 MBit/s 218.7 MBit/s +25.0% 1518 byte 150.0 MBit/s 181.2 MBit/s +20.8% The throughput for the largest frame size is lower than for the previous size because the IP packets need to be fragmented in this case to make there way through the IPsec tunnel. Signed-off-by: Mathias Krause <minipli@googlemail.com> Cc: Maxim Locktyukhin <maxim.locktyukhin@intel.com> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2011-08-04 20:19:25 +02:00
return false;
}
return true;
}
crypto: x86/sha1 - re-enable the AVX variant Commit 7c1da8d0d0 "crypto: sha - SHA1 transform x86_64 AVX2" accidentally disabled the AVX variant by making the avx_usable() test not only fail in case the CPU doesn't support AVX or OSXSAVE but also if it doesn't support AVX2. Fix that regression by splitting up the AVX/AVX2 test into two functions. Also test for the BMI1 extension in the avx2_usable() test as the AVX2 implementation not only makes use of BMI2 but also BMI1 instructions. Cc: Chandramouli Narayanan <mouli@linux.intel.com> Signed-off-by: Mathias Krause <minipli@googlemail.com> Reviewed-by: H. Peter Anvin <hpa@linux.intel.com> Reviewed-by: Marek Vasut <marex@denx.de> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2014-03-24 17:10:37 +01:00
crypto: x86/sha - Restructure x86 sha1 glue code to expose all the available sha1 transforms Restructure the x86 sha1 glue code so we will expose sha1 transforms based on SSSE3, AVX, AVX2 or SHA-NI extension as separate individual drivers when cpu provides such support. This will make it easy for alternative algorithms to be used if desired and makes the code cleaner and easier to maintain. Signed-off-by: Tim Chen <tim.c.chen@linux.intel.com> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2015-09-17 01:34:53 +02:00
static int register_sha1_avx(void)
{
if (avx_usable())
return crypto_register_shash(&sha1_avx_alg);
return 0;
}
static void unregister_sha1_avx(void)
crypto: x86/sha1 - re-enable the AVX variant Commit 7c1da8d0d0 "crypto: sha - SHA1 transform x86_64 AVX2" accidentally disabled the AVX variant by making the avx_usable() test not only fail in case the CPU doesn't support AVX or OSXSAVE but also if it doesn't support AVX2. Fix that regression by splitting up the AVX/AVX2 test into two functions. Also test for the BMI1 extension in the avx2_usable() test as the AVX2 implementation not only makes use of BMI2 but also BMI1 instructions. Cc: Chandramouli Narayanan <mouli@linux.intel.com> Signed-off-by: Mathias Krause <minipli@googlemail.com> Reviewed-by: H. Peter Anvin <hpa@linux.intel.com> Reviewed-by: Marek Vasut <marex@denx.de> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2014-03-24 17:10:37 +01:00
{
crypto: x86/sha - Restructure x86 sha1 glue code to expose all the available sha1 transforms Restructure the x86 sha1 glue code so we will expose sha1 transforms based on SSSE3, AVX, AVX2 or SHA-NI extension as separate individual drivers when cpu provides such support. This will make it easy for alternative algorithms to be used if desired and makes the code cleaner and easier to maintain. Signed-off-by: Tim Chen <tim.c.chen@linux.intel.com> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2015-09-17 01:34:53 +02:00
if (avx_usable())
crypto_unregister_shash(&sha1_avx_alg);
}
#else /* CONFIG_AS_AVX */
static inline int register_sha1_avx(void) { return 0; }
static inline void unregister_sha1_avx(void) { }
#endif /* CONFIG_AS_AVX */
#if defined(CONFIG_AS_AVX2) && (CONFIG_AS_AVX)
#define SHA1_AVX2_BLOCK_OPTSIZE 4 /* optimal 4*64 bytes of SHA1 blocks */
asmlinkage void sha1_transform_avx2(u32 *digest, const char *data,
unsigned int rounds);
static bool avx2_usable(void)
crypto: x86/sha1 - re-enable the AVX variant Commit 7c1da8d0d0 "crypto: sha - SHA1 transform x86_64 AVX2" accidentally disabled the AVX variant by making the avx_usable() test not only fail in case the CPU doesn't support AVX or OSXSAVE but also if it doesn't support AVX2. Fix that regression by splitting up the AVX/AVX2 test into two functions. Also test for the BMI1 extension in the avx2_usable() test as the AVX2 implementation not only makes use of BMI2 but also BMI1 instructions. Cc: Chandramouli Narayanan <mouli@linux.intel.com> Signed-off-by: Mathias Krause <minipli@googlemail.com> Reviewed-by: H. Peter Anvin <hpa@linux.intel.com> Reviewed-by: Marek Vasut <marex@denx.de> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2014-03-24 17:10:37 +01:00
{
crypto: x86/sha - Restructure x86 sha1 glue code to expose all the available sha1 transforms Restructure the x86 sha1 glue code so we will expose sha1 transforms based on SSSE3, AVX, AVX2 or SHA-NI extension as separate individual drivers when cpu provides such support. This will make it easy for alternative algorithms to be used if desired and makes the code cleaner and easier to maintain. Signed-off-by: Tim Chen <tim.c.chen@linux.intel.com> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2015-09-17 01:34:53 +02:00
if (avx_usable() && boot_cpu_has(X86_FEATURE_AVX2)
&& boot_cpu_has(X86_FEATURE_BMI1)
&& boot_cpu_has(X86_FEATURE_BMI2))
crypto: x86/sha1 - re-enable the AVX variant Commit 7c1da8d0d0 "crypto: sha - SHA1 transform x86_64 AVX2" accidentally disabled the AVX variant by making the avx_usable() test not only fail in case the CPU doesn't support AVX or OSXSAVE but also if it doesn't support AVX2. Fix that regression by splitting up the AVX/AVX2 test into two functions. Also test for the BMI1 extension in the avx2_usable() test as the AVX2 implementation not only makes use of BMI2 but also BMI1 instructions. Cc: Chandramouli Narayanan <mouli@linux.intel.com> Signed-off-by: Mathias Krause <minipli@googlemail.com> Reviewed-by: H. Peter Anvin <hpa@linux.intel.com> Reviewed-by: Marek Vasut <marex@denx.de> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2014-03-24 17:10:37 +01:00
return true;
return false;
}
crypto: sha1 - SSSE3 based SHA1 implementation for x86-64 This is an assembler implementation of the SHA1 algorithm using the Supplemental SSE3 (SSSE3) instructions or, when available, the Advanced Vector Extensions (AVX). Testing with the tcrypt module shows the raw hash performance is up to 2.3 times faster than the C implementation, using 8k data blocks on a Core 2 Duo T5500. For the smalest data set (16 byte) it is still 25% faster. Since this implementation uses SSE/YMM registers it cannot safely be used in every situation, e.g. while an IRQ interrupts a kernel thread. The implementation falls back to the generic SHA1 variant, if using the SSE/YMM registers is not possible. With this algorithm I was able to increase the throughput of a single IPsec link from 344 Mbit/s to 464 Mbit/s on a Core 2 Quad CPU using the SSSE3 variant -- a speedup of +34.8%. Saving and restoring SSE/YMM state might make the actual throughput fluctuate when there are FPU intensive userland applications running. For example, meassuring the performance using iperf2 directly on the machine under test gives wobbling numbers because iperf2 uses the FPU for each packet to check if the reporting interval has expired (in the above test I got min/max/avg: 402/484/464 MBit/s). Using this algorithm on a IPsec gateway gives much more reasonable and stable numbers, albeit not as high as in the directly connected case. Here is the result from an RFC 2544 test run with a EXFO Packet Blazer FTB-8510: frame size sha1-generic sha1-ssse3 delta 64 byte 37.5 MBit/s 37.5 MBit/s 0.0% 128 byte 56.3 MBit/s 62.5 MBit/s +11.0% 256 byte 87.5 MBit/s 100.0 MBit/s +14.3% 512 byte 131.3 MBit/s 150.0 MBit/s +14.2% 1024 byte 162.5 MBit/s 193.8 MBit/s +19.3% 1280 byte 175.0 MBit/s 212.5 MBit/s +21.4% 1420 byte 175.0 MBit/s 218.7 MBit/s +25.0% 1518 byte 150.0 MBit/s 181.2 MBit/s +20.8% The throughput for the largest frame size is lower than for the previous size because the IP packets need to be fragmented in this case to make there way through the IPsec tunnel. Signed-off-by: Mathias Krause <minipli@googlemail.com> Cc: Maxim Locktyukhin <maxim.locktyukhin@intel.com> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2011-08-04 20:19:25 +02:00
crypto: x86/sha - Restructure x86 sha1 glue code to expose all the available sha1 transforms Restructure the x86 sha1 glue code so we will expose sha1 transforms based on SSSE3, AVX, AVX2 or SHA-NI extension as separate individual drivers when cpu provides such support. This will make it easy for alternative algorithms to be used if desired and makes the code cleaner and easier to maintain. Signed-off-by: Tim Chen <tim.c.chen@linux.intel.com> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2015-09-17 01:34:53 +02:00
static void sha1_apply_transform_avx2(u32 *digest, const char *data,
unsigned int rounds)
crypto: sha1 - SSSE3 based SHA1 implementation for x86-64 This is an assembler implementation of the SHA1 algorithm using the Supplemental SSE3 (SSSE3) instructions or, when available, the Advanced Vector Extensions (AVX). Testing with the tcrypt module shows the raw hash performance is up to 2.3 times faster than the C implementation, using 8k data blocks on a Core 2 Duo T5500. For the smalest data set (16 byte) it is still 25% faster. Since this implementation uses SSE/YMM registers it cannot safely be used in every situation, e.g. while an IRQ interrupts a kernel thread. The implementation falls back to the generic SHA1 variant, if using the SSE/YMM registers is not possible. With this algorithm I was able to increase the throughput of a single IPsec link from 344 Mbit/s to 464 Mbit/s on a Core 2 Quad CPU using the SSSE3 variant -- a speedup of +34.8%. Saving and restoring SSE/YMM state might make the actual throughput fluctuate when there are FPU intensive userland applications running. For example, meassuring the performance using iperf2 directly on the machine under test gives wobbling numbers because iperf2 uses the FPU for each packet to check if the reporting interval has expired (in the above test I got min/max/avg: 402/484/464 MBit/s). Using this algorithm on a IPsec gateway gives much more reasonable and stable numbers, albeit not as high as in the directly connected case. Here is the result from an RFC 2544 test run with a EXFO Packet Blazer FTB-8510: frame size sha1-generic sha1-ssse3 delta 64 byte 37.5 MBit/s 37.5 MBit/s 0.0% 128 byte 56.3 MBit/s 62.5 MBit/s +11.0% 256 byte 87.5 MBit/s 100.0 MBit/s +14.3% 512 byte 131.3 MBit/s 150.0 MBit/s +14.2% 1024 byte 162.5 MBit/s 193.8 MBit/s +19.3% 1280 byte 175.0 MBit/s 212.5 MBit/s +21.4% 1420 byte 175.0 MBit/s 218.7 MBit/s +25.0% 1518 byte 150.0 MBit/s 181.2 MBit/s +20.8% The throughput for the largest frame size is lower than for the previous size because the IP packets need to be fragmented in this case to make there way through the IPsec tunnel. Signed-off-by: Mathias Krause <minipli@googlemail.com> Cc: Maxim Locktyukhin <maxim.locktyukhin@intel.com> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2011-08-04 20:19:25 +02:00
{
crypto: x86/sha - Restructure x86 sha1 glue code to expose all the available sha1 transforms Restructure the x86 sha1 glue code so we will expose sha1 transforms based on SSSE3, AVX, AVX2 or SHA-NI extension as separate individual drivers when cpu provides such support. This will make it easy for alternative algorithms to be used if desired and makes the code cleaner and easier to maintain. Signed-off-by: Tim Chen <tim.c.chen@linux.intel.com> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2015-09-17 01:34:53 +02:00
/* Select the optimal transform based on data block size */
if (rounds >= SHA1_AVX2_BLOCK_OPTSIZE)
sha1_transform_avx2(digest, data, rounds);
else
sha1_transform_avx(digest, data, rounds);
}
crypto: x86/sha1 - re-enable the AVX variant Commit 7c1da8d0d0 "crypto: sha - SHA1 transform x86_64 AVX2" accidentally disabled the AVX variant by making the avx_usable() test not only fail in case the CPU doesn't support AVX or OSXSAVE but also if it doesn't support AVX2. Fix that regression by splitting up the AVX/AVX2 test into two functions. Also test for the BMI1 extension in the avx2_usable() test as the AVX2 implementation not only makes use of BMI2 but also BMI1 instructions. Cc: Chandramouli Narayanan <mouli@linux.intel.com> Signed-off-by: Mathias Krause <minipli@googlemail.com> Reviewed-by: H. Peter Anvin <hpa@linux.intel.com> Reviewed-by: Marek Vasut <marex@denx.de> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2014-03-24 17:10:37 +01:00
crypto: x86/sha - Restructure x86 sha1 glue code to expose all the available sha1 transforms Restructure the x86 sha1 glue code so we will expose sha1 transforms based on SSSE3, AVX, AVX2 or SHA-NI extension as separate individual drivers when cpu provides such support. This will make it easy for alternative algorithms to be used if desired and makes the code cleaner and easier to maintain. Signed-off-by: Tim Chen <tim.c.chen@linux.intel.com> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2015-09-17 01:34:53 +02:00
static int sha1_avx2_update(struct shash_desc *desc, const u8 *data,
unsigned int len)
{
return sha1_update(desc, data, len,
(sha1_transform_fn *) sha1_apply_transform_avx2);
}
crypto: sha1 - SSSE3 based SHA1 implementation for x86-64 This is an assembler implementation of the SHA1 algorithm using the Supplemental SSE3 (SSSE3) instructions or, when available, the Advanced Vector Extensions (AVX). Testing with the tcrypt module shows the raw hash performance is up to 2.3 times faster than the C implementation, using 8k data blocks on a Core 2 Duo T5500. For the smalest data set (16 byte) it is still 25% faster. Since this implementation uses SSE/YMM registers it cannot safely be used in every situation, e.g. while an IRQ interrupts a kernel thread. The implementation falls back to the generic SHA1 variant, if using the SSE/YMM registers is not possible. With this algorithm I was able to increase the throughput of a single IPsec link from 344 Mbit/s to 464 Mbit/s on a Core 2 Quad CPU using the SSSE3 variant -- a speedup of +34.8%. Saving and restoring SSE/YMM state might make the actual throughput fluctuate when there are FPU intensive userland applications running. For example, meassuring the performance using iperf2 directly on the machine under test gives wobbling numbers because iperf2 uses the FPU for each packet to check if the reporting interval has expired (in the above test I got min/max/avg: 402/484/464 MBit/s). Using this algorithm on a IPsec gateway gives much more reasonable and stable numbers, albeit not as high as in the directly connected case. Here is the result from an RFC 2544 test run with a EXFO Packet Blazer FTB-8510: frame size sha1-generic sha1-ssse3 delta 64 byte 37.5 MBit/s 37.5 MBit/s 0.0% 128 byte 56.3 MBit/s 62.5 MBit/s +11.0% 256 byte 87.5 MBit/s 100.0 MBit/s +14.3% 512 byte 131.3 MBit/s 150.0 MBit/s +14.2% 1024 byte 162.5 MBit/s 193.8 MBit/s +19.3% 1280 byte 175.0 MBit/s 212.5 MBit/s +21.4% 1420 byte 175.0 MBit/s 218.7 MBit/s +25.0% 1518 byte 150.0 MBit/s 181.2 MBit/s +20.8% The throughput for the largest frame size is lower than for the previous size because the IP packets need to be fragmented in this case to make there way through the IPsec tunnel. Signed-off-by: Mathias Krause <minipli@googlemail.com> Cc: Maxim Locktyukhin <maxim.locktyukhin@intel.com> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2011-08-04 20:19:25 +02:00
crypto: x86/sha - Restructure x86 sha1 glue code to expose all the available sha1 transforms Restructure the x86 sha1 glue code so we will expose sha1 transforms based on SSSE3, AVX, AVX2 or SHA-NI extension as separate individual drivers when cpu provides such support. This will make it easy for alternative algorithms to be used if desired and makes the code cleaner and easier to maintain. Signed-off-by: Tim Chen <tim.c.chen@linux.intel.com> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2015-09-17 01:34:53 +02:00
static int sha1_avx2_finup(struct shash_desc *desc, const u8 *data,
unsigned int len, u8 *out)
{
return sha1_finup(desc, data, len, out,
(sha1_transform_fn *) sha1_apply_transform_avx2);
}
static int sha1_avx2_final(struct shash_desc *desc, u8 *out)
{
return sha1_avx2_finup(desc, NULL, 0, out);
}
static struct shash_alg sha1_avx2_alg = {
.digestsize = SHA1_DIGEST_SIZE,
.init = sha1_base_init,
.update = sha1_avx2_update,
.final = sha1_avx2_final,
.finup = sha1_avx2_finup,
.descsize = sizeof(struct sha1_state),
.base = {
.cra_name = "sha1",
.cra_driver_name = "sha1-avx2",
.cra_priority = 170,
.cra_flags = CRYPTO_ALG_TYPE_SHASH,
.cra_blocksize = SHA1_BLOCK_SIZE,
.cra_module = THIS_MODULE,
crypto: sha - SHA1 transform x86_64 AVX2 This git patch adds x86_64 AVX2 optimization of SHA1 transform to crypto support. The patch has been tested with 3.14.0-rc1 kernel. On a Haswell desktop, with turbo disabled and all cpus running at maximum frequency, tcrypt shows AVX2 performance improvement from 3% for 256 bytes update to 16% for 1024 bytes update over AVX implementation. This patch adds sha1_avx2_transform(), the glue, build and configuration changes needed for AVX2 optimization of SHA1 transform to crypto support. sha1-ssse3 is one module which adds the necessary optimization support (SSSE3/AVX/AVX2) for the low-level SHA1 transform function. With better optimization support, transform function is overridden as the case may be. In the case of AVX2, due to performance reasons across datablock sizes, the AVX or AVX2 transform function is used at run-time as it suits best. The Makefile change therefore appends the necessary objects to the linkage. Due to this, the patch merely appends AVX2 transform to the existing build mix and Kconfig support and leaves the configuration build support as is. Signed-off-by: Chandramouli Narayanan <mouli@linux.intel.com> Reviewed-by: Marek Vasut <marex@denx.de> Acked-by: H. Peter Anvin <hpa@linux.intel.com> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2014-03-20 23:14:00 +01:00
}
crypto: x86/sha - Restructure x86 sha1 glue code to expose all the available sha1 transforms Restructure the x86 sha1 glue code so we will expose sha1 transforms based on SSSE3, AVX, AVX2 or SHA-NI extension as separate individual drivers when cpu provides such support. This will make it easy for alternative algorithms to be used if desired and makes the code cleaner and easier to maintain. Signed-off-by: Tim Chen <tim.c.chen@linux.intel.com> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2015-09-17 01:34:53 +02:00
};
static int register_sha1_avx2(void)
{
if (avx2_usable())
return crypto_register_shash(&sha1_avx2_alg);
return 0;
}
static void unregister_sha1_avx2(void)
{
if (avx2_usable())
crypto_unregister_shash(&sha1_avx2_alg);
}
#else
static inline int register_sha1_avx2(void) { return 0; }
static inline void unregister_sha1_avx2(void) { }
crypto: x86/sha1 - re-enable the AVX variant Commit 7c1da8d0d0 "crypto: sha - SHA1 transform x86_64 AVX2" accidentally disabled the AVX variant by making the avx_usable() test not only fail in case the CPU doesn't support AVX or OSXSAVE but also if it doesn't support AVX2. Fix that regression by splitting up the AVX/AVX2 test into two functions. Also test for the BMI1 extension in the avx2_usable() test as the AVX2 implementation not only makes use of BMI2 but also BMI1 instructions. Cc: Chandramouli Narayanan <mouli@linux.intel.com> Signed-off-by: Mathias Krause <minipli@googlemail.com> Reviewed-by: H. Peter Anvin <hpa@linux.intel.com> Reviewed-by: Marek Vasut <marex@denx.de> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2014-03-24 17:10:37 +01:00
#endif
crypto: x86/sha - Restructure x86 sha1 glue code to expose all the available sha1 transforms Restructure the x86 sha1 glue code so we will expose sha1 transforms based on SSSE3, AVX, AVX2 or SHA-NI extension as separate individual drivers when cpu provides such support. This will make it easy for alternative algorithms to be used if desired and makes the code cleaner and easier to maintain. Signed-off-by: Tim Chen <tim.c.chen@linux.intel.com> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2015-09-17 01:34:53 +02:00
crypto: x86/sha - glue code for Intel SHA extensions optimized SHA1 & SHA256 This patch adds the glue code to detect and utilize the Intel SHA extensions optimized SHA1 and SHA256 update transforms when available. This code has been tested on Broxton for functionality. Originally-by: Chandramouli Narayanan <mouli_7982@yahoo.com> Signed-off-by: Tim Chen <tim.c.chen@linux.intel.com> Acked-by: David S. Miller <davem@davemloft.net> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2015-09-11 00:27:20 +02:00
#ifdef CONFIG_AS_SHA1_NI
crypto: x86/sha - Restructure x86 sha1 glue code to expose all the available sha1 transforms Restructure the x86 sha1 glue code so we will expose sha1 transforms based on SSSE3, AVX, AVX2 or SHA-NI extension as separate individual drivers when cpu provides such support. This will make it easy for alternative algorithms to be used if desired and makes the code cleaner and easier to maintain. Signed-off-by: Tim Chen <tim.c.chen@linux.intel.com> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2015-09-17 01:34:53 +02:00
asmlinkage void sha1_ni_transform(u32 *digest, const char *data,
unsigned int rounds);
static int sha1_ni_update(struct shash_desc *desc, const u8 *data,
unsigned int len)
{
return sha1_update(desc, data, len,
(sha1_transform_fn *) sha1_ni_transform);
}
static int sha1_ni_finup(struct shash_desc *desc, const u8 *data,
unsigned int len, u8 *out)
{
return sha1_finup(desc, data, len, out,
(sha1_transform_fn *) sha1_ni_transform);
}
static int sha1_ni_final(struct shash_desc *desc, u8 *out)
{
return sha1_ni_finup(desc, NULL, 0, out);
}
static struct shash_alg sha1_ni_alg = {
.digestsize = SHA1_DIGEST_SIZE,
.init = sha1_base_init,
.update = sha1_ni_update,
.final = sha1_ni_final,
.finup = sha1_ni_finup,
.descsize = sizeof(struct sha1_state),
.base = {
.cra_name = "sha1",
.cra_driver_name = "sha1-ni",
.cra_priority = 250,
.cra_flags = CRYPTO_ALG_TYPE_SHASH,
.cra_blocksize = SHA1_BLOCK_SIZE,
.cra_module = THIS_MODULE,
crypto: x86/sha - glue code for Intel SHA extensions optimized SHA1 & SHA256 This patch adds the glue code to detect and utilize the Intel SHA extensions optimized SHA1 and SHA256 update transforms when available. This code has been tested on Broxton for functionality. Originally-by: Chandramouli Narayanan <mouli_7982@yahoo.com> Signed-off-by: Tim Chen <tim.c.chen@linux.intel.com> Acked-by: David S. Miller <davem@davemloft.net> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2015-09-11 00:27:20 +02:00
}
crypto: x86/sha - Restructure x86 sha1 glue code to expose all the available sha1 transforms Restructure the x86 sha1 glue code so we will expose sha1 transforms based on SSSE3, AVX, AVX2 or SHA-NI extension as separate individual drivers when cpu provides such support. This will make it easy for alternative algorithms to be used if desired and makes the code cleaner and easier to maintain. Signed-off-by: Tim Chen <tim.c.chen@linux.intel.com> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2015-09-17 01:34:53 +02:00
};
static int register_sha1_ni(void)
{
if (boot_cpu_has(X86_FEATURE_SHA_NI))
return crypto_register_shash(&sha1_ni_alg);
return 0;
}
static void unregister_sha1_ni(void)
{
if (boot_cpu_has(X86_FEATURE_SHA_NI))
crypto_unregister_shash(&sha1_ni_alg);
}
#else
static inline int register_sha1_ni(void) { return 0; }
static inline void unregister_sha1_ni(void) { }
crypto: sha1 - SSSE3 based SHA1 implementation for x86-64 This is an assembler implementation of the SHA1 algorithm using the Supplemental SSE3 (SSSE3) instructions or, when available, the Advanced Vector Extensions (AVX). Testing with the tcrypt module shows the raw hash performance is up to 2.3 times faster than the C implementation, using 8k data blocks on a Core 2 Duo T5500. For the smalest data set (16 byte) it is still 25% faster. Since this implementation uses SSE/YMM registers it cannot safely be used in every situation, e.g. while an IRQ interrupts a kernel thread. The implementation falls back to the generic SHA1 variant, if using the SSE/YMM registers is not possible. With this algorithm I was able to increase the throughput of a single IPsec link from 344 Mbit/s to 464 Mbit/s on a Core 2 Quad CPU using the SSSE3 variant -- a speedup of +34.8%. Saving and restoring SSE/YMM state might make the actual throughput fluctuate when there are FPU intensive userland applications running. For example, meassuring the performance using iperf2 directly on the machine under test gives wobbling numbers because iperf2 uses the FPU for each packet to check if the reporting interval has expired (in the above test I got min/max/avg: 402/484/464 MBit/s). Using this algorithm on a IPsec gateway gives much more reasonable and stable numbers, albeit not as high as in the directly connected case. Here is the result from an RFC 2544 test run with a EXFO Packet Blazer FTB-8510: frame size sha1-generic sha1-ssse3 delta 64 byte 37.5 MBit/s 37.5 MBit/s 0.0% 128 byte 56.3 MBit/s 62.5 MBit/s +11.0% 256 byte 87.5 MBit/s 100.0 MBit/s +14.3% 512 byte 131.3 MBit/s 150.0 MBit/s +14.2% 1024 byte 162.5 MBit/s 193.8 MBit/s +19.3% 1280 byte 175.0 MBit/s 212.5 MBit/s +21.4% 1420 byte 175.0 MBit/s 218.7 MBit/s +25.0% 1518 byte 150.0 MBit/s 181.2 MBit/s +20.8% The throughput for the largest frame size is lower than for the previous size because the IP packets need to be fragmented in this case to make there way through the IPsec tunnel. Signed-off-by: Mathias Krause <minipli@googlemail.com> Cc: Maxim Locktyukhin <maxim.locktyukhin@intel.com> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2011-08-04 20:19:25 +02:00
#endif
static int __init sha1_ssse3_mod_init(void)
{
crypto: x86/sha - Restructure x86 sha1 glue code to expose all the available sha1 transforms Restructure the x86 sha1 glue code so we will expose sha1 transforms based on SSSE3, AVX, AVX2 or SHA-NI extension as separate individual drivers when cpu provides such support. This will make it easy for alternative algorithms to be used if desired and makes the code cleaner and easier to maintain. Signed-off-by: Tim Chen <tim.c.chen@linux.intel.com> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2015-09-17 01:34:53 +02:00
if (register_sha1_ssse3())
goto fail;
crypto: x86/sha1 - re-enable the AVX variant Commit 7c1da8d0d0 "crypto: sha - SHA1 transform x86_64 AVX2" accidentally disabled the AVX variant by making the avx_usable() test not only fail in case the CPU doesn't support AVX or OSXSAVE but also if it doesn't support AVX2. Fix that regression by splitting up the AVX/AVX2 test into two functions. Also test for the BMI1 extension in the avx2_usable() test as the AVX2 implementation not only makes use of BMI2 but also BMI1 instructions. Cc: Chandramouli Narayanan <mouli@linux.intel.com> Signed-off-by: Mathias Krause <minipli@googlemail.com> Reviewed-by: H. Peter Anvin <hpa@linux.intel.com> Reviewed-by: Marek Vasut <marex@denx.de> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2014-03-24 17:10:37 +01:00
crypto: x86/sha - Restructure x86 sha1 glue code to expose all the available sha1 transforms Restructure the x86 sha1 glue code so we will expose sha1 transforms based on SSSE3, AVX, AVX2 or SHA-NI extension as separate individual drivers when cpu provides such support. This will make it easy for alternative algorithms to be used if desired and makes the code cleaner and easier to maintain. Signed-off-by: Tim Chen <tim.c.chen@linux.intel.com> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2015-09-17 01:34:53 +02:00
if (register_sha1_avx()) {
unregister_sha1_ssse3();
goto fail;
crypto: sha - SHA1 transform x86_64 AVX2 This git patch adds x86_64 AVX2 optimization of SHA1 transform to crypto support. The patch has been tested with 3.14.0-rc1 kernel. On a Haswell desktop, with turbo disabled and all cpus running at maximum frequency, tcrypt shows AVX2 performance improvement from 3% for 256 bytes update to 16% for 1024 bytes update over AVX implementation. This patch adds sha1_avx2_transform(), the glue, build and configuration changes needed for AVX2 optimization of SHA1 transform to crypto support. sha1-ssse3 is one module which adds the necessary optimization support (SSSE3/AVX/AVX2) for the low-level SHA1 transform function. With better optimization support, transform function is overridden as the case may be. In the case of AVX2, due to performance reasons across datablock sizes, the AVX or AVX2 transform function is used at run-time as it suits best. The Makefile change therefore appends the necessary objects to the linkage. Due to this, the patch merely appends AVX2 transform to the existing build mix and Kconfig support and leaves the configuration build support as is. Signed-off-by: Chandramouli Narayanan <mouli@linux.intel.com> Reviewed-by: Marek Vasut <marex@denx.de> Acked-by: H. Peter Anvin <hpa@linux.intel.com> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2014-03-20 23:14:00 +01:00
}
crypto: sha1 - SSSE3 based SHA1 implementation for x86-64 This is an assembler implementation of the SHA1 algorithm using the Supplemental SSE3 (SSSE3) instructions or, when available, the Advanced Vector Extensions (AVX). Testing with the tcrypt module shows the raw hash performance is up to 2.3 times faster than the C implementation, using 8k data blocks on a Core 2 Duo T5500. For the smalest data set (16 byte) it is still 25% faster. Since this implementation uses SSE/YMM registers it cannot safely be used in every situation, e.g. while an IRQ interrupts a kernel thread. The implementation falls back to the generic SHA1 variant, if using the SSE/YMM registers is not possible. With this algorithm I was able to increase the throughput of a single IPsec link from 344 Mbit/s to 464 Mbit/s on a Core 2 Quad CPU using the SSSE3 variant -- a speedup of +34.8%. Saving and restoring SSE/YMM state might make the actual throughput fluctuate when there are FPU intensive userland applications running. For example, meassuring the performance using iperf2 directly on the machine under test gives wobbling numbers because iperf2 uses the FPU for each packet to check if the reporting interval has expired (in the above test I got min/max/avg: 402/484/464 MBit/s). Using this algorithm on a IPsec gateway gives much more reasonable and stable numbers, albeit not as high as in the directly connected case. Here is the result from an RFC 2544 test run with a EXFO Packet Blazer FTB-8510: frame size sha1-generic sha1-ssse3 delta 64 byte 37.5 MBit/s 37.5 MBit/s 0.0% 128 byte 56.3 MBit/s 62.5 MBit/s +11.0% 256 byte 87.5 MBit/s 100.0 MBit/s +14.3% 512 byte 131.3 MBit/s 150.0 MBit/s +14.2% 1024 byte 162.5 MBit/s 193.8 MBit/s +19.3% 1280 byte 175.0 MBit/s 212.5 MBit/s +21.4% 1420 byte 175.0 MBit/s 218.7 MBit/s +25.0% 1518 byte 150.0 MBit/s 181.2 MBit/s +20.8% The throughput for the largest frame size is lower than for the previous size because the IP packets need to be fragmented in this case to make there way through the IPsec tunnel. Signed-off-by: Mathias Krause <minipli@googlemail.com> Cc: Maxim Locktyukhin <maxim.locktyukhin@intel.com> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2011-08-04 20:19:25 +02:00
crypto: x86/sha - Restructure x86 sha1 glue code to expose all the available sha1 transforms Restructure the x86 sha1 glue code so we will expose sha1 transforms based on SSSE3, AVX, AVX2 or SHA-NI extension as separate individual drivers when cpu provides such support. This will make it easy for alternative algorithms to be used if desired and makes the code cleaner and easier to maintain. Signed-off-by: Tim Chen <tim.c.chen@linux.intel.com> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2015-09-17 01:34:53 +02:00
if (register_sha1_avx2()) {
unregister_sha1_avx();
unregister_sha1_ssse3();
goto fail;
crypto: sha - SHA1 transform x86_64 AVX2 This git patch adds x86_64 AVX2 optimization of SHA1 transform to crypto support. The patch has been tested with 3.14.0-rc1 kernel. On a Haswell desktop, with turbo disabled and all cpus running at maximum frequency, tcrypt shows AVX2 performance improvement from 3% for 256 bytes update to 16% for 1024 bytes update over AVX implementation. This patch adds sha1_avx2_transform(), the glue, build and configuration changes needed for AVX2 optimization of SHA1 transform to crypto support. sha1-ssse3 is one module which adds the necessary optimization support (SSSE3/AVX/AVX2) for the low-level SHA1 transform function. With better optimization support, transform function is overridden as the case may be. In the case of AVX2, due to performance reasons across datablock sizes, the AVX or AVX2 transform function is used at run-time as it suits best. The Makefile change therefore appends the necessary objects to the linkage. Due to this, the patch merely appends AVX2 transform to the existing build mix and Kconfig support and leaves the configuration build support as is. Signed-off-by: Chandramouli Narayanan <mouli@linux.intel.com> Reviewed-by: Marek Vasut <marex@denx.de> Acked-by: H. Peter Anvin <hpa@linux.intel.com> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2014-03-20 23:14:00 +01:00
}
crypto: sha1 - SSSE3 based SHA1 implementation for x86-64 This is an assembler implementation of the SHA1 algorithm using the Supplemental SSE3 (SSSE3) instructions or, when available, the Advanced Vector Extensions (AVX). Testing with the tcrypt module shows the raw hash performance is up to 2.3 times faster than the C implementation, using 8k data blocks on a Core 2 Duo T5500. For the smalest data set (16 byte) it is still 25% faster. Since this implementation uses SSE/YMM registers it cannot safely be used in every situation, e.g. while an IRQ interrupts a kernel thread. The implementation falls back to the generic SHA1 variant, if using the SSE/YMM registers is not possible. With this algorithm I was able to increase the throughput of a single IPsec link from 344 Mbit/s to 464 Mbit/s on a Core 2 Quad CPU using the SSSE3 variant -- a speedup of +34.8%. Saving and restoring SSE/YMM state might make the actual throughput fluctuate when there are FPU intensive userland applications running. For example, meassuring the performance using iperf2 directly on the machine under test gives wobbling numbers because iperf2 uses the FPU for each packet to check if the reporting interval has expired (in the above test I got min/max/avg: 402/484/464 MBit/s). Using this algorithm on a IPsec gateway gives much more reasonable and stable numbers, albeit not as high as in the directly connected case. Here is the result from an RFC 2544 test run with a EXFO Packet Blazer FTB-8510: frame size sha1-generic sha1-ssse3 delta 64 byte 37.5 MBit/s 37.5 MBit/s 0.0% 128 byte 56.3 MBit/s 62.5 MBit/s +11.0% 256 byte 87.5 MBit/s 100.0 MBit/s +14.3% 512 byte 131.3 MBit/s 150.0 MBit/s +14.2% 1024 byte 162.5 MBit/s 193.8 MBit/s +19.3% 1280 byte 175.0 MBit/s 212.5 MBit/s +21.4% 1420 byte 175.0 MBit/s 218.7 MBit/s +25.0% 1518 byte 150.0 MBit/s 181.2 MBit/s +20.8% The throughput for the largest frame size is lower than for the previous size because the IP packets need to be fragmented in this case to make there way through the IPsec tunnel. Signed-off-by: Mathias Krause <minipli@googlemail.com> Cc: Maxim Locktyukhin <maxim.locktyukhin@intel.com> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2011-08-04 20:19:25 +02:00
crypto: x86/sha - Restructure x86 sha1 glue code to expose all the available sha1 transforms Restructure the x86 sha1 glue code so we will expose sha1 transforms based on SSSE3, AVX, AVX2 or SHA-NI extension as separate individual drivers when cpu provides such support. This will make it easy for alternative algorithms to be used if desired and makes the code cleaner and easier to maintain. Signed-off-by: Tim Chen <tim.c.chen@linux.intel.com> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2015-09-17 01:34:53 +02:00
if (register_sha1_ni()) {
unregister_sha1_avx2();
unregister_sha1_avx();
unregister_sha1_ssse3();
goto fail;
crypto: sha1 - SSSE3 based SHA1 implementation for x86-64 This is an assembler implementation of the SHA1 algorithm using the Supplemental SSE3 (SSSE3) instructions or, when available, the Advanced Vector Extensions (AVX). Testing with the tcrypt module shows the raw hash performance is up to 2.3 times faster than the C implementation, using 8k data blocks on a Core 2 Duo T5500. For the smalest data set (16 byte) it is still 25% faster. Since this implementation uses SSE/YMM registers it cannot safely be used in every situation, e.g. while an IRQ interrupts a kernel thread. The implementation falls back to the generic SHA1 variant, if using the SSE/YMM registers is not possible. With this algorithm I was able to increase the throughput of a single IPsec link from 344 Mbit/s to 464 Mbit/s on a Core 2 Quad CPU using the SSSE3 variant -- a speedup of +34.8%. Saving and restoring SSE/YMM state might make the actual throughput fluctuate when there are FPU intensive userland applications running. For example, meassuring the performance using iperf2 directly on the machine under test gives wobbling numbers because iperf2 uses the FPU for each packet to check if the reporting interval has expired (in the above test I got min/max/avg: 402/484/464 MBit/s). Using this algorithm on a IPsec gateway gives much more reasonable and stable numbers, albeit not as high as in the directly connected case. Here is the result from an RFC 2544 test run with a EXFO Packet Blazer FTB-8510: frame size sha1-generic sha1-ssse3 delta 64 byte 37.5 MBit/s 37.5 MBit/s 0.0% 128 byte 56.3 MBit/s 62.5 MBit/s +11.0% 256 byte 87.5 MBit/s 100.0 MBit/s +14.3% 512 byte 131.3 MBit/s 150.0 MBit/s +14.2% 1024 byte 162.5 MBit/s 193.8 MBit/s +19.3% 1280 byte 175.0 MBit/s 212.5 MBit/s +21.4% 1420 byte 175.0 MBit/s 218.7 MBit/s +25.0% 1518 byte 150.0 MBit/s 181.2 MBit/s +20.8% The throughput for the largest frame size is lower than for the previous size because the IP packets need to be fragmented in this case to make there way through the IPsec tunnel. Signed-off-by: Mathias Krause <minipli@googlemail.com> Cc: Maxim Locktyukhin <maxim.locktyukhin@intel.com> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2011-08-04 20:19:25 +02:00
}
crypto: x86/sha - Restructure x86 sha1 glue code to expose all the available sha1 transforms Restructure the x86 sha1 glue code so we will expose sha1 transforms based on SSSE3, AVX, AVX2 or SHA-NI extension as separate individual drivers when cpu provides such support. This will make it easy for alternative algorithms to be used if desired and makes the code cleaner and easier to maintain. Signed-off-by: Tim Chen <tim.c.chen@linux.intel.com> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2015-09-17 01:34:53 +02:00
return 0;
fail:
crypto: sha1 - SSSE3 based SHA1 implementation for x86-64 This is an assembler implementation of the SHA1 algorithm using the Supplemental SSE3 (SSSE3) instructions or, when available, the Advanced Vector Extensions (AVX). Testing with the tcrypt module shows the raw hash performance is up to 2.3 times faster than the C implementation, using 8k data blocks on a Core 2 Duo T5500. For the smalest data set (16 byte) it is still 25% faster. Since this implementation uses SSE/YMM registers it cannot safely be used in every situation, e.g. while an IRQ interrupts a kernel thread. The implementation falls back to the generic SHA1 variant, if using the SSE/YMM registers is not possible. With this algorithm I was able to increase the throughput of a single IPsec link from 344 Mbit/s to 464 Mbit/s on a Core 2 Quad CPU using the SSSE3 variant -- a speedup of +34.8%. Saving and restoring SSE/YMM state might make the actual throughput fluctuate when there are FPU intensive userland applications running. For example, meassuring the performance using iperf2 directly on the machine under test gives wobbling numbers because iperf2 uses the FPU for each packet to check if the reporting interval has expired (in the above test I got min/max/avg: 402/484/464 MBit/s). Using this algorithm on a IPsec gateway gives much more reasonable and stable numbers, albeit not as high as in the directly connected case. Here is the result from an RFC 2544 test run with a EXFO Packet Blazer FTB-8510: frame size sha1-generic sha1-ssse3 delta 64 byte 37.5 MBit/s 37.5 MBit/s 0.0% 128 byte 56.3 MBit/s 62.5 MBit/s +11.0% 256 byte 87.5 MBit/s 100.0 MBit/s +14.3% 512 byte 131.3 MBit/s 150.0 MBit/s +14.2% 1024 byte 162.5 MBit/s 193.8 MBit/s +19.3% 1280 byte 175.0 MBit/s 212.5 MBit/s +21.4% 1420 byte 175.0 MBit/s 218.7 MBit/s +25.0% 1518 byte 150.0 MBit/s 181.2 MBit/s +20.8% The throughput for the largest frame size is lower than for the previous size because the IP packets need to be fragmented in this case to make there way through the IPsec tunnel. Signed-off-by: Mathias Krause <minipli@googlemail.com> Cc: Maxim Locktyukhin <maxim.locktyukhin@intel.com> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2011-08-04 20:19:25 +02:00
return -ENODEV;
}
static void __exit sha1_ssse3_mod_fini(void)
{
crypto: x86/sha - Restructure x86 sha1 glue code to expose all the available sha1 transforms Restructure the x86 sha1 glue code so we will expose sha1 transforms based on SSSE3, AVX, AVX2 or SHA-NI extension as separate individual drivers when cpu provides such support. This will make it easy for alternative algorithms to be used if desired and makes the code cleaner and easier to maintain. Signed-off-by: Tim Chen <tim.c.chen@linux.intel.com> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2015-09-17 01:34:53 +02:00
unregister_sha1_ni();
unregister_sha1_avx2();
unregister_sha1_avx();
unregister_sha1_ssse3();
crypto: sha1 - SSSE3 based SHA1 implementation for x86-64 This is an assembler implementation of the SHA1 algorithm using the Supplemental SSE3 (SSSE3) instructions or, when available, the Advanced Vector Extensions (AVX). Testing with the tcrypt module shows the raw hash performance is up to 2.3 times faster than the C implementation, using 8k data blocks on a Core 2 Duo T5500. For the smalest data set (16 byte) it is still 25% faster. Since this implementation uses SSE/YMM registers it cannot safely be used in every situation, e.g. while an IRQ interrupts a kernel thread. The implementation falls back to the generic SHA1 variant, if using the SSE/YMM registers is not possible. With this algorithm I was able to increase the throughput of a single IPsec link from 344 Mbit/s to 464 Mbit/s on a Core 2 Quad CPU using the SSSE3 variant -- a speedup of +34.8%. Saving and restoring SSE/YMM state might make the actual throughput fluctuate when there are FPU intensive userland applications running. For example, meassuring the performance using iperf2 directly on the machine under test gives wobbling numbers because iperf2 uses the FPU for each packet to check if the reporting interval has expired (in the above test I got min/max/avg: 402/484/464 MBit/s). Using this algorithm on a IPsec gateway gives much more reasonable and stable numbers, albeit not as high as in the directly connected case. Here is the result from an RFC 2544 test run with a EXFO Packet Blazer FTB-8510: frame size sha1-generic sha1-ssse3 delta 64 byte 37.5 MBit/s 37.5 MBit/s 0.0% 128 byte 56.3 MBit/s 62.5 MBit/s +11.0% 256 byte 87.5 MBit/s 100.0 MBit/s +14.3% 512 byte 131.3 MBit/s 150.0 MBit/s +14.2% 1024 byte 162.5 MBit/s 193.8 MBit/s +19.3% 1280 byte 175.0 MBit/s 212.5 MBit/s +21.4% 1420 byte 175.0 MBit/s 218.7 MBit/s +25.0% 1518 byte 150.0 MBit/s 181.2 MBit/s +20.8% The throughput for the largest frame size is lower than for the previous size because the IP packets need to be fragmented in this case to make there way through the IPsec tunnel. Signed-off-by: Mathias Krause <minipli@googlemail.com> Cc: Maxim Locktyukhin <maxim.locktyukhin@intel.com> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2011-08-04 20:19:25 +02:00
}
module_init(sha1_ssse3_mod_init);
module_exit(sha1_ssse3_mod_fini);
MODULE_LICENSE("GPL");
MODULE_DESCRIPTION("SHA1 Secure Hash Algorithm, Supplemental SSE3 accelerated");
crypto: prefix module autoloading with "crypto-" This prefixes all crypto module loading with "crypto-" so we never run the risk of exposing module auto-loading to userspace via a crypto API, as demonstrated by Mathias Krause: https://lkml.org/lkml/2013/3/4/70 Signed-off-by: Kees Cook <keescook@chromium.org> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2014-11-21 02:05:53 +01:00
MODULE_ALIAS_CRYPTO("sha1");
crypto: sha-ssse3 - add MODULE_ALIAS Add the MODULE_ALIAS for the cra_driver_name of the different ciphers to allow an automated loading if a driver name is used. Signed-off-by: Stephan Mueller <smueller@chronox.de> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2016-05-13 14:02:00 +02:00
MODULE_ALIAS_CRYPTO("sha1-ssse3");
MODULE_ALIAS_CRYPTO("sha1-avx");
MODULE_ALIAS_CRYPTO("sha1-avx2");
#ifdef CONFIG_AS_SHA1_NI
MODULE_ALIAS_CRYPTO("sha1-ni");
#endif