WSL2-Linux-Kernel/arch/arm/crypto/curve25519-glue.c

137 строки
3.6 KiB
C
Исходник Обычный вид История

crypto: arm/curve25519 - wire up NEON implementation This ports the SUPERCOP implementation for usage in kernel space. In addition to the usual header, macro, and style changes required for kernel space, it makes a few small changes to the code: - The stack alignment is relaxed to 16 bytes. - Superfluous mov statements have been removed. - ldr for constants has been replaced with movw. - ldreq has been replaced with moveq. - The str epilogue has been made more idiomatic. - SIMD registers are not pushed and popped at the beginning and end. - The prologue and epilogue have been made idiomatic. - A hole has been removed from the stack, saving 32 bytes. - We write-back the base register whenever possible for vld1.8. - Some multiplications have been reordered for better A7 performance. There are more opportunities for cleanup, since this code is from qhasm, which doesn't always do the most opportune thing. But even prior to extensive hand optimizations, this code delivers significant performance improvements (given in get_cycles() per call): ----------- ------------- | generic C | this commit | ------------ ----------- ------------- | Cortex-A7 | 49136 | 22395 | ------------ ----------- ------------- | Cortex-A17 | 17326 | 4983 | ------------ ----------- ------------- Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com> [ardb: - move to arch/arm/crypto - wire into lib/crypto framework - implement crypto API KPP hooks ] Signed-off-by: Ard Biesheuvel <ardb@kernel.org> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2019-11-08 15:22:38 +03:00
// SPDX-License-Identifier: GPL-2.0 OR MIT
/*
* Copyright (C) 2015-2019 Jason A. Donenfeld <Jason@zx2c4.com>. All Rights Reserved.
*
* Based on public domain code from Daniel J. Bernstein and Peter Schwabe. This
* began from SUPERCOP's curve25519/neon2/scalarmult.s, but has subsequently been
* manually reworked for use in kernel space.
*/
#include <asm/hwcap.h>
#include <asm/neon.h>
#include <asm/simd.h>
#include <crypto/internal/kpp.h>
#include <crypto/internal/simd.h>
#include <linux/types.h>
#include <linux/module.h>
#include <linux/init.h>
#include <linux/jump_label.h>
#include <linux/scatterlist.h>
crypto: arm/curve25519 - wire up NEON implementation This ports the SUPERCOP implementation for usage in kernel space. In addition to the usual header, macro, and style changes required for kernel space, it makes a few small changes to the code: - The stack alignment is relaxed to 16 bytes. - Superfluous mov statements have been removed. - ldr for constants has been replaced with movw. - ldreq has been replaced with moveq. - The str epilogue has been made more idiomatic. - SIMD registers are not pushed and popped at the beginning and end. - The prologue and epilogue have been made idiomatic. - A hole has been removed from the stack, saving 32 bytes. - We write-back the base register whenever possible for vld1.8. - Some multiplications have been reordered for better A7 performance. There are more opportunities for cleanup, since this code is from qhasm, which doesn't always do the most opportune thing. But even prior to extensive hand optimizations, this code delivers significant performance improvements (given in get_cycles() per call): ----------- ------------- | generic C | this commit | ------------ ----------- ------------- | Cortex-A7 | 49136 | 22395 | ------------ ----------- ------------- | Cortex-A17 | 17326 | 4983 | ------------ ----------- ------------- Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com> [ardb: - move to arch/arm/crypto - wire into lib/crypto framework - implement crypto API KPP hooks ] Signed-off-by: Ard Biesheuvel <ardb@kernel.org> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2019-11-08 15:22:38 +03:00
#include <crypto/curve25519.h>
asmlinkage void curve25519_neon(u8 mypublic[CURVE25519_KEY_SIZE],
const u8 secret[CURVE25519_KEY_SIZE],
const u8 basepoint[CURVE25519_KEY_SIZE]);
static __ro_after_init DEFINE_STATIC_KEY_FALSE(have_neon);
void curve25519_arch(u8 out[CURVE25519_KEY_SIZE],
const u8 scalar[CURVE25519_KEY_SIZE],
const u8 point[CURVE25519_KEY_SIZE])
{
if (static_branch_likely(&have_neon) && crypto_simd_usable()) {
kernel_neon_begin();
curve25519_neon(out, scalar, point);
kernel_neon_end();
} else {
curve25519_generic(out, scalar, point);
}
}
EXPORT_SYMBOL(curve25519_arch);
void curve25519_base_arch(u8 pub[CURVE25519_KEY_SIZE],
const u8 secret[CURVE25519_KEY_SIZE])
{
return curve25519_arch(pub, secret, curve25519_base_point);
}
EXPORT_SYMBOL(curve25519_base_arch);
crypto: arm/curve25519 - wire up NEON implementation This ports the SUPERCOP implementation for usage in kernel space. In addition to the usual header, macro, and style changes required for kernel space, it makes a few small changes to the code: - The stack alignment is relaxed to 16 bytes. - Superfluous mov statements have been removed. - ldr for constants has been replaced with movw. - ldreq has been replaced with moveq. - The str epilogue has been made more idiomatic. - SIMD registers are not pushed and popped at the beginning and end. - The prologue and epilogue have been made idiomatic. - A hole has been removed from the stack, saving 32 bytes. - We write-back the base register whenever possible for vld1.8. - Some multiplications have been reordered for better A7 performance. There are more opportunities for cleanup, since this code is from qhasm, which doesn't always do the most opportune thing. But even prior to extensive hand optimizations, this code delivers significant performance improvements (given in get_cycles() per call): ----------- ------------- | generic C | this commit | ------------ ----------- ------------- | Cortex-A7 | 49136 | 22395 | ------------ ----------- ------------- | Cortex-A17 | 17326 | 4983 | ------------ ----------- ------------- Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com> [ardb: - move to arch/arm/crypto - wire into lib/crypto framework - implement crypto API KPP hooks ] Signed-off-by: Ard Biesheuvel <ardb@kernel.org> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2019-11-08 15:22:38 +03:00
static int curve25519_set_secret(struct crypto_kpp *tfm, const void *buf,
unsigned int len)
{
u8 *secret = kpp_tfm_ctx(tfm);
if (!len)
curve25519_generate_secret(secret);
else if (len == CURVE25519_KEY_SIZE &&
crypto_memneq(buf, curve25519_null_point, CURVE25519_KEY_SIZE))
memcpy(secret, buf, CURVE25519_KEY_SIZE);
else
return -EINVAL;
return 0;
}
static int curve25519_compute_value(struct kpp_request *req)
{
struct crypto_kpp *tfm = crypto_kpp_reqtfm(req);
const u8 *secret = kpp_tfm_ctx(tfm);
u8 public_key[CURVE25519_KEY_SIZE];
u8 buf[CURVE25519_KEY_SIZE];
int copied, nbytes;
u8 const *bp;
if (req->src) {
copied = sg_copy_to_buffer(req->src,
sg_nents_for_len(req->src,
CURVE25519_KEY_SIZE),
public_key, CURVE25519_KEY_SIZE);
if (copied != CURVE25519_KEY_SIZE)
return -EINVAL;
bp = public_key;
} else {
bp = curve25519_base_point;
}
curve25519_arch(buf, secret, bp);
/* might want less than we've got */
nbytes = min_t(size_t, CURVE25519_KEY_SIZE, req->dst_len);
copied = sg_copy_from_buffer(req->dst, sg_nents_for_len(req->dst,
nbytes),
buf, nbytes);
if (copied != nbytes)
return -EINVAL;
return 0;
}
static unsigned int curve25519_max_size(struct crypto_kpp *tfm)
{
return CURVE25519_KEY_SIZE;
}
static struct kpp_alg curve25519_alg = {
.base.cra_name = "curve25519",
.base.cra_driver_name = "curve25519-neon",
.base.cra_priority = 200,
.base.cra_module = THIS_MODULE,
.base.cra_ctxsize = CURVE25519_KEY_SIZE,
.set_secret = curve25519_set_secret,
.generate_public_key = curve25519_compute_value,
.compute_shared_secret = curve25519_compute_value,
.max_size = curve25519_max_size,
};
static int __init arm_curve25519_init(void)
crypto: arm/curve25519 - wire up NEON implementation This ports the SUPERCOP implementation for usage in kernel space. In addition to the usual header, macro, and style changes required for kernel space, it makes a few small changes to the code: - The stack alignment is relaxed to 16 bytes. - Superfluous mov statements have been removed. - ldr for constants has been replaced with movw. - ldreq has been replaced with moveq. - The str epilogue has been made more idiomatic. - SIMD registers are not pushed and popped at the beginning and end. - The prologue and epilogue have been made idiomatic. - A hole has been removed from the stack, saving 32 bytes. - We write-back the base register whenever possible for vld1.8. - Some multiplications have been reordered for better A7 performance. There are more opportunities for cleanup, since this code is from qhasm, which doesn't always do the most opportune thing. But even prior to extensive hand optimizations, this code delivers significant performance improvements (given in get_cycles() per call): ----------- ------------- | generic C | this commit | ------------ ----------- ------------- | Cortex-A7 | 49136 | 22395 | ------------ ----------- ------------- | Cortex-A17 | 17326 | 4983 | ------------ ----------- ------------- Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com> [ardb: - move to arch/arm/crypto - wire into lib/crypto framework - implement crypto API KPP hooks ] Signed-off-by: Ard Biesheuvel <ardb@kernel.org> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2019-11-08 15:22:38 +03:00
{
if (elf_hwcap & HWCAP_NEON) {
static_branch_enable(&have_neon);
return IS_REACHABLE(CONFIG_CRYPTO_KPP) ?
crypto_register_kpp(&curve25519_alg) : 0;
crypto: arm/curve25519 - wire up NEON implementation This ports the SUPERCOP implementation for usage in kernel space. In addition to the usual header, macro, and style changes required for kernel space, it makes a few small changes to the code: - The stack alignment is relaxed to 16 bytes. - Superfluous mov statements have been removed. - ldr for constants has been replaced with movw. - ldreq has been replaced with moveq. - The str epilogue has been made more idiomatic. - SIMD registers are not pushed and popped at the beginning and end. - The prologue and epilogue have been made idiomatic. - A hole has been removed from the stack, saving 32 bytes. - We write-back the base register whenever possible for vld1.8. - Some multiplications have been reordered for better A7 performance. There are more opportunities for cleanup, since this code is from qhasm, which doesn't always do the most opportune thing. But even prior to extensive hand optimizations, this code delivers significant performance improvements (given in get_cycles() per call): ----------- ------------- | generic C | this commit | ------------ ----------- ------------- | Cortex-A7 | 49136 | 22395 | ------------ ----------- ------------- | Cortex-A17 | 17326 | 4983 | ------------ ----------- ------------- Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com> [ardb: - move to arch/arm/crypto - wire into lib/crypto framework - implement crypto API KPP hooks ] Signed-off-by: Ard Biesheuvel <ardb@kernel.org> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2019-11-08 15:22:38 +03:00
}
return 0;
}
static void __exit arm_curve25519_exit(void)
crypto: arm/curve25519 - wire up NEON implementation This ports the SUPERCOP implementation for usage in kernel space. In addition to the usual header, macro, and style changes required for kernel space, it makes a few small changes to the code: - The stack alignment is relaxed to 16 bytes. - Superfluous mov statements have been removed. - ldr for constants has been replaced with movw. - ldreq has been replaced with moveq. - The str epilogue has been made more idiomatic. - SIMD registers are not pushed and popped at the beginning and end. - The prologue and epilogue have been made idiomatic. - A hole has been removed from the stack, saving 32 bytes. - We write-back the base register whenever possible for vld1.8. - Some multiplications have been reordered for better A7 performance. There are more opportunities for cleanup, since this code is from qhasm, which doesn't always do the most opportune thing. But even prior to extensive hand optimizations, this code delivers significant performance improvements (given in get_cycles() per call): ----------- ------------- | generic C | this commit | ------------ ----------- ------------- | Cortex-A7 | 49136 | 22395 | ------------ ----------- ------------- | Cortex-A17 | 17326 | 4983 | ------------ ----------- ------------- Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com> [ardb: - move to arch/arm/crypto - wire into lib/crypto framework - implement crypto API KPP hooks ] Signed-off-by: Ard Biesheuvel <ardb@kernel.org> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2019-11-08 15:22:38 +03:00
{
if (IS_REACHABLE(CONFIG_CRYPTO_KPP) && elf_hwcap & HWCAP_NEON)
crypto: arm/curve25519 - wire up NEON implementation This ports the SUPERCOP implementation for usage in kernel space. In addition to the usual header, macro, and style changes required for kernel space, it makes a few small changes to the code: - The stack alignment is relaxed to 16 bytes. - Superfluous mov statements have been removed. - ldr for constants has been replaced with movw. - ldreq has been replaced with moveq. - The str epilogue has been made more idiomatic. - SIMD registers are not pushed and popped at the beginning and end. - The prologue and epilogue have been made idiomatic. - A hole has been removed from the stack, saving 32 bytes. - We write-back the base register whenever possible for vld1.8. - Some multiplications have been reordered for better A7 performance. There are more opportunities for cleanup, since this code is from qhasm, which doesn't always do the most opportune thing. But even prior to extensive hand optimizations, this code delivers significant performance improvements (given in get_cycles() per call): ----------- ------------- | generic C | this commit | ------------ ----------- ------------- | Cortex-A7 | 49136 | 22395 | ------------ ----------- ------------- | Cortex-A17 | 17326 | 4983 | ------------ ----------- ------------- Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com> [ardb: - move to arch/arm/crypto - wire into lib/crypto framework - implement crypto API KPP hooks ] Signed-off-by: Ard Biesheuvel <ardb@kernel.org> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2019-11-08 15:22:38 +03:00
crypto_unregister_kpp(&curve25519_alg);
}
module_init(arm_curve25519_init);
module_exit(arm_curve25519_exit);
crypto: arm/curve25519 - wire up NEON implementation This ports the SUPERCOP implementation for usage in kernel space. In addition to the usual header, macro, and style changes required for kernel space, it makes a few small changes to the code: - The stack alignment is relaxed to 16 bytes. - Superfluous mov statements have been removed. - ldr for constants has been replaced with movw. - ldreq has been replaced with moveq. - The str epilogue has been made more idiomatic. - SIMD registers are not pushed and popped at the beginning and end. - The prologue and epilogue have been made idiomatic. - A hole has been removed from the stack, saving 32 bytes. - We write-back the base register whenever possible for vld1.8. - Some multiplications have been reordered for better A7 performance. There are more opportunities for cleanup, since this code is from qhasm, which doesn't always do the most opportune thing. But even prior to extensive hand optimizations, this code delivers significant performance improvements (given in get_cycles() per call): ----------- ------------- | generic C | this commit | ------------ ----------- ------------- | Cortex-A7 | 49136 | 22395 | ------------ ----------- ------------- | Cortex-A17 | 17326 | 4983 | ------------ ----------- ------------- Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com> [ardb: - move to arch/arm/crypto - wire into lib/crypto framework - implement crypto API KPP hooks ] Signed-off-by: Ard Biesheuvel <ardb@kernel.org> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2019-11-08 15:22:38 +03:00
MODULE_ALIAS_CRYPTO("curve25519");
MODULE_ALIAS_CRYPTO("curve25519-neon");
MODULE_LICENSE("GPL v2");