linux/lib/crypto/s390/sha3.h
Eric Biggers 862445d3b9 lib/crypto: s390/sha3: Add optimized one-shot SHA-3 digest functions
Some z/Architecture processors can compute a SHA-3 digest in a single
instruction.  arch/s390/crypto/ already uses this capability to optimize
the SHA-3 crypto_shash algorithms.

Use this capability to implement the sha3_224(), sha3_256(), sha3_384(),
and sha3_512() library functions too.

SHA3-256 benchmark results provided by Harald Freudenberger
(https://lore.kernel.org/r/4188d18bfcc8a64941c5ebd8de10ede2@linux.ibm.com/)
on a z/Architecture machine with "facility 86" (MSA level 12):

    Length (bytes)    Before (MB/s)   After (MB/s)
    ==============    =============   ============
          16                212             225
          64                820             915
         256               1850            3350
        1024               5400            8300
        4096              11200           11300

Note: the original data from Harald was given in the form of a graph for
each length, showing the distribution of throughputs from 500 runs.  I
guesstimated the peak of each one.

Harald also reported that the generic SHA-3 code was at most 259 MB/s
(https://lore.kernel.org/r/c39f6b6c110def0095e5da5becc12085@linux.ibm.com/).
So as expected, the earlier commit that optimized sha3_absorb_blocks()
and sha3_keccakf() is the more important one; it optimized the Keccak
permutation which is the most performance-critical part of SHA-3.
Still, this additional commit does notably improve performance further
on some lengths.

Reviewed-by: Ard Biesheuvel <ardb@kernel.org>
Tested-by: Harald Freudenberger <freude@linux.ibm.com>
Link: https://lore.kernel.org/r/20251026055032.1413733-13-ebiggers@kernel.org
Signed-off-by: Eric Biggers <ebiggers@kernel.org>
2025-11-05 20:30:41 -08:00

151 lines
4.1 KiB
C

/* SPDX-License-Identifier: GPL-2.0-or-later */
/*
* SHA-3 optimized using the CP Assist for Cryptographic Functions (CPACF)
*
* Copyright 2025 Google LLC
*/
#include <asm/cpacf.h>
#include <linux/cpufeature.h>
static __ro_after_init DEFINE_STATIC_KEY_FALSE(have_sha3);
static __ro_after_init DEFINE_STATIC_KEY_FALSE(have_sha3_init_optim);
static void sha3_absorb_blocks(struct sha3_state *state, const u8 *data,
size_t nblocks, size_t block_size)
{
if (static_branch_likely(&have_sha3)) {
/*
* Note that KIMD assumes little-endian order of the state
* words. sha3_state already uses that order, though, so
* there's no need for a byteswap.
*/
switch (block_size) {
case SHA3_224_BLOCK_SIZE:
cpacf_kimd(CPACF_KIMD_SHA3_224, state,
data, nblocks * block_size);
return;
case SHA3_256_BLOCK_SIZE:
/*
* This case handles both SHA3-256 and SHAKE256, since
* they have the same block size.
*/
cpacf_kimd(CPACF_KIMD_SHA3_256, state,
data, nblocks * block_size);
return;
case SHA3_384_BLOCK_SIZE:
cpacf_kimd(CPACF_KIMD_SHA3_384, state,
data, nblocks * block_size);
return;
case SHA3_512_BLOCK_SIZE:
cpacf_kimd(CPACF_KIMD_SHA3_512, state,
data, nblocks * block_size);
return;
}
}
sha3_absorb_blocks_generic(state, data, nblocks, block_size);
}
static void sha3_keccakf(struct sha3_state *state)
{
if (static_branch_likely(&have_sha3)) {
/*
* Passing zeroes into any of CPACF_KIMD_SHA3_* gives the plain
* Keccak-f permutation, which is what we want here. Use
* SHA3-512 since it has the smallest block size.
*/
static const u8 zeroes[SHA3_512_BLOCK_SIZE];
cpacf_kimd(CPACF_KIMD_SHA3_512, state, zeroes, sizeof(zeroes));
} else {
sha3_keccakf_generic(state);
}
}
static inline bool s390_sha3(int func, const u8 *in, size_t in_len,
u8 *out, size_t out_len)
{
struct sha3_state state;
if (!static_branch_likely(&have_sha3))
return false;
if (static_branch_likely(&have_sha3_init_optim))
func |= CPACF_KLMD_NIP | CPACF_KLMD_DUFOP;
else
memset(&state, 0, sizeof(state));
cpacf_klmd(func, &state, in, in_len);
if (static_branch_likely(&have_sha3_init_optim))
kmsan_unpoison_memory(&state, out_len);
memcpy(out, &state, out_len);
memzero_explicit(&state, sizeof(state));
return true;
}
#define sha3_224_arch sha3_224_arch
static bool sha3_224_arch(const u8 *in, size_t in_len,
u8 out[SHA3_224_DIGEST_SIZE])
{
return s390_sha3(CPACF_KLMD_SHA3_224, in, in_len,
out, SHA3_224_DIGEST_SIZE);
}
#define sha3_256_arch sha3_256_arch
static bool sha3_256_arch(const u8 *in, size_t in_len,
u8 out[SHA3_256_DIGEST_SIZE])
{
return s390_sha3(CPACF_KLMD_SHA3_256, in, in_len,
out, SHA3_256_DIGEST_SIZE);
}
#define sha3_384_arch sha3_384_arch
static bool sha3_384_arch(const u8 *in, size_t in_len,
u8 out[SHA3_384_DIGEST_SIZE])
{
return s390_sha3(CPACF_KLMD_SHA3_384, in, in_len,
out, SHA3_384_DIGEST_SIZE);
}
#define sha3_512_arch sha3_512_arch
static bool sha3_512_arch(const u8 *in, size_t in_len,
u8 out[SHA3_512_DIGEST_SIZE])
{
return s390_sha3(CPACF_KLMD_SHA3_512, in, in_len,
out, SHA3_512_DIGEST_SIZE);
}
#define sha3_mod_init_arch sha3_mod_init_arch
static void sha3_mod_init_arch(void)
{
int num_present = 0;
int num_possible = 0;
if (!cpu_have_feature(S390_CPU_FEATURE_MSA))
return;
/*
* Since all the SHA-3 functions are in Message-Security-Assist
* Extension 6, just treat them as all or nothing. This way we need
* only one static_key.
*/
#define QUERY(opcode, func) \
({ num_present += !!cpacf_query_func(opcode, func); num_possible++; })
QUERY(CPACF_KIMD, CPACF_KIMD_SHA3_224);
QUERY(CPACF_KIMD, CPACF_KIMD_SHA3_256);
QUERY(CPACF_KIMD, CPACF_KIMD_SHA3_384);
QUERY(CPACF_KIMD, CPACF_KIMD_SHA3_512);
QUERY(CPACF_KLMD, CPACF_KLMD_SHA3_224);
QUERY(CPACF_KLMD, CPACF_KLMD_SHA3_256);
QUERY(CPACF_KLMD, CPACF_KLMD_SHA3_384);
QUERY(CPACF_KLMD, CPACF_KLMD_SHA3_512);
#undef QUERY
if (num_present == num_possible) {
static_branch_enable(&have_sha3);
if (test_facility(86))
static_branch_enable(&have_sha3_init_optim);
} else if (num_present != 0) {
pr_warn("Unsupported combination of SHA-3 facilities\n");
}
}