diff options
| author | Eric Biggers <ebiggers@kernel.org> | 2025-08-05 15:28:53 -0700 |
|---|---|---|
| committer | Eric Biggers <ebiggers@kernel.org> | 2025-08-26 12:52:28 -0400 |
| commit | a1848f6e382145e0200843549f24f5af4b5c8136 (patch) | |
| tree | d6f657a24ecf402db112399d3adb0bd477b79cd9 /lib/crypto/sparc | |
| parent | 09371e1349c9bb34ac030973c7867016a8a8914d (diff) | |
lib/crypto: sparc/md5: Migrate optimized code into library
Instead of exposing the sparc-optimized MD5 code via sparc-specific
crypto_shash algorithms, instead just implement the md5_blocks() library
function. This is much simpler, it makes the MD5 library functions be
sparc-optimized, and it fixes the longstanding issue where the
sparc-optimized MD5 code was disabled by default. MD5 still remains
available through crypto_shash, but individual architectures no longer
need to handle it.
Note: to see the diff from arch/sparc/crypto/md5_glue.c to
lib/crypto/sparc/md5.h, view this commit with 'git show -M10'.
Link: https://lore.kernel.org/r/20250805222855.10362-6-ebiggers@kernel.org
Signed-off-by: Eric Biggers <ebiggers@kernel.org>
Diffstat (limited to 'lib/crypto/sparc')
| -rw-r--r-- | lib/crypto/sparc/md5.h | 48 | ||||
| -rw-r--r-- | lib/crypto/sparc/md5_asm.S | 70 |
2 files changed, 118 insertions, 0 deletions
diff --git a/lib/crypto/sparc/md5.h b/lib/crypto/sparc/md5.h new file mode 100644 index 000000000000..3f1b0ed8c0b3 --- /dev/null +++ b/lib/crypto/sparc/md5.h @@ -0,0 +1,48 @@ +/* SPDX-License-Identifier: GPL-2.0-only */ +/* + * MD5 accelerated using the sparc64 crypto opcodes + * + * 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> + * Copyright (c) Cryptoapi developers. + * Copyright (c) 2002 James Morris <jmorris@intercode.com.au> + */ + +#include <asm/elf.h> +#include <asm/opcodes.h> +#include <asm/pstate.h> + +static __ro_after_init DEFINE_STATIC_KEY_FALSE(have_md5_opcodes); + +asmlinkage void md5_sparc64_transform(struct md5_block_state *state, + const u8 *data, size_t nblocks); + +static void md5_blocks(struct md5_block_state *state, + const u8 *data, size_t nblocks) +{ + if (static_branch_likely(&have_md5_opcodes)) { + cpu_to_le32_array(state->h, ARRAY_SIZE(state->h)); + md5_sparc64_transform(state, data, nblocks); + le32_to_cpu_array(state->h, ARRAY_SIZE(state->h)); + } else { + md5_blocks_generic(state, data, nblocks); + } +} + +#define md5_mod_init_arch md5_mod_init_arch +static inline void md5_mod_init_arch(void) +{ + unsigned long cfr; + + if (!(sparc64_elf_hwcap & HWCAP_SPARC_CRYPTO)) + return; + + __asm__ __volatile__("rd %%asr26, %0" : "=r" (cfr)); + if (!(cfr & CFR_MD5)) + return; + + static_branch_enable(&have_md5_opcodes); + pr_info("Using sparc64 md5 opcode optimized MD5 implementation\n"); +} diff --git a/lib/crypto/sparc/md5_asm.S b/lib/crypto/sparc/md5_asm.S new file mode 100644 index 000000000000..60b544e4d205 --- /dev/null +++ b/lib/crypto/sparc/md5_asm.S @@ -0,0 +1,70 @@ +/* SPDX-License-Identifier: GPL-2.0 */ +#include <linux/linkage.h> +#include <asm/opcodes.h> +#include <asm/visasm.h> + +ENTRY(md5_sparc64_transform) + /* %o0 = digest, %o1 = data, %o2 = rounds */ + VISEntryHalf + ld [%o0 + 0x00], %f0 + ld [%o0 + 0x04], %f1 + andcc %o1, 0x7, %g0 + ld [%o0 + 0x08], %f2 + bne,pn %xcc, 10f + ld [%o0 + 0x0c], %f3 + +1: + ldd [%o1 + 0x00], %f8 + ldd [%o1 + 0x08], %f10 + ldd [%o1 + 0x10], %f12 + ldd [%o1 + 0x18], %f14 + ldd [%o1 + 0x20], %f16 + ldd [%o1 + 0x28], %f18 + ldd [%o1 + 0x30], %f20 + ldd [%o1 + 0x38], %f22 + + MD5 + + subcc %o2, 1, %o2 + bne,pt %xcc, 1b + add %o1, 0x40, %o1 + +5: + st %f0, [%o0 + 0x00] + st %f1, [%o0 + 0x04] + st %f2, [%o0 + 0x08] + st %f3, [%o0 + 0x0c] + retl + VISExitHalf +10: + alignaddr %o1, %g0, %o1 + + ldd [%o1 + 0x00], %f10 +1: + ldd [%o1 + 0x08], %f12 + ldd [%o1 + 0x10], %f14 + ldd [%o1 + 0x18], %f16 + ldd [%o1 + 0x20], %f18 + ldd [%o1 + 0x28], %f20 + ldd [%o1 + 0x30], %f22 + ldd [%o1 + 0x38], %f24 + ldd [%o1 + 0x40], %f26 + + faligndata %f10, %f12, %f8 + faligndata %f12, %f14, %f10 + faligndata %f14, %f16, %f12 + faligndata %f16, %f18, %f14 + faligndata %f18, %f20, %f16 + faligndata %f20, %f22, %f18 + faligndata %f22, %f24, %f20 + faligndata %f24, %f26, %f22 + + MD5 + + subcc %o2, 1, %o2 + fsrc2 %f26, %f10 + bne,pt %xcc, 1b + add %o1, 0x40, %o1 + + ba,a,pt %xcc, 5b +ENDPROC(md5_sparc64_transform) |