linux/arch/arm/crypto/crct10dif-ce-core.S

382 lines
12 KiB
ArmAsm

//
// Accelerated CRC-T10DIF using ARM NEON and Crypto Extensions instructions
//
// Copyright (C) 2016 Linaro Ltd <ard.biesheuvel@linaro.org>
// Copyright (C) 2019 Google LLC <ebiggers@google.com>
//
// This program is free software; you can redistribute it and/or modify
// it under the terms of the GNU General Public License version 2 as
// published by the Free Software Foundation.
//
// Derived from the x86 version:
//
// Implement fast CRC-T10DIF computation with SSE and PCLMULQDQ instructions
//
// Copyright (c) 2013, Intel Corporation
//
// Authors:
// Erdinc Ozturk <erdinc.ozturk@intel.com>
// Vinodh Gopal <vinodh.gopal@intel.com>
// James Guilford <james.guilford@intel.com>
// Tim Chen <tim.c.chen@linux.intel.com>
//
// This software is available to you under a choice of one of two
// licenses. You may choose to be licensed under the terms of the GNU
// General Public License (GPL) Version 2, available from the file
// COPYING in the main directory of this source tree, or the
// OpenIB.org BSD license below:
//
// Redistribution and use in source and binary forms, with or without
// modification, are permitted provided that the following conditions are
// met:
//
// * Redistributions of source code must retain the above copyright
// notice, this list of conditions and the following disclaimer.
//
// * Redistributions in binary form must reproduce the above copyright
// notice, this list of conditions and the following disclaimer in the
// documentation and/or other materials provided with the
// distribution.
//
// * Neither the name of the Intel Corporation nor the names of its
// contributors may be used to endorse or promote products derived from
// this software without specific prior written permission.
//
//
// THIS SOFTWARE IS PROVIDED BY INTEL CORPORATION ""AS IS"" AND ANY
// EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
// IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
// PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL INTEL CORPORATION OR
// CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL,
// EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO,
// PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR
// PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
// LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
// NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
// SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
//
// Reference paper titled "Fast CRC Computation for Generic
// Polynomials Using PCLMULQDQ Instruction"
// URL: http://www.intel.com/content/dam/www/public/us/en/documents
// /white-papers/fast-crc-computation-generic-polynomials-pclmulqdq-paper.pdf
//
#include <linux/linkage.h>
#include <asm/assembler.h>
#ifdef CONFIG_CPU_ENDIAN_BE8
#define CPU_LE(code...)
#else
#define CPU_LE(code...) code
#endif
.text
.arch armv8-a
.fpu crypto-neon-fp-armv8
init_crc .req r0
buf .req r1
len .req r2
fold_consts_ptr .req ip
q0l .req d0
q0h .req d1
q1l .req d2
q1h .req d3
q2l .req d4
q2h .req d5
q3l .req d6
q3h .req d7
q4l .req d8
q4h .req d9
q5l .req d10
q5h .req d11
q6l .req d12
q6h .req d13
q7l .req d14
q7h .req d15
q8l .req d16
q8h .req d17
q9l .req d18
q9h .req d19
q10l .req d20
q10h .req d21
q11l .req d22
q11h .req d23
q12l .req d24
q12h .req d25
FOLD_CONSTS .req q10
FOLD_CONST_L .req q10l
FOLD_CONST_H .req q10h
// Fold reg1, reg2 into the next 32 data bytes, storing the result back
// into reg1, reg2.
.macro fold_32_bytes, reg1, reg2
vld1.64 {q11-q12}, [buf]!
vmull.p64 q8, \reg1\()h, FOLD_CONST_H
vmull.p64 \reg1, \reg1\()l, FOLD_CONST_L
vmull.p64 q9, \reg2\()h, FOLD_CONST_H
vmull.p64 \reg2, \reg2\()l, FOLD_CONST_L
CPU_LE( vrev64.8 q11, q11 )
CPU_LE( vrev64.8 q12, q12 )
vswp q11l, q11h
vswp q12l, q12h
veor.8 \reg1, \reg1, q8
veor.8 \reg2, \reg2, q9
veor.8 \reg1, \reg1, q11
veor.8 \reg2, \reg2, q12
.endm
// Fold src_reg into dst_reg, optionally loading the next fold constants
.macro fold_16_bytes, src_reg, dst_reg, load_next_consts
vmull.p64 q8, \src_reg\()l, FOLD_CONST_L
vmull.p64 \src_reg, \src_reg\()h, FOLD_CONST_H
.ifnb \load_next_consts
vld1.64 {FOLD_CONSTS}, [fold_consts_ptr, :128]!
.endif
veor.8 \dst_reg, \dst_reg, q8
veor.8 \dst_reg, \dst_reg, \src_reg
.endm
.macro __adrl, out, sym
movw \out, #:lower16:\sym
movt \out, #:upper16:\sym
.endm
//
// u16 crc_t10dif_pmull(u16 init_crc, const u8 *buf, size_t len);
//
// Assumes len >= 16.
//
ENTRY(crc_t10dif_pmull)
// For sizes less than 256 bytes, we can't fold 128 bytes at a time.
cmp len, #256
blt .Lless_than_256_bytes
__adrl fold_consts_ptr, .Lfold_across_128_bytes_consts
// Load the first 128 data bytes. Byte swapping is necessary to make
// the bit order match the polynomial coefficient order.
vld1.64 {q0-q1}, [buf]!
vld1.64 {q2-q3}, [buf]!
vld1.64 {q4-q5}, [buf]!
vld1.64 {q6-q7}, [buf]!
CPU_LE( vrev64.8 q0, q0 )
CPU_LE( vrev64.8 q1, q1 )
CPU_LE( vrev64.8 q2, q2 )
CPU_LE( vrev64.8 q3, q3 )
CPU_LE( vrev64.8 q4, q4 )
CPU_LE( vrev64.8 q5, q5 )
CPU_LE( vrev64.8 q6, q6 )
CPU_LE( vrev64.8 q7, q7 )
vswp q0l, q0h
vswp q1l, q1h
vswp q2l, q2h
vswp q3l, q3h
vswp q4l, q4h
vswp q5l, q5h
vswp q6l, q6h
vswp q7l, q7h
// XOR the first 16 data *bits* with the initial CRC value.
vmov.i8 q8h, #0
vmov.u16 q8h[3], init_crc
veor q0h, q0h, q8h
// Load the constants for folding across 128 bytes.
vld1.64 {FOLD_CONSTS}, [fold_consts_ptr, :128]!
// Subtract 128 for the 128 data bytes just consumed. Subtract another
// 128 to simplify the termination condition of the following loop.
sub len, len, #256
// While >= 128 data bytes remain (not counting q0-q7), fold the 128
// bytes q0-q7 into them, storing the result back into q0-q7.
.Lfold_128_bytes_loop:
fold_32_bytes q0, q1
fold_32_bytes q2, q3
fold_32_bytes q4, q5
fold_32_bytes q6, q7
subs len, len, #128
bge .Lfold_128_bytes_loop
// Now fold the 112 bytes in q0-q6 into the 16 bytes in q7.
// Fold across 64 bytes.
vld1.64 {FOLD_CONSTS}, [fold_consts_ptr, :128]!
fold_16_bytes q0, q4
fold_16_bytes q1, q5
fold_16_bytes q2, q6
fold_16_bytes q3, q7, 1
// Fold across 32 bytes.
fold_16_bytes q4, q6
fold_16_bytes q5, q7, 1
// Fold across 16 bytes.
fold_16_bytes q6, q7
// Add 128 to get the correct number of data bytes remaining in 0...127
// (not counting q7), following the previous extra subtraction by 128.
// Then subtract 16 to simplify the termination condition of the
// following loop.
adds len, len, #(128-16)
// While >= 16 data bytes remain (not counting q7), fold the 16 bytes q7
// into them, storing the result back into q7.
blt .Lfold_16_bytes_loop_done
.Lfold_16_bytes_loop:
vmull.p64 q8, q7l, FOLD_CONST_L
vmull.p64 q7, q7h, FOLD_CONST_H
veor.8 q7, q7, q8
vld1.64 {q0}, [buf]!
CPU_LE( vrev64.8 q0, q0 )
vswp q0l, q0h
veor.8 q7, q7, q0
subs len, len, #16
bge .Lfold_16_bytes_loop
.Lfold_16_bytes_loop_done:
// Add 16 to get the correct number of data bytes remaining in 0...15
// (not counting q7), following the previous extra subtraction by 16.
adds len, len, #16
beq .Lreduce_final_16_bytes
.Lhandle_partial_segment:
// Reduce the last '16 + len' bytes where 1 <= len <= 15 and the first
// 16 bytes are in q7 and the rest are the remaining data in 'buf'. To
// do this without needing a fold constant for each possible 'len',
// redivide the bytes into a first chunk of 'len' bytes and a second
// chunk of 16 bytes, then fold the first chunk into the second.
// q0 = last 16 original data bytes
add buf, buf, len
sub buf, buf, #16
vld1.64 {q0}, [buf]
CPU_LE( vrev64.8 q0, q0 )
vswp q0l, q0h
// q1 = high order part of second chunk: q7 left-shifted by 'len' bytes.
__adrl r3, .Lbyteshift_table + 16
sub r3, r3, len
vld1.8 {q2}, [r3]
vtbl.8 q1l, {q7l-q7h}, q2l
vtbl.8 q1h, {q7l-q7h}, q2h
// q3 = first chunk: q7 right-shifted by '16-len' bytes.
vmov.i8 q3, #0x80
veor.8 q2, q2, q3
vtbl.8 q3l, {q7l-q7h}, q2l
vtbl.8 q3h, {q7l-q7h}, q2h
// Convert to 8-bit masks: 'len' 0x00 bytes, then '16-len' 0xff bytes.
vshr.s8 q2, q2, #7
// q2 = second chunk: 'len' bytes from q0 (low-order bytes),
// then '16-len' bytes from q1 (high-order bytes).
vbsl.8 q2, q1, q0
// Fold the first chunk into the second chunk, storing the result in q7.
vmull.p64 q0, q3l, FOLD_CONST_L
vmull.p64 q7, q3h, FOLD_CONST_H
veor.8 q7, q7, q0
veor.8 q7, q7, q2
.Lreduce_final_16_bytes:
// Reduce the 128-bit value M(x), stored in q7, to the final 16-bit CRC.
// Load 'x^48 * (x^48 mod G(x))' and 'x^48 * (x^80 mod G(x))'.
vld1.64 {FOLD_CONSTS}, [fold_consts_ptr, :128]!
// Fold the high 64 bits into the low 64 bits, while also multiplying by
// x^64. This produces a 128-bit value congruent to x^64 * M(x) and
// whose low 48 bits are 0.
vmull.p64 q0, q7h, FOLD_CONST_H // high bits * x^48 * (x^80 mod G(x))
veor.8 q0h, q0h, q7l // + low bits * x^64
// Fold the high 32 bits into the low 96 bits. This produces a 96-bit
// value congruent to x^64 * M(x) and whose low 48 bits are 0.
vmov.i8 q1, #0
vmov s4, s3 // extract high 32 bits
vmov s3, s5 // zero high 32 bits
vmull.p64 q1, q1l, FOLD_CONST_L // high 32 bits * x^48 * (x^48 mod G(x))
veor.8 q0, q0, q1 // + low bits
// Load G(x) and floor(x^48 / G(x)).
vld1.64 {FOLD_CONSTS}, [fold_consts_ptr, :128]
// Use Barrett reduction to compute the final CRC value.
vmull.p64 q1, q0h, FOLD_CONST_H // high 32 bits * floor(x^48 / G(x))
vshr.u64 q1l, q1l, #32 // /= x^32
vmull.p64 q1, q1l, FOLD_CONST_L // *= G(x)
vshr.u64 q0l, q0l, #48
veor.8 q0l, q0l, q1l // + low 16 nonzero bits
// Final CRC value (x^16 * M(x)) mod G(x) is in low 16 bits of q0.
vmov.u16 r0, q0l[0]
bx lr
.Lless_than_256_bytes:
// Checksumming a buffer of length 16...255 bytes
__adrl fold_consts_ptr, .Lfold_across_16_bytes_consts
// Load the first 16 data bytes.
vld1.64 {q7}, [buf]!
CPU_LE( vrev64.8 q7, q7 )
vswp q7l, q7h
// XOR the first 16 data *bits* with the initial CRC value.
vmov.i8 q0h, #0
vmov.u16 q0h[3], init_crc
veor.8 q7h, q7h, q0h
// Load the fold-across-16-bytes constants.
vld1.64 {FOLD_CONSTS}, [fold_consts_ptr, :128]!
cmp len, #16
beq .Lreduce_final_16_bytes // len == 16
subs len, len, #32
addlt len, len, #16
blt .Lhandle_partial_segment // 17 <= len <= 31
b .Lfold_16_bytes_loop // 32 <= len <= 255
ENDPROC(crc_t10dif_pmull)
.section ".rodata", "a"
.align 4
// Fold constants precomputed from the polynomial 0x18bb7
// G(x) = x^16 + x^15 + x^11 + x^9 + x^8 + x^7 + x^5 + x^4 + x^2 + x^1 + x^0
.Lfold_across_128_bytes_consts:
.quad 0x0000000000006123 // x^(8*128) mod G(x)
.quad 0x0000000000002295 // x^(8*128+64) mod G(x)
// .Lfold_across_64_bytes_consts:
.quad 0x0000000000001069 // x^(4*128) mod G(x)
.quad 0x000000000000dd31 // x^(4*128+64) mod G(x)
// .Lfold_across_32_bytes_consts:
.quad 0x000000000000857d // x^(2*128) mod G(x)
.quad 0x0000000000007acc // x^(2*128+64) mod G(x)
.Lfold_across_16_bytes_consts:
.quad 0x000000000000a010 // x^(1*128) mod G(x)
.quad 0x0000000000001faa // x^(1*128+64) mod G(x)
// .Lfinal_fold_consts:
.quad 0x1368000000000000 // x^48 * (x^48 mod G(x))
.quad 0x2d56000000000000 // x^48 * (x^80 mod G(x))
// .Lbarrett_reduction_consts:
.quad 0x0000000000018bb7 // G(x)
.quad 0x00000001f65a57f8 // floor(x^48 / G(x))
// For 1 <= len <= 15, the 16-byte vector beginning at &byteshift_table[16 -
// len] is the index vector to shift left by 'len' bytes, and is also {0x80,
// ..., 0x80} XOR the index vector to shift right by '16 - len' bytes.
.Lbyteshift_table:
.byte 0x0, 0x81, 0x82, 0x83, 0x84, 0x85, 0x86, 0x87
.byte 0x88, 0x89, 0x8a, 0x8b, 0x8c, 0x8d, 0x8e, 0x8f
.byte 0x0, 0x1, 0x2, 0x3, 0x4, 0x5, 0x6, 0x7
.byte 0x8, 0x9, 0xa, 0xb, 0xc, 0xd, 0xe , 0x0