^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1) /* SPDX-License-Identifier: GPL-2.0 */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3) * Hardware-accelerated CRC-32 variants for Linux on z Systems
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5) * Use the z/Architecture Vector Extension Facility to accelerate the
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6) * computing of CRC-32 checksums.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8) * This CRC-32 implementation algorithm processes the most-significant
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9) * bit first (BE).
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11) * Copyright IBM Corp. 2015
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 12) * Author(s): Hendrik Brueckner <brueckner@linux.vnet.ibm.com>
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 13) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 14)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 15) #include <linux/linkage.h>
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 16) #include <asm/nospec-insn.h>
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 17) #include <asm/vx-insn.h>
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 18)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 19) /* Vector register range containing CRC-32 constants */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 20) #define CONST_R1R2 %v9
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 21) #define CONST_R3R4 %v10
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 22) #define CONST_R5 %v11
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 23) #define CONST_R6 %v12
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 24) #define CONST_RU_POLY %v13
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 25) #define CONST_CRC_POLY %v14
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 26)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 27) .data
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 28) .align 8
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 29)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 30) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 31) * The CRC-32 constant block contains reduction constants to fold and
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 32) * process particular chunks of the input data stream in parallel.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 33) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 34) * For the CRC-32 variants, the constants are precomputed according to
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 35) * these defintions:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 36) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 37) * R1 = x4*128+64 mod P(x)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 38) * R2 = x4*128 mod P(x)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 39) * R3 = x128+64 mod P(x)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 40) * R4 = x128 mod P(x)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 41) * R5 = x96 mod P(x)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 42) * R6 = x64 mod P(x)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 43) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 44) * Barret reduction constant, u, is defined as floor(x**64 / P(x)).
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 45) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 46) * where P(x) is the polynomial in the normal domain and the P'(x) is the
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 47) * polynomial in the reversed (bitreflected) domain.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 48) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 49) * Note that the constant definitions below are extended in order to compute
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 50) * intermediate results with a single VECTOR GALOIS FIELD MULTIPLY instruction.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 51) * The righmost doubleword can be 0 to prevent contribution to the result or
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 52) * can be multiplied by 1 to perform an XOR without the need for a separate
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 53) * VECTOR EXCLUSIVE OR instruction.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 54) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 55) * CRC-32 (IEEE 802.3 Ethernet, ...) polynomials:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 56) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 57) * P(x) = 0x04C11DB7
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 58) * P'(x) = 0xEDB88320
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 59) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 60)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 61) .Lconstants_CRC_32_BE:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 62) .quad 0x08833794c, 0x0e6228b11 # R1, R2
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 63) .quad 0x0c5b9cd4c, 0x0e8a45605 # R3, R4
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 64) .quad 0x0f200aa66, 1 << 32 # R5, x32
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 65) .quad 0x0490d678d, 1 # R6, 1
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 66) .quad 0x104d101df, 0 # u
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 67) .quad 0x104C11DB7, 0 # P(x)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 68)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 69) .previous
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 70)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 71) GEN_BR_THUNK %r14
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 72)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 73) .text
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 74) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 75) * The CRC-32 function(s) use these calling conventions:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 76) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 77) * Parameters:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 78) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 79) * %r2: Initial CRC value, typically ~0; and final CRC (return) value.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 80) * %r3: Input buffer pointer, performance might be improved if the
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 81) * buffer is on a doubleword boundary.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 82) * %r4: Length of the buffer, must be 64 bytes or greater.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 83) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 84) * Register usage:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 85) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 86) * %r5: CRC-32 constant pool base pointer.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 87) * V0: Initial CRC value and intermediate constants and results.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 88) * V1..V4: Data for CRC computation.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 89) * V5..V8: Next data chunks that are fetched from the input buffer.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 90) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 91) * V9..V14: CRC-32 constants.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 92) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 93) ENTRY(crc32_be_vgfm_16)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 94) /* Load CRC-32 constants */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 95) larl %r5,.Lconstants_CRC_32_BE
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 96) VLM CONST_R1R2,CONST_CRC_POLY,0,%r5
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 97)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 98) /* Load the initial CRC value into the leftmost word of V0. */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 99) VZERO %v0
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 100) VLVGF %v0,%r2,0
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 101)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 102) /* Load a 64-byte data chunk and XOR with CRC */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 103) VLM %v1,%v4,0,%r3 /* 64-bytes into V1..V4 */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 104) VX %v1,%v0,%v1 /* V1 ^= CRC */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 105) aghi %r3,64 /* BUF = BUF + 64 */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 106) aghi %r4,-64 /* LEN = LEN - 64 */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 107)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 108) /* Check remaining buffer size and jump to proper folding method */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 109) cghi %r4,64
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 110) jl .Lless_than_64bytes
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 111)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 112) .Lfold_64bytes_loop:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 113) /* Load the next 64-byte data chunk into V5 to V8 */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 114) VLM %v5,%v8,0,%r3
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 115)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 116) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 117) * Perform a GF(2) multiplication of the doublewords in V1 with
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 118) * the reduction constants in V0. The intermediate result is
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 119) * then folded (accumulated) with the next data chunk in V5 and
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 120) * stored in V1. Repeat this step for the register contents
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 121) * in V2, V3, and V4 respectively.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 122) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 123) VGFMAG %v1,CONST_R1R2,%v1,%v5
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 124) VGFMAG %v2,CONST_R1R2,%v2,%v6
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 125) VGFMAG %v3,CONST_R1R2,%v3,%v7
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 126) VGFMAG %v4,CONST_R1R2,%v4,%v8
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 127)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 128) /* Adjust buffer pointer and length for next loop */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 129) aghi %r3,64 /* BUF = BUF + 64 */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 130) aghi %r4,-64 /* LEN = LEN - 64 */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 131)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 132) cghi %r4,64
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 133) jnl .Lfold_64bytes_loop
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 134)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 135) .Lless_than_64bytes:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 136) /* Fold V1 to V4 into a single 128-bit value in V1 */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 137) VGFMAG %v1,CONST_R3R4,%v1,%v2
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 138) VGFMAG %v1,CONST_R3R4,%v1,%v3
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 139) VGFMAG %v1,CONST_R3R4,%v1,%v4
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 140)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 141) /* Check whether to continue with 64-bit folding */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 142) cghi %r4,16
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 143) jl .Lfinal_fold
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 144)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 145) .Lfold_16bytes_loop:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 146)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 147) VL %v2,0,,%r3 /* Load next data chunk */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 148) VGFMAG %v1,CONST_R3R4,%v1,%v2 /* Fold next data chunk */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 149)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 150) /* Adjust buffer pointer and size for folding next data chunk */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 151) aghi %r3,16
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 152) aghi %r4,-16
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 153)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 154) /* Process remaining data chunks */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 155) cghi %r4,16
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 156) jnl .Lfold_16bytes_loop
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 157)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 158) .Lfinal_fold:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 159) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 160) * The R5 constant is used to fold a 128-bit value into an 96-bit value
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 161) * that is XORed with the next 96-bit input data chunk. To use a single
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 162) * VGFMG instruction, multiply the rightmost 64-bit with x^32 (1<<32) to
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 163) * form an intermediate 96-bit value (with appended zeros) which is then
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 164) * XORed with the intermediate reduction result.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 165) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 166) VGFMG %v1,CONST_R5,%v1
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 167)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 168) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 169) * Further reduce the remaining 96-bit value to a 64-bit value using a
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 170) * single VGFMG, the rightmost doubleword is multiplied with 0x1. The
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 171) * intermediate result is then XORed with the product of the leftmost
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 172) * doubleword with R6. The result is a 64-bit value and is subject to
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 173) * the Barret reduction.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 174) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 175) VGFMG %v1,CONST_R6,%v1
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 176)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 177) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 178) * The input values to the Barret reduction are the degree-63 polynomial
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 179) * in V1 (R(x)), degree-32 generator polynomial, and the reduction
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 180) * constant u. The Barret reduction result is the CRC value of R(x) mod
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 181) * P(x).
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 182) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 183) * The Barret reduction algorithm is defined as:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 184) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 185) * 1. T1(x) = floor( R(x) / x^32 ) GF2MUL u
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 186) * 2. T2(x) = floor( T1(x) / x^32 ) GF2MUL P(x)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 187) * 3. C(x) = R(x) XOR T2(x) mod x^32
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 188) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 189) * Note: To compensate the division by x^32, use the vector unpack
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 190) * instruction to move the leftmost word into the leftmost doubleword
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 191) * of the vector register. The rightmost doubleword is multiplied
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 192) * with zero to not contribute to the intermedate results.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 193) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 194)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 195) /* T1(x) = floor( R(x) / x^32 ) GF2MUL u */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 196) VUPLLF %v2,%v1
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 197) VGFMG %v2,CONST_RU_POLY,%v2
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 198)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 199) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 200) * Compute the GF(2) product of the CRC polynomial in VO with T1(x) in
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 201) * V2 and XOR the intermediate result, T2(x), with the value in V1.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 202) * The final result is in the rightmost word of V2.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 203) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 204) VUPLLF %v2,%v2
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 205) VGFMAG %v2,CONST_CRC_POLY,%v2,%v1
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 206)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 207) .Ldone:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 208) VLGVF %r2,%v2,3
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 209) BR_EX %r14
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 210) ENDPROC(crc32_be_vgfm_16)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 211)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 212) .previous