diff options
Diffstat (limited to 'arch/x86/crypto/twofish-i586-asm_32.S')
| -rw-r--r-- | arch/x86/crypto/twofish-i586-asm_32.S | 321 |
1 files changed, 321 insertions, 0 deletions
diff --git a/arch/x86/crypto/twofish-i586-asm_32.S b/arch/x86/crypto/twofish-i586-asm_32.S new file mode 100644 index 000000000..3abcad661 --- /dev/null +++ b/arch/x86/crypto/twofish-i586-asm_32.S @@ -0,0 +1,321 @@ +/* SPDX-License-Identifier: GPL-2.0-or-later */ +/*************************************************************************** +* Copyright (C) 2006 by Joachim Fritschi, <jfritschi@freenet.de> * +* * +***************************************************************************/ + +.file "twofish-i586-asm.S" +.text + +#include <linux/linkage.h> +#include <asm/asm-offsets.h> + +/* return address at 0 */ + +#define in_blk 12 /* input byte array address parameter*/ +#define out_blk 8 /* output byte array address parameter*/ +#define ctx 4 /* Twofish context structure */ + +#define a_offset 0 +#define b_offset 4 +#define c_offset 8 +#define d_offset 12 + +/* Structure of the crypto context struct*/ + +#define s0 0 /* S0 Array 256 Words each */ +#define s1 1024 /* S1 Array */ +#define s2 2048 /* S2 Array */ +#define s3 3072 /* S3 Array */ +#define w 4096 /* 8 whitening keys (word) */ +#define k 4128 /* key 1-32 ( word ) */ + +/* define a few register aliases to allow macro substitution */ + +#define R0D %eax +#define R0B %al +#define R0H %ah + +#define R1D %ebx +#define R1B %bl +#define R1H %bh + +#define R2D %ecx +#define R2B %cl +#define R2H %ch + +#define R3D %edx +#define R3B %dl +#define R3H %dh + + +/* performs input whitening */ +#define input_whitening(src,context,offset)\ + xor w+offset(context), src; + +/* performs input whitening */ +#define output_whitening(src,context,offset)\ + xor w+16+offset(context), src; + +/* + * a input register containing a (rotated 16) + * b input register containing b + * c input register containing c + * d input register containing d (already rol $1) + * operations on a and b are interleaved to increase performance + */ +#define encrypt_round(a,b,c,d,round)\ + push d ## D;\ + movzx b ## B, %edi;\ + mov s1(%ebp,%edi,4),d ## D;\ + movzx a ## B, %edi;\ + mov s2(%ebp,%edi,4),%esi;\ + movzx b ## H, %edi;\ + ror $16, b ## D;\ + xor s2(%ebp,%edi,4),d ## D;\ + movzx a ## H, %edi;\ + ror $16, a ## D;\ + xor s3(%ebp,%edi,4),%esi;\ + movzx b ## B, %edi;\ + xor s3(%ebp,%edi,4),d ## D;\ + movzx a ## B, %edi;\ + xor (%ebp,%edi,4), %esi;\ + movzx b ## H, %edi;\ + ror $15, b ## D;\ + xor (%ebp,%edi,4), d ## D;\ + movzx a ## H, %edi;\ + xor s1(%ebp,%edi,4),%esi;\ + pop %edi;\ + add d ## D, %esi;\ + add %esi, d ## D;\ + add k+round(%ebp), %esi;\ + xor %esi, c ## D;\ + rol $15, c ## D;\ + add k+4+round(%ebp),d ## D;\ + xor %edi, d ## D; + +/* + * a input register containing a (rotated 16) + * b input register containing b + * c input register containing c + * d input register containing d (already rol $1) + * operations on a and b are interleaved to increase performance + * last round has different rotations for the output preparation + */ +#define encrypt_last_round(a,b,c,d,round)\ + push d ## D;\ + movzx b ## B, %edi;\ + mov s1(%ebp,%edi,4),d ## D;\ + movzx a ## B, %edi;\ + mov s2(%ebp,%edi,4),%esi;\ + movzx b ## H, %edi;\ + ror $16, b ## D;\ + xor s2(%ebp,%edi,4),d ## D;\ + movzx a ## H, %edi;\ + ror $16, a ## D;\ + xor s3(%ebp,%edi,4),%esi;\ + movzx b ## B, %edi;\ + xor s3(%ebp,%edi,4),d ## D;\ + movzx a ## B, %edi;\ + xor (%ebp,%edi,4), %esi;\ + movzx b ## H, %edi;\ + ror $16, b ## D;\ + xor (%ebp,%edi,4), d ## D;\ + movzx a ## H, %edi;\ + xor s1(%ebp,%edi,4),%esi;\ + pop %edi;\ + add d ## D, %esi;\ + add %esi, d ## D;\ + add k+round(%ebp), %esi;\ + xor %esi, c ## D;\ + ror $1, c ## D;\ + add k+4+round(%ebp),d ## D;\ + xor %edi, d ## D; + +/* + * a input register containing a + * b input register containing b (rotated 16) + * c input register containing c + * d input register containing d (already rol $1) + * operations on a and b are interleaved to increase performance + */ +#define decrypt_round(a,b,c,d,round)\ + push c ## D;\ + movzx a ## B, %edi;\ + mov (%ebp,%edi,4), c ## D;\ + movzx b ## B, %edi;\ + mov s3(%ebp,%edi,4),%esi;\ + movzx a ## H, %edi;\ + ror $16, a ## D;\ + xor s1(%ebp,%edi,4),c ## D;\ + movzx b ## H, %edi;\ + ror $16, b ## D;\ + xor (%ebp,%edi,4), %esi;\ + movzx a ## B, %edi;\ + xor s2(%ebp,%edi,4),c ## D;\ + movzx b ## B, %edi;\ + xor s1(%ebp,%edi,4),%esi;\ + movzx a ## H, %edi;\ + ror $15, a ## D;\ + xor s3(%ebp,%edi,4),c ## D;\ + movzx b ## H, %edi;\ + xor s2(%ebp,%edi,4),%esi;\ + pop %edi;\ + add %esi, c ## D;\ + add c ## D, %esi;\ + add k+round(%ebp), c ## D;\ + xor %edi, c ## D;\ + add k+4+round(%ebp),%esi;\ + xor %esi, d ## D;\ + rol $15, d ## D; + +/* + * a input register containing a + * b input register containing b (rotated 16) + * c input register containing c + * d input register containing d (already rol $1) + * operations on a and b are interleaved to increase performance + * last round has different rotations for the output preparation + */ +#define decrypt_last_round(a,b,c,d,round)\ + push c ## D;\ + movzx a ## B, %edi;\ + mov (%ebp,%edi,4), c ## D;\ + movzx b ## B, %edi;\ + mov s3(%ebp,%edi,4),%esi;\ + movzx a ## H, %edi;\ + ror $16, a ## D;\ + xor s1(%ebp,%edi,4),c ## D;\ + movzx b ## H, %edi;\ + ror $16, b ## D;\ + xor (%ebp,%edi,4), %esi;\ + movzx a ## B, %edi;\ + xor s2(%ebp,%edi,4),c ## D;\ + movzx b ## B, %edi;\ + xor s1(%ebp,%edi,4),%esi;\ + movzx a ## H, %edi;\ + ror $16, a ## D;\ + xor s3(%ebp,%edi,4),c ## D;\ + movzx b ## H, %edi;\ + xor s2(%ebp,%edi,4),%esi;\ + pop %edi;\ + add %esi, c ## D;\ + add c ## D, %esi;\ + add k+round(%ebp), c ## D;\ + xor %edi, c ## D;\ + add k+4+round(%ebp),%esi;\ + xor %esi, d ## D;\ + ror $1, d ## D; + +SYM_FUNC_START(twofish_enc_blk) + push %ebp /* save registers according to calling convention*/ + push %ebx + push %esi + push %edi + + mov ctx + 16(%esp), %ebp /* abuse the base pointer: set new base + * pointer to the ctx address */ + mov in_blk+16(%esp),%edi /* input address in edi */ + + mov (%edi), %eax + mov b_offset(%edi), %ebx + mov c_offset(%edi), %ecx + mov d_offset(%edi), %edx + input_whitening(%eax,%ebp,a_offset) + ror $16, %eax + input_whitening(%ebx,%ebp,b_offset) + input_whitening(%ecx,%ebp,c_offset) + input_whitening(%edx,%ebp,d_offset) + rol $1, %edx + + encrypt_round(R0,R1,R2,R3,0); + encrypt_round(R2,R3,R0,R1,8); + encrypt_round(R0,R1,R2,R3,2*8); + encrypt_round(R2,R3,R0,R1,3*8); + encrypt_round(R0,R1,R2,R3,4*8); + encrypt_round(R2,R3,R0,R1,5*8); + encrypt_round(R0,R1,R2,R3,6*8); + encrypt_round(R2,R3,R0,R1,7*8); + encrypt_round(R0,R1,R2,R3,8*8); + encrypt_round(R2,R3,R0,R1,9*8); + encrypt_round(R0,R1,R2,R3,10*8); + encrypt_round(R2,R3,R0,R1,11*8); + encrypt_round(R0,R1,R2,R3,12*8); + encrypt_round(R2,R3,R0,R1,13*8); + encrypt_round(R0,R1,R2,R3,14*8); + encrypt_last_round(R2,R3,R0,R1,15*8); + + output_whitening(%eax,%ebp,c_offset) + output_whitening(%ebx,%ebp,d_offset) + output_whitening(%ecx,%ebp,a_offset) + output_whitening(%edx,%ebp,b_offset) + mov out_blk+16(%esp),%edi; + mov %eax, c_offset(%edi) + mov %ebx, d_offset(%edi) + mov %ecx, (%edi) + mov %edx, b_offset(%edi) + + pop %edi + pop %esi + pop %ebx + pop %ebp + mov $1, %eax + RET +SYM_FUNC_END(twofish_enc_blk) + +SYM_FUNC_START(twofish_dec_blk) + push %ebp /* save registers according to calling convention*/ + push %ebx + push %esi + push %edi + + + mov ctx + 16(%esp), %ebp /* abuse the base pointer: set new base + * pointer to the ctx address */ + mov in_blk+16(%esp),%edi /* input address in edi */ + + mov (%edi), %eax + mov b_offset(%edi), %ebx + mov c_offset(%edi), %ecx + mov d_offset(%edi), %edx + output_whitening(%eax,%ebp,a_offset) + output_whitening(%ebx,%ebp,b_offset) + ror $16, %ebx + output_whitening(%ecx,%ebp,c_offset) + output_whitening(%edx,%ebp,d_offset) + rol $1, %ecx + + decrypt_round(R0,R1,R2,R3,15*8); + decrypt_round(R2,R3,R0,R1,14*8); + decrypt_round(R0,R1,R2,R3,13*8); + decrypt_round(R2,R3,R0,R1,12*8); + decrypt_round(R0,R1,R2,R3,11*8); + decrypt_round(R2,R3,R0,R1,10*8); + decrypt_round(R0,R1,R2,R3,9*8); + decrypt_round(R2,R3,R0,R1,8*8); + decrypt_round(R0,R1,R2,R3,7*8); + decrypt_round(R2,R3,R0,R1,6*8); + decrypt_round(R0,R1,R2,R3,5*8); + decrypt_round(R2,R3,R0,R1,4*8); + decrypt_round(R0,R1,R2,R3,3*8); + decrypt_round(R2,R3,R0,R1,2*8); + decrypt_round(R0,R1,R2,R3,1*8); + decrypt_last_round(R2,R3,R0,R1,0); + + input_whitening(%eax,%ebp,c_offset) + input_whitening(%ebx,%ebp,d_offset) + input_whitening(%ecx,%ebp,a_offset) + input_whitening(%edx,%ebp,b_offset) + mov out_blk+16(%esp),%edi; + mov %eax, c_offset(%edi) + mov %ebx, d_offset(%edi) + mov %ecx, (%edi) + mov %edx, b_offset(%edi) + + pop %edi + pop %esi + pop %ebx + pop %ebp + mov $1, %eax + RET +SYM_FUNC_END(twofish_dec_blk) |