Subject: Twofish encryption for loop device for old S.u.S.E. crypto partitions From: kraxel@suse.de Patch-mainline: not yet See $subject, used up to 9.2 on new installs. --- drivers/block/Kconfig | 6 drivers/block/Makefile | 2 drivers/block/loop_fish2.c | 625 +++++++++++++++++++++++++++++++++++++++++++++ 3 files changed, 633 insertions(+) --- a/drivers/block/Kconfig +++ b/drivers/block/Kconfig @@ -452,6 +452,12 @@ config SUNVDC Support for virtual disk devices as a client under Sun Logical Domains. +config CIPHER_TWOFISH + tristate "Twofish encryption for loop device for old S.u.S.E. crypto partitions" + depends on BLK_DEV_LOOP + help + Say Y here if you want to support old S.u.S.E. crypto partitions. + source "drivers/s390/block/Kconfig" config XILINX_SYSACE --- a/drivers/block/Makefile +++ b/drivers/block/Makefile @@ -38,4 +38,6 @@ obj-$(CONFIG_BLK_DEV_HD) += hd.o obj-$(CONFIG_XEN_BLKDEV_FRONTEND) += xen-blkfront.o obj-$(CONFIG_BLK_DEV_DRBD) += drbd/ +obj-$(CONFIG_CIPHER_TWOFISH) += loop_fish2.o + swim_mod-objs := swim.o swim_asm.o --- /dev/null +++ b/drivers/block/loop_fish2.c @@ -0,0 +1,625 @@ +#include +#include +#include +#include +#include +#include +#include +#include +#include + +#define ROL(x,c) (((x) << (c)) | ((x) >> (32-(c)))) +#define ROR(x,c) (((x) >> (c)) | ((x) << (32-(c)))) +#define Bswap(x) __le32_to_cpu(x) + +#define DWORD __u32 +#define BYTE unsigned char + +typedef struct fish2_key +{ int keyLen; /* Key Length in Bit */ + DWORD sboxKeys[4]; + DWORD subKeys[40]; + BYTE key[32]; + DWORD sbox_full[1024]; /* This have to be 1024 DWORDs */ +} fish2_key; + + +/* Mul_5B[i] is 0x5B * i in GF(256), whatever that means... */ + +static unsigned char Mul_5B[256] = { + 0x00,0x5B,0xB6,0xED,0x05,0x5E,0xB3,0xE8, + 0x0A,0x51,0xBC,0xE7,0x0F,0x54,0xB9,0xE2, + 0x14,0x4F,0xA2,0xF9,0x11,0x4A,0xA7,0xFC, + 0x1E,0x45,0xA8,0xF3,0x1B,0x40,0xAD,0xF6, + 0x28,0x73,0x9E,0xC5,0x2D,0x76,0x9B,0xC0, + 0x22,0x79,0x94,0xCF,0x27,0x7C,0x91,0xCA, + 0x3C,0x67,0x8A,0xD1,0x39,0x62,0x8F,0xD4, + 0x36,0x6D,0x80,0xDB,0x33,0x68,0x85,0xDE, + 0x50,0x0B,0xE6,0xBD,0x55,0x0E,0xE3,0xB8, + 0x5A,0x01,0xEC,0xB7,0x5F,0x04,0xE9,0xB2, + 0x44,0x1F,0xF2,0xA9,0x41,0x1A,0xF7,0xAC, + 0x4E,0x15,0xF8,0xA3,0x4B,0x10,0xFD,0xA6, + 0x78,0x23,0xCE,0x95,0x7D,0x26,0xCB,0x90, + 0x72,0x29,0xC4,0x9F,0x77,0x2C,0xC1,0x9A, + 0x6C,0x37,0xDA,0x81,0x69,0x32,0xDF,0x84, + 0x66,0x3D,0xD0,0x8B,0x63,0x38,0xD5,0x8E, + 0xA0,0xFB,0x16,0x4D,0xA5,0xFE,0x13,0x48, + 0xAA,0xF1,0x1C,0x47,0xAF,0xF4,0x19,0x42, + 0xB4,0xEF,0x02,0x59,0xB1,0xEA,0x07,0x5C, + 0xBE,0xE5,0x08,0x53,0xBB,0xE0,0x0D,0x56, + 0x88,0xD3,0x3E,0x65,0x8D,0xD6,0x3B,0x60, + 0x82,0xD9,0x34,0x6F,0x87,0xDC,0x31,0x6A, + 0x9C,0xC7,0x2A,0x71,0x99,0xC2,0x2F,0x74, + 0x96,0xCD,0x20,0x7B,0x93,0xC8,0x25,0x7E, + 0xF0,0xAB,0x46,0x1D,0xF5,0xAE,0x43,0x18, + 0xFA,0xA1,0x4C,0x17,0xFF,0xA4,0x49,0x12, + 0xE4,0xBF,0x52,0x09,0xE1,0xBA,0x57,0x0C, + 0xEE,0xB5,0x58,0x03,0xEB,0xB0,0x5D,0x06, + 0xD8,0x83,0x6E,0x35,0xDD,0x86,0x6B,0x30, + 0xD2,0x89,0x64,0x3F,0xD7,0x8C,0x61,0x3A, + 0xCC,0x97,0x7A,0x21,0xC9,0x92,0x7F,0x24, + 0xC6,0x9D,0x70,0x2B,0xC3,0x98,0x75,0x2E }; + + +/* Mul_EF[i] is 0xEF * i in GF(256), whatever that means... */ + +static unsigned char Mul_EF[256] = { + 0x00,0xEF,0xB7,0x58,0x07,0xE8,0xB0,0x5F, + 0x0E,0xE1,0xB9,0x56,0x09,0xE6,0xBE,0x51, + 0x1C,0xF3,0xAB,0x44,0x1B,0xF4,0xAC,0x43, + 0x12,0xFD,0xA5,0x4A,0x15,0xFA,0xA2,0x4D, + 0x38,0xD7,0x8F,0x60,0x3F,0xD0,0x88,0x67, + 0x36,0xD9,0x81,0x6E,0x31,0xDE,0x86,0x69, + 0x24,0xCB,0x93,0x7C,0x23,0xCC,0x94,0x7B, + 0x2A,0xC5,0x9D,0x72,0x2D,0xC2,0x9A,0x75, + 0x70,0x9F,0xC7,0x28,0x77,0x98,0xC0,0x2F, + 0x7E,0x91,0xC9,0x26,0x79,0x96,0xCE,0x21, + 0x6C,0x83,0xDB,0x34,0x6B,0x84,0xDC,0x33, + 0x62,0x8D,0xD5,0x3A,0x65,0x8A,0xD2,0x3D, + 0x48,0xA7,0xFF,0x10,0x4F,0xA0,0xF8,0x17, + 0x46,0xA9,0xF1,0x1E,0x41,0xAE,0xF6,0x19, + 0x54,0xBB,0xE3,0x0C,0x53,0xBC,0xE4,0x0B, + 0x5A,0xB5,0xED,0x02,0x5D,0xB2,0xEA,0x05, + 0xE0,0x0F,0x57,0xB8,0xE7,0x08,0x50,0xBF, + 0xEE,0x01,0x59,0xB6,0xE9,0x06,0x5E,0xB1, + 0xFC,0x13,0x4B,0xA4,0xFB,0x14,0x4C,0xA3, + 0xF2,0x1D,0x45,0xAA,0xF5,0x1A,0x42,0xAD, + 0xD8,0x37,0x6F,0x80,0xDF,0x30,0x68,0x87, + 0xD6,0x39,0x61,0x8E,0xD1,0x3E,0x66,0x89, + 0xC4,0x2B,0x73,0x9C,0xC3,0x2C,0x74,0x9B, + 0xCA,0x25,0x7D,0x92,0xCD,0x22,0x7A,0x95, + 0x90,0x7F,0x27,0xC8,0x97,0x78,0x20,0xCF, + 0x9E,0x71,0x29,0xC6,0x99,0x76,0x2E,0xC1, + 0x8C,0x63,0x3B,0xD4,0x8B,0x64,0x3C,0xD3, + 0x82,0x6D,0x35,0xDA,0x85,0x6A,0x32,0xDD, + 0xA8,0x47,0x1F,0xF0,0xAF,0x40,0x18,0xF7, + 0xA6,0x49,0x11,0xFE,0xA1,0x4E,0x16,0xF9, + 0xB4,0x5B,0x03,0xEC,0xB3,0x5C,0x04,0xEB, + 0xBA,0x55,0x0D,0xE2,0xBD,0x52,0x0A,0xE5 }; + +static inline DWORD mds_mul(BYTE *y) +{ DWORD z; + + z=Mul_EF[y[0]] ^ y[1] ^ Mul_EF[y[2]] ^ Mul_5B[y[3]]; + z<<=8; + z|=Mul_EF[y[0]] ^ Mul_5B[y[1]] ^ y[2] ^ Mul_EF[y[3]]; + z<<=8; + z|=Mul_5B[y[0]] ^ Mul_EF[y[1]] ^ Mul_EF[y[2]] ^ y[3]; + z<<=8; + z|=y[0] ^ Mul_EF[y[1]] ^ Mul_5B[y[2]] ^ Mul_5B[y[3]]; + + return z; +} + +/* q0 and q1 are the lookup substitutions done in twofish */ + +static unsigned char q0[256] = +{ 0xA9, 0x67, 0xB3, 0xE8, 0x04, 0xFD, 0xA3, 0x76, + 0x9A, 0x92, 0x80, 0x78, 0xE4, 0xDD, 0xD1, 0x38, + 0x0D, 0xC6, 0x35, 0x98, 0x18, 0xF7, 0xEC, 0x6C, + 0x43, 0x75, 0x37, 0x26, 0xFA, 0x13, 0x94, 0x48, + 0xF2, 0xD0, 0x8B, 0x30, 0x84, 0x54, 0xDF, 0x23, + 0x19, 0x5B, 0x3D, 0x59, 0xF3, 0xAE, 0xA2, 0x82, + 0x63, 0x01, 0x83, 0x2E, 0xD9, 0x51, 0x9B, 0x7C, + 0xA6, 0xEB, 0xA5, 0xBE, 0x16, 0x0C, 0xE3, 0x61, + 0xC0, 0x8C, 0x3A, 0xF5, 0x73, 0x2C, 0x25, 0x0B, + 0xBB, 0x4E, 0x89, 0x6B, 0x53, 0x6A, 0xB4, 0xF1, + 0xE1, 0xE6, 0xBD, 0x45, 0xE2, 0xF4, 0xB6, 0x66, + 0xCC, 0x95, 0x03, 0x56, 0xD4, 0x1C, 0x1E, 0xD7, + 0xFB, 0xC3, 0x8E, 0xB5, 0xE9, 0xCF, 0xBF, 0xBA, + 0xEA, 0x77, 0x39, 0xAF, 0x33, 0xC9, 0x62, 0x71, + 0x81, 0x79, 0x09, 0xAD, 0x24, 0xCD, 0xF9, 0xD8, + 0xE5, 0xC5, 0xB9, 0x4D, 0x44, 0x08, 0x86, 0xE7, + 0xA1, 0x1D, 0xAA, 0xED, 0x06, 0x70, 0xB2, 0xD2, + 0x41, 0x7B, 0xA0, 0x11, 0x31, 0xC2, 0x27, 0x90, + 0x20, 0xF6, 0x60, 0xFF, 0x96, 0x5C, 0xB1, 0xAB, + 0x9E, 0x9C, 0x52, 0x1B, 0x5F, 0x93, 0x0A, 0xEF, + 0x91, 0x85, 0x49, 0xEE, 0x2D, 0x4F, 0x8F, 0x3B, + 0x47, 0x87, 0x6D, 0x46, 0xD6, 0x3E, 0x69, 0x64, + 0x2A, 0xCE, 0xCB, 0x2F, 0xFC, 0x97, 0x05, 0x7A, + 0xAC, 0x7F, 0xD5, 0x1A, 0x4B, 0x0E, 0xA7, 0x5A, + 0x28, 0x14, 0x3F, 0x29, 0x88, 0x3C, 0x4C, 0x02, + 0xB8, 0xDA, 0xB0, 0x17, 0x55, 0x1F, 0x8A, 0x7D, + 0x57, 0xC7, 0x8D, 0x74, 0xB7, 0xC4, 0x9F, 0x72, + 0x7E, 0x15, 0x22, 0x12, 0x58, 0x07, 0x99, 0x34, + 0x6E, 0x50, 0xDE, 0x68, 0x65, 0xBC, 0xDB, 0xF8, + 0xC8, 0xA8, 0x2B, 0x40, 0xDC, 0xFE, 0x32, 0xA4, + 0xCA, 0x10, 0x21, 0xF0, 0xD3, 0x5D, 0x0F, 0x00, + 0x6F, 0x9D, 0x36, 0x42, 0x4A, 0x5E, 0xC1, 0xE0}; + +static unsigned char q1[256] = +{ 0x75, 0xF3, 0xC6, 0xF4, 0xDB, 0x7B, 0xFB, 0xC8, + 0x4A, 0xD3, 0xE6, 0x6B, 0x45, 0x7D, 0xE8, 0x4B, + 0xD6, 0x32, 0xD8, 0xFD, 0x37, 0x71, 0xF1, 0xE1, + 0x30, 0x0F, 0xF8, 0x1B, 0x87, 0xFA, 0x06, 0x3F, + 0x5E, 0xBA, 0xAE, 0x5B, 0x8A, 0x00, 0xBC, 0x9D, + 0x6D, 0xC1, 0xB1, 0x0E, 0x80, 0x5D, 0xD2, 0xD5, + 0xA0, 0x84, 0x07, 0x14, 0xB5, 0x90, 0x2C, 0xA3, + 0xB2, 0x73, 0x4C, 0x54, 0x92, 0x74, 0x36, 0x51, + 0x38, 0xB0, 0xBD, 0x5A, 0xFC, 0x60, 0x62, 0x96, + 0x6C, 0x42, 0xF7, 0x10, 0x7C, 0x28, 0x27, 0x8C, + 0x13, 0x95, 0x9C, 0xC7, 0x24, 0x46, 0x3B, 0x70, + 0xCA, 0xE3, 0x85, 0xCB, 0x11, 0xD0, 0x93, 0xB8, + 0xA6, 0x83, 0x20, 0xFF, 0x9F, 0x77, 0xC3, 0xCC, + 0x03, 0x6F, 0x08, 0xBF, 0x40, 0xE7, 0x2B, 0xE2, + 0x79, 0x0C, 0xAA, 0x82, 0x41, 0x3A, 0xEA, 0xB9, + 0xE4, 0x9A, 0xA4, 0x97, 0x7E, 0xDA, 0x7A, 0x17, + 0x66, 0x94, 0xA1, 0x1D, 0x3D, 0xF0, 0xDE, 0xB3, + 0x0B, 0x72, 0xA7, 0x1C, 0xEF, 0xD1, 0x53, 0x3E, + 0x8F, 0x33, 0x26, 0x5F, 0xEC, 0x76, 0x2A, 0x49, + 0x81, 0x88, 0xEE, 0x21, 0xC4, 0x1A, 0xEB, 0xD9, + 0xC5, 0x39, 0x99, 0xCD, 0xAD, 0x31, 0x8B, 0x01, + 0x18, 0x23, 0xDD, 0x1F, 0x4E, 0x2D, 0xF9, 0x48, + 0x4F, 0xF2, 0x65, 0x8E, 0x78, 0x5C, 0x58, 0x19, + 0x8D, 0xE5, 0x98, 0x57, 0x67, 0x7F, 0x05, 0x64, + 0xAF, 0x63, 0xB6, 0xFE, 0xF5, 0xB7, 0x3C, 0xA5, + 0xCE, 0xE9, 0x68, 0x44, 0xE0, 0x4D, 0x43, 0x69, + 0x29, 0x2E, 0xAC, 0x15, 0x59, 0xA8, 0x0A, 0x9E, + 0x6E, 0x47, 0xDF, 0x34, 0x35, 0x6A, 0xCF, 0xDC, + 0x22, 0xC9, 0xC0, 0x9B, 0x89, 0xD4, 0xED, 0xAB, + 0x12, 0xA2, 0x0D, 0x52, 0xBB, 0x02, 0x2F, 0xA9, + 0xD7, 0x61, 0x1E, 0xB4, 0x50, 0x04, 0xF6, 0xC2, + 0x16, 0x25, 0x86, 0x56, 0x55, 0x09, 0xBE, 0x91 + }; + + +static DWORD f32(DWORD x, const DWORD * k32, int keyLen) +{ + BYTE b[4]; + + /* Run each byte thru 8x8 S-boxes, xoring with key byte at each stage. */ + /* Note that each byte goes through a different combination of S-boxes. */ + + *((DWORD *) b) = Bswap(x); /* make b[0] = LSB, b[3] = MSB */ + + switch (((keyLen + 63) / 64) & 3) + { + case 0: /* 256 bits of key */ + b[0] = q1[b[0]]; + b[1] = q0[b[1]]; + b[2] = q0[b[2]]; + b[3] = q1[b[3]]; + + *((DWORD *) b) ^= k32[3]; + + /* fall thru, having pre-processed b[0]..b[3] with k32[3] */ + case 3: /* 192 bits of key */ + b[0] = q1[b[0]]; + b[1] = q1[b[1]]; + b[2] = q0[b[2]]; + b[3] = q0[b[3]]; + + *((DWORD *) b) ^= k32[2]; + + /* fall thru, having pre-processed b[0]..b[3] with k32[2] */ + case 2: /* 128 bits of key */ + b[0] = q0[b[0]]; + b[1] = q1[b[1]]; + b[2] = q0[b[2]]; + b[3] = q1[b[3]]; + + *((DWORD *) b) ^= k32[1]; + + b[0] = q0[b[0]]; + b[1] = q0[b[1]]; + b[2] = q1[b[2]]; + b[3] = q1[b[3]]; + + *((DWORD *) b) ^= k32[0]; + + b[0] = q1[b[0]]; + b[1] = q0[b[1]]; + b[2] = q1[b[2]]; + b[3] = q0[b[3]]; + } + + + /* Now perform the MDS matrix multiply inline. */ + return mds_mul(b); +} + + +static void init_sbox(fish2_key *key) +{ DWORD x,*sbox,z,*k32; + int i,keyLen; + BYTE b[4]; + + k32=key->sboxKeys; + keyLen=key->keyLen; + sbox=key->sbox_full; + + x=0; + for (i=0;i<256;i++,x+=0x01010101) + { + *((DWORD *) b) = Bswap(x); /* make b[0] = LSB, b[3] = MSB */ + + switch (((keyLen + 63) / 64) & 3) + { + case 0: /* 256 bits of key */ + b[0] = q1[b[0]]; + b[1] = q0[b[1]]; + b[2] = q0[b[2]]; + b[3] = q1[b[3]]; + + *((DWORD *) b) ^= k32[3]; + + /* fall thru, having pre-processed b[0]..b[3] with k32[3] */ + case 3: /* 192 bits of key */ + b[0] = q1[b[0]]; + b[1] = q1[b[1]]; + b[2] = q0[b[2]]; + b[3] = q0[b[3]]; + + *((DWORD *) b) ^= k32[2]; + + /* fall thru, having pre-processed b[0]..b[3] with k32[2] */ + case 2: /* 128 bits of key */ + b[0] = q0[b[0]]; + b[1] = q1[b[1]]; + b[2] = q0[b[2]]; + b[3] = q1[b[3]]; + + *((DWORD *) b) ^= k32[1]; + + b[0] = q0[b[0]]; + b[1] = q0[b[1]]; + b[2] = q1[b[2]]; + b[3] = q1[b[3]]; + + *((DWORD *) b) ^= k32[0]; + + b[0] = q1[b[0]]; + b[1] = q0[b[1]]; + b[2] = q1[b[2]]; + b[3] = q0[b[3]]; + } + + z=Mul_EF[b[0]]; + z<<=8; + z|=Mul_EF[b[0]]; + z<<=8; + z|=Mul_5B[b[0]]; + z<<=8; + z|=b[0]; + + sbox[i]=z; + + z=b[1]; + z<<=8; + z|=Mul_5B[b[1]]; + z<<=8; + z|=Mul_EF[b[1]]; + z<<=8; + z|=Mul_EF[b[1]]; + + sbox[i+256]=z; + + z=Mul_EF[b[2]]; + z<<=8; + z|=b[2]; + z<<=8; + z|=Mul_EF[b[2]]; + z<<=8; + z|=Mul_5B[b[2]]; + + sbox[i+512]=z; + + z=Mul_5B[b[3]]; + z<<=8; + z|=Mul_EF[b[3]]; + z<<=8; + z|=b[3]; + z<<=8; + z|=Mul_5B[b[3]]; + + sbox[i+768]=z; + } +} + + +/* Reed-Solomon code parameters: (12,8) reversible code + g(x) = x**4 + (a + 1/a) x**3 + a x**2 + (a + 1/a) x + 1 + where a = primitive root of field generator 0x14D */ +#define RS_GF_FDBK 0x14D /* field generator */ +#define RS_rem(x) \ + { BYTE b = x >> 24; \ + DWORD g2 = ((b << 1) ^ ((b & 0x80) ? RS_GF_FDBK : 0 )) & 0xFF; \ + DWORD g3 = ((b >> 1) & 0x7F) ^ ((b & 1) ? RS_GF_FDBK >> 1 : 0 ) ^ g2 ; \ + x = (x << 8) ^ (g3 << 24) ^ (g2 << 16) ^ (g3 << 8) ^ b; \ + } + +static DWORD rs_mds(DWORD k0, DWORD k1) +{ + int i, j; + DWORD r; + + for (i = r = 0; i < 2; i++) + { + r ^= (i) ? k0 : k1; /* merge in 32 more key bits */ + for (j = 0; j < 4; j++) /* shift one byte at a time */ + RS_rem(r); + } + return r; +} + + +#define INPUT_WHITEN 0 /* subkey array indices */ +#define OUTPUT_WHITEN 4 +#define ROUND_SUBKEYS 8 /* use 2 * (# rounds) */ +#define TOTAL_SUBKEYS 40 + +static void init_key(fish2_key * key) +{ + int i, k64Cnt; + int keyLen = key->keyLen; + int subkeyCnt = TOTAL_SUBKEYS; + DWORD A, B; + DWORD k32e[4], k32o[4]; /* even/odd key dwords */ + + k64Cnt = (keyLen + 63) / 64; /* round up to next multiple of 64 bits */ + for (i = 0; i < k64Cnt; i++) + { /* split into even/odd key dwords */ + k32e[i] = ((DWORD *)key->key)[2 * i]; + k32o[i] = ((DWORD *)key->key)[2 * i + 1]; + /* compute S-box keys using (12,8) Reed-Solomon code over GF(256) */ + /* store in reverse order */ + key->sboxKeys[k64Cnt - 1 - i] = + Bswap(rs_mds(Bswap(k32e[i]), Bswap(k32o[i]))); + + } + + for (i = 0; i < subkeyCnt / 2; i++) /* compute round subkeys for PHT */ + { + A = f32(i * 0x02020202, k32e, keyLen); /* A uses even key dwords */ + B = f32(i * 0x02020202 + 0x01010101, k32o, keyLen); /* B uses odd key + dwords */ + B = ROL(B, 8); + key->subKeys[2 * i] = A + B; /* combine with a PHT */ + key->subKeys[2 * i + 1] = ROL(A + 2 * B, 9); + } + + init_sbox(key); +} + + +static inline DWORD f32_sbox(DWORD x,DWORD *sbox) +{ + /* Run each byte thru 8x8 S-boxes, xoring with key byte at each stage. */ + /* Note that each byte goes through a different combination of S-boxes. */ + + return (sbox[ (x) &0xff]^ + sbox[256 + (((x)>> 8)&0xff)]^ + sbox[512 + (((x)>>16)&0xff)]^ + sbox[768 + (((x)>>24)&0xff)]); +} + +#define roundE_m(x0,x1,x2,x3,rnd) \ + t0 = f32_sbox( x0, key->sbox_full ) ; \ + t1 = f32_sbox( ROL(x1,8), key->sbox_full ); \ + x2 ^= t0 + t1 + key->subKeys[2*rnd+8]; \ + x3 = ROL(x3,1); \ + x3 ^= t0 + 2*t1 + key->subKeys[2*rnd+9]; \ + x2 = ROR(x2,1); + + +static int blockEncrypt_CBC(fish2_key *key,BYTE *src,BYTE *dst,int len) +{ DWORD xx0,xx1,xx2,xx3,t0,t1,iv0,iv1,iv2,iv3; + + if (len & 0xF) return -1; + + iv0=0; + iv1=0; + iv2=0; + iv3=0; + for (;len>=16;len-=16) + + { + if ( ( len & 0x1FF) == 0) + { iv0=0; + iv1=0; + iv2=0; + iv3=0; + } + + xx0=Bswap(((DWORD *)src)[0]) ^ key->subKeys[0] ^ iv0; + xx1=Bswap(((DWORD *)src)[1]) ^ key->subKeys[1] ^ iv1; + xx2=Bswap(((DWORD *)src)[2]) ^ key->subKeys[2] ^ iv2; + xx3=Bswap(((DWORD *)src)[3]) ^ key->subKeys[3] ^ iv3; + + src+=16; + + roundE_m(xx0,xx1,xx2,xx3,0); + roundE_m(xx2,xx3,xx0,xx1,1); + roundE_m(xx0,xx1,xx2,xx3,2); + roundE_m(xx2,xx3,xx0,xx1,3); + roundE_m(xx0,xx1,xx2,xx3,4); + roundE_m(xx2,xx3,xx0,xx1,5); + roundE_m(xx0,xx1,xx2,xx3,6); + roundE_m(xx2,xx3,xx0,xx1,7); + roundE_m(xx0,xx1,xx2,xx3,8); + roundE_m(xx2,xx3,xx0,xx1,9); + roundE_m(xx0,xx1,xx2,xx3,10); + roundE_m(xx2,xx3,xx0,xx1,11); + roundE_m(xx0,xx1,xx2,xx3,12); + roundE_m(xx2,xx3,xx0,xx1,13); + roundE_m(xx0,xx1,xx2,xx3,14); + roundE_m(xx2,xx3,xx0,xx1,15); + + iv0=xx2 ^ key->subKeys[4]; + iv1=xx3 ^ key->subKeys[5]; + iv2=xx0 ^ key->subKeys[6]; + iv3=xx1 ^ key->subKeys[7]; + + ((DWORD *)dst)[0] = Bswap(iv0); + ((DWORD *)dst)[1] = Bswap(iv1); + ((DWORD *)dst)[2] = Bswap(iv2); + ((DWORD *)dst)[3] = Bswap(iv3); + dst+=16; + } + return len; +} + +#define roundD_m(x0,x1,x2,x3,rnd) \ + t0 = f32_sbox( x0, key->sbox_full); \ + t1 = f32_sbox( ROL(x1,8),key->sbox_full); \ + x2 = ROL(x2,1); \ + x3 ^= t0 + 2*t1 + key->subKeys[rnd*2+9]; \ + x3 = ROR(x3,1); \ + x2 ^= t0 + t1 + key->subKeys[rnd*2+8]; + + +static int blockDecrypt_CBC(fish2_key *key,BYTE *src,BYTE *dst,int len) +{ DWORD xx0,xx1,xx2,xx3,t0,t1,lx0,lx1,lx2,lx3,iv0,iv1,iv2,iv3; + + if (len & 0xF) return -1; + + iv0=0; + iv1=0; + iv2=0; + iv3=0; + + for (;len>=16;len-=16) + { + if ( ( len & 0x1FF) == 0) + { iv0=0; + iv1=0; + iv2=0; + iv3=0; + } + + lx0=iv0;iv0=Bswap(((DWORD *)src)[0]);xx0=iv0 ^ key->subKeys[4]; + lx1=iv1;iv1=Bswap(((DWORD *)src)[1]);xx1=iv1 ^ key->subKeys[5]; + lx2=iv2;iv2=Bswap(((DWORD *)src)[2]);xx2=iv2 ^ key->subKeys[6]; + lx3=iv3;iv3=Bswap(((DWORD *)src)[3]);xx3=iv3 ^ key->subKeys[7]; + src+=16; + + roundD_m(xx0,xx1,xx2,xx3,15); + roundD_m(xx2,xx3,xx0,xx1,14); + roundD_m(xx0,xx1,xx2,xx3,13); + roundD_m(xx2,xx3,xx0,xx1,12); + roundD_m(xx0,xx1,xx2,xx3,11); + roundD_m(xx2,xx3,xx0,xx1,10); + roundD_m(xx0,xx1,xx2,xx3,9); + roundD_m(xx2,xx3,xx0,xx1,8); + roundD_m(xx0,xx1,xx2,xx3,7); + roundD_m(xx2,xx3,xx0,xx1,6); + roundD_m(xx0,xx1,xx2,xx3,5); + roundD_m(xx2,xx3,xx0,xx1,4); + roundD_m(xx0,xx1,xx2,xx3,3); + roundD_m(xx2,xx3,xx0,xx1,2); + roundD_m(xx0,xx1,xx2,xx3,1); + roundD_m(xx2,xx3,xx0,xx1,0); + + ((DWORD *)dst)[0] = Bswap(xx2 ^ key->subKeys[0] ^ lx0); + ((DWORD *)dst)[1] = Bswap(xx3 ^ key->subKeys[1] ^ lx1); + ((DWORD *)dst)[2] = Bswap(xx0 ^ key->subKeys[2] ^ lx2); + ((DWORD *)dst)[3] = Bswap(xx1 ^ key->subKeys[3] ^ lx3); + dst+=16; + } + return len; +} + + +int transfer_fish2(struct loop_device *lo, int cmd, + struct page *raw_page, unsigned raw_off, + struct page *loop_page, unsigned loop_off, + int size, sector_t IV) +{ + char *raw_buf = kmap_atomic(raw_page, KM_USER0) + raw_off; + char *loop_buf = kmap_atomic(loop_page, KM_USER1) + loop_off; + + if (cmd == READ) + blockDecrypt_CBC((fish2_key *)lo->key_data,raw_buf,loop_buf,size); + else + blockEncrypt_CBC((fish2_key *)lo->key_data,loop_buf,raw_buf,size); + + kunmap_atomic(raw_buf, KM_USER0); + kunmap_atomic(loop_buf, KM_USER1); + cond_resched(); + + return 0; +} + +int fish2_init(struct loop_device *lo,const struct loop_info64 *info) +{ fish2_key *key; + + if (info->lo_encrypt_key_size<16 || info->lo_encrypt_key_size>32) + return -EINVAL; + + key=(fish2_key *)kmalloc(sizeof(fish2_key),GFP_KERNEL); + + if (key==NULL) + return -ENOMEM; + + lo->key_data=key; + + memset(key->key,0,32); + + key->keyLen=info->lo_encrypt_key_size << 3; + memcpy(key->key,info->lo_encrypt_key,info->lo_encrypt_key_size); + + init_key(key); + + return 0; +} + +static int fish2_release(struct loop_device *lo) +{ if (lo->key_data!=NULL) + { + kfree(lo->key_data); + lo->key_data=NULL; + } + return(0); +} + +static struct loop_func_table fish2_funcs = +{ .number = LO_CRYPT_FISH2, + .transfer = transfer_fish2, + .init = fish2_init, + .release = fish2_release, + .owner = THIS_MODULE +}; + +int __init loop_fish2_init(void) +{ + int err; + + if ((err=loop_register_transfer(&fish2_funcs))) + { + printk(KERN_WARNING "Couldn't register Twofish encryption\n"); + return err; + } + printk(KERN_INFO "loop: registered Twofish encryption \n"); + return 0; +} + +void __exit loop_fish2_exit(void) +{ + if (loop_unregister_transfer(LO_CRYPT_FISH2)) + printk(KERN_WARNING "Couldn't unregister Twofish encryption\n"); + printk(KERN_INFO "loop: unregistered Twofish encryption \n"); +} + +module_init(loop_fish2_init); +module_exit(loop_fish2_exit); +MODULE_LICENSE("GPL");