[apple/xnu.git] / EXTERNAL_HEADERS / corecrypto / cc_priv.h

/*
 *  cc_priv.h
 *  corecrypto
 *
 *  Created on 12/01/2010
 *
 *  Copyright (c) 2010,2011,2012,2014,2015 Apple Inc. All rights reserved.
 *
 */

#ifndef _CORECRYPTO_CC_PRIV_H_
#define _CORECRYPTO_CC_PRIV_H_

#include <corecrypto/cc.h>
#include <stdint.h>

/* defines the following macros :

 CC_MEMCPY  : optimized memcpy.
 CC_MEMMOVE : optimized memmove.
 CC_MEMSET  : optimized memset.

 CC_STORE32_BE : store 32 bit value in big endian in unaligned buffer.
 CC_STORE32_LE : store 32 bit value in little endian in unaligned buffer.
 CC_STORE64_BE : store 64 bit value in big endian in unaligned buffer.
 CC_STORE64_LE : store 64 bit value in little endian in unaligned buffer.

 CC_LOAD32_BE : load 32 bit value in big endian from unaligned buffer.
 CC_LOAD32_LE : load 32 bit value in little endian from unaligned buffer.
 CC_LOAD64_BE : load 64 bit value in big endian from unaligned buffer.
 CC_LOAD64_LE : load 64 bit value in little endian from unaligned buffer.

 CC_ROR  : Rotate Right 32 bits. Rotate count can be a variable.
 CC_ROL  : Rotate Left 32 bits. Rotate count can be a variable.
 CC_RORc : Rotate Right 32 bits. Rotate count must be a constant.
 CC_ROLc : Rotate Left 32 bits. Rotate count must be a constant.

 CC_ROR64  : Rotate Right 64 bits. Rotate count can be a variable.
 CC_ROL64  : Rotate Left 64 bits. Rotate count can be a variable.
 CC_ROR64c : Rotate Right 64 bits. Rotate count must be a constant.
 CC_ROL64c : Rotate Left 64 bits. Rotate count must be a constant.

 CC_BSWAP  : byte swap a 32 bits variable.

 CC_H2BE32 : convert a 32 bits value between host and big endian order.
 CC_H2LE32 : convert a 32 bits value between host and little endian order.

The following are not defined yet... define them if needed.

 CC_BSWAPc   : byte swap a 32 bits constant

 CC_BSWAP64  : byte swap a 64 bits variable
 CC_BSWAP64c : byte swap a 64 bits constant

 CC_READ_LE32 : read a 32 bits little endian value
 CC_READ_LE64 : read a 64 bits little endian value
 CC_READ_BE32 : read a 32 bits big endian value
 CC_READ_BE64 : read a 64 bits big endian value

 CC_WRITE_LE32 : write a 32 bits little endian value
 CC_WRITE_LE64 : write a 64 bits little endian value
 CC_WRITE_BE32 : write a 32 bits big endian value
 CC_WRITE_BE64 : write a 64 bits big endian value

 CC_H2BE64 : convert a 64 bits value between host and big endian order
 CC_H2LE64 : convert a 64 bits value between host and little endian order

*/

/* TODO: optimized versions */
#define CC_MEMCPY(D,S,L) memcpy((D),(S),(L))
#define CC_MEMMOVE(D,S,L) memmove((D),(S),(L))
#define CC_MEMSET(D,V,L) memset((D),(V),(L))

// MARK: - Loads and Store

// MARK: -- 32 bits - little endian

// MARK: --- Default version

#define	CC_STORE32_LE(x, y) do {                                    \
    ((unsigned char *)(y))[3] = (unsigned char)(((x)>>24)&255);		\
    ((unsigned char *)(y))[2] = (unsigned char)(((x)>>16)&255);		\
    ((unsigned char *)(y))[1] = (unsigned char)(((x)>>8)&255);		\
    ((unsigned char *)(y))[0] = (unsigned char)((x)&255);			\
} while(0)

#define	CC_LOAD32_LE(x, y) do {                                     \
x = ((uint32_t)(((const unsigned char *)(y))[3] & 255)<<24) |			    \
    ((uint32_t)(((const unsigned char *)(y))[2] & 255)<<16) |			    \
    ((uint32_t)(((const unsigned char *)(y))[1] & 255)<<8)  |			    \
    ((uint32_t)(((const unsigned char *)(y))[0] & 255));				    \
} while(0)

// MARK: -- 64 bits - little endian

#define	CC_STORE64_LE(x, y) do {                                    \
    ((unsigned char *)(y))[7] = (unsigned char)(((x)>>56)&255);     \
    ((unsigned char *)(y))[6] = (unsigned char)(((x)>>48)&255);		\
    ((unsigned char *)(y))[5] = (unsigned char)(((x)>>40)&255);		\
    ((unsigned char *)(y))[4] = (unsigned char)(((x)>>32)&255);		\
    ((unsigned char *)(y))[3] = (unsigned char)(((x)>>24)&255);		\
    ((unsigned char *)(y))[2] = (unsigned char)(((x)>>16)&255);		\
    ((unsigned char *)(y))[1] = (unsigned char)(((x)>>8)&255);		\
    ((unsigned char *)(y))[0] = (unsigned char)((x)&255);			\
} while(0)

#define	CC_LOAD64_LE(x, y) do {                                     \
x = (((uint64_t)(((const unsigned char *)(y))[7] & 255))<<56) |           \
    (((uint64_t)(((const unsigned char *)(y))[6] & 255))<<48) |           \
    (((uint64_t)(((const unsigned char *)(y))[5] & 255))<<40) |           \
    (((uint64_t)(((const unsigned char *)(y))[4] & 255))<<32) |           \
    (((uint64_t)(((const unsigned char *)(y))[3] & 255))<<24) |           \
    (((uint64_t)(((const unsigned char *)(y))[2] & 255))<<16) |           \
    (((uint64_t)(((const unsigned char *)(y))[1] & 255))<<8)  |           \
    (((uint64_t)(((const unsigned char *)(y))[0] & 255)));                \
} while(0)

// MARK: -- 32 bits - big endian
// MARK: --- intel version

#if (defined(__i386__) || defined(__x86_64__)) && !defined(_MSC_VER)

#define CC_STORE32_BE(x, y)     \
    __asm__ __volatile__ (      \
    "bswapl %0     \n\t"        \
    "movl   %0,(%1)\n\t"        \
    "bswapl %0     \n\t"        \
    ::"r"(x), "r"(y))

#define CC_LOAD32_BE(x, y)      \
    __asm__ __volatile__ (      \
    "movl (%1),%0\n\t"          \
    "bswapl %0\n\t"             \
    :"=r"(x): "r"(y))

#else
// MARK: --- default version
#define	CC_STORE32_BE(x, y) do {                                \
    ((unsigned char *)(y))[0] = (unsigned char)(((x)>>24)&255);	\
    ((unsigned char *)(y))[1] = (unsigned char)(((x)>>16)&255);	\
    ((unsigned char *)(y))[2] = (unsigned char)(((x)>>8)&255);	\
    ((unsigned char *)(y))[3] = (unsigned char)((x)&255);       \
} while(0)

#define	CC_LOAD32_BE(x, y) do {                             \
x = ((uint32_t)(((const unsigned char *)(y))[0] & 255)<<24) |	    \
    ((uint32_t)(((const unsigned char *)(y))[1] & 255)<<16) |		\
    ((uint32_t)(((const unsigned char *)(y))[2] & 255)<<8)  |		\
    ((uint32_t)(((const unsigned char *)(y))[3] & 255));          \
} while(0)

#endif

// MARK: -- 64 bits - big endian

// MARK: --- intel 64 bits version

#if defined(__x86_64__) && !defined (_MSC_VER)

#define	CC_STORE64_BE(x, y)   \
__asm__ __volatile__ (        \
"bswapq %0     \n\t"          \
"movq   %0,(%1)\n\t"          \
"bswapq %0     \n\t"          \
::"r"(x), "r"(y))

#define	CC_LOAD64_BE(x, y)    \
__asm__ __volatile__ (        \
"movq (%1),%0\n\t"            \
"bswapq %0\n\t"               \
:"=r"(x): "r"(y))

#else

// MARK: --- default version

#define CC_STORE64_BE(x, y) do {                                    \
    ((unsigned char *)(y))[0] = (unsigned char)(((x)>>56)&255);		\
    ((unsigned char *)(y))[1] = (unsigned char)(((x)>>48)&255);		\
    ((unsigned char *)(y))[2] = (unsigned char)(((x)>>40)&255);		\
    ((unsigned char *)(y))[3] = (unsigned char)(((x)>>32)&255);		\
    ((unsigned char *)(y))[4] = (unsigned char)(((x)>>24)&255);		\
    ((unsigned char *)(y))[5] = (unsigned char)(((x)>>16)&255);		\
    ((unsigned char *)(y))[6] = (unsigned char)(((x)>>8)&255);		\
    ((unsigned char *)(y))[7] = (unsigned char)((x)&255);			\
} while(0)

#define	CC_LOAD64_BE(x, y) do {                                     \
x = (((uint64_t)(((const unsigned char *)(y))[0] & 255))<<56) |           \
    (((uint64_t)(((const unsigned char *)(y))[1] & 255))<<48) |           \
    (((uint64_t)(((const unsigned char *)(y))[2] & 255))<<40) |           \
    (((uint64_t)(((const unsigned char *)(y))[3] & 255))<<32) |           \
    (((uint64_t)(((const unsigned char *)(y))[4] & 255))<<24) |           \
    (((uint64_t)(((const unsigned char *)(y))[5] & 255))<<16) |           \
    (((uint64_t)(((const unsigned char *)(y))[6] & 255))<<8)  |          	\
    (((uint64_t)(((const unsigned char *)(y))[7] & 255)));	            \
} while(0)

#endif

// MARK: - 32-bit Rotates

#if defined(_MSC_VER)
// MARK: -- MSVC version

#include <stdlib.h>
#if !defined(__clang__)
 #pragma intrinsic(_lrotr,_lrotl)
#endif
#define	CC_ROR(x,n) _lrotr(x,n)
#define	CC_ROL(x,n) _lrotl(x,n)
#define	CC_RORc(x,n) _lrotr(x,n)
#define	CC_ROLc(x,n) _lrotl(x,n)

#elif (defined(__i386__) || defined(__x86_64__))
// MARK: -- intel asm version

CC_INLINE uint32_t CC_ROL(uint32_t word, int i)
{
    __asm__ ("roll %%cl,%0"
         :"=r" (word)
         :"0" (word),"c" (i));
    return word;
}

CC_INLINE uint32_t CC_ROR(uint32_t word, int i)
{
    __asm__ ("rorl %%cl,%0"
         :"=r" (word)
         :"0" (word),"c" (i));
    return word;
}

/* Need to be a macro here, because 'i' is an immediate (constant) */
#define CC_ROLc(word, i)                \
({  uint32_t _word=(word);              \
    __asm__ __volatile__ ("roll %2,%0"  \
        :"=r" (_word)                   \
        :"0" (_word),"I" (i));          \
    _word;                              \
})


#define CC_RORc(word, i)                \
({  uint32_t _word=(word);              \
    __asm__ __volatile__ ("rorl %2,%0"  \
        :"=r" (_word)                   \
        :"0" (_word),"I" (i));          \
    _word;                              \
})

#else

// MARK: -- default version

CC_INLINE uint32_t CC_ROL(uint32_t word, int i)
{
    return ( (word<<(i&31)) | (word>>(32-(i&31))) );
}

CC_INLINE uint32_t CC_ROR(uint32_t word, int i)
{
    return ( (word>>(i&31)) | (word<<(32-(i&31))) );
}

#define	CC_ROLc(x, y) CC_ROL(x, y)
#define	CC_RORc(x, y) CC_ROR(x, y)

#endif

// MARK: - 64 bits rotates

#if defined(__x86_64__) && !defined(_MSC_VER) //clang _MSVC doesn't support GNU-style inline assembly
// MARK: -- intel 64 asm version

CC_INLINE uint64_t CC_ROL64(uint64_t word, int i)
{
    __asm__("rolq %%cl,%0"
        :"=r" (word)
        :"0" (word),"c" (i));
    return word;
}

CC_INLINE uint64_t CC_ROR64(uint64_t word, int i)
{
    __asm__("rorq %%cl,%0"
        :"=r" (word)
        :"0" (word),"c" (i));
    return word;
}

/* Need to be a macro here, because 'i' is an immediate (constant) */
#define CC_ROL64c(word, i)      \
({                              \
    uint64_t _word=(word);      \
    __asm__("rolq %2,%0"        \
        :"=r" (_word)           \
        :"0" (_word),"J" (i));  \
    _word;                      \
})

#define CC_ROR64c(word, i)      \
({                              \
    uint64_t _word=(word);      \
    __asm__("rorq %2,%0"        \
        :"=r" (_word)           \
        :"0" (_word),"J" (i));  \
    _word;                      \
})


#else /* Not x86_64  */

// MARK: -- default C version

CC_INLINE uint64_t CC_ROL64(uint64_t word, int i)
{
    return ( (word<<(i&63)) | (word>>(64-(i&63))) );
}

CC_INLINE uint64_t CC_ROR64(uint64_t word, int i)
{
    return ( (word>>(i&63)) | (word<<(64-(i&63))) );
}

#define	CC_ROL64c(x, y) CC_ROL64(x, y)
#define	CC_ROR64c(x, y) CC_ROR64(x, y)

#endif


// MARK: - Byte Swaps

CC_INLINE uint32_t CC_BSWAP(uint32_t x)
{
    return (
        ((x>>24)&0x000000FF) |
        ((x<<24)&0xFF000000) |
        ((x>>8) &0x0000FF00) |
        ((x<<8) &0x00FF0000)
    );
}

#define CC_BSWAP64(x) \
((uint64_t)((((uint64_t)(x) & 0xff00000000000000ULL) >> 56) | \
(((uint64_t)(x) & 0x00ff000000000000ULL) >> 40) | \
(((uint64_t)(x) & 0x0000ff0000000000ULL) >> 24) | \
(((uint64_t)(x) & 0x000000ff00000000ULL) >>  8) | \
(((uint64_t)(x) & 0x00000000ff000000ULL) <<  8) | \
(((uint64_t)(x) & 0x0000000000ff0000ULL) << 24) | \
(((uint64_t)(x) & 0x000000000000ff00ULL) << 40) | \
(((uint64_t)(x) & 0x00000000000000ffULL) << 56)))

#ifdef __LITTLE_ENDIAN__
#define CC_H2BE32(x) CC_BSWAP(x)
#define CC_H2LE32(x) (x)
#else
#define CC_H2BE32(x) (x)
#define CC_H2LE32(x) CC_BSWAP(x)
#endif


/* extract a byte portably */
#ifdef _MSC_VER
#define cc_byte(x, n) ((unsigned char)((x) >> (8 * (n))))
#else
#define cc_byte(x, n) (((x) >> (8 * (n))) & 255)
#endif

/* HEAVISIDE_STEP (shifted by one)
   function f(x): x->0, when x=0 
                  x->1, when x>0
   Can also be seen as a bitwise operation: 
      f(x): x -> y
        y[0]=(OR x[i]) for all i (all bits)
        y[i]=0 for all i>0
   Run in constant time (log2(<bitsize of x>))  
   Useful to run constant time checks
*/
#define HEAVISIDE_STEP_UINT64(r,s) {uint64_t _t=s; \
    _t=(((_t)>>32) | (_t)); \
    _t=(0xFFFFFFFF + (_t & 0xFFFFFFFF)); \
    r=_t >> 32;}

#define HEAVISIDE_STEP_UINT32(r,s) {uint32_t _t=s; \
    _t=(((_t)>>16) | (_t)); \
    _t=(0xFFFF + (_t & 0xFFFF)); \
    r=_t >> 16;}

#define HEAVISIDE_STEP_UINT16(r,s) {uint32_t _t=s; \
    _t=(0xFFFF + ((_t) & 0xFFFF)); \
    r=_t >> 16;}

#define HEAVISIDE_STEP_UINT8(r,s) {uint16_t _t=s; \
    _t=(0xFF + ((_t) & 0xFF)); \
    r=_t >> 8;}

#define CC_HEAVISIDE_STEP(r,s) { \
    if (sizeof(s) == 1)      {HEAVISIDE_STEP_UINT8(r,s);}  \
    else if (sizeof(s) == 2) {HEAVISIDE_STEP_UINT16(r,s);} \
    else if (sizeof(s) == 4) {HEAVISIDE_STEP_UINT32(r,s);} \
    else if (sizeof(s) == 8) {HEAVISIDE_STEP_UINT64(r,s);} \
    else {r=(((s)==0)?0:1);} \
    }

/* Return 1 if x mod 4 =1,2,3, 0 otherwise */
#define CC_CARRY_2BITS(x) (((x>>1) | x) & 0x1)
#define CC_CARRY_3BITS(x) (((x>>2) | (x>>1) | x) & 0x1)

/* Set a variable to the biggest power of 2 which can be represented */ 
#define MAX_POWER_OF_2(x)   ((__typeof__(x))1<<(8*sizeof(x)-1))
#define cc_ceiling(a,b)  (((a)+((b)-1))/(b))
#define CC_BITLEN_TO_BYTELEN(x) cc_ceiling((x), 8)

//cc_abort() is implemented to comply with FIPS 140-2. See radar 19129408
void cc_abort(const char * msg , ...);

/*!
 @brief     cc_muxp(s, a, b) is equivalent to z = s ? a : b, but it executes in constant time
 @param a	input pointer
 @param b	input pointer
 @param s	The selection parameter s must be 0 or 1. if s is integer 1 a is returned. If s is integer 0, b is returned. Otherwise, the output is undefined.
 @return    Returns a, if s is 1 and b if s is 0
 */
void *cc_muxp(int s, const void *a, const void *b);

/*!
 @brief     cc_mux2p
 @param a	input pointer
 @param b	input pointer
 @param r_true	output pointer: if s is integer 1 r_true=a  is returned, otherwise r_true=b
 @param r_false	output pointer: if s is integer 1 r_false=b is returned, otherwise r_false=a
 @param s	The selection parameter s must be 0 or 1.
 @discussion Executes in constant time
 */
void cc_mux2p(int s, void **r_true, void **r_false, const void *a, const void *b);

/*!
 @brief     CC_MUXU(s, a, b) is equivalent to z = s ? a : b, but it executes in constant time
 @param a	input unsigned type
 @param b	input unsigned type
 @param s	The selection parameter s must be 0 or 1. if s is integer 1 a is returned. If s is integer 0, b is returned. Otherwise, the output is undefined.
 @param r	output
 @return    r = a, if s is 1 and b if s is 0
 */
#define CC_MUXU(r, s, a, b)   \
{                       \
    __typeof__(r) _cond = ((__typeof__(r))(s)-(__typeof__(r))1); \
    r = (~_cond&(a))|(_cond&(b)); \
}

int cc_is_compiled_with_tu(void);

#endif /* _CORECRYPTO_CC_PRIV_H_ */
Commit	Line	Data
316670eb A	1	/*
	2	* cc_priv.h
	3	* corecrypto
	4	*
3e170ce0 A	5	* Created on 12/01/2010
	6	*
	7	* Copyright (c) 2010,2011,2012,2014,2015 Apple Inc. All rights reserved.
316670eb A	8	*
	9	*/
	10
	11	#ifndef _CORECRYPTO_CC_PRIV_H_
	12	#define _CORECRYPTO_CC_PRIV_H_
	13
	14	#include <corecrypto/cc.h>
	15	#include <stdint.h>
	16
	17	/* defines the following macros :
	18
	19	CC_MEMCPY : optimized memcpy.
	20	CC_MEMMOVE : optimized memmove.
	21	CC_MEMSET : optimized memset.
316670eb A	22
	23	CC_STORE32_BE : store 32 bit value in big endian in unaligned buffer.
	24	CC_STORE32_LE : store 32 bit value in little endian in unaligned buffer.
	25	CC_STORE64_BE : store 64 bit value in big endian in unaligned buffer.
	26	CC_STORE64_LE : store 64 bit value in little endian in unaligned buffer.
	27
	28	CC_LOAD32_BE : load 32 bit value in big endian from unaligned buffer.
	29	CC_LOAD32_LE : load 32 bit value in little endian from unaligned buffer.
	30	CC_LOAD64_BE : load 64 bit value in big endian from unaligned buffer.
	31	CC_LOAD64_LE : load 64 bit value in little endian from unaligned buffer.
	32
	33	CC_ROR : Rotate Right 32 bits. Rotate count can be a variable.
	34	CC_ROL : Rotate Left 32 bits. Rotate count can be a variable.
	35	CC_RORc : Rotate Right 32 bits. Rotate count must be a constant.
	36	CC_ROLc : Rotate Left 32 bits. Rotate count must be a constant.
	37
	38	CC_ROR64 : Rotate Right 64 bits. Rotate count can be a variable.
	39	CC_ROL64 : Rotate Left 64 bits. Rotate count can be a variable.
	40	CC_ROR64c : Rotate Right 64 bits. Rotate count must be a constant.
	41	CC_ROL64c : Rotate Left 64 bits. Rotate count must be a constant.
	42
	43	CC_BSWAP : byte swap a 32 bits variable.
	44
	45	CC_H2BE32 : convert a 32 bits value between host and big endian order.
	46	CC_H2LE32 : convert a 32 bits value between host and little endian order.
	47
	48	The following are not defined yet... define them if needed.
	49
	50	CC_BSWAPc : byte swap a 32 bits constant
	51
	52	CC_BSWAP64 : byte swap a 64 bits variable
	53	CC_BSWAP64c : byte swap a 64 bits constant
	54
	55	CC_READ_LE32 : read a 32 bits little endian value
	56	CC_READ_LE64 : read a 64 bits little endian value
	57	CC_READ_BE32 : read a 32 bits big endian value
	58	CC_READ_BE64 : read a 64 bits big endian value
	59
	60	CC_WRITE_LE32 : write a 32 bits little endian value
	61	CC_WRITE_LE64 : write a 64 bits little endian value
	62	CC_WRITE_BE32 : write a 32 bits big endian value
	63	CC_WRITE_BE64 : write a 64 bits big endian value
	64
	65	CC_H2BE64 : convert a 64 bits value between host and big endian order
	66	CC_H2LE64 : convert a 64 bits value between host and little endian order
fe8ab488	67
316670eb A	68	*/
	69
	70	/* TODO: optimized versions */
	71	#define CC_MEMCPY(D,S,L) memcpy((D),(S),(L))
	72	#define CC_MEMMOVE(D,S,L) memmove((D),(S),(L))
	73	#define CC_MEMSET(D,V,L) memset((D),(V),(L))
316670eb	74
fe8ab488	75	// MARK: - Loads and Store
316670eb	76
fe8ab488	77	// MARK: -- 32 bits - little endian
316670eb	78
fe8ab488	79	// MARK: --- Default version
316670eb A	80
	81	#define CC_STORE32_LE(x, y) do { \
	82	((unsigned char *)(y))[3] = (unsigned char)(((x)>>24)&255); \
	83	((unsigned char *)(y))[2] = (unsigned char)(((x)>>16)&255); \
	84	((unsigned char *)(y))[1] = (unsigned char)(((x)>>8)&255); \
	85	((unsigned char *)(y))[0] = (unsigned char)((x)&255); \
	86	} while(0)
	87
	88	#define CC_LOAD32_LE(x, y) do { \
3e170ce0 A	89	x = ((uint32_t)(((const unsigned char *)(y))[3] & 255)<<24) \| \
	90	((uint32_t)(((const unsigned char *)(y))[2] & 255)<<16) \| \
	91	((uint32_t)(((const unsigned char *)(y))[1] & 255)<<8) \| \
	92	((uint32_t)(((const unsigned char *)(y))[0] & 255)); \
316670eb A	93	} while(0)
316670eb A	94
fe8ab488	95	// MARK: -- 64 bits - little endian
316670eb A	96
	97	#define CC_STORE64_LE(x, y) do { \
	98	((unsigned char *)(y))[7] = (unsigned char)(((x)>>56)&255); \
	99	((unsigned char *)(y))[6] = (unsigned char)(((x)>>48)&255); \
	100	((unsigned char *)(y))[5] = (unsigned char)(((x)>>40)&255); \
	101	((unsigned char *)(y))[4] = (unsigned char)(((x)>>32)&255); \
	102	((unsigned char *)(y))[3] = (unsigned char)(((x)>>24)&255); \
	103	((unsigned char *)(y))[2] = (unsigned char)(((x)>>16)&255); \
	104	((unsigned char *)(y))[1] = (unsigned char)(((x)>>8)&255); \
	105	((unsigned char *)(y))[0] = (unsigned char)((x)&255); \
	106	} while(0)
	107
	108	#define CC_LOAD64_LE(x, y) do { \
3e170ce0 A	109	x = (((uint64_t)(((const unsigned char *)(y))[7] & 255))<<56) \| \
	110	(((uint64_t)(((const unsigned char *)(y))[6] & 255))<<48) \| \
	111	(((uint64_t)(((const unsigned char *)(y))[5] & 255))<<40) \| \
	112	(((uint64_t)(((const unsigned char *)(y))[4] & 255))<<32) \| \
	113	(((uint64_t)(((const unsigned char *)(y))[3] & 255))<<24) \| \
	114	(((uint64_t)(((const unsigned char *)(y))[2] & 255))<<16) \| \
	115	(((uint64_t)(((const unsigned char *)(y))[1] & 255))<<8) \| \
	116	(((uint64_t)(((const unsigned char *)(y))[0] & 255))); \
316670eb A	117	} while(0)
316670eb A	118
fe8ab488 A	119	// MARK: -- 32 bits - big endian
fe8ab488 A	120	// MARK: --- intel version
316670eb	121
39037602	122	#if (defined(__i386__) \|\| defined(__x86_64__)) && !defined(_MSC_VER)
316670eb A	123
	124	#define CC_STORE32_BE(x, y) \
	125	__asm__ __volatile__ ( \
	126	"bswapl %0 \n\t" \
	127	"movl %0,(%1)\n\t" \
	128	"bswapl %0 \n\t" \
	129	::"r"(x), "r"(y))
	130
	131	#define CC_LOAD32_BE(x, y) \
	132	__asm__ __volatile__ ( \
	133	"movl (%1),%0\n\t" \
	134	"bswapl %0\n\t" \
	135	:"=r"(x): "r"(y))
	136
	137	#else
fe8ab488	138	// MARK: --- default version
316670eb A	139	#define CC_STORE32_BE(x, y) do { \
	140	((unsigned char *)(y))[0] = (unsigned char)(((x)>>24)&255); \
	141	((unsigned char *)(y))[1] = (unsigned char)(((x)>>16)&255); \
	142	((unsigned char *)(y))[2] = (unsigned char)(((x)>>8)&255); \
	143	((unsigned char *)(y))[3] = (unsigned char)((x)&255); \
	144	} while(0)
	145
	146	#define CC_LOAD32_BE(x, y) do { \
3e170ce0 A	147	x = ((uint32_t)(((const unsigned char *)(y))[0] & 255)<<24) \| \
	148	((uint32_t)(((const unsigned char *)(y))[1] & 255)<<16) \| \
	149	((uint32_t)(((const unsigned char *)(y))[2] & 255)<<8) \| \
	150	((uint32_t)(((const unsigned char *)(y))[3] & 255)); \
316670eb A	151	} while(0)
	152
	153	#endif
	154
fe8ab488	155	// MARK: -- 64 bits - big endian
316670eb	156
fe8ab488	157	// MARK: --- intel 64 bits version
316670eb	158
39037602	159	#if defined(__x86_64__) && !defined (_MSC_VER)
316670eb A	160
	161	#define CC_STORE64_BE(x, y) \
	162	__asm__ __volatile__ ( \
	163	"bswapq %0 \n\t" \
	164	"movq %0,(%1)\n\t" \
	165	"bswapq %0 \n\t" \
	166	::"r"(x), "r"(y))
	167
	168	#define CC_LOAD64_BE(x, y) \
	169	__asm__ __volatile__ ( \
	170	"movq (%1),%0\n\t" \
	171	"bswapq %0\n\t" \
	172	:"=r"(x): "r"(y))
	173
	174	#else
	175
fe8ab488	176	// MARK: --- default version
316670eb A	177
	178	#define CC_STORE64_BE(x, y) do { \
	179	((unsigned char *)(y))[0] = (unsigned char)(((x)>>56)&255); \
	180	((unsigned char *)(y))[1] = (unsigned char)(((x)>>48)&255); \
	181	((unsigned char *)(y))[2] = (unsigned char)(((x)>>40)&255); \
	182	((unsigned char *)(y))[3] = (unsigned char)(((x)>>32)&255); \
	183	((unsigned char *)(y))[4] = (unsigned char)(((x)>>24)&255); \
	184	((unsigned char *)(y))[5] = (unsigned char)(((x)>>16)&255); \
	185	((unsigned char *)(y))[6] = (unsigned char)(((x)>>8)&255); \
	186	((unsigned char *)(y))[7] = (unsigned char)((x)&255); \
	187	} while(0)
	188
	189	#define CC_LOAD64_BE(x, y) do { \
3e170ce0 A	190	x = (((uint64_t)(((const unsigned char *)(y))[0] & 255))<<56) \| \
	191	(((uint64_t)(((const unsigned char *)(y))[1] & 255))<<48) \| \
	192	(((uint64_t)(((const unsigned char *)(y))[2] & 255))<<40) \| \
	193	(((uint64_t)(((const unsigned char *)(y))[3] & 255))<<32) \| \
	194	(((uint64_t)(((const unsigned char *)(y))[4] & 255))<<24) \| \
	195	(((uint64_t)(((const unsigned char *)(y))[5] & 255))<<16) \| \
	196	(((uint64_t)(((const unsigned char *)(y))[6] & 255))<<8) \| \
	197	(((uint64_t)(((const unsigned char *)(y))[7] & 255))); \
316670eb A	198	} while(0)
	199
	200	#endif
	201
fe8ab488	202	// MARK: - 32-bit Rotates
316670eb A	203
316670eb A	204	#if defined(_MSC_VER)
fe8ab488	205	// MARK: -- MSVC version
316670eb A	206
316670eb A	207	#include <stdlib.h>
39037602 A	208	#if !defined(__clang__)
	209	#pragma intrinsic(_lrotr,_lrotl)
	210	#endif
316670eb A	211	#define CC_ROR(x,n) _lrotr(x,n)
	212	#define CC_ROL(x,n) _lrotl(x,n)
	213	#define CC_RORc(x,n) _lrotr(x,n)
	214	#define CC_ROLc(x,n) _lrotl(x,n)
	215
	216	#elif (defined(__i386__) \|\| defined(__x86_64__))
fe8ab488	217	// MARK: -- intel asm version
316670eb	218
39037602	219	CC_INLINE uint32_t CC_ROL(uint32_t word, int i)
316670eb A	220	{
	221	__asm__ ("roll %%cl,%0"
	222	:"=r" (word)
	223	:"0" (word),"c" (i));
	224	return word;
	225	}
	226
39037602	227	CC_INLINE uint32_t CC_ROR(uint32_t word, int i)
316670eb A	228	{
	229	__asm__ ("rorl %%cl,%0"
	230	:"=r" (word)
	231	:"0" (word),"c" (i));
	232	return word;
	233	}
	234
	235	/* Need to be a macro here, because 'i' is an immediate (constant) */
	236	#define CC_ROLc(word, i) \
	237	({ uint32_t _word=(word); \
	238	__asm__ __volatile__ ("roll %2,%0" \
	239	:"=r" (_word) \
	240	:"0" (_word),"I" (i)); \
	241	_word; \
	242	})
	243
	244
	245	#define CC_RORc(word, i) \
	246	({ uint32_t _word=(word); \
	247	__asm__ __volatile__ ("rorl %2,%0" \
	248	:"=r" (_word) \
	249	:"0" (_word),"I" (i)); \
	250	_word; \
	251	})
	252
	253	#else
	254
fe8ab488	255	// MARK: -- default version
316670eb	256
39037602	257	CC_INLINE uint32_t CC_ROL(uint32_t word, int i)
316670eb A	258	{
	259	return ( (word<<(i&31)) \| (word>>(32-(i&31))) );
	260	}
	261
39037602	262	CC_INLINE uint32_t CC_ROR(uint32_t word, int i)
316670eb A	263	{
	264	return ( (word>>(i&31)) \| (word<<(32-(i&31))) );
	265	}
	266
	267	#define CC_ROLc(x, y) CC_ROL(x, y)
	268	#define CC_RORc(x, y) CC_ROR(x, y)
	269
	270	#endif
	271
fe8ab488	272	// MARK: - 64 bits rotates
316670eb	273
39037602	274	#if defined(__x86_64__) && !defined(_MSC_VER) //clang _MSVC doesn't support GNU-style inline assembly
fe8ab488	275	// MARK: -- intel 64 asm version
316670eb	276
39037602	277	CC_INLINE uint64_t CC_ROL64(uint64_t word, int i)
316670eb A	278	{
	279	__asm__("rolq %%cl,%0"
	280	:"=r" (word)
	281	:"0" (word),"c" (i));
	282	return word;
	283	}
	284
39037602	285	CC_INLINE uint64_t CC_ROR64(uint64_t word, int i)
316670eb A	286	{
	287	__asm__("rorq %%cl,%0"
	288	:"=r" (word)
	289	:"0" (word),"c" (i));
	290	return word;
	291	}
	292
	293	/* Need to be a macro here, because 'i' is an immediate (constant) */
	294	#define CC_ROL64c(word, i) \
	295	({ \
	296	uint64_t _word=(word); \
	297	__asm__("rolq %2,%0" \
	298	:"=r" (_word) \
	299	:"0" (_word),"J" (i)); \
	300	_word; \
	301	})
	302
	303	#define CC_ROR64c(word, i) \
	304	({ \
	305	uint64_t _word=(word); \
	306	__asm__("rorq %2,%0" \
	307	:"=r" (_word) \
	308	:"0" (_word),"J" (i)); \
	309	_word; \
	310	})
	311
	312
	313	#else /* Not x86_64 */
	314
fe8ab488	315	// MARK: -- default C version
316670eb	316
39037602	317	CC_INLINE uint64_t CC_ROL64(uint64_t word, int i)
316670eb A	318	{
	319	return ( (word<<(i&63)) \| (word>>(64-(i&63))) );
	320	}
	321
39037602	322	CC_INLINE uint64_t CC_ROR64(uint64_t word, int i)
316670eb A	323	{
	324	return ( (word>>(i&63)) \| (word<<(64-(i&63))) );
	325	}
	326
	327	#define CC_ROL64c(x, y) CC_ROL64(x, y)
	328	#define CC_ROR64c(x, y) CC_ROR64(x, y)
	329
	330	#endif
	331
	332
fe8ab488	333	// MARK: - Byte Swaps
316670eb	334
39037602	335	CC_INLINE uint32_t CC_BSWAP(uint32_t x)
316670eb A	336	{
	337	return (
	338	((x>>24)&0x000000FF) \|
	339	((x<<24)&0xFF000000) \|
	340	((x>>8) &0x0000FF00) \|
	341	((x<<8) &0x00FF0000)
	342	);
	343	}
	344
fe8ab488 A	345	#define CC_BSWAP64(x) \
	346	((uint64_t)((((uint64_t)(x) & 0xff00000000000000ULL) >> 56) \| \
	347	(((uint64_t)(x) & 0x00ff000000000000ULL) >> 40) \| \
	348	(((uint64_t)(x) & 0x0000ff0000000000ULL) >> 24) \| \
	349	(((uint64_t)(x) & 0x000000ff00000000ULL) >> 8) \| \
	350	(((uint64_t)(x) & 0x00000000ff000000ULL) << 8) \| \
	351	(((uint64_t)(x) & 0x0000000000ff0000ULL) << 24) \| \
	352	(((uint64_t)(x) & 0x000000000000ff00ULL) << 40) \| \
	353	(((uint64_t)(x) & 0x00000000000000ffULL) << 56)))
	354
316670eb A	355	#ifdef __LITTLE_ENDIAN__
	356	#define CC_H2BE32(x) CC_BSWAP(x)
	357	#define CC_H2LE32(x) (x)
	358	#else
316670eb A	359	#define CC_H2BE32(x) (x)
	360	#define CC_H2LE32(x) CC_BSWAP(x)
	361	#endif
	362
	363
	364	/* extract a byte portably */
	365	#ifdef _MSC_VER
	366	#define cc_byte(x, n) ((unsigned char)((x) >> (8 * (n))))
	367	#else
	368	#define cc_byte(x, n) (((x) >> (8 * (n))) & 255)
	369	#endif
	370
fe8ab488 A	371	/* HEAVISIDE_STEP (shifted by one)
	372	function f(x): x->0, when x=0
	373	x->1, when x>0
	374	Can also be seen as a bitwise operation:
	375	f(x): x -> y
	376	y[0]=(OR x[i]) for all i (all bits)
	377	y[i]=0 for all i>0
	378	Run in constant time (log2(<bitsize of x>))
	379	Useful to run constant time checks
	380	*/
39037602 A	381	#define HEAVISIDE_STEP_UINT64(r,s) {uint64_t _t=s; \
39037602 A	382	_t=(((_t)>>32) \| (_t)); \
3e170ce0	383	_t=(0xFFFFFFFF + (_t & 0xFFFFFFFF)); \
39037602	384	r=_t >> 32;}
3e170ce0	385
39037602 A	386	#define HEAVISIDE_STEP_UINT32(r,s) {uint32_t _t=s; \
39037602 A	387	_t=(((_t)>>16) \| (_t)); \
3e170ce0	388	_t=(0xFFFF + (_t & 0xFFFF)); \
39037602 A	389	r=_t >> 16;}
	390
	391	#define HEAVISIDE_STEP_UINT16(r,s) {uint32_t _t=s; \
	392	_t=(0xFFFF + ((_t) & 0xFFFF)); \
	393	r=_t >> 16;}
	394
	395	#define HEAVISIDE_STEP_UINT8(r,s) {uint16_t _t=s; \
	396	_t=(0xFF + ((_t) & 0xFF)); \
	397	r=_t >> 8;}
	398
	399	#define CC_HEAVISIDE_STEP(r,s) { \
	400	if (sizeof(s) == 1) {HEAVISIDE_STEP_UINT8(r,s);} \
	401	else if (sizeof(s) == 2) {HEAVISIDE_STEP_UINT16(r,s);} \
	402	else if (sizeof(s) == 4) {HEAVISIDE_STEP_UINT32(r,s);} \
	403	else if (sizeof(s) == 8) {HEAVISIDE_STEP_UINT64(r,s);} \
	404	else {r=(((s)==0)?0:1);} \
fe8ab488 A	405	}
fe8ab488 A	406
3e170ce0 A	407	/* Return 1 if x mod 4 =1,2,3, 0 otherwise */
	408	#define CC_CARRY_2BITS(x) (((x>>1) \| x) & 0x1)
	409	#define CC_CARRY_3BITS(x) (((x>>2) \| (x>>1) \| x) & 0x1)
fe8ab488 A	410
	411	/* Set a variable to the biggest power of 2 which can be represented */
	412	#define MAX_POWER_OF_2(x) ((__typeof__(x))1<<(8*sizeof(x)-1))
3e170ce0 A	413	#define cc_ceiling(a,b) (((a)+((b)-1))/(b))
3e170ce0 A	414	#define CC_BITLEN_TO_BYTELEN(x) cc_ceiling((x), 8)
fe8ab488	415
39037602 A	416	//cc_abort() is implemented to comply with FIPS 140-2. See radar 19129408
	417	void cc_abort(const char * msg , ...);
	418
	419	/*!
	420	@brief cc_muxp(s, a, b) is equivalent to z = s ? a : b, but it executes in constant time
	421	@param a input pointer
	422	@param b input pointer
	423	@param s The selection parameter s must be 0 or 1. if s is integer 1 a is returned. If s is integer 0, b is returned. Otherwise, the output is undefined.
	424	@return Returns a, if s is 1 and b if s is 0
	425	*/
	426	void cc_muxp(int s, const void a, const void *b);
	427
	428	/*!
	429	@brief cc_mux2p
	430	@param a input pointer
	431	@param b input pointer
	432	@param r_true output pointer: if s is integer 1 r_true=a is returned, otherwise r_true=b
	433	@param r_false output pointer: if s is integer 1 r_false=b is returned, otherwise r_false=a
	434	@param s The selection parameter s must be 0 or 1.
	435	@discussion Executes in constant time
	436	*/
	437	void cc_mux2p(int s, void r_true, void r_false, const void a, const void b);
	438
	439	/*!
	440	@brief CC_MUXU(s, a, b) is equivalent to z = s ? a : b, but it executes in constant time
	441	@param a input unsigned type
	442	@param b input unsigned type
	443	@param s The selection parameter s must be 0 or 1. if s is integer 1 a is returned. If s is integer 0, b is returned. Otherwise, the output is undefined.
	444	@param r output
	445	@return r = a, if s is 1 and b if s is 0
	446	*/
	447	#define CC_MUXU(r, s, a, b) \
	448	{ \
	449	__typeof__(r) _cond = ((__typeof__(r))(s)-(__typeof__(r))1); \
	450	r = (~_cond&(a))\|(_cond&(b)); \
	451	}
	452
	453	int cc_is_compiled_with_tu(void);
	454
316670eb	455	#endif /* _CORECRYPTO_CC_PRIV_H_ */