andre@0: /* arcfour.c - the arc four algorithm.
andre@0:  *
andre@0:  * This Source Code Form is subject to the terms of the Mozilla Public
andre@0:  * License, v. 2.0. If a copy of the MPL was not distributed with this
andre@0:  * file, You can obtain one at http://mozilla.org/MPL/2.0/. */
andre@0: 
andre@0: #ifdef FREEBL_NO_DEPEND
andre@0: #include "stubs.h"
andre@0: #endif
andre@0: 
andre@0: #include "prerr.h"
andre@0: #include "secerr.h"
andre@0: 
andre@0: #include "prtypes.h"
andre@0: #include "blapi.h"
andre@0: 
andre@0: /* Architecture-dependent defines */
andre@0: 
andre@0: #if defined(SOLARIS) || defined(HPUX) || defined(NSS_X86) || \
andre@0:     defined(_WIN64)
andre@0: /* Convert the byte-stream to a word-stream */
andre@0: #define CONVERT_TO_WORDS
andre@0: #endif
andre@0: 
andre@0: #if defined(AIX) || defined(OSF1) || defined(NSS_BEVAND_ARCFOUR)
andre@0: /* Treat array variables as words, not bytes, on CPUs that take 
andre@0:  * much longer to write bytes than to write words, or when using 
andre@0:  * assembler code that required it.
andre@0:  */
andre@0: #define USE_WORD
andre@0: #endif
andre@0: 
andre@0: #if defined(IS_64) || defined(NSS_BEVAND_ARCFOUR)
andre@0: typedef PRUint64 WORD;
andre@0: #else
andre@0: typedef PRUint32 WORD;
andre@0: #endif
andre@0: #define WORDSIZE sizeof(WORD)
andre@0: 
andre@0: #if defined(USE_WORD)
andre@0: typedef WORD Stype;
andre@0: #else
andre@0: typedef PRUint8 Stype;
andre@0: #endif
andre@0: 
andre@0: #define ARCFOUR_STATE_SIZE 256
andre@0: 
andre@0: #define MASK1BYTE (WORD)(0xff)
andre@0: 
andre@0: #define SWAP(a, b) \
andre@0: 	tmp = a; \
andre@0: 	a = b; \
andre@0: 	b = tmp;
andre@0: 
andre@0: /*
andre@0:  * State information for stream cipher.
andre@0:  */
andre@0: struct RC4ContextStr
andre@0: {
andre@0: #if defined(NSS_ARCFOUR_IJ_B4_S) || defined(NSS_BEVAND_ARCFOUR)
andre@0: 	Stype i;
andre@0: 	Stype j;
andre@0: 	Stype S[ARCFOUR_STATE_SIZE];
andre@0: #else
andre@0: 	Stype S[ARCFOUR_STATE_SIZE];
andre@0: 	Stype i;
andre@0: 	Stype j;
andre@0: #endif
andre@0: };
andre@0: 
andre@0: /*
andre@0:  * array indices [0..255] to initialize cx->S array (faster than loop).
andre@0:  */
andre@0: static const Stype Kinit[256] = {
andre@0: 	0x00, 0x01, 0x02, 0x03, 0x04, 0x05, 0x06, 0x07,
andre@0: 	0x08, 0x09, 0x0a, 0x0b, 0x0c, 0x0d, 0x0e, 0x0f,
andre@0: 	0x10, 0x11, 0x12, 0x13, 0x14, 0x15, 0x16, 0x17,
andre@0: 	0x18, 0x19, 0x1a, 0x1b, 0x1c, 0x1d, 0x1e, 0x1f,
andre@0: 	0x20, 0x21, 0x22, 0x23, 0x24, 0x25, 0x26, 0x27,
andre@0: 	0x28, 0x29, 0x2a, 0x2b, 0x2c, 0x2d, 0x2e, 0x2f,
andre@0: 	0x30, 0x31, 0x32, 0x33, 0x34, 0x35, 0x36, 0x37,
andre@0: 	0x38, 0x39, 0x3a, 0x3b, 0x3c, 0x3d, 0x3e, 0x3f,
andre@0: 	0x40, 0x41, 0x42, 0x43, 0x44, 0x45, 0x46, 0x47,
andre@0: 	0x48, 0x49, 0x4a, 0x4b, 0x4c, 0x4d, 0x4e, 0x4f,
andre@0: 	0x50, 0x51, 0x52, 0x53, 0x54, 0x55, 0x56, 0x57,
andre@0: 	0x58, 0x59, 0x5a, 0x5b, 0x5c, 0x5d, 0x5e, 0x5f,
andre@0: 	0x60, 0x61, 0x62, 0x63, 0x64, 0x65, 0x66, 0x67,
andre@0: 	0x68, 0x69, 0x6a, 0x6b, 0x6c, 0x6d, 0x6e, 0x6f,
andre@0: 	0x70, 0x71, 0x72, 0x73, 0x74, 0x75, 0x76, 0x77,
andre@0: 	0x78, 0x79, 0x7a, 0x7b, 0x7c, 0x7d, 0x7e, 0x7f,
andre@0: 	0x80, 0x81, 0x82, 0x83, 0x84, 0x85, 0x86, 0x87,
andre@0: 	0x88, 0x89, 0x8a, 0x8b, 0x8c, 0x8d, 0x8e, 0x8f,
andre@0: 	0x90, 0x91, 0x92, 0x93, 0x94, 0x95, 0x96, 0x97,
andre@0: 	0x98, 0x99, 0x9a, 0x9b, 0x9c, 0x9d, 0x9e, 0x9f,
andre@0: 	0xa0, 0xa1, 0xa2, 0xa3, 0xa4, 0xa5, 0xa6, 0xa7,
andre@0: 	0xa8, 0xa9, 0xaa, 0xab, 0xac, 0xad, 0xae, 0xaf,
andre@0: 	0xb0, 0xb1, 0xb2, 0xb3, 0xb4, 0xb5, 0xb6, 0xb7,
andre@0: 	0xb8, 0xb9, 0xba, 0xbb, 0xbc, 0xbd, 0xbe, 0xbf,
andre@0: 	0xc0, 0xc1, 0xc2, 0xc3, 0xc4, 0xc5, 0xc6, 0xc7,
andre@0: 	0xc8, 0xc9, 0xca, 0xcb, 0xcc, 0xcd, 0xce, 0xcf,
andre@0: 	0xd0, 0xd1, 0xd2, 0xd3, 0xd4, 0xd5, 0xd6, 0xd7,
andre@0: 	0xd8, 0xd9, 0xda, 0xdb, 0xdc, 0xdd, 0xde, 0xdf,
andre@0: 	0xe0, 0xe1, 0xe2, 0xe3, 0xe4, 0xe5, 0xe6, 0xe7,
andre@0: 	0xe8, 0xe9, 0xea, 0xeb, 0xec, 0xed, 0xee, 0xef,
andre@0: 	0xf0, 0xf1, 0xf2, 0xf3, 0xf4, 0xf5, 0xf6, 0xf7,
andre@0: 	0xf8, 0xf9, 0xfa, 0xfb, 0xfc, 0xfd, 0xfe, 0xff
andre@0: };
andre@0: 
andre@0: RC4Context *
andre@0: RC4_AllocateContext(void)
andre@0: {
andre@0:     return PORT_ZNew(RC4Context);
andre@0: }
andre@0: 
andre@0: SECStatus   
andre@0: RC4_InitContext(RC4Context *cx, const unsigned char *key, unsigned int len,
andre@0: 	        const unsigned char * unused1, int unused2, 
andre@0: 		unsigned int unused3, unsigned int unused4)
andre@0: {
andre@0: 	unsigned int i;
andre@0: 	PRUint8 j, tmp;
andre@0: 	PRUint8 K[256];
andre@0: 	PRUint8 *L;
andre@0: 
andre@0: 	/* verify the key length. */
andre@0: 	PORT_Assert(len > 0 && len < ARCFOUR_STATE_SIZE);
andre@0: 	if (len == 0 || len >= ARCFOUR_STATE_SIZE) {
andre@0: 		PORT_SetError(SEC_ERROR_BAD_KEY);
andre@0: 		return SECFailure;
andre@0: 	}
andre@0: 	if (cx == NULL) {
andre@0: 	    PORT_SetError(SEC_ERROR_INVALID_ARGS);
andre@0: 	    return SECFailure;
andre@0: 	}
andre@0: 	/* Initialize the state using array indices. */
andre@0: 	memcpy(cx->S, Kinit, sizeof cx->S);
andre@0: 	/* Fill in K repeatedly with values from key. */
andre@0: 	L = K;
andre@0: 	for (i = sizeof K; i > len; i-= len) {
andre@0: 		memcpy(L, key, len);
andre@0: 		L += len;
andre@0: 	}
andre@0: 	memcpy(L, key, i);
andre@0: 	/* Stir the state of the generator.  At this point it is assumed
andre@0: 	 * that the key is the size of the state buffer.  If this is not
andre@0: 	 * the case, the key bytes are repeated to fill the buffer.
andre@0: 	 */
andre@0: 	j = 0;
andre@0: #define ARCFOUR_STATE_STIR(ii) \
andre@0: 	j = j + cx->S[ii] + K[ii]; \
andre@0: 	SWAP(cx->S[ii], cx->S[j]);
andre@0: 	for (i=0; i<ARCFOUR_STATE_SIZE; i++) {
andre@0: 		ARCFOUR_STATE_STIR(i);
andre@0: 	}
andre@0: 	cx->i = 0;
andre@0: 	cx->j = 0;
andre@0: 	return SECSuccess;
andre@0: }
andre@0: 
andre@0: 
andre@0: /*
andre@0:  * Initialize a new generator.
andre@0:  */
andre@0: RC4Context *
andre@0: RC4_CreateContext(const unsigned char *key, int len)
andre@0: {
andre@0:     RC4Context *cx = RC4_AllocateContext();
andre@0:     if (cx) {
andre@0: 	SECStatus rv = RC4_InitContext(cx, key, len, NULL, 0, 0, 0);
andre@0: 	if (rv != SECSuccess) {
andre@0: 	    PORT_ZFree(cx, sizeof(*cx));
andre@0: 	    cx = NULL;
andre@0: 	}
andre@0:     }
andre@0:     return cx;
andre@0: }
andre@0: 
andre@0: void 
andre@0: RC4_DestroyContext(RC4Context *cx, PRBool freeit)
andre@0: {
andre@0: 	if (freeit)
andre@0: 		PORT_ZFree(cx, sizeof(*cx));
andre@0: }
andre@0: 
andre@0: #if defined(NSS_BEVAND_ARCFOUR)
andre@0: extern void ARCFOUR(RC4Context *cx, WORD inputLen, 
andre@0: 	const unsigned char *input, unsigned char *output);
andre@0: #else
andre@0: /*
andre@0:  * Generate the next byte in the stream.
andre@0:  */
andre@0: #define ARCFOUR_NEXT_BYTE() \
andre@0: 	tmpSi = cx->S[++tmpi]; \
andre@0: 	tmpj += tmpSi; \
andre@0: 	tmpSj = cx->S[tmpj]; \
andre@0: 	cx->S[tmpi] = tmpSj; \
andre@0: 	cx->S[tmpj] = tmpSi; \
andre@0: 	t = tmpSi + tmpSj;
andre@0: 
andre@0: #ifdef CONVERT_TO_WORDS
andre@0: /*
andre@0:  * Straight ARCFOUR op.  No optimization.
andre@0:  */
andre@0: static SECStatus 
andre@0: rc4_no_opt(RC4Context *cx, unsigned char *output,
andre@0:            unsigned int *outputLen, unsigned int maxOutputLen,
andre@0:            const unsigned char *input, unsigned int inputLen)
andre@0: {
andre@0:     PRUint8 t;
andre@0: 	Stype tmpSi, tmpSj;
andre@0: 	register PRUint8 tmpi = cx->i;
andre@0: 	register PRUint8 tmpj = cx->j;
andre@0: 	unsigned int index;
andre@0: 	PORT_Assert(maxOutputLen >= inputLen);
andre@0: 	if (maxOutputLen < inputLen) {
andre@0: 		PORT_SetError(SEC_ERROR_OUTPUT_LEN);
andre@0: 		return SECFailure;
andre@0: 	}
andre@0: 	for (index=0; index < inputLen; index++) {
andre@0: 		/* Generate next byte from stream. */
andre@0: 		ARCFOUR_NEXT_BYTE();
andre@0: 		/* output = next stream byte XOR next input byte */
andre@0: 		output[index] = cx->S[t] ^ input[index];
andre@0: 	}
andre@0: 	*outputLen = inputLen;
andre@0: 	cx->i = tmpi;
andre@0: 	cx->j = tmpj;
andre@0: 	return SECSuccess;
andre@0: }
andre@0: 
andre@0: #else
andre@0: /* !CONVERT_TO_WORDS */
andre@0: 
andre@0: /*
andre@0:  * Byte-at-a-time ARCFOUR, unrolling the loop into 8 pieces.
andre@0:  */
andre@0: static SECStatus 
andre@0: rc4_unrolled(RC4Context *cx, unsigned char *output,
andre@0:              unsigned int *outputLen, unsigned int maxOutputLen,
andre@0:              const unsigned char *input, unsigned int inputLen)
andre@0: {
andre@0: 	PRUint8 t;
andre@0: 	Stype tmpSi, tmpSj;
andre@0: 	register PRUint8 tmpi = cx->i;
andre@0: 	register PRUint8 tmpj = cx->j;
andre@0: 	int index;
andre@0: 	PORT_Assert(maxOutputLen >= inputLen);
andre@0: 	if (maxOutputLen < inputLen) {
andre@0: 		PORT_SetError(SEC_ERROR_OUTPUT_LEN);
andre@0: 		return SECFailure;
andre@0: 	}
andre@0: 	for (index = inputLen / 8; index-- > 0; input += 8, output += 8) {
andre@0: 		ARCFOUR_NEXT_BYTE();
andre@0: 		output[0] = cx->S[t] ^ input[0];
andre@0: 		ARCFOUR_NEXT_BYTE();
andre@0: 		output[1] = cx->S[t] ^ input[1];
andre@0: 		ARCFOUR_NEXT_BYTE();
andre@0: 		output[2] = cx->S[t] ^ input[2];
andre@0: 		ARCFOUR_NEXT_BYTE();
andre@0: 		output[3] = cx->S[t] ^ input[3];
andre@0: 		ARCFOUR_NEXT_BYTE();
andre@0: 		output[4] = cx->S[t] ^ input[4];
andre@0: 		ARCFOUR_NEXT_BYTE();
andre@0: 		output[5] = cx->S[t] ^ input[5];
andre@0: 		ARCFOUR_NEXT_BYTE();
andre@0: 		output[6] = cx->S[t] ^ input[6];
andre@0: 		ARCFOUR_NEXT_BYTE();
andre@0: 		output[7] = cx->S[t] ^ input[7];
andre@0: 	}
andre@0: 	index = inputLen % 8;
andre@0: 	if (index) {
andre@0: 		input += index;
andre@0: 		output += index;
andre@0: 		switch (index) {
andre@0: 		case 7:
andre@0: 			ARCFOUR_NEXT_BYTE();
andre@0: 			output[-7] = cx->S[t] ^ input[-7]; /* FALLTHRU */
andre@0: 		case 6:
andre@0: 			ARCFOUR_NEXT_BYTE();
andre@0: 			output[-6] = cx->S[t] ^ input[-6]; /* FALLTHRU */
andre@0: 		case 5:
andre@0: 			ARCFOUR_NEXT_BYTE();
andre@0: 			output[-5] = cx->S[t] ^ input[-5]; /* FALLTHRU */
andre@0: 		case 4:
andre@0: 			ARCFOUR_NEXT_BYTE();
andre@0: 			output[-4] = cx->S[t] ^ input[-4]; /* FALLTHRU */
andre@0: 		case 3:
andre@0: 			ARCFOUR_NEXT_BYTE();
andre@0: 			output[-3] = cx->S[t] ^ input[-3]; /* FALLTHRU */
andre@0: 		case 2:
andre@0: 			ARCFOUR_NEXT_BYTE();
andre@0: 			output[-2] = cx->S[t] ^ input[-2]; /* FALLTHRU */
andre@0: 		case 1:
andre@0: 			ARCFOUR_NEXT_BYTE();
andre@0: 			output[-1] = cx->S[t] ^ input[-1]; /* FALLTHRU */
andre@0: 		default:
andre@0: 			/* FALLTHRU */
andre@0: 			; /* hp-ux build breaks without this */
andre@0: 		}
andre@0: 	}
andre@0: 	cx->i = tmpi;
andre@0: 	cx->j = tmpj;
andre@0: 	*outputLen = inputLen;
andre@0: 	return SECSuccess;
andre@0: }
andre@0: #endif
andre@0: 
andre@0: #ifdef IS_LITTLE_ENDIAN
andre@0: #define ARCFOUR_NEXT4BYTES_L(n) \
andre@0: 	ARCFOUR_NEXT_BYTE(); streamWord |= (WORD)cx->S[t] << (n     ); \
andre@0: 	ARCFOUR_NEXT_BYTE(); streamWord |= (WORD)cx->S[t] << (n +  8); \
andre@0: 	ARCFOUR_NEXT_BYTE(); streamWord |= (WORD)cx->S[t] << (n + 16); \
andre@0: 	ARCFOUR_NEXT_BYTE(); streamWord |= (WORD)cx->S[t] << (n + 24);
andre@0: #else
andre@0: #define ARCFOUR_NEXT4BYTES_B(n) \
andre@0: 	ARCFOUR_NEXT_BYTE(); streamWord |= (WORD)cx->S[t] << (n + 24); \
andre@0: 	ARCFOUR_NEXT_BYTE(); streamWord |= (WORD)cx->S[t] << (n + 16); \
andre@0: 	ARCFOUR_NEXT_BYTE(); streamWord |= (WORD)cx->S[t] << (n +  8); \
andre@0: 	ARCFOUR_NEXT_BYTE(); streamWord |= (WORD)cx->S[t] << (n     );
andre@0: #endif
andre@0: 
andre@0: #if (defined(IS_64) && !defined(__sparc)) || defined(NSS_USE_64)
andre@0: /* 64-bit wordsize */
andre@0: #ifdef IS_LITTLE_ENDIAN
andre@0: #define ARCFOUR_NEXT_WORD() \
andre@0: 	{ streamWord = 0; ARCFOUR_NEXT4BYTES_L(0); ARCFOUR_NEXT4BYTES_L(32); }
andre@0: #else
andre@0: #define ARCFOUR_NEXT_WORD() \
andre@0: 	{ streamWord = 0; ARCFOUR_NEXT4BYTES_B(32); ARCFOUR_NEXT4BYTES_B(0); }
andre@0: #endif
andre@0: #else
andre@0: /* 32-bit wordsize */
andre@0: #ifdef IS_LITTLE_ENDIAN
andre@0: #define ARCFOUR_NEXT_WORD() \
andre@0: 	{ streamWord = 0; ARCFOUR_NEXT4BYTES_L(0); }
andre@0: #else
andre@0: #define ARCFOUR_NEXT_WORD() \
andre@0: 	{ streamWord = 0; ARCFOUR_NEXT4BYTES_B(0); }
andre@0: #endif
andre@0: #endif
andre@0: 
andre@0: #ifdef IS_LITTLE_ENDIAN
andre@0: #define RSH <<
andre@0: #define LSH >>
andre@0: #else
andre@0: #define RSH >>
andre@0: #define LSH <<
andre@0: #endif
andre@0: 
andre@0: #ifdef IS_LITTLE_ENDIAN
andre@0: #define LEFTMOST_BYTE_SHIFT 0
andre@0: #define NEXT_BYTE_SHIFT(shift) shift + 8
andre@0: #else
andre@0: #define LEFTMOST_BYTE_SHIFT 8*(WORDSIZE - 1)
andre@0: #define NEXT_BYTE_SHIFT(shift) shift - 8
andre@0: #endif
andre@0: 
andre@0: #ifdef CONVERT_TO_WORDS
andre@0: static SECStatus 
andre@0: rc4_wordconv(RC4Context *cx, unsigned char *output,
andre@0:              unsigned int *outputLen, unsigned int maxOutputLen,
andre@0:              const unsigned char *input, unsigned int inputLen)
andre@0: {
andre@0: 	PR_STATIC_ASSERT(sizeof(PRUword) == sizeof(ptrdiff_t));
andre@0: 	unsigned int inOffset = (PRUword)input % WORDSIZE;
andre@0: 	unsigned int outOffset = (PRUword)output % WORDSIZE;
andre@0: 	register WORD streamWord;
andre@0: 	register const WORD *pInWord;
andre@0: 	register WORD *pOutWord;
andre@0: 	register WORD inWord, nextInWord;
andre@0: 	PRUint8 t;
andre@0: 	register Stype tmpSi, tmpSj;
andre@0: 	register PRUint8 tmpi = cx->i;
andre@0: 	register PRUint8 tmpj = cx->j;
andre@0: 	unsigned int bufShift, invBufShift;
andre@0: 	unsigned int i;
andre@0: 	const unsigned char *finalIn;
andre@0: 	unsigned char *finalOut;
andre@0: 
andre@0: 	PORT_Assert(maxOutputLen >= inputLen);
andre@0: 	if (maxOutputLen < inputLen) {
andre@0: 		PORT_SetError(SEC_ERROR_OUTPUT_LEN);
andre@0: 		return SECFailure;
andre@0: 	}
andre@0: 	if (inputLen < 2*WORDSIZE) {
andre@0: 		/* Ignore word conversion, do byte-at-a-time */
andre@0: 		return rc4_no_opt(cx, output, outputLen, maxOutputLen, input, inputLen);
andre@0: 	}
andre@0: 	*outputLen = inputLen;
andre@0: 	pInWord = (const WORD *)(input - inOffset);
andre@0: 	pOutWord = (WORD *)(output - outOffset);
andre@0: 	if (inOffset <= outOffset) {
andre@0: 		bufShift = 8*(outOffset - inOffset);
andre@0: 		invBufShift = 8*WORDSIZE - bufShift;
andre@0: 	} else {
andre@0: 		invBufShift = 8*(inOffset - outOffset);
andre@0: 		bufShift = 8*WORDSIZE - invBufShift;
andre@0: 	}
andre@0: 	/*****************************************************************/
andre@0: 	/* Step 1:                                                       */
andre@0: 	/* If the first output word is partial, consume the bytes in the */
andre@0: 	/* first partial output word by loading one or two words of      */
andre@0: 	/* input and shifting them accordingly.  Otherwise, just load    */
andre@0: 	/* in the first word of input.  At the end of this block, at     */
andre@0: 	/* least one partial word of input should ALWAYS be loaded.      */
andre@0: 	/*****************************************************************/
andre@0: 	if (outOffset) {
andre@0: 		unsigned int byteCount = WORDSIZE - outOffset; 
andre@0: 		for (i = 0; i < byteCount; i++) {
andre@0: 			ARCFOUR_NEXT_BYTE();
andre@0: 			output[i] = cx->S[t] ^ input[i];
andre@0: 		}
andre@0: 		/* Consumed byteCount bytes of input */
andre@0: 		inputLen -= byteCount;
andre@0: 		pInWord++;
andre@0: 
andre@0: 		/* move to next word of output */
andre@0: 		pOutWord++;
andre@0: 
andre@0: 		/* If buffers are relatively misaligned, shift the bytes in inWord
andre@0: 		 * to be aligned to the output buffer.
andre@0: 		 */
andre@0: 		if (inOffset < outOffset) {
andre@0: 			/* The first input word (which may be partial) has more bytes
andre@0: 			 * than needed.  Copy the remainder to inWord.
andre@0: 			 */
andre@0: 			unsigned int shift = LEFTMOST_BYTE_SHIFT;
andre@0: 			inWord = 0;
andre@0: 			for (i = 0; i < outOffset - inOffset; i++) {
andre@0: 				inWord |= (WORD)input[byteCount + i] << shift;
andre@0: 				shift = NEXT_BYTE_SHIFT(shift);
andre@0: 			}
andre@0: 		} else if (inOffset > outOffset) {
andre@0: 			/* Consumed some bytes in the second input word.  Copy the
andre@0: 			 * remainder to inWord.
andre@0: 			 */
andre@0: 			inWord = *pInWord++;
andre@0: 			inWord = inWord LSH invBufShift;
andre@0: 		} else {
andre@0: 			inWord = 0;
andre@0: 		}
andre@0: 	} else {
andre@0: 		/* output is word-aligned */
andre@0: 		if (inOffset) {
andre@0: 			/* Input is not word-aligned.  The first word load of input 
andre@0: 			 * will not produce a full word of input bytes, so one word
andre@0: 			 * must be pre-loaded.  The main loop below will load in the
andre@0: 			 * next input word and shift some of its bytes into inWord
andre@0: 			 * in order to create a full input word.  Note that the main
andre@0: 			 * loop must execute at least once because the input must
andre@0: 			 * be at least two words.
andre@0: 			 */
andre@0: 			unsigned int shift = LEFTMOST_BYTE_SHIFT;
andre@0: 			inWord = 0;
andre@0: 			for (i = 0; i < WORDSIZE - inOffset; i++) {
andre@0: 				inWord |= (WORD)input[i] << shift;
andre@0: 				shift = NEXT_BYTE_SHIFT(shift);
andre@0: 			}
andre@0: 			pInWord++;
andre@0: 		} else {
andre@0: 			/* Input is word-aligned.  The first word load of input 
andre@0: 			 * will produce a full word of input bytes, so nothing
andre@0: 			 * needs to be loaded here.
andre@0: 			 */
andre@0: 			inWord = 0;
andre@0: 		}
andre@0: 	}
andre@0: 	/*****************************************************************/
andre@0: 	/* Step 2: main loop                                             */
andre@0: 	/* At this point the output buffer is word-aligned.  Any unused  */
andre@0: 	/* bytes from above will be in inWord (shifted correctly).  If   */
andre@0: 	/* the input buffer is unaligned relative to the output buffer,  */
andre@0: 	/* shifting has to be done.                                      */
andre@0: 	/*****************************************************************/
andre@0: 	if (bufShift) {
andre@0: 		/* preloadedByteCount is the number of input bytes pre-loaded
andre@0: 		 * in inWord.
andre@0: 		 */
andre@0: 		unsigned int preloadedByteCount = bufShift/8;
andre@0: 		for (; inputLen >= preloadedByteCount + WORDSIZE;
andre@0: 		     inputLen -= WORDSIZE) {
andre@0: 			nextInWord = *pInWord++;
andre@0: 			inWord |= nextInWord RSH bufShift;
andre@0: 			nextInWord = nextInWord LSH invBufShift;
andre@0: 			ARCFOUR_NEXT_WORD();
andre@0: 			*pOutWord++ = inWord ^ streamWord;
andre@0: 			inWord = nextInWord;
andre@0: 		}
andre@0: 		if (inputLen == 0) {
andre@0: 			/* Nothing left to do. */
andre@0: 			cx->i = tmpi;
andre@0: 			cx->j = tmpj;
andre@0: 			return SECSuccess;
andre@0: 		}
andre@0: 		finalIn = (const unsigned char *)pInWord - preloadedByteCount;
andre@0: 	} else {
andre@0: 		for (; inputLen >= WORDSIZE; inputLen -= WORDSIZE) {
andre@0: 			inWord = *pInWord++;
andre@0: 			ARCFOUR_NEXT_WORD();
andre@0: 			*pOutWord++ = inWord ^ streamWord;
andre@0: 		}
andre@0: 		if (inputLen == 0) {
andre@0: 			/* Nothing left to do. */
andre@0: 			cx->i = tmpi;
andre@0: 			cx->j = tmpj;
andre@0: 			return SECSuccess;
andre@0: 		}
andre@0: 		finalIn = (const unsigned char *)pInWord;
andre@0: 	}
andre@0: 	/*****************************************************************/
andre@0: 	/* Step 3:                                                       */
andre@0: 	/* Do the remaining partial word of input one byte at a time.    */
andre@0: 	/*****************************************************************/
andre@0: 	finalOut = (unsigned char *)pOutWord;
andre@0: 	for (i = 0; i < inputLen; i++) {
andre@0: 		ARCFOUR_NEXT_BYTE();
andre@0: 		finalOut[i] = cx->S[t] ^ finalIn[i];
andre@0: 	}
andre@0: 	cx->i = tmpi;
andre@0: 	cx->j = tmpj;
andre@0: 	return SECSuccess;
andre@0: }
andre@0: #endif
andre@0: #endif /* NSS_BEVAND_ARCFOUR */
andre@0: 
andre@0: SECStatus 
andre@0: RC4_Encrypt(RC4Context *cx, unsigned char *output,
andre@0:             unsigned int *outputLen, unsigned int maxOutputLen,
andre@0:             const unsigned char *input, unsigned int inputLen)
andre@0: {
andre@0: 	PORT_Assert(maxOutputLen >= inputLen);
andre@0: 	if (maxOutputLen < inputLen) {
andre@0: 		PORT_SetError(SEC_ERROR_OUTPUT_LEN);
andre@0: 		return SECFailure;
andre@0: 	}
andre@0: #if defined(NSS_BEVAND_ARCFOUR)
andre@0: 	ARCFOUR(cx, inputLen, input, output);
andre@0:         *outputLen = inputLen;
andre@0: 	return SECSuccess;
andre@0: #elif defined( CONVERT_TO_WORDS )
andre@0: 	/* Convert the byte-stream to a word-stream */
andre@0: 	return rc4_wordconv(cx, output, outputLen, maxOutputLen, input, inputLen);
andre@0: #else
andre@0: 	/* Operate on bytes, but unroll the main loop */
andre@0: 	return rc4_unrolled(cx, output, outputLen, maxOutputLen, input, inputLen);
andre@0: #endif
andre@0: }
andre@0: 
andre@0: SECStatus RC4_Decrypt(RC4Context *cx, unsigned char *output,
andre@0:                       unsigned int *outputLen, unsigned int maxOutputLen,
andre@0:                       const unsigned char *input, unsigned int inputLen)
andre@0: {
andre@0: 	PORT_Assert(maxOutputLen >= inputLen);
andre@0: 	if (maxOutputLen < inputLen) {
andre@0: 		PORT_SetError(SEC_ERROR_OUTPUT_LEN);
andre@0: 		return SECFailure;
andre@0: 	}
andre@0: 	/* decrypt and encrypt are same operation. */
andre@0: #if defined(NSS_BEVAND_ARCFOUR)
andre@0: 	ARCFOUR(cx, inputLen, input, output);
andre@0:         *outputLen = inputLen;
andre@0: 	return SECSuccess;
andre@0: #elif defined( CONVERT_TO_WORDS )
andre@0: 	/* Convert the byte-stream to a word-stream */
andre@0: 	return rc4_wordconv(cx, output, outputLen, maxOutputLen, input, inputLen);
andre@0: #else
andre@0: 	/* Operate on bytes, but unroll the main loop */
andre@0: 	return rc4_unrolled(cx, output, outputLen, maxOutputLen, input, inputLen);
andre@0: #endif
andre@0: }
andre@0: 
andre@0: #undef CONVERT_TO_WORDS
andre@0: #undef USE_WORD