Replace murmurhash algorithm with Robert Jenkin's hash algorithm Upstream-Status: Pending From test result, murmurhash algorithm does not work in big endian processor, so replace it with Robert Jenkin's hash which has worked in linux kernel for many years and has more adaptability. Signed-off-by: Roy.Li --- libfdproto/ostr.c | 192 +++++++++++++++++++++-------------------------------- 1 file changed, 74 insertions(+), 118 deletions(-) diff --git a/libfdproto/ostr.c b/libfdproto/ostr.c index 8f29b48..ce1f4dd 100644 --- a/libfdproto/ostr.c +++ b/libfdproto/ostr.c @@ -430,128 +430,84 @@ after_proto: /********************************************************************************************************/ -/* Hash function -- credits to Austin Appleby, thank you ^^ */ -/* See http://murmurhash.googlepages.com for more information on this function */ - -/* the strings are NOT always aligned properly (ex: received in RADIUS message), so we use the aligned MurmurHash2 function as needed */ -#define _HASH_MIX(h,k,m) { k *= m; k ^= k >> r; k *= m; h *= m; h ^= k; } -uint32_t fd_os_hash ( uint8_t * string, size_t len ) +/* + * Robert Jenkin's hash function. + * http://burtleburtle.net/bob/hash/evahash.html + * This is in the public domain. + */ +#define mix(a, b, c) \ + do { \ + a = a - b; a = a - c; a = a ^ (c >> 13); \ + b = b - c; b = b - a; b = b ^ (a << 8); \ + c = c - a; c = c - b; c = c ^ (b >> 13); \ + a = a - b; a = a - c; a = a ^ (c >> 12); \ + b = b - c; b = b - a; b = b ^ (a << 16); \ + c = c - a; c = c - b; c = c ^ (b >> 5); \ + a = a - b; a = a - c; a = a ^ (c >> 3); \ + b = b - c; b = b - a; b = b ^ (a << 10); \ + c = c - a; c = c - b; c = c ^ (b >> 15); \ + } while (0) + +unsigned hash_rjenkins(const char *str, unsigned length) { - uint32_t hash = len; - uint8_t * data = string; - - const unsigned int m = 0x5bd1e995; - const int r = 24; - int align = (long)string & 3; - - if (!align || (len < 4)) { - /* In case data is aligned, MurmurHash2 function */ - while(len >= 4) - { - /* Mix 4 bytes at a time into the hash */ - uint32_t k = *(uint32_t *)data; /* We don't care about the byte order */ - - _HASH_MIX(hash, k, m); - - data += 4; - len -= 4; - } - - /* Handle the last few bytes of the input */ - switch(len) { - case 3: hash ^= data[2] << 16; - case 2: hash ^= data[1] << 8; - case 1: hash ^= data[0]; - hash *= m; - } - - } else { - /* Unaligned data, use alignment-safe slower version */ - - /* Pre-load the temp registers */ - uint32_t t = 0, d = 0; - switch(align) - { - case 1: t |= data[2] << 16; - case 2: t |= data[1] << 8; - case 3: t |= data[0]; - } - t <<= (8 * align); - - data += 4-align; - len -= 4-align; - - /* From this point, "data" can be read by chunks of 4 bytes */ - - int sl = 8 * (4-align); - int sr = 8 * align; - - /* Mix */ - while(len >= 4) - { - uint32_t k; - - d = *(unsigned int *)data; - k = (t >> sr) | (d << sl); - - _HASH_MIX(hash, k, m); - - t = d; - - data += 4; - len -= 4; - } - - /* Handle leftover data in temp registers */ - d = 0; - if(len >= align) - { - uint32_t k; - - switch(align) - { - case 3: d |= data[2] << 16; - case 2: d |= data[1] << 8; - case 1: d |= data[0]; - } - - k = (t >> sr) | (d << sl); - _HASH_MIX(hash, k, m); - - data += align; - len -= align; - - /* Handle tail bytes */ - - switch(len) - { - case 3: hash ^= data[2] << 16; - case 2: hash ^= data[1] << 8; - case 1: hash ^= data[0]; - hash *= m; - }; - } - else - { - switch(len) - { - case 3: d |= data[2] << 16; - case 2: d |= data[1] << 8; - case 1: d |= data[0]; - case 0: hash ^= (t >> sr) | (d << sl); - hash *= m; - } - } - + const unsigned char *k = (const unsigned char *)str; + uint32_t a, b, c; /* the internal state */ + uint32_t len; /* how many key bytes still need mixing */ + + /* Set up the internal state */ + len = length; + a = 0x9e3779b9; /* the golden ratio; an arbitrary value */ + b = a; + c = 0; /* variable initialization of internal state */ + + /* handle most of the key */ + while (len >= 12) { + a = a + (k[0] + ((uint32_t)k[1] << 8) + ((uint32_t)k[2] << 16) + + ((uint32_t)k[3] << 24)); + b = b + (k[4] + ((uint32_t)k[5] << 8) + ((uint32_t)k[6] << 16) + + ((uint32_t)k[7] << 24)); + c = c + (k[8] + ((uint32_t)k[9] << 8) + ((uint32_t)k[10] << 16) + + ((uint32_t)k[11] << 24)); + mix(a, b, c); + k = k + 12; + len = len - 12; + } + /* handle the last 11 bytes */ + c = c + length; + switch (len) { /* all the case statements fall through */ + case 11: + c = c + ((uint32_t)k[10] << 24); + case 10: + c = c + ((uint32_t)k[9] << 16); + case 9: + c = c + ((uint32_t)k[8] << 8); + /* the first byte of c is reserved for the length */ + case 8: + b = b + ((uint32_t)k[7] << 24); + case 7: + b = b + ((uint32_t)k[6] << 16); + case 6: + b = b + ((uint32_t)k[5] << 8); + case 5: + b = b + k[4]; + case 4: + a = a + ((uint32_t)k[3] << 24); + case 3: + a = a + ((uint32_t)k[2] << 16); + case 2: + a = a + ((uint32_t)k[1] << 8); + case 1: + a = a + k[0]; + /* case 0: nothing left to add */ } + mix(a, b, c); - /* Do a few final mixes of the hash to ensure the last few - bytes are well-incorporated. */ - hash ^= hash >> 13; - hash *= m; - hash ^= hash >> 15; + return c; +} - return hash; +uint32_t fd_os_hash ( uint8_t * string, size_t len ) +{ + return hash_rjenkins(string, len); } -- 1.7.10.4