From 456120cfac29cf67e039868762864f0cf4d0e7e8 Mon Sep 17 00:00:00 2001 From: Simon Tatham Date: Thu, 24 Dec 2020 17:08:55 +0000 Subject: [PATCH] Rewrite MD5 implementation in my modern style. MD5 is structurally very similar to all the SHA-1 and SHA-2 hashes (with the main difference being that the message schedule consists of just repeating the 32-bit words of the message four times in different permutations, instead of transforming them via an LFSR-style process). So it helps legibility and maintainability if all the implementations of these hashes are coded in a similar style - for example, that way, the next time I need to make a change to the ssh_hash API, I can do it the same way in all these modules without having to think everything out again. After the SHA-512 rewrite earlier today, all the hashes in that family had been updated to a consistent new style as a side effect of adding optional hardware acceleration, except for MD5, because there's no hardware-accelerated version of it. (And not much chance of anyone ever needing one, I hope!) So this is a purely stylistic update which reworks MD5 so that it looks just like all the SHA-1 and SHA-2 hash implementations. No functional change. --- sshmd5.c | 434 ++++++++++++++++++++++++++----------------------------- 1 file changed, 204 insertions(+), 230 deletions(-) diff --git a/sshmd5.c b/sshmd5.c index 903aef5d..9155c99e 100644 --- a/sshmd5.c +++ b/sshmd5.c @@ -1,262 +1,236 @@ -#include -#include "ssh.h" - /* * MD5 implementation for PuTTY. Written directly from the spec by * Simon Tatham. */ -typedef struct { - uint32_t h[4]; -} MD5_Core_State; +#include +#include "ssh.h" -struct MD5Context { - MD5_Core_State core; - unsigned char block[64]; - int blkused; +static const uint32_t md5_initial_state[] = { + 0x67452301, 0xefcdab89, 0x98badcfe, 0x10325476, +}; + +static const struct md5_round_constant { + uint32_t addition, rotation, msg_index; +} md5_round_constants[] = { + { 0xd76aa478, 7, 0 }, { 0xe8c7b756, 12, 1 }, + { 0x242070db, 17, 2 }, { 0xc1bdceee, 22, 3 }, + { 0xf57c0faf, 7, 4 }, { 0x4787c62a, 12, 5 }, + { 0xa8304613, 17, 6 }, { 0xfd469501, 22, 7 }, + { 0x698098d8, 7, 8 }, { 0x8b44f7af, 12, 9 }, + { 0xffff5bb1, 17, 10 }, { 0x895cd7be, 22, 11 }, + { 0x6b901122, 7, 12 }, { 0xfd987193, 12, 13 }, + { 0xa679438e, 17, 14 }, { 0x49b40821, 22, 15 }, + { 0xf61e2562, 5, 1 }, { 0xc040b340, 9, 6 }, + { 0x265e5a51, 14, 11 }, { 0xe9b6c7aa, 20, 0 }, + { 0xd62f105d, 5, 5 }, { 0x02441453, 9, 10 }, + { 0xd8a1e681, 14, 15 }, { 0xe7d3fbc8, 20, 4 }, + { 0x21e1cde6, 5, 9 }, { 0xc33707d6, 9, 14 }, + { 0xf4d50d87, 14, 3 }, { 0x455a14ed, 20, 8 }, + { 0xa9e3e905, 5, 13 }, { 0xfcefa3f8, 9, 2 }, + { 0x676f02d9, 14, 7 }, { 0x8d2a4c8a, 20, 12 }, + { 0xfffa3942, 4, 5 }, { 0x8771f681, 11, 8 }, + { 0x6d9d6122, 16, 11 }, { 0xfde5380c, 23, 14 }, + { 0xa4beea44, 4, 1 }, { 0x4bdecfa9, 11, 4 }, + { 0xf6bb4b60, 16, 7 }, { 0xbebfbc70, 23, 10 }, + { 0x289b7ec6, 4, 13 }, { 0xeaa127fa, 11, 0 }, + { 0xd4ef3085, 16, 3 }, { 0x04881d05, 23, 6 }, + { 0xd9d4d039, 4, 9 }, { 0xe6db99e5, 11, 12 }, + { 0x1fa27cf8, 16, 15 }, { 0xc4ac5665, 23, 2 }, + { 0xf4292244, 6, 0 }, { 0x432aff97, 10, 7 }, + { 0xab9423a7, 15, 14 }, { 0xfc93a039, 21, 5 }, + { 0x655b59c3, 6, 12 }, { 0x8f0ccc92, 10, 3 }, + { 0xffeff47d, 15, 10 }, { 0x85845dd1, 21, 1 }, + { 0x6fa87e4f, 6, 8 }, { 0xfe2ce6e0, 10, 15 }, + { 0xa3014314, 15, 6 }, { 0x4e0811a1, 21, 13 }, + { 0xf7537e82, 6, 4 }, { 0xbd3af235, 10, 11 }, + { 0x2ad7d2bb, 15, 2 }, { 0xeb86d391, 21, 9 }, +}; + +typedef struct md5_block md5_block; +struct md5_block { + uint8_t block[64]; + size_t used; uint64_t len; +}; + +static inline void md5_block_setup(md5_block *blk) +{ + blk->used = 0; + blk->len = 0; +} + +static inline bool md5_block_write( + md5_block *blk, const void **vdata, size_t *len) +{ + size_t blkleft = sizeof(blk->block) - blk->used; + size_t chunk = *len < blkleft ? *len : blkleft; + + const uint8_t *p = *vdata; + memcpy(blk->block + blk->used, p, chunk); + *vdata = p + chunk; + *len -= chunk; + blk->used += chunk; + blk->len += chunk; + + if (blk->used == sizeof(blk->block)) { + blk->used = 0; + return true; + } + + return false; +} + +static inline void md5_block_pad(md5_block *blk, BinarySink *bs) +{ + uint64_t final_len = blk->len << 3; + size_t pad = 63 & (55 - blk->used); + + put_byte(bs, 0x80); + put_padding(bs, pad, 0); + + unsigned char buf[8]; + PUT_64BIT_LSB_FIRST(buf, final_len); + put_data(bs, buf, 8); + smemclr(buf, 8); + + assert(blk->used == 0 && "Should have exactly hit a block boundary"); +} + +static inline uint32_t rol(uint32_t x, unsigned y) +{ + return (x << (31 & y)) | (x >> (31 & -y)); +} + +static inline uint32_t Ch(uint32_t ctrl, uint32_t if1, uint32_t if0) +{ + return if0 ^ (ctrl & (if1 ^ if0)); +} + +/* Parameter functions for the four MD5 round types */ +static inline uint32_t F(uint32_t x, uint32_t y, uint32_t z) +{ return Ch(x, y, z); } +static inline uint32_t G(uint32_t x, uint32_t y, uint32_t z) +{ return Ch(z, x, y); } +static inline uint32_t H(uint32_t x, uint32_t y, uint32_t z) +{ return x ^ y ^ z; } +static inline uint32_t I(uint32_t x, uint32_t y, uint32_t z) +{ return y ^ (x | ~z); } + +static inline void md5_round( + unsigned round_index, const uint32_t *message, + uint32_t *a, uint32_t *b, uint32_t *c, uint32_t *d, + uint32_t (*f)(uint32_t, uint32_t, uint32_t)) +{ + struct md5_round_constant rc = md5_round_constants[round_index]; + + *a = *b + rol(*a + f(*b, *c, *d) + message[rc.msg_index] + rc.addition, + rc.rotation); +} + +static void md5_do_block(uint32_t *core, const uint8_t *block) +{ + uint32_t message_words[16]; + for (size_t i = 0; i < 16; i++) + message_words[i] = GET_32BIT_LSB_FIRST(block + 4*i); + + uint32_t a = core[0], b = core[1], c = core[2], d = core[3]; + + size_t t = 0; + for (size_t u = 0; u < 4; u++) { + md5_round(t++, message_words, &a, &b, &c, &d, F); + md5_round(t++, message_words, &d, &a, &b, &c, F); + md5_round(t++, message_words, &c, &d, &a, &b, F); + md5_round(t++, message_words, &b, &c, &d, &a, F); + } + for (size_t u = 0; u < 4; u++) { + md5_round(t++, message_words, &a, &b, &c, &d, G); + md5_round(t++, message_words, &d, &a, &b, &c, G); + md5_round(t++, message_words, &c, &d, &a, &b, G); + md5_round(t++, message_words, &b, &c, &d, &a, G); + } + for (size_t u = 0; u < 4; u++) { + md5_round(t++, message_words, &a, &b, &c, &d, H); + md5_round(t++, message_words, &d, &a, &b, &c, H); + md5_round(t++, message_words, &c, &d, &a, &b, H); + md5_round(t++, message_words, &b, &c, &d, &a, H); + } + for (size_t u = 0; u < 4; u++) { + md5_round(t++, message_words, &a, &b, &c, &d, I); + md5_round(t++, message_words, &d, &a, &b, &c, I); + md5_round(t++, message_words, &c, &d, &a, &b, I); + md5_round(t++, message_words, &b, &c, &d, &a, I); + } + + core[0] += a; + core[1] += b; + core[2] += c; + core[3] += d; + + smemclr(message_words, sizeof(message_words)); +} + +typedef struct md5 { + uint32_t core[4]; + md5_block blk; BinarySink_IMPLEMENTATION; -}; - -/* ---------------------------------------------------------------------- - * Core MD5 algorithm: processes 16-word blocks into a message digest. - */ - -#define F(x,y,z) ( ((x) & (y)) | ((~(x)) & (z)) ) -#define G(x,y,z) ( ((x) & (z)) | ((~(z)) & (y)) ) -#define H(x,y,z) ( (x) ^ (y) ^ (z) ) -#define I(x,y,z) ( (y) ^ ( (x) | ~(z) ) ) - -#define rol(x,y) ( ((x) << (y)) | (((uint32_t)x) >> (32-y)) ) - -#define subround(f,w,x,y,z,k,s,ti) \ - w = x + rol(w + f(x,y,z) + block[k] + ti, s) - -static void MD5_Core_Init(MD5_Core_State * s) -{ - s->h[0] = 0x67452301; - s->h[1] = 0xefcdab89; - s->h[2] = 0x98badcfe; - s->h[3] = 0x10325476; -} - -static void MD5_Block(MD5_Core_State *s, uint32_t *block) -{ - uint32_t a, b, c, d; - - a = s->h[0]; - b = s->h[1]; - c = s->h[2]; - d = s->h[3]; - - subround(F, a, b, c, d, 0, 7, 0xd76aa478); - subround(F, d, a, b, c, 1, 12, 0xe8c7b756); - subround(F, c, d, a, b, 2, 17, 0x242070db); - subround(F, b, c, d, a, 3, 22, 0xc1bdceee); - subround(F, a, b, c, d, 4, 7, 0xf57c0faf); - subround(F, d, a, b, c, 5, 12, 0x4787c62a); - subround(F, c, d, a, b, 6, 17, 0xa8304613); - subround(F, b, c, d, a, 7, 22, 0xfd469501); - subround(F, a, b, c, d, 8, 7, 0x698098d8); - subround(F, d, a, b, c, 9, 12, 0x8b44f7af); - subround(F, c, d, a, b, 10, 17, 0xffff5bb1); - subround(F, b, c, d, a, 11, 22, 0x895cd7be); - subround(F, a, b, c, d, 12, 7, 0x6b901122); - subround(F, d, a, b, c, 13, 12, 0xfd987193); - subround(F, c, d, a, b, 14, 17, 0xa679438e); - subround(F, b, c, d, a, 15, 22, 0x49b40821); - subround(G, a, b, c, d, 1, 5, 0xf61e2562); - subround(G, d, a, b, c, 6, 9, 0xc040b340); - subround(G, c, d, a, b, 11, 14, 0x265e5a51); - subround(G, b, c, d, a, 0, 20, 0xe9b6c7aa); - subround(G, a, b, c, d, 5, 5, 0xd62f105d); - subround(G, d, a, b, c, 10, 9, 0x02441453); - subround(G, c, d, a, b, 15, 14, 0xd8a1e681); - subround(G, b, c, d, a, 4, 20, 0xe7d3fbc8); - subround(G, a, b, c, d, 9, 5, 0x21e1cde6); - subround(G, d, a, b, c, 14, 9, 0xc33707d6); - subround(G, c, d, a, b, 3, 14, 0xf4d50d87); - subround(G, b, c, d, a, 8, 20, 0x455a14ed); - subround(G, a, b, c, d, 13, 5, 0xa9e3e905); - subround(G, d, a, b, c, 2, 9, 0xfcefa3f8); - subround(G, c, d, a, b, 7, 14, 0x676f02d9); - subround(G, b, c, d, a, 12, 20, 0x8d2a4c8a); - subround(H, a, b, c, d, 5, 4, 0xfffa3942); - subround(H, d, a, b, c, 8, 11, 0x8771f681); - subround(H, c, d, a, b, 11, 16, 0x6d9d6122); - subround(H, b, c, d, a, 14, 23, 0xfde5380c); - subround(H, a, b, c, d, 1, 4, 0xa4beea44); - subround(H, d, a, b, c, 4, 11, 0x4bdecfa9); - subround(H, c, d, a, b, 7, 16, 0xf6bb4b60); - subround(H, b, c, d, a, 10, 23, 0xbebfbc70); - subround(H, a, b, c, d, 13, 4, 0x289b7ec6); - subround(H, d, a, b, c, 0, 11, 0xeaa127fa); - subround(H, c, d, a, b, 3, 16, 0xd4ef3085); - subround(H, b, c, d, a, 6, 23, 0x04881d05); - subround(H, a, b, c, d, 9, 4, 0xd9d4d039); - subround(H, d, a, b, c, 12, 11, 0xe6db99e5); - subround(H, c, d, a, b, 15, 16, 0x1fa27cf8); - subround(H, b, c, d, a, 2, 23, 0xc4ac5665); - subround(I, a, b, c, d, 0, 6, 0xf4292244); - subround(I, d, a, b, c, 7, 10, 0x432aff97); - subround(I, c, d, a, b, 14, 15, 0xab9423a7); - subround(I, b, c, d, a, 5, 21, 0xfc93a039); - subround(I, a, b, c, d, 12, 6, 0x655b59c3); - subround(I, d, a, b, c, 3, 10, 0x8f0ccc92); - subround(I, c, d, a, b, 10, 15, 0xffeff47d); - subround(I, b, c, d, a, 1, 21, 0x85845dd1); - subround(I, a, b, c, d, 8, 6, 0x6fa87e4f); - subround(I, d, a, b, c, 15, 10, 0xfe2ce6e0); - subround(I, c, d, a, b, 6, 15, 0xa3014314); - subround(I, b, c, d, a, 13, 21, 0x4e0811a1); - subround(I, a, b, c, d, 4, 6, 0xf7537e82); - subround(I, d, a, b, c, 11, 10, 0xbd3af235); - subround(I, c, d, a, b, 2, 15, 0x2ad7d2bb); - subround(I, b, c, d, a, 9, 21, 0xeb86d391); - - s->h[0] += a; - s->h[1] += b; - s->h[2] += c; - s->h[3] += d; -} - -/* ---------------------------------------------------------------------- - * Outer MD5 algorithm: take an arbitrary length byte string, - * convert it into 16-word blocks with the prescribed padding at - * the end, and pass those blocks to the core MD5 algorithm. - */ - -#define BLKSIZE 64 - -static void MD5_BinarySink_write(BinarySink *bs, const void *data, size_t len); - -static void MD5Init(struct MD5Context *s) -{ - MD5_Core_Init(&s->core); - s->blkused = 0; - s->len = 0; - BinarySink_INIT(s, MD5_BinarySink_write); -} - -static void MD5_BinarySink_write(BinarySink *bs, const void *data, size_t len) -{ - struct MD5Context *s = BinarySink_DOWNCAST(bs, struct MD5Context); - const unsigned char *q = (const unsigned char *)data; - uint32_t wordblock[16]; - uint32_t lenw = len; - int i; - - assert(lenw == len); - - /* - * Update the length field. - */ - s->len += lenw; - - if (s->blkused + len < BLKSIZE) { - /* - * Trivial case: just add to the block. - */ - memcpy(s->block + s->blkused, q, len); - s->blkused += len; - } else { - /* - * We must complete and process at least one block. - */ - while (s->blkused + len >= BLKSIZE) { - memcpy(s->block + s->blkused, q, BLKSIZE - s->blkused); - q += BLKSIZE - s->blkused; - len -= BLKSIZE - s->blkused; - /* Now process the block. Gather bytes little-endian into words */ - for (i = 0; i < 16; i++) { - wordblock[i] = - (((uint32_t) s->block[i * 4 + 3]) << 24) | - (((uint32_t) s->block[i * 4 + 2]) << 16) | - (((uint32_t) s->block[i * 4 + 1]) << 8) | - (((uint32_t) s->block[i * 4 + 0]) << 0); - } - MD5_Block(&s->core, wordblock); - s->blkused = 0; - } - memcpy(s->block, q, len); - s->blkused = len; - } -} - -static void MD5Final(unsigned char output[16], struct MD5Context *s) -{ - int i; - unsigned pad; - unsigned char c[64]; - uint64_t len; - - if (s->blkused >= 56) - pad = 56 + 64 - s->blkused; - else - pad = 56 - s->blkused; - - len = (s->len << 3); - - memset(c, 0, pad); - c[0] = 0x80; - put_data(s, c, pad); - - PUT_64BIT_LSB_FIRST(c, len); - - put_data(s, c, 8); - - for (i = 0; i < 4; i++) { - output[4 * i + 3] = (s->core.h[i] >> 24) & 0xFF; - output[4 * i + 2] = (s->core.h[i] >> 16) & 0xFF; - output[4 * i + 1] = (s->core.h[i] >> 8) & 0xFF; - output[4 * i + 0] = (s->core.h[i] >> 0) & 0xFF; - } -} - -/* ---------------------------------------------------------------------- - * Thin abstraction for things where hashes are pluggable. - */ - -struct md5_hash { - struct MD5Context state; ssh_hash hash; -}; +} md5; + +static void md5_write(BinarySink *bs, const void *vp, size_t len); static ssh_hash *md5_new(const ssh_hashalg *alg) { - struct md5_hash *h = snew(struct md5_hash); - h->hash.vt = alg; - BinarySink_DELEGATE_INIT(&h->hash, &h->state); - return &h->hash; + md5 *s = snew(md5); + + s->hash.vt = alg; + BinarySink_INIT(s, md5_write); + BinarySink_DELEGATE_INIT(&s->hash, s); + return &s->hash; } static void md5_reset(ssh_hash *hash) { - struct md5_hash *h = container_of(hash, struct md5_hash, hash); - MD5Init(&h->state); + md5 *s = container_of(hash, md5, hash); + + memcpy(s->core, md5_initial_state, sizeof(s->core)); + md5_block_setup(&s->blk); } static void md5_copyfrom(ssh_hash *hcopy, ssh_hash *horig) { - struct md5_hash *copy = container_of(hcopy, struct md5_hash, hash); - struct md5_hash *orig = container_of(horig, struct md5_hash, hash); + md5 *copy = container_of(hcopy, md5, hash); + md5 *orig = container_of(horig, md5, hash); - copy->state = orig->state; - BinarySink_COPIED(©->state); + memcpy(copy, orig, sizeof(*copy)); + BinarySink_COPIED(copy); + BinarySink_DELEGATE_INIT(©->hash, copy); } static void md5_free(ssh_hash *hash) { - struct md5_hash *h = container_of(hash, struct md5_hash, hash); + md5 *s = container_of(hash, md5, hash); - smemclr(h, sizeof(*h)); - sfree(h); + smemclr(s, sizeof(*s)); + sfree(s); } -static void md5_digest(ssh_hash *hash, unsigned char *output) +static void md5_write(BinarySink *bs, const void *vp, size_t len) { - struct md5_hash *h = container_of(hash, struct md5_hash, hash); - MD5Final(output, &h->state); + md5 *s = BinarySink_DOWNCAST(bs, md5); + + while (len > 0) + if (md5_block_write(&s->blk, &vp, &len)) + md5_do_block(s->core, s->blk.block); +} + +static void md5_digest(ssh_hash *hash, uint8_t *digest) +{ + md5 *s = container_of(hash, md5, hash); + + md5_block_pad(&s->blk, BinarySink_UPCAST(s)); + for (size_t i = 0; i < 4; i++) + PUT_32BIT_LSB_FIRST(digest + 4*i, s->core[i]); } const ssh_hashalg ssh_md5 = {