Demacro MD5 and improve data loading.

Use static inline functions instead of macros and improve handling of
aligned data. Also number rounds as per RFC 1321.

ok tb@

1 files changed, 154 insertions, 116 deletions

diff --git a/src/lib/libcrypto/md5/md5.c b/src/lib/libcrypto/md5/md5.c
index c2ee2958df..cb1a9a3a09 100644
--- a/src/lib/libcrypto/md5/md5.c
+++ b/src/lib/libcrypto/md5/md5.c
@@ -1,4 +1,4 @@
-/* $OpenBSD: md5.c,v 1.18 2023/08/15 08:39:27 jsing Exp $ */
+/* $OpenBSD: md5.c,v 1.19 2024/03/26 05:21:35 jsing Exp $ */
 /* Copyright (C) 1995-1998 Eric Young (eay@cryptsoft.com)
  * All rights reserved.
  *
@@ -90,47 +90,64 @@ void md5_block_asm_data_order(MD5_CTX *c, const void *p, size_t num);
 
 #include "md32_common.h"
 
-/*
-#define F(x,y,z)        (((x) & (y))  |  ((~(x)) & (z)))
-#define G(x,y,z)        (((x) & (z))  |  ((y) & (~(z))))
-*/
+#ifndef MD5_ASM
+static inline uint32_t
+md5_F(uint32_t x, uint32_t y, uint32_t z)
+{
+        return (x & y) | (~x & z);
+}
 
-/* As pointed out by Wei Dai <weidai@eskimo.com>, the above can be
- * simplified to the code below.  Wei attributes these optimizations
- * to Peter Gutmann's SHS code, and he attributes it to Rich Schroeppel.
- */
-#define F(b,c,d)        ((((c) ^ (d)) & (b)) ^ (d))
-#define G(b,c,d)        ((((b) ^ (c)) & (d)) ^ (c))
-#define H(b,c,d)        ((b) ^ (c) ^ (d))
-#define I(b,c,d)        (((~(d)) | (b)) ^ (c))
+static inline uint32_t
+md5_G(uint32_t x, uint32_t y, uint32_t z)
+{
+        return (x & z) | (y & ~z);
+}
 
-#define R0(a,b,c,d,k,s,t) { \
-        a+=((k)+(t)+F((b),(c),(d))); \
-        a=ROTATE(a,s); \
-        a+=b; };\
+static inline uint32_t
+md5_H(uint32_t x, uint32_t y, uint32_t z)
+{
+        return x ^ y ^ z;
+}
+
+static inline uint32_t
+md5_I(uint32_t x, uint32_t y, uint32_t z)
+{
+        return y ^ (x | ~z);
+}
 
-#define R1(a,b,c,d,k,s,t) { \
-        a+=((k)+(t)+G((b),(c),(d))); \
-        a=ROTATE(a,s); \
-        a+=b; };
+static inline void
+md5_round1(uint32_t *a, uint32_t b, uint32_t c, uint32_t d, uint32_t x,
+    uint32_t t, uint32_t s)
+{
+        *a = b + crypto_rol_u32(*a + md5_F(b, c, d) + x + t, s);
+}
 
-#define R2(a,b,c,d,k,s,t) { \
-        a+=((k)+(t)+H((b),(c),(d))); \
-        a=ROTATE(a,s); \
-        a+=b; };
+static inline void
+md5_round2(uint32_t *a, uint32_t b, uint32_t c, uint32_t d, uint32_t x,
+    uint32_t t, uint32_t s)
+{
+        *a = b + crypto_rol_u32(*a + md5_G(b, c, d) + x + t, s);
+}
 
-#define R3(a,b,c,d,k,s,t) { \
-        a+=((k)+(t)+I((b),(c),(d))); \
-        a=ROTATE(a,s); \
-        a+=b; };
+static inline void
+md5_round3(uint32_t *a, uint32_t b, uint32_t c, uint32_t d, uint32_t x,
+    uint32_t t, uint32_t s)
+{
+        *a = b + crypto_rol_u32(*a + md5_H(b, c, d) + x + t, s);
+}
 
-/* Implemented from RFC1321 The MD5 Message-Digest Algorithm. */
+static inline void
+md5_round4(uint32_t *a, uint32_t b, uint32_t c, uint32_t d, uint32_t x,
+    uint32_t t, uint32_t s)
+{
+        *a = b + crypto_rol_u32(*a + md5_I(b, c, d) + x + t, s);
+}
 
-#ifndef MD5_ASM
 static void
 md5_block_data_order(MD5_CTX *c, const void *_in, size_t num)
 {
         const uint8_t *in = _in;
+        const MD5_LONG *in32;
         MD5_LONG A, B, C, D;
         MD5_LONG X0, X1, X2, X3, X4, X5, X6, X7,
             X8, X9, X10, X11, X12, X13, X14, X15;
@@ -140,93 +157,114 @@ md5_block_data_order(MD5_CTX *c, const void *_in, size_t num)
         C = c->C;
         D = c->D;
 
-        for (; num--; ) {
-                X0 = crypto_load_le32toh(&in[0 * 4]);
-                X1 = crypto_load_le32toh(&in[1 * 4]);
-                X2 = crypto_load_le32toh(&in[2 * 4]);
-                X3 = crypto_load_le32toh(&in[3 * 4]);
-                X4 = crypto_load_le32toh(&in[4 * 4]);
-                X5 = crypto_load_le32toh(&in[5 * 4]);
-                X6 = crypto_load_le32toh(&in[6 * 4]);
-                X7 = crypto_load_le32toh(&in[7 * 4]);
-                X8 = crypto_load_le32toh(&in[8 * 4]);
-                X9 = crypto_load_le32toh(&in[9 * 4]);
-                X10 = crypto_load_le32toh(&in[10 * 4]);
-                X11 = crypto_load_le32toh(&in[11 * 4]);
-                X12 = crypto_load_le32toh(&in[12 * 4]);
-                X13 = crypto_load_le32toh(&in[13 * 4]);
-                X14 = crypto_load_le32toh(&in[14 * 4]);
-                X15 = crypto_load_le32toh(&in[15 * 4]);
+        while (num-- > 0) {
+                if ((uintptr_t)in % 4 == 0) {
+                        /* Input is 32 bit aligned. */
+                        in32 = (const MD5_LONG *)in;
+                        X0 = le32toh(in32[0]);
+                        X1 = le32toh(in32[1]);
+                        X2 = le32toh(in32[2]);
+                        X3 = le32toh(in32[3]);
+                        X4 = le32toh(in32[4]);
+                        X5 = le32toh(in32[5]);
+                        X6 = le32toh(in32[6]);
+                        X7 = le32toh(in32[7]);
+                        X8 = le32toh(in32[8]);
+                        X9 = le32toh(in32[9]);
+                        X10 = le32toh(in32[10]);
+                        X11 = le32toh(in32[11]);
+                        X12 = le32toh(in32[12]);
+                        X13 = le32toh(in32[13]);
+                        X14 = le32toh(in32[14]);
+                        X15 = le32toh(in32[15]);
+                } else {
+                        /* Input is not 32 bit aligned. */
+                        X0 = crypto_load_le32toh(&in[0 * 4]);
+                        X1 = crypto_load_le32toh(&in[1 * 4]);
+                        X2 = crypto_load_le32toh(&in[2 * 4]);
+                        X3 = crypto_load_le32toh(&in[3 * 4]);
+                        X4 = crypto_load_le32toh(&in[4 * 4]);
+                        X5 = crypto_load_le32toh(&in[5 * 4]);
+                        X6 = crypto_load_le32toh(&in[6 * 4]);
+                        X7 = crypto_load_le32toh(&in[7 * 4]);
+                        X8 = crypto_load_le32toh(&in[8 * 4]);
+                        X9 = crypto_load_le32toh(&in[9 * 4]);
+                        X10 = crypto_load_le32toh(&in[10 * 4]);
+                        X11 = crypto_load_le32toh(&in[11 * 4]);
+                        X12 = crypto_load_le32toh(&in[12 * 4]);
+                        X13 = crypto_load_le32toh(&in[13 * 4]);
+                        X14 = crypto_load_le32toh(&in[14 * 4]);
+                        X15 = crypto_load_le32toh(&in[15 * 4]);
+                }
                 in += MD5_CBLOCK;
 
-                /* Round 0 */
-                R0(A, B, C, D, X0, 7, 0xd76aa478L);
-                R0(D, A, B, C, X1, 12, 0xe8c7b756L);
-                R0(C, D, A, B, X2, 17, 0x242070dbL);
-                R0(B, C, D, A, X3, 22, 0xc1bdceeeL);
-                R0(A, B, C, D, X4, 7, 0xf57c0fafL);
-                R0(D, A, B, C, X5, 12, 0x4787c62aL);
-                R0(C, D, A, B, X6, 17, 0xa8304613L);
-                R0(B, C, D, A, X7, 22, 0xfd469501L);
-                R0(A, B, C, D, X8, 7, 0x698098d8L);
-                R0(D, A, B, C, X9, 12, 0x8b44f7afL);
-                R0(C, D, A, B, X10, 17, 0xffff5bb1L);
-                R0(B, C, D, A, X11, 22, 0x895cd7beL);
-                R0(A, B, C, D, X12, 7, 0x6b901122L);
-                R0(D, A, B, C, X13, 12, 0xfd987193L);
-                R0(C, D, A, B, X14, 17, 0xa679438eL);
-                R0(B, C, D, A, X15, 22, 0x49b40821L);
-                /* Round 1 */
-                R1(A, B, C, D, X1, 5, 0xf61e2562L);
-                R1(D, A, B, C, X6, 9, 0xc040b340L);
-                R1(C, D, A, B, X11, 14, 0x265e5a51L);
-                R1(B, C, D, A, X0, 20, 0xe9b6c7aaL);
-                R1(A, B, C, D, X5, 5, 0xd62f105dL);
-                R1(D, A, B, C, X10, 9, 0x02441453L);
-                R1(C, D, A, B, X15, 14, 0xd8a1e681L);
-                R1(B, C, D, A, X4, 20, 0xe7d3fbc8L);
-                R1(A, B, C, D, X9, 5, 0x21e1cde6L);
-                R1(D, A, B, C, X14, 9, 0xc33707d6L);
-                R1(C, D, A, B, X3, 14, 0xf4d50d87L);
-                R1(B, C, D, A, X8, 20, 0x455a14edL);
-                R1(A, B, C, D, X13, 5, 0xa9e3e905L);
-                R1(D, A, B, C, X2, 9, 0xfcefa3f8L);
-                R1(C, D, A, B, X7, 14, 0x676f02d9L);
-                R1(B, C, D, A, X12, 20, 0x8d2a4c8aL);
-                /* Round 2 */
-                R2(A, B, C, D, X5, 4, 0xfffa3942L);
-                R2(D, A, B, C, X8, 11, 0x8771f681L);
-                R2(C, D, A, B, X11, 16, 0x6d9d6122L);
-                R2(B, C, D, A, X14, 23, 0xfde5380cL);
-                R2(A, B, C, D, X1, 4, 0xa4beea44L);
-                R2(D, A, B, C, X4, 11, 0x4bdecfa9L);
-                R2(C, D, A, B, X7, 16, 0xf6bb4b60L);
-                R2(B, C, D, A, X10, 23, 0xbebfbc70L);
-                R2(A, B, C, D, X13, 4, 0x289b7ec6L);
-                R2(D, A, B, C, X0, 11, 0xeaa127faL);
-                R2(C, D, A, B, X3, 16, 0xd4ef3085L);
-                R2(B, C, D, A, X6, 23, 0x04881d05L);
-                R2(A, B, C, D, X9, 4, 0xd9d4d039L);
-                R2(D, A, B, C, X12, 11, 0xe6db99e5L);
-                R2(C, D, A, B, X15, 16, 0x1fa27cf8L);
-                R2(B, C, D, A, X2, 23, 0xc4ac5665L);
-                /* Round 3 */
-                R3(A, B, C, D, X0, 6, 0xf4292244L);
-                R3(D, A, B, C, X7, 10, 0x432aff97L);
-                R3(C, D, A, B, X14, 15, 0xab9423a7L);
-                R3(B, C, D, A, X5, 21, 0xfc93a039L);
-                R3(A, B, C, D, X12, 6, 0x655b59c3L);
-                R3(D, A, B, C, X3, 10, 0x8f0ccc92L);
-                R3(C, D, A, B, X10, 15, 0xffeff47dL);
-                R3(B, C, D, A, X1, 21, 0x85845dd1L);
-                R3(A, B, C, D, X8, 6, 0x6fa87e4fL);
-                R3(D, A, B, C, X15, 10, 0xfe2ce6e0L);
-                R3(C, D, A, B, X6, 15, 0xa3014314L);
-                R3(B, C, D, A, X13, 21, 0x4e0811a1L);
-                R3(A, B, C, D, X4, 6, 0xf7537e82L);
-                R3(D, A, B, C, X11, 10, 0xbd3af235L);
-                R3(C, D, A, B, X2, 15, 0x2ad7d2bbL);
-                R3(B, C, D, A, X9, 21, 0xeb86d391L);
+                md5_round1(&A, B, C, D, X0, 0xd76aa478L, 7);
+                md5_round1(&D, A, B, C, X1, 0xe8c7b756L, 12);
+                md5_round1(&C, D, A, B, X2, 0x242070dbL, 17);
+                md5_round1(&B, C, D, A, X3, 0xc1bdceeeL, 22);
+                md5_round1(&A, B, C, D, X4, 0xf57c0fafL, 7);
+                md5_round1(&D, A, B, C, X5, 0x4787c62aL, 12);
+                md5_round1(&C, D, A, B, X6, 0xa8304613L, 17);
+                md5_round1(&B, C, D, A, X7, 0xfd469501L, 22);
+                md5_round1(&A, B, C, D, X8, 0x698098d8L, 7);
+                md5_round1(&D, A, B, C, X9, 0x8b44f7afL, 12);
+                md5_round1(&C, D, A, B, X10, 0xffff5bb1L, 17);
+                md5_round1(&B, C, D, A, X11, 0x895cd7beL, 22);
+                md5_round1(&A, B, C, D, X12, 0x6b901122L, 7);
+                md5_round1(&D, A, B, C, X13, 0xfd987193L, 12);
+                md5_round1(&C, D, A, B, X14, 0xa679438eL, 17);
+                md5_round1(&B, C, D, A, X15, 0x49b40821L, 22);
+
+                md5_round2(&A, B, C, D, X1, 0xf61e2562L, 5);
+                md5_round2(&D, A, B, C, X6, 0xc040b340L, 9);
+                md5_round2(&C, D, A, B, X11, 0x265e5a51L, 14);
+                md5_round2(&B, C, D, A, X0, 0xe9b6c7aaL, 20);
+                md5_round2(&A, B, C, D, X5, 0xd62f105dL, 5);
+                md5_round2(&D, A, B, C, X10, 0x02441453L, 9);
+                md5_round2(&C, D, A, B, X15, 0xd8a1e681L, 14);
+                md5_round2(&B, C, D, A, X4, 0xe7d3fbc8L, 20);
+                md5_round2(&A, B, C, D, X9, 0x21e1cde6L, 5);
+                md5_round2(&D, A, B, C, X14, 0xc33707d6L, 9);
+                md5_round2(&C, D, A, B, X3, 0xf4d50d87L, 14);
+                md5_round2(&B, C, D, A, X8, 0x455a14edL, 20);
+                md5_round2(&A, B, C, D, X13, 0xa9e3e905L, 5);
+                md5_round2(&D, A, B, C, X2, 0xfcefa3f8L, 9);
+                md5_round2(&C, D, A, B, X7, 0x676f02d9L, 14);
+                md5_round2(&B, C, D, A, X12, 0x8d2a4c8aL, 20);
+
+                md5_round3(&A, B, C, D, X5, 0xfffa3942L, 4);
+                md5_round3(&D, A, B, C, X8, 0x8771f681L, 11);
+                md5_round3(&C, D, A, B, X11, 0x6d9d6122L, 16);
+                md5_round3(&B, C, D, A, X14, 0xfde5380cL, 23);
+                md5_round3(&A, B, C, D, X1, 0xa4beea44L, 4);
+                md5_round3(&D, A, B, C, X4, 0x4bdecfa9L, 11);
+                md5_round3(&C, D, A, B, X7, 0xf6bb4b60L, 16);
+                md5_round3(&B, C, D, A, X10, 0xbebfbc70L, 23);
+                md5_round3(&A, B, C, D, X13, 0x289b7ec6L, 4);
+                md5_round3(&D, A, B, C, X0, 0xeaa127faL, 11);
+                md5_round3(&C, D, A, B, X3, 0xd4ef3085L, 16);
+                md5_round3(&B, C, D, A, X6, 0x04881d05L, 23);
+                md5_round3(&A, B, C, D, X9, 0xd9d4d039L, 4);
+                md5_round3(&D, A, B, C, X12, 0xe6db99e5L, 11);
+                md5_round3(&C, D, A, B, X15, 0x1fa27cf8L, 16);
+                md5_round3(&B, C, D, A, X2, 0xc4ac5665L, 23);
+
+                md5_round4(&A, B, C, D, X0, 0xf4292244L, 6);
+                md5_round4(&D, A, B, C, X7, 0x432aff97L, 10);
+                md5_round4(&C, D, A, B, X14, 0xab9423a7L, 15);
+                md5_round4(&B, C, D, A, X5, 0xfc93a039L, 21);
+                md5_round4(&A, B, C, D, X12, 0x655b59c3L, 6);
+                md5_round4(&D, A, B, C, X3, 0x8f0ccc92L, 10);
+                md5_round4(&C, D, A, B, X10, 0xffeff47dL, 15);
+                md5_round4(&B, C, D, A, X1, 0x85845dd1L, 21);
+                md5_round4(&A, B, C, D, X8, 0x6fa87e4fL, 6);
+                md5_round4(&D, A, B, C, X15, 0xfe2ce6e0L, 10);
+                md5_round4(&C, D, A, B, X6, 0xa3014314L, 15);
+                md5_round4(&B, C, D, A, X13, 0x4e0811a1L, 21);
+                md5_round4(&A, B, C, D, X4, 0xf7537e82L, 6);
+                md5_round4(&D, A, B, C, X11, 0xbd3af235L, 10);
+                md5_round4(&C, D, A, B, X2, 0x2ad7d2bbL, 15);
+                md5_round4(&B, C, D, A, X9, 0xeb86d391L, 21);
 
                 A = c->A += A;
                 B = c->B += B;