crypto: switch poly1305 to incremental implementation

This code is taken from poly1305-donna-32.h, poly1305-donna.h, poly1305-donna.c
from https://github.com/floodyberry/poly1305-donna, commit
e6ad6e091d30d7f4ec2d4f978be1fcfcbce72781, with the following modifications:

* Coding style (braces around one-line indented if/for loops).
* Rename unsigned long (long) to uint32_t and uint64_t.
* Rename poly1305_block_size to POLY1305_BLOCK_SIZE.
* Adding noexcept to functions.
* Merging poly1305_state_internal_t and poly1305_context types.
* Merging code from multiple files.
* Place all imported code in the poly1305_donna namespace.

2 files changed, 233 insertions, 122 deletions

diff --git a/src/crypto/poly1305.cpp b/src/crypto/poly1305.cpp
index 8a86c9601c..c89383c73a 100644
--- a/src/crypto/poly1305.cpp
+++ b/src/crypto/poly1305.cpp
@@ -2,140 +2,230 @@
 // Distributed under the MIT software license, see the accompanying
 // file COPYING or http://www.opensource.org/licenses/mit-license.php.
 
-// Based on the public domain implementation by Andrew Moon
-// poly1305-donna-unrolled.c from https://github.com/floodyberry/poly1305-donna
-
 #include <crypto/common.h>
 #include <crypto/poly1305.h>
 
 #include <string.h>
 
-#define mul32x32_64(a,b) ((uint64_t)(a) * (b))
+namespace poly1305_donna {
 
-void poly1305_auth(unsigned char out[POLY1305_TAGLEN], const unsigned char *m, size_t inlen, const unsigned char key[POLY1305_KEYLEN]) {
-    uint32_t t0,t1,t2,t3;
-    uint32_t h0,h1,h2,h3,h4;
+// Based on the public domain implementation by Andrew Moon
+// poly1305-donna-32.h from https://github.com/floodyberry/poly1305-donna
+
+void poly1305_init(poly1305_context *st, const unsigned char key[32]) noexcept {
+    /* r &= 0xffffffc0ffffffc0ffffffc0fffffff */
+    st->r[0] = (ReadLE32(&key[ 0])     ) & 0x3ffffff;
+    st->r[1] = (ReadLE32(&key[ 3]) >> 2) & 0x3ffff03;
+    st->r[2] = (ReadLE32(&key[ 6]) >> 4) & 0x3ffc0ff;
+    st->r[3] = (ReadLE32(&key[ 9]) >> 6) & 0x3f03fff;
+    st->r[4] = (ReadLE32(&key[12]) >> 8) & 0x00fffff;
+
+    /* h = 0 */
+    st->h[0] = 0;
+    st->h[1] = 0;
+    st->h[2] = 0;
+    st->h[3] = 0;
+    st->h[4] = 0;
+
+    /* save pad for later */
+    st->pad[0] = ReadLE32(&key[16]);
+    st->pad[1] = ReadLE32(&key[20]);
+    st->pad[2] = ReadLE32(&key[24]);
+    st->pad[3] = ReadLE32(&key[28]);
+
+    st->leftover = 0;
+    st->final = 0;
+}
+
+static void poly1305_blocks(poly1305_context *st, const unsigned char *m, size_t bytes) noexcept {
+    const uint32_t hibit = (st->final) ? 0 : (1UL << 24); /* 1 << 128 */
     uint32_t r0,r1,r2,r3,r4;
     uint32_t s1,s2,s3,s4;
-    uint32_t b, nb;
-    size_t j;
-    uint64_t t[5];
-    uint64_t f0,f1,f2,f3;
-    uint64_t g0,g1,g2,g3,g4;
-    uint64_t c;
-    unsigned char mp[16];
-
-    /* clamp key */
-    t0 = ReadLE32(key+0);
-    t1 = ReadLE32(key+4);
-    t2 = ReadLE32(key+8);
-    t3 = ReadLE32(key+12);
-
-    /* precompute multipliers */
-    r0 = t0 & 0x3ffffff; t0 >>= 26; t0 |= t1 << 6;
-    r1 = t0 & 0x3ffff03; t1 >>= 20; t1 |= t2 << 12;
-    r2 = t1 & 0x3ffc0ff; t2 >>= 14; t2 |= t3 << 18;
-    r3 = t2 & 0x3f03fff; t3 >>= 8;
-    r4 = t3 & 0x00fffff;
+    uint32_t h0,h1,h2,h3,h4;
+    uint64_t d0,d1,d2,d3,d4;
+    uint32_t c;
+
+    r0 = st->r[0];
+    r1 = st->r[1];
+    r2 = st->r[2];
+    r3 = st->r[3];
+    r4 = st->r[4];
 
     s1 = r1 * 5;
     s2 = r2 * 5;
     s3 = r3 * 5;
     s4 = r4 * 5;
 
-    /* init state */
-    h0 = 0;
-    h1 = 0;
-    h2 = 0;
-    h3 = 0;
-    h4 = 0;
-
-    /* full blocks */
-    if (inlen < 16) goto poly1305_donna_atmost15bytes;
-poly1305_donna_16bytes:
-    m += 16;
-    inlen -= 16;
-
-    t0 = ReadLE32(m-16);
-    t1 = ReadLE32(m-12);
-    t2 = ReadLE32(m-8);
-    t3 = ReadLE32(m-4);
-
-    h0 += t0 & 0x3ffffff;
-    h1 += ((((uint64_t)t1 << 32) | t0) >> 26) & 0x3ffffff;
-    h2 += ((((uint64_t)t2 << 32) | t1) >> 20) & 0x3ffffff;
-    h3 += ((((uint64_t)t3 << 32) | t2) >> 14) & 0x3ffffff;
-    h4 += (t3 >> 8) | (1 << 24);
-
-
-poly1305_donna_mul:
-    t[0]  = mul32x32_64(h0,r0) + mul32x32_64(h1,s4) + mul32x32_64(h2,s3) + mul32x32_64(h3,s2) + mul32x32_64(h4,s1);
-    t[1]  = mul32x32_64(h0,r1) + mul32x32_64(h1,r0) + mul32x32_64(h2,s4) + mul32x32_64(h3,s3) + mul32x32_64(h4,s2);
-    t[2]  = mul32x32_64(h0,r2) + mul32x32_64(h1,r1) + mul32x32_64(h2,r0) + mul32x32_64(h3,s4) + mul32x32_64(h4,s3);
-    t[3]  = mul32x32_64(h0,r3) + mul32x32_64(h1,r2) + mul32x32_64(h2,r1) + mul32x32_64(h3,r0) + mul32x32_64(h4,s4);
-    t[4]  = mul32x32_64(h0,r4) + mul32x32_64(h1,r3) + mul32x32_64(h2,r2) + mul32x32_64(h3,r1) + mul32x32_64(h4,r0);
-
-                    h0 = (uint32_t)t[0] & 0x3ffffff; c =           (t[0] >> 26);
-    t[1] += c;      h1 = (uint32_t)t[1] & 0x3ffffff; b = (uint32_t)(t[1] >> 26);
-    t[2] += b;      h2 = (uint32_t)t[2] & 0x3ffffff; b = (uint32_t)(t[2] >> 26);
-    t[3] += b;      h3 = (uint32_t)t[3] & 0x3ffffff; b = (uint32_t)(t[3] >> 26);
-    t[4] += b;      h4 = (uint32_t)t[4] & 0x3ffffff; b = (uint32_t)(t[4] >> 26);
-    h0 += b * 5;
-
-    if (inlen >= 16) goto poly1305_donna_16bytes;
-
-    /* final bytes */
-poly1305_donna_atmost15bytes:
-    if (!inlen) goto poly1305_donna_finish;
-
-    for (j = 0; j < inlen; j++) mp[j] = m[j];
-    mp[j++] = 1;
-    for (; j < 16; j++) mp[j] = 0;
-    inlen = 0;
-
-    t0 = ReadLE32(mp+0);
-    t1 = ReadLE32(mp+4);
-    t2 = ReadLE32(mp+8);
-    t3 = ReadLE32(mp+12);
-
-    h0 += t0 & 0x3ffffff;
-    h1 += ((((uint64_t)t1 << 32) | t0) >> 26) & 0x3ffffff;
-    h2 += ((((uint64_t)t2 << 32) | t1) >> 20) & 0x3ffffff;
-    h3 += ((((uint64_t)t3 << 32) | t2) >> 14) & 0x3ffffff;
-    h4 += (t3 >> 8);
-
-    goto poly1305_donna_mul;
-
-poly1305_donna_finish:
-                 b = h0 >> 26; h0 = h0 & 0x3ffffff;
-    h1 +=     b; b = h1 >> 26; h1 = h1 & 0x3ffffff;
-    h2 +=     b; b = h2 >> 26; h2 = h2 & 0x3ffffff;
-    h3 +=     b; b = h3 >> 26; h3 = h3 & 0x3ffffff;
-    h4 +=     b; b = h4 >> 26; h4 = h4 & 0x3ffffff;
-    h0 += b * 5; b = h0 >> 26; h0 = h0 & 0x3ffffff;
-    h1 +=     b;
-
-    g0 = h0 + 5; b = g0 >> 26; g0 &= 0x3ffffff;
-    g1 = h1 + b; b = g1 >> 26; g1 &= 0x3ffffff;
-    g2 = h2 + b; b = g2 >> 26; g2 &= 0x3ffffff;
-    g3 = h3 + b; b = g3 >> 26; g3 &= 0x3ffffff;
-    g4 = h4 + b - (1 << 26);
-
-    b = (g4 >> 31) - 1;
-    nb = ~b;
-    h0 = (h0 & nb) | (g0 & b);
-    h1 = (h1 & nb) | (g1 & b);
-    h2 = (h2 & nb) | (g2 & b);
-    h3 = (h3 & nb) | (g3 & b);
-    h4 = (h4 & nb) | (g4 & b);
-
-    f0 = ((h0      ) | (h1 << 26)) + (uint64_t)ReadLE32(&key[16]);
-    f1 = ((h1 >>  6) | (h2 << 20)) + (uint64_t)ReadLE32(&key[20]);
-    f2 = ((h2 >> 12) | (h3 << 14)) + (uint64_t)ReadLE32(&key[24]);
-    f3 = ((h3 >> 18) | (h4 <<  8)) + (uint64_t)ReadLE32(&key[28]);
-
-    WriteLE32(&out[ 0], f0); f1 += (f0 >> 32);
-    WriteLE32(&out[ 4], f1); f2 += (f1 >> 32);
-    WriteLE32(&out[ 8], f2); f3 += (f2 >> 32);
-    WriteLE32(&out[12], f3);
+    h0 = st->h[0];
+    h1 = st->h[1];
+    h2 = st->h[2];
+    h3 = st->h[3];
+    h4 = st->h[4];
+
+    while (bytes >= POLY1305_BLOCK_SIZE) {
+        /* h += m[i] */
+        h0 += (ReadLE32(m+ 0)     ) & 0x3ffffff;
+        h1 += (ReadLE32(m+ 3) >> 2) & 0x3ffffff;
+        h2 += (ReadLE32(m+ 6) >> 4) & 0x3ffffff;
+        h3 += (ReadLE32(m+ 9) >> 6) & 0x3ffffff;
+        h4 += (ReadLE32(m+12) >> 8) | hibit;
+
+        /* h *= r */
+        d0 = ((uint64_t)h0 * r0) + ((uint64_t)h1 * s4) + ((uint64_t)h2 * s3) + ((uint64_t)h3 * s2) + ((uint64_t)h4 * s1);
+        d1 = ((uint64_t)h0 * r1) + ((uint64_t)h1 * r0) + ((uint64_t)h2 * s4) + ((uint64_t)h3 * s3) + ((uint64_t)h4 * s2);
+        d2 = ((uint64_t)h0 * r2) + ((uint64_t)h1 * r1) + ((uint64_t)h2 * r0) + ((uint64_t)h3 * s4) + ((uint64_t)h4 * s3);
+        d3 = ((uint64_t)h0 * r3) + ((uint64_t)h1 * r2) + ((uint64_t)h2 * r1) + ((uint64_t)h3 * r0) + ((uint64_t)h4 * s4);
+        d4 = ((uint64_t)h0 * r4) + ((uint64_t)h1 * r3) + ((uint64_t)h2 * r2) + ((uint64_t)h3 * r1) + ((uint64_t)h4 * r0);
+
+        /* (partial) h %= p */
+                      c = (uint32_t)(d0 >> 26); h0 = (uint32_t)d0 & 0x3ffffff;
+        d1 += c;      c = (uint32_t)(d1 >> 26); h1 = (uint32_t)d1 & 0x3ffffff;
+        d2 += c;      c = (uint32_t)(d2 >> 26); h2 = (uint32_t)d2 & 0x3ffffff;
+        d3 += c;      c = (uint32_t)(d3 >> 26); h3 = (uint32_t)d3 & 0x3ffffff;
+        d4 += c;      c = (uint32_t)(d4 >> 26); h4 = (uint32_t)d4 & 0x3ffffff;
+        h0 += c * 5;  c =           (h0 >> 26); h0 =           h0 & 0x3ffffff;
+        h1 += c;
+
+        m += POLY1305_BLOCK_SIZE;
+        bytes -= POLY1305_BLOCK_SIZE;
+    }
+
+    st->h[0] = h0;
+    st->h[1] = h1;
+    st->h[2] = h2;
+    st->h[3] = h3;
+    st->h[4] = h4;
+}
+
+void poly1305_finish(poly1305_context *st, unsigned char mac[16]) noexcept {
+    uint32_t h0,h1,h2,h3,h4,c;
+    uint32_t g0,g1,g2,g3,g4;
+    uint64_t f;
+    uint32_t mask;
+
+    /* process the remaining block */
+    if (st->leftover) {
+        size_t i = st->leftover;
+        st->buffer[i++] = 1;
+        for (; i < POLY1305_BLOCK_SIZE; i++) {
+            st->buffer[i] = 0;
+        }
+        st->final = 1;
+        poly1305_blocks(st, st->buffer, POLY1305_BLOCK_SIZE);
+    }
+
+    /* fully carry h */
+    h0 = st->h[0];
+    h1 = st->h[1];
+    h2 = st->h[2];
+    h3 = st->h[3];
+    h4 = st->h[4];
+
+                 c = h1 >> 26; h1 = h1 & 0x3ffffff;
+    h2 +=     c; c = h2 >> 26; h2 = h2 & 0x3ffffff;
+    h3 +=     c; c = h3 >> 26; h3 = h3 & 0x3ffffff;
+    h4 +=     c; c = h4 >> 26; h4 = h4 & 0x3ffffff;
+    h0 += c * 5; c = h0 >> 26; h0 = h0 & 0x3ffffff;
+    h1 +=     c;
+
+    /* compute h + -p */
+    g0 = h0 + 5; c = g0 >> 26; g0 &= 0x3ffffff;
+    g1 = h1 + c; c = g1 >> 26; g1 &= 0x3ffffff;
+    g2 = h2 + c; c = g2 >> 26; g2 &= 0x3ffffff;
+    g3 = h3 + c; c = g3 >> 26; g3 &= 0x3ffffff;
+    g4 = h4 + c - (1UL << 26);
+
+    /* select h if h < p, or h + -p if h >= p */
+    mask = (g4 >> ((sizeof(uint32_t) * 8) - 1)) - 1;
+    g0 &= mask;
+    g1 &= mask;
+    g2 &= mask;
+    g3 &= mask;
+    g4 &= mask;
+    mask = ~mask;
+    h0 = (h0 & mask) | g0;
+    h1 = (h1 & mask) | g1;
+    h2 = (h2 & mask) | g2;
+    h3 = (h3 & mask) | g3;
+    h4 = (h4 & mask) | g4;
+
+    /* h = h % (2^128) */
+    h0 = ((h0      ) | (h1 << 26)) & 0xffffffff;
+    h1 = ((h1 >>  6) | (h2 << 20)) & 0xffffffff;
+    h2 = ((h2 >> 12) | (h3 << 14)) & 0xffffffff;
+    h3 = ((h3 >> 18) | (h4 <<  8)) & 0xffffffff;
+
+    /* mac = (h + pad) % (2^128) */
+    f = (uint64_t)h0 + st->pad[0]            ; h0 = (uint32_t)f;
+    f = (uint64_t)h1 + st->pad[1] + (f >> 32); h1 = (uint32_t)f;
+    f = (uint64_t)h2 + st->pad[2] + (f >> 32); h2 = (uint32_t)f;
+    f = (uint64_t)h3 + st->pad[3] + (f >> 32); h3 = (uint32_t)f;
+
+    WriteLE32(mac +  0, h0);
+    WriteLE32(mac +  4, h1);
+    WriteLE32(mac +  8, h2);
+    WriteLE32(mac + 12, h3);
+
+    /* zero out the state */
+    st->h[0] = 0;
+    st->h[1] = 0;
+    st->h[2] = 0;
+    st->h[3] = 0;
+    st->h[4] = 0;
+    st->r[0] = 0;
+    st->r[1] = 0;
+    st->r[2] = 0;
+    st->r[3] = 0;
+    st->r[4] = 0;
+    st->pad[0] = 0;
+    st->pad[1] = 0;
+    st->pad[2] = 0;
+    st->pad[3] = 0;
+}
+
+void poly1305_update(poly1305_context *st, const unsigned char *m, size_t bytes) noexcept {
+    size_t i;
+
+    /* handle leftover */
+    if (st->leftover) {
+        size_t want = (POLY1305_BLOCK_SIZE - st->leftover);
+        if (want > bytes) {
+            want = bytes;
+        }
+        for (i = 0; i < want; i++) {
+            st->buffer[st->leftover + i] = m[i];
+        }
+        bytes -= want;
+        m += want;
+        st->leftover += want;
+        if (st->leftover < POLY1305_BLOCK_SIZE) return;
+        poly1305_blocks(st, st->buffer, POLY1305_BLOCK_SIZE);
+        st->leftover = 0;
+    }
+
+    /* process full blocks */
+    if (bytes >= POLY1305_BLOCK_SIZE) {
+        size_t want = (bytes & ~(POLY1305_BLOCK_SIZE - 1));
+        poly1305_blocks(st, m, want);
+        m += want;
+        bytes -= want;
+    }
+
+    /* store leftover */
+    if (bytes) {
+        for (i = 0; i < bytes; i++) {
+            st->buffer[st->leftover + i] = m[i];
+        }
+        st->leftover += bytes;
+    }
+}
+
+}  // namespace poly1305_donna
+
+void poly1305_auth(unsigned char mac[16], const unsigned char *m, size_t bytes, const unsigned char key[32]) {
+    using namespace poly1305_donna;
+    poly1305_context ctx;
+    poly1305_init(&ctx, key);
+    poly1305_update(&ctx, m, bytes);
+    poly1305_finish(&ctx, mac);
 }
diff --git a/src/crypto/poly1305.h b/src/crypto/poly1305.h
index 650e35bbca..39b69e1cd4 100644
--- a/src/crypto/poly1305.h
+++ b/src/crypto/poly1305.h
@@ -10,6 +10,27 @@
 
 #define POLY1305_KEYLEN 32
 #define POLY1305_TAGLEN 16
+#define POLY1305_BLOCK_SIZE 16
+
+namespace poly1305_donna {
+
+// Based on the public domain implementation by Andrew Moon
+// poly1305-donna-32.h from https://github.com/floodyberry/poly1305-donna
+
+typedef struct {
+    uint32_t r[5];
+    uint32_t h[5];
+    uint32_t pad[4];
+    size_t leftover;
+    unsigned char buffer[POLY1305_BLOCK_SIZE];
+    unsigned char final;
+} poly1305_context;
+
+void poly1305_init(poly1305_context *st, const unsigned char key[32]) noexcept;
+void poly1305_update(poly1305_context *st, const unsigned char *m, size_t bytes) noexcept;
+void poly1305_finish(poly1305_context *st, unsigned char mac[16]) noexcept;
+
+}  // namespace poly1305_donna
 
 void poly1305_auth(unsigned char out[POLY1305_TAGLEN], const unsigned char *m, size_t inlen,
     const unsigned char key[POLY1305_KEYLEN]);