Merge #15512: Add ChaCha20 encryption option (XOR)

2dfe27517 Add ChaCha20 bench (Jonas Schnelli)
2bc2b8b49 Add ChaCha20 encryption option (XOR) (Jonas Schnelli)

Pull request description:

  The current ChaCha20 implementation does not support message encryption (it can only output the keystream which is sufficient for the RNG).

  This PR adds the actual XORing of the `plaintext` with the `keystream` in order to return the desired `ciphertext`.

  Required for v2 message transport protocol.

ACKs for commit 2dfe27:
  jnewbery:
    Looks good. utACK 2dfe2751713c814aea53b5a7563eb74ad1baea00.
  jnewbery:
    utACK 2dfe2751713c814aea53b5a7563eb74ad1baea00
  sipa:
    utACK 2dfe2751713c814aea53b5a7563eb74ad1baea00
  ryanofsky:
    utACK 2dfe2751713c814aea53b5a7563eb74ad1baea00. Changes since last review are just renaming the Crypt method, adding comments, and simplifying the benchmark.

Tree-SHA512: 84bb234da2ca9fdc44bc29a786d9dd215520f81245270c1aef801ef66b6091b7793e2eb38ad6dbb084925245065c5dce9e5582f2d0fa220ab3e182d43412d5b5

2 files changed, 144 insertions, 6 deletions

diff --git a/src/crypto/chacha20.cpp b/src/crypto/chacha20.cpp
index ac4470f04f..42a17f02ff 100644
--- a/src/crypto/chacha20.cpp
+++ b/src/crypto/chacha20.cpp
@@ -71,7 +71,7 @@ void ChaCha20::Seek(uint64_t pos)
     input[13] = pos >> 32;
 }
 
-void ChaCha20::Output(unsigned char* c, size_t bytes)
+void ChaCha20::Keystream(unsigned char* c, size_t bytes)
 {
     uint32_t x0, x1, x2, x3, x4, x5, x6, x7, x8, x9, x10, x11, x12, x13, x14, x15;
     uint32_t j0, j1, j2, j3, j4, j5, j6, j7, j8, j9, j10, j11, j12, j13, j14, j15;
@@ -178,3 +178,133 @@ void ChaCha20::Output(unsigned char* c, size_t bytes)
         c += 64;
     }
 }
+
+void ChaCha20::Crypt(const unsigned char* m, unsigned char* c, size_t bytes)
+{
+    uint32_t x0, x1, x2, x3, x4, x5, x6, x7, x8, x9, x10, x11, x12, x13, x14, x15;
+    uint32_t j0, j1, j2, j3, j4, j5, j6, j7, j8, j9, j10, j11, j12, j13, j14, j15;
+    unsigned char *ctarget = nullptr;
+    unsigned char tmp[64];
+    unsigned int i;
+
+    if (!bytes) return;
+
+    j0 = input[0];
+    j1 = input[1];
+    j2 = input[2];
+    j3 = input[3];
+    j4 = input[4];
+    j5 = input[5];
+    j6 = input[6];
+    j7 = input[7];
+    j8 = input[8];
+    j9 = input[9];
+    j10 = input[10];
+    j11 = input[11];
+    j12 = input[12];
+    j13 = input[13];
+    j14 = input[14];
+    j15 = input[15];
+
+    for (;;) {
+        if (bytes < 64) {
+            // if m has fewer than 64 bytes available, copy m to tmp and
+            // read from tmp instead
+            for (i = 0;i < bytes;++i) tmp[i] = m[i];
+            m = tmp;
+            ctarget = c;
+            c = tmp;
+        }
+        x0 = j0;
+        x1 = j1;
+        x2 = j2;
+        x3 = j3;
+        x4 = j4;
+        x5 = j5;
+        x6 = j6;
+        x7 = j7;
+        x8 = j8;
+        x9 = j9;
+        x10 = j10;
+        x11 = j11;
+        x12 = j12;
+        x13 = j13;
+        x14 = j14;
+        x15 = j15;
+        for (i = 20;i > 0;i -= 2) {
+            QUARTERROUND( x0, x4, x8,x12)
+            QUARTERROUND( x1, x5, x9,x13)
+            QUARTERROUND( x2, x6,x10,x14)
+            QUARTERROUND( x3, x7,x11,x15)
+            QUARTERROUND( x0, x5,x10,x15)
+            QUARTERROUND( x1, x6,x11,x12)
+            QUARTERROUND( x2, x7, x8,x13)
+            QUARTERROUND( x3, x4, x9,x14)
+        }
+        x0 += j0;
+        x1 += j1;
+        x2 += j2;
+        x3 += j3;
+        x4 += j4;
+        x5 += j5;
+        x6 += j6;
+        x7 += j7;
+        x8 += j8;
+        x9 += j9;
+        x10 += j10;
+        x11 += j11;
+        x12 += j12;
+        x13 += j13;
+        x14 += j14;
+        x15 += j15;
+
+        x0 ^= ReadLE32(m + 0);
+        x1 ^= ReadLE32(m + 4);
+        x2 ^= ReadLE32(m + 8);
+        x3 ^= ReadLE32(m + 12);
+        x4 ^= ReadLE32(m + 16);
+        x5 ^= ReadLE32(m + 20);
+        x6 ^= ReadLE32(m + 24);
+        x7 ^= ReadLE32(m + 28);
+        x8 ^= ReadLE32(m + 32);
+        x9 ^= ReadLE32(m + 36);
+        x10 ^= ReadLE32(m + 40);
+        x11 ^= ReadLE32(m + 44);
+        x12 ^= ReadLE32(m + 48);
+        x13 ^= ReadLE32(m + 52);
+        x14 ^= ReadLE32(m + 56);
+        x15 ^= ReadLE32(m + 60);
+
+        ++j12;
+        if (!j12) ++j13;
+
+        WriteLE32(c + 0, x0);
+        WriteLE32(c + 4, x1);
+        WriteLE32(c + 8, x2);
+        WriteLE32(c + 12, x3);
+        WriteLE32(c + 16, x4);
+        WriteLE32(c + 20, x5);
+        WriteLE32(c + 24, x6);
+        WriteLE32(c + 28, x7);
+        WriteLE32(c + 32, x8);
+        WriteLE32(c + 36, x9);
+        WriteLE32(c + 40, x10);
+        WriteLE32(c + 44, x11);
+        WriteLE32(c + 48, x12);
+        WriteLE32(c + 52, x13);
+        WriteLE32(c + 56, x14);
+        WriteLE32(c + 60, x15);
+
+        if (bytes <= 64) {
+            if (bytes < 64) {
+                for (i = 0;i < bytes;++i) ctarget[i] = c[i];
+            }
+            input[12] = j12;
+            input[13] = j13;
+            return;
+        }
+        bytes -= 64;
+        c += 64;
+        m += 64;
+    }
+}
diff --git a/src/crypto/chacha20.h b/src/crypto/chacha20.h
index a305977bcd..5a4674f4a8 100644
--- a/src/crypto/chacha20.h
+++ b/src/crypto/chacha20.h
@@ -8,7 +8,8 @@
 #include <stdint.h>
 #include <stdlib.h>
 
-/** A PRNG class for ChaCha20. */
+/** A class for ChaCha20 256-bit stream cipher developed by Daniel J. Bernstein
+    https://cr.yp.to/chacha/chacha-20080128.pdf */
 class ChaCha20
 {
 private:
@@ -17,10 +18,17 @@ private:
 public:
     ChaCha20();
     ChaCha20(const unsigned char* key, size_t keylen);
-    void SetKey(const unsigned char* key, size_t keylen);
-    void SetIV(uint64_t iv);
-    void Seek(uint64_t pos);
-    void Output(unsigned char* output, size_t bytes);
+    void SetKey(const unsigned char* key, size_t keylen); //!< set key with flexible keylength; 256bit recommended */
+    void SetIV(uint64_t iv); // set the 64bit nonce
+    void Seek(uint64_t pos); // set the 64bit block counter
+
+    /** outputs the keystream of size <bytes> into <c> */
+    void Keystream(unsigned char* c, size_t bytes);
+
+    /** enciphers the message <input> of length <bytes> and write the enciphered representation into <output>
+     *  Used for encryption and decryption (XOR)
+     */
+    void Crypt(const unsigned char* input, unsigned char* output, size_t bytes);
 };
 
 #endif // BITCOIN_CRYPTO_CHACHA20_H