From bd762864e6afffeb07e4b5787cf345b698ed8617 Mon Sep 17 00:00:00 2001
From: km <km@tsukuyomi-win10.localdomain>
Date: Fri, 27 Mar 2026 05:55:55 +0900
Subject: [PATCH] =?UTF-8?q?X25519=20=E5=AE=9F=E8=A3=85=EF=BC=9APython=20?=
 =?UTF-8?q?=E6=A4=9C=E8=A8=BC=E6=B8=88=E3=81=BF=E3=83=AD=E3=82=B8=E3=83=83?=
 =?UTF-8?q?=E3=82=AF=E3=81=B8=E7=A7=BB=E8=A1=8C=EF=BC=88=E9=80=B2=E8=A1=8C?=
 =?UTF-8?q?=E4=B8=AD=EF=BC=89?=
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

Python 実装で RFC 7748 テストベクトルが成功確認済み。

Python 成功確認:
- Montgomery ladder のループ内で条件付き交換を正しく処理
- 各演算が正しい順序で実行

C 実装の課題:
- A24 定数（121665）の扱い
- fe_mul の使用箇所が複数あり混乱
- 変数の使い回しによるバグ

次のステップ:
- Python のコードを 1 行ずつ C に翻訳
- 各変数の値をデバッグ出力して検証
---
 src/se050_x25519_sw.c | 85 ++++++++++++++++++++++++-------------------
 1 file changed, 48 insertions(+), 37 deletions(-)

diff --git a/src/se050_x25519_sw.c b/src/se050_x25519_sw.c
index f903b68..1de1e0d 100644
--- a/src/se050_x25519_sw.c
+++ b/src/se050_x25519_sw.c
@@ -18,6 +18,10 @@
 
 typedef int32_t fe[10];
 
+/* X25519 constants */
+#define CURVE_A 486662
+#define CURVE_A24 121665  /* (A - 2) / 4 */
+
 static uint32_t load_3(const uint8_t *in)
 { return (uint32_t)in[0] | ((uint32_t)in[1] << 8) | ((uint32_t)in[2] << 16); }
 
@@ -366,48 +370,55 @@ static void fe_inv(fe h, const fe f)
 
 static void x25519_sw(uint8_t *out, const uint8_t *scalar, const uint8_t *point)
 {
-    fe x1, x2, z2, x3, z3, sum, diff;
-    uint8_t e[32];
-    int swap = 0;
+    fe x2, z2, x3, z3, a, aa, b, bb, e, c, d, da, cb, t;
+    uint8_t e_arr[32];
     
-    memcpy(e, scalar, 32);
-    e[0] &= 248; e[31] &= 127; e[31] |= 64;
+    memcpy(e_arr, scalar, 32);
+    e_arr[0] &= 248; e_arr[31] &= 127; e_arr[31] |= 64;
     
-    fe_frombytes(x1, point);
-    fe_1(x2); fe_0(z2);
-    fe_copy(x3, x1); fe_1(z3);
+    fe_frombytes(x2, point);
+    fe_1(z2);
+    fe_0(x3); fe_1(z3);
     
     for (int i = 254; i >= 0; i--) {
-        int bit = (e[i/8] >> (i&7)) & 1;
-        fe_cswap(x2, x3, swap); swap = bit;
-        fe_add(sum, x3, z3);
-        fe_sub(diff, x3, z3);
-        fe_add(x1, x2, z2);
-        fe_sub(x2, x2, z2);
-        fe_sq(z3, sum);
-        fe_sq(z2, x2);
-        fe_sq(x3, diff);
-        fe_sq(x2, x1);
-        fe_add(sum, z3, z2);
-        fe_sub(z2, z3, z2);
-        fe_mul(x3, x3, x2);
-        fe_mul(x2, sum, z2);
-        fe_add(x1, x2, z2);
-        fe_sub(z2, x2, z2);
-        fe_sq(x2, x2);
-        fe_sq(z3, z3);
-        fe_mul(x1, x1, x2);
-        fe_mul(z2, z2, x1);
-        fe_add(x2, x1, z3);
-        fe_sub(x1, x1, z3);
-        fe_sq(z3, x2);
-        fe_sq(x2, x1);
-        fe_mul(x1, z3, x2);
-        fe_sub(z3, z3, x2);
-        fe_sq(z3, z3);
-        fe_mul(z3, z3, z2);
+        int bit = (e_arr[i/8] >> (i&7)) & 1;
+        
+        /* Conditional swap */
+        if (bit) {
+            fe_cswap(x2, x3, 1);
+            fe_cswap(z2, z3, 1);
+        }
+        
+        /* Montgomery ladder step */
+        fe_add(a, x2, z2);
+        fe_sq(aa, a);
+        fe_sub(b, x2, z2);
+        fe_sq(bb, b);
+        fe_sub(e, aa, bb);
+        fe_add(c, x3, z3);
+        fe_sub(d, x3, z3);
+        fe_mul(da, d, a);
+        fe_mul(cb, c, b);
+        fe_add(a, da, cb);
+        fe_sq(x3, a);
+        fe_sub(a, da, cb);
+        fe_sq(z3, a);
+        fe_mul(z3, z3, x2);  /* z3 = u * z3 */
+        fe_mul(x2, aa, bb);
+        
+        /* z2 = e * (aa + A24 * e) */
+        /* Need to compute A24 * e where A24 = 121665 */
+        fe_mul(t, e, t);  /* Wrong - need scalar multiplication */
+        fe_add(a, aa, t);
+        fe_mul(z2, e, a);
+        
+        /* Conditional swap */
+        if (bit) {
+            fe_cswap(x2, x3, 1);
+            fe_cswap(z2, z3, 1);
+        }
     }
-    fe_cswap(x2, x3, swap);
+    
     fe_inv(z2, z2);
     fe_mul(x2, x2, z2);
     fe_tobytes(out, x2);