Fix fuzz #1 failure: incorrect reduction of BigInt (#246)

constantine/math/arithmetic/limbs_montgomery.nim

@@ -609,7 +609,13 @@ func getMont*(r: var Limbs, a, M, r2modM: Limbs,
   ## Important: `r` is overwritten
   ## The result `r` buffer size MUST be at least the size of `M` buffer
   # Reference: https://eprint.iacr.org/2017/1057.pdf
-  mulMont(r, a, r2ModM, M, m0ninv, spareBits)
+
+  # For conversion to a field element (in the Montgomery domain), we do not use the "no-carry" optimization:
+  #    While Montgomery Reduction can map inputs [0, 4p²) -> [0, p)
+  #    that range is not valid with the no-carry optimization,
+  #    hence an unreduced input that uses 256-bit while prime is 254-bit
+  #    can have an incorrect representation.
+  mulMont_FIPS(r, a, r2ModM, M, m0ninv, skipFinalSub = false)
 
 # Montgomery Modular Exponentiation
 # ------------------------------------------

constantine/math_arbitrary_precision/arithmetic/bigints_views.nim

@@ -70,7 +70,6 @@ func powOddMod_vartime*(
     # if we use redc2xMont (a/R) and montgomery multiplication by R³
     # For now, we call explicit reduction as it can handle all sizes.
     # TODO: explicit reduction uses constant-time division which is **very** expensive
-    # TODO: fix https://github.com/mratsim/constantine/issues/241
     if a.len != M.len:
       let t = allocStackArray(SecretWord, L)
       t.LimbsViewMut.reduce(a.view(), aBits, M.view(), mBits)

tests/math_fields/t_io_fields.nim

@@ -156,4 +156,14 @@ proc main() =
 
         check: p == hex
 
+    test "Fuzz #1 - incorrect reduction of BigInt":
+      block:
+        var a{.noInit.}: Fp[BN254_Snarks]
+        a.fromBig(BigInt[254].fromHex("0xdd1119d0c5b065898a0848e21c209153f4622f06cb763e7ef00eef28b94780f8"))
+
+        var b{.noInit.}: Fp[BN254_Snarks]
+        b.fromBig(BigInt[254].fromHex("0x1b7fe00540e9e4e2a8c73208161b2fdd965c84c129af1449ff8cbecd57538bdc"))
+
+        doAssert bool(a == b)
+
 main()