crypto/elliptic: fix carry bug in x86-64 P-256 implementation.

Patch from Vlad Krasnov and confirmed to be under CLA.

Fixes #20040.

Change-Id: Ieb8436c4dcb6669a1620f1e0d257efd047b1b87c
Reviewed-on: https://go-review.googlesource.com/41070
Run-TryBot: Adam Langley <agl@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>

src/crypto/elliptic/elliptic_test.go

@@ -300,6 +300,29 @@ var p224BaseMultTests = []baseMultTest{
 	},
 }
 
+type scalarMultTest struct {
+	k          string
+	xIn, yIn   string
+	xOut, yOut string
+}
+
+var p256MultTests = []scalarMultTest{
+	{
+		"2a265f8bcbdcaf94d58519141e578124cb40d64a501fba9c11847b28965bc737",
+		"023819813ac969847059028ea88a1f30dfbcde03fc791d3a252c6b41211882ea",
+		"f93e4ae433cc12cf2a43fc0ef26400c0e125508224cdb649380f25479148a4ad",
+		"4d4de80f1534850d261075997e3049321a0864082d24a917863366c0724f5ae3",
+		"a22d2b7f7818a3563e0f7a76c9bf0921ac55e06e2e4d11795b233824b1db8cc0",
+	},
+	{
+		"313f72ff9fe811bf573176231b286a3bdb6f1b14e05c40146590727a71c3bccd",
+		"cc11887b2d66cbae8f4d306627192522932146b42f01d3c6f92bd5c8ba739b06",
+		"a2f08a029cd06b46183085bae9248b0ed15b70280c7ef13a457f5af382426031",
+		"831c3f6b5f762d2f461901577af41354ac5f228c2591f84f8a6e51e2e3f17991",
+		"93f90934cd0ef2c698cc471c60a93524e87ab31ca2412252337f364513e43684",
+	},
+}
+
 func TestBaseMult(t *testing.T) {
 	p224 := P224()
 	for i, e := range p224BaseMultTests {
@@ -379,6 +402,19 @@ func TestP256Mult(t *testing.T) {
 			break
 		}
 	}
+
+	for i, e := range p256MultTests {
+		x, _ := new(big.Int).SetString(e.xIn, 16)
+		y, _ := new(big.Int).SetString(e.yIn, 16)
+		k, _ := new(big.Int).SetString(e.k, 16)
+		expectedX, _ := new(big.Int).SetString(e.xOut, 16)
+		expectedY, _ := new(big.Int).SetString(e.yOut, 16)
+
+		xx, yy := p256.ScalarMult(x, y, k.Bytes())
+		if xx.Cmp(expectedX) != 0 || yy.Cmp(expectedY) != 0 {
+			t.Errorf("#%d: got (%x, %x), want (%x, %x)", i, xx, yy, expectedX, expectedY)
+		}
+	}
 }
 
 func TestInfinity(t *testing.T) {

src/crypto/elliptic/p256_asm_amd64.s

@@ -1314,12 +1314,12 @@ TEXT p256SubInternal(SB),NOSPLIT,$0
 	ADCQ p256const0<>(SB), acc5
 	ADCQ $0, acc6
 	ADCQ p256const1<>(SB), acc7
-	ADCQ $0, mul0
+	ANDQ $1, mul0
 
-	CMOVQNE acc0, acc4
-	CMOVQNE acc1, acc5
-	CMOVQNE acc2, acc6
-	CMOVQNE acc3, acc7
+	CMOVQEQ acc0, acc4
+	CMOVQEQ acc1, acc5
+	CMOVQEQ acc2, acc6
+	CMOVQEQ acc3, acc7
 
 	RET
 /* ---------------------------------------*/