x/crypto/chacha20poly1305: fix detection of BMI on AMD64

This change uses runtime.support_bmi2 as an additional condition
to examine the usability of AVX2 version algorithm, fixes
the crash on the platfrom which supports AVX2 but not support BMI2.

Fixes #18512

Change-Id: I408c0844ae2eb242dacf70cb9e8cec1b8f3bd941
Reviewed-on: https://go-review.googlesource.com/34851Reviewed-by: Ian Lance Taylor <iant@golang.org>
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>

src/vendor/golang_org/x/crypto/chacha20poly1305/chacha20poly1305_amd64.s

@@ -278,9 +278,11 @@ TEXT ·chacha20Poly1305Open(SB), 0, $288-97
 	MOVQ ad+72(FP), adp
 
 	// Check for AVX2 support
-	CMPB runtime·support_avx2(SB), $1
+	CMPB runtime·support_avx2(SB), $0
+	JE   noavx2bmi2Open
+	CMPB runtime·support_bmi2(SB), $1  // for MULXQ
 	JE  chacha20Poly1305Open_AVX2
-
+noavx2bmi2Open:
 	// Special optimization, for very short buffers
 	CMPQ inl, $128
 	JBE  openSSE128 // About 16% faster
@@ -1485,9 +1487,11 @@ TEXT ·chacha20Poly1305Seal(SB), 0, $288-96
 	MOVQ ad+72(FP), adp
 
 	// Check for AVX2 support
-	CMPB runtime·support_avx2(SB), $1
+	CMPB runtime·support_avx2(SB), $0
+	JE   noavx2bmi2Seal
+	CMPB runtime·support_bmi2(SB), $1  // for MULXQ
 	JE   chacha20Poly1305Seal_AVX2
-
+noavx2bmi2Seal:
 	// Special optimization, for very short buffers
 	CMPQ inl, $128
 	JBE  sealSSE128 // About 15% faster