runtime: correct the TLS base offset on Android

CL 170955 set tlsg to the Android Q free TLS slot offset in the linker
data (16 on amd64, 8 on 386), offsetting all TLS relative access.
We need the 0'th slot (TLS_SLOT_SELF) at initialization, so
compensate with a corresponding negative offset.

Fixes the android/386 and android/amd64 builders broken by CL 170955.

Change-Id: I9882088c0c8bc6a777d2aabc9404cb76f02b6cea
Reviewed-on: https://go-review.googlesource.com/c/go/+/170956
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Cherry Zhang <cherryyz@google.com>

src/runtime/asm_386.s

@@ -172,8 +172,10 @@ nocpuinfo:
 	TESTL	AX, AX
 	JZ	needtls
 #ifdef GOOS_android
-	MOVL	0(TLS), BX
-	MOVL	BX, 12(SP)	// arg 4: TLS base, stored in the first slot (TLS_SLOT_SELF).
+	// arg 4: TLS base, stored in slot 0 (Android's TLS_SLOT_SELF).
+	// Compensate for tls_g (+8).
+	MOVL	-8(TLS), BX
+	MOVL	BX, 12(SP)
 	MOVL	$runtime·tls_g(SB), 8(SP)	// arg 3: &tls_g
 #else
 	MOVL	$0, BX

src/runtime/asm_amd64.s

@@ -136,7 +136,9 @@ nocpuinfo:
 	MOVQ	$setg_gcc<>(SB), SI // arg 2: setg_gcc
 #ifdef GOOS_android
 	MOVQ	$runtime·tls_g(SB), DX 	// arg 3: &tls_g
-	MOVQ	0(TLS), CX	// arg 4: TLS base, stored in the first slot (TLS_SLOT_SELF).
+	// arg 4: TLS base, stored in slot 0 (Android's TLS_SLOT_SELF).
+	// Compensate for tls_g (+16).
+	MOVQ	-16(TLS), CX
 #else
 	MOVQ	$0, DX	// arg 3, 4: not used when using platform's TLS
 	MOVQ	$0, CX