internal/poly1305: extend ppc64le support to ppc64

The cipher needs to load the stream in LE order. Use the byte
reversing loads on BE.

Also, remove the unused variable poly1305Mask in the PPC64
asm file too.

Change-Id: Ie90fe7bb0ea7a3bcb76583e0cf9c1e4133499541
Reviewed-on: https://go-review.googlesource.com/c/crypto/+/614298
Reviewed-by: Michael Knyszek <[email protected]>
Reviewed-by: Archana Ravindar <[email protected]>
LUCI-TryBot-Result: Go LUCI <[email protected]>
Reviewed-by: David Chase <[email protected]>

Author: pmur

Diff for: internal/poly1305/mac_noasm.go

@@ -2,7 +2,7 @@
 // Use of this source code is governed by a BSD-style
 // license that can be found in the LICENSE file.
 
-//go:build (!amd64 && !ppc64le && !s390x) || !gc || purego
+//go:build (!amd64 && !ppc64le && !ppc64 && !s390x) || !gc || purego
 
 package poly1305
 

Diff for: internal/poly1305/sum_ppc64le.go renamed to internal/poly1305/sum_ppc64x.go

@@ -2,7 +2,7 @@
 // Use of this source code is governed by a BSD-style
 // license that can be found in the LICENSE file.
 
-//go:build gc && !purego
+//go:build gc && !purego && (ppc64 || ppc64le)
 
 package poly1305
 

Diff for: internal/poly1305/sum_ppc64le.s renamed to internal/poly1305/sum_ppc64x.s

@@ -2,15 +2,25 @@
 // Use of this source code is governed by a BSD-style
 // license that can be found in the LICENSE file.
 
-//go:build gc && !purego
+//go:build gc && !purego && (ppc64 || ppc64le)
 
 #include "textflag.h"
 
 // This was ported from the amd64 implementation.
 
+#ifdef GOARCH_ppc64le
+#define LE_MOVD MOVD
+#define LE_MOVWZ MOVWZ
+#define LE_MOVHZ MOVHZ
+#else
+#define LE_MOVD MOVDBR
+#define LE_MOVWZ MOVWBR
+#define LE_MOVHZ MOVHBR
+#endif
+
 #define POLY1305_ADD(msg, h0, h1, h2, t0, t1, t2) \
-	MOVD (msg), t0;  \
-	MOVD 8(msg), t1; \
+	LE_MOVD (msg)( R0), t0; \
+	LE_MOVD (msg)(R24), t1; \
 	MOVD $1, t2;     \
 	ADDC t0, h0, h0; \
 	ADDE t1, h1, h1; \
@@ -50,10 +60,6 @@
 	ADDE   t3, h1, h1;  \
 	ADDZE  h2
 
-DATA ·poly1305Mask<>+0x00(SB)/8, $0x0FFFFFFC0FFFFFFF
-DATA ·poly1305Mask<>+0x08(SB)/8, $0x0FFFFFFC0FFFFFFC
-GLOBL ·poly1305Mask<>(SB), RODATA, $16
-
 // func update(state *[7]uint64, msg []byte)
 TEXT ·update(SB), $0-32
 	MOVD state+0(FP), R3
@@ -66,6 +72,8 @@ TEXT ·update(SB), $0-32
 	MOVD 24(R3), R11 // r0
 	MOVD 32(R3), R12 // r1
 
+	MOVD $8, R24
+
 	CMP R5, $16
 	BLT bytes_between_0_and_15
 
@@ -94,7 +102,7 @@ flush_buffer:
 
 	// Greater than 8 -- load the rightmost remaining bytes in msg
 	// and put into R17 (h1)
-	MOVD (R4)(R21), R17
+	LE_MOVD (R4)(R21), R17
 	MOVD $16, R22
 
 	// Find the offset to those bytes
@@ -118,7 +126,7 @@ just1:
 	BLT less8
 
 	// Exactly 8
-	MOVD (R4), R16
+	LE_MOVD (R4), R16
 
 	CMP R17, $0
 
@@ -133,15 +141,15 @@ less8:
 	MOVD  $0, R22   // shift count
 	CMP   R5, $4
 	BLT   less4
-	MOVWZ (R4), R16
+	LE_MOVWZ (R4), R16
 	ADD   $4, R4
 	ADD   $-4, R5
 	MOVD  $32, R22
 
 less4:
 	CMP   R5, $2
 	BLT   less2
-	MOVHZ (R4), R21
+	LE_MOVHZ (R4), R21
 	SLD   R22, R21, R21
 	OR    R16, R21, R16
 	ADD   $16, R22