arm64: fix inline asm in load_unaligned_zeropad()

The inline asm's addr operand is marked as input-only, however in
the case where an exception is taken it may be modified by the BIC
instruction on the exception path. Fix the problem by using a temporary
register as the destination register for the BIC instruction.
Signed-off-by: Peter Collingbourne <pcc@google.com>
Cc: stable@vger.kernel.org
Link: https://linux-review.googlesource.com/id/I84538c8a2307d567b4f45bb20b715451005f9617
Link: https://lore.kernel.org/r/20210401165110.3952103-1-pcc@google.comSigned-off-by: Will Deacon <will@kernel.org>

arch/arm64/include/asm/word-at-a-time.h

@@ -53,7 +53,7 @@ static inline unsigned long find_zero(unsigned long mask)
  */
 static inline unsigned long load_unaligned_zeropad(const void *addr)
 {
-	unsigned long ret, offset;
+	unsigned long ret, tmp;
 
 	/* Load word from unaligned pointer addr */
 	asm(
@@ -61,9 +61,9 @@ static inline unsigned long load_unaligned_zeropad(const void *addr)
 	"2:\n"
 	"	.pushsection .fixup,\"ax\"\n"
 	"	.align 2\n"
-	"3:	and	%1, %2, #0x7\n"
-	"	bic	%2, %2, #0x7\n"
-	"	ldr	%0, [%2]\n"
+	"3:	bic	%1, %2, #0x7\n"
+	"	ldr	%0, [%1]\n"
+	"	and	%1, %2, #0x7\n"
 	"	lsl	%1, %1, #0x3\n"
 #ifndef __AARCH64EB__
 	"	lsr	%0, %0, %1\n"
@@ -73,7 +73,7 @@ static inline unsigned long load_unaligned_zeropad(const void *addr)
 	"	b	2b\n"
 	"	.popsection\n"
 	_ASM_EXTABLE(1b, 3b)
-	: "=&r" (ret), "=&r" (offset)
+	: "=&r" (ret), "=&r" (tmp)
 	: "r" (addr), "Q" (*(unsigned long *)addr));
 
 	return ret;