ARM: xscale: fix multi-cpu compilation

[ Upstream commit c7b6804 ]

Building a combined ARMv4+XScale kernel produces these
and other build failures:

/tmp/copypage-xscale-3aa821.s: Assembler messages:
/tmp/copypage-xscale-3aa821.s:167: Error: selected processor does not support `pld [r7,#0]' in ARM mode
/tmp/copypage-xscale-3aa821.s:168: Error: selected processor does not support `pld [r7,#32]' in ARM mode
/tmp/copypage-xscale-3aa821.s:169: Error: selected processor does not support `pld [r1,#0]' in ARM mode
/tmp/copypage-xscale-3aa821.s:170: Error: selected processor does not support `pld [r1,#32]' in ARM mode
/tmp/copypage-xscale-3aa821.s:171: Error: selected processor does not support `pld [r7,#64]' in ARM mode
/tmp/copypage-xscale-3aa821.s:176: Error: selected processor does not support `ldrd r4,r5,[r7],#8' in ARM mode
/tmp/copypage-xscale-3aa821.s:180: Error: selected processor does not support `strd r4,r5,[r1],#8' in ARM mode

Add an explict .arch armv5 in the inline assembly to allow the ARMv5
specific instructions regardless of the compiler -march= target.

Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Arnd Bergmann <[email protected]>
Signed-off-by: Sasha Levin <[email protected]>

Author: arndb

arch/arm/mm/copypage-xscale.c

@@ -42,6 +42,7 @@ static void mc_copy_user_page(void *from, void *to)
 	 * when prefetching destination as well.  (NP)
 	 */
 	asm volatile ("\
+.arch xscale					\n\
 	pld	[%0, #0]			\n\
 	pld	[%0, #32]			\n\
 	pld	[%1, #0]			\n\
@@ -106,8 +107,9 @@ void
 xscale_mc_clear_user_highpage(struct page *page, unsigned long vaddr)
 {
 	void *ptr, *kaddr = kmap_atomic(page);
-	asm volatile(
-	"mov	r1, %2				\n\
+	asm volatile("\
+.arch xscale					\n\
+	mov	r1, %2				\n\
 	mov	r2, #0				\n\
 	mov	r3, #0				\n\
 1:	mov	ip, %0				\n\