s390/lib: use expoline for all bcr instructions

The memove, memset, memcpy, __memset16, __memset32 and __memset64
function have an additional indirect return branch in form of a
"bzr" instruction. These need to use expolines as well.

Cc: <stable@vger.kernel.org> # v4.17+
Fixes: 97489e06 ("s390/lib: use expoline for indirect branches")
Reviewed-by: Heiko Carstens <heiko.carstens@de.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>

arch/s390/lib/mem.S

@@ -17,7 +17,7 @@
 ENTRY(memmove)
 	ltgr	%r4,%r4
 	lgr	%r1,%r2
-	bzr	%r14
+	jz	.Lmemmove_exit
 	aghi	%r4,-1
 	clgr	%r2,%r3
 	jnh	.Lmemmove_forward
@@ -36,6 +36,7 @@ ENTRY(memmove)
 .Lmemmove_forward_remainder:
 	larl	%r5,.Lmemmove_mvc
 	ex	%r4,0(%r5)
+.Lmemmove_exit:
 	BR_EX	%r14
 .Lmemmove_reverse:
 	ic	%r0,0(%r4,%r3)
@@ -65,7 +66,7 @@ EXPORT_SYMBOL(memmove)
  */
 ENTRY(memset)
 	ltgr	%r4,%r4
-	bzr	%r14
+	jz	.Lmemset_exit
 	ltgr	%r3,%r3
 	jnz	.Lmemset_fill
 	aghi	%r4,-1
@@ -80,6 +81,7 @@ ENTRY(memset)
 .Lmemset_clear_remainder:
 	larl	%r3,.Lmemset_xc
 	ex	%r4,0(%r3)
+.Lmemset_exit:
 	BR_EX	%r14
 .Lmemset_fill:
 	cghi	%r4,1
@@ -115,7 +117,7 @@ EXPORT_SYMBOL(memset)
  */
 ENTRY(memcpy)
 	ltgr	%r4,%r4
-	bzr	%r14
+	jz	.Lmemcpy_exit
 	aghi	%r4,-1
 	srlg	%r5,%r4,8
 	ltgr	%r5,%r5
@@ -124,6 +126,7 @@ ENTRY(memcpy)
 .Lmemcpy_remainder:
 	larl	%r5,.Lmemcpy_mvc
 	ex	%r4,0(%r5)
+.Lmemcpy_exit:
 	BR_EX	%r14
 .Lmemcpy_loop:
 	mvc	0(256,%r1),0(%r3)
@@ -145,9 +148,9 @@ EXPORT_SYMBOL(memcpy)
 .macro __MEMSET bits,bytes,insn
 ENTRY(__memset\bits)
 	ltgr	%r4,%r4
-	bzr	%r14
+	jz	.L__memset_exit\bits
 	cghi	%r4,\bytes
-	je	.L__memset_exit\bits
+	je	.L__memset_store\bits
 	aghi	%r4,-(\bytes+1)
 	srlg	%r5,%r4,8
 	ltgr	%r5,%r5
@@ -163,8 +166,9 @@ ENTRY(__memset\bits)
 	larl	%r5,.L__memset_mvc\bits
 	ex	%r4,0(%r5)
 	BR_EX	%r14
-.L__memset_exit\bits:
+.L__memset_store\bits:
 	\insn	%r3,0(%r2)
+.L__memset_exit\bits:
 	BR_EX	%r14
 .L__memset_mvc\bits:
 	mvc	\bytes(1,%r1),0(%r1)