MIPS: memset: Limit excessive `noreorder' assembly mode use

Rewrite to use the `reorder' assembly mode and remove manually scheduled
delay slots except where GAS cannot schedule a delay-slot instruction
due to a data dependency or a section switch (as is the case with the EX
macro). No change in machine code produced.

Signed-off-by: Maciej W. Rozycki <[e-mail]>
[[e-mail]:
Fix conflict with commit 932afdeec18b ("MIPS: Add Kconfig variable for
CPUs with unaligned load/store instructions")]
Signed-off-by: Paul Burton <[e-mail]>
Patchwork: https://patchwork.linux-mips.org/patch/20834/
Cc: Ralf Baechle <[e-mail]>

arch/mips/lib/memset.S

···
 #endif
 	.endm
 
-	.set	noreorder
 	.align	5
 
 	/*
···
 	.endif
 
 	sltiu		t0, a2, STORSIZE	/* very small region? */
+	.set		noreorder
 	bnez		t0, .Lsmall_memset\@
 	 andi		t0, a0, STORMASK	/* aligned? */
+	.set		reorder
 
 #ifdef CONFIG_CPU_MICROMIPS
 	move		t8, a1			/* used by 'swp' instruction */
 	move		t9, a1
 #endif
+	.set		noreorder
 #ifndef CONFIG_CPU_DADDI_WORKAROUNDS
 	beqz		t0, 1f
 	 PTR_SUBU	t0, STORSIZE		/* alignment in bytes */
···
 	 PTR_SUBU	t0, AT			/* alignment in bytes */
 	.set		at
 #endif
+	.set		reorder
 
 #ifdef CONFIG_CPU_HAS_LOAD_STORE_LR
 	R10KCBARRIER(0(ra))
···
 #else /* !CONFIG_CPU_HAS_LOAD_STORE_LR */
 #define STORE_BYTE(N)				\
 	EX(sb, a1, N(a0), .Lbyte_fixup\@);	\
+	.set		noreorder;		\
 	beqz		t0, 0f;			\
-	PTR_ADDU	t0, 1;
+	 PTR_ADDU	t0, 1;			\
+	.set		reorder;
 
 	PTR_ADDU	a2, t0			/* correct size */
 	PTR_ADDU	t0, 1
···
 #endif /* !CONFIG_CPU_HAS_LOAD_STORE_LR */
 1:	ori		t1, a2, 0x3f		/* # of full blocks */
 	xori		t1, 0x3f
+	andi		t0, a2, 0x40-STORSIZE
 	beqz		t1, .Lmemset_partial\@	/* no block to fill */
-	 andi		t0, a2, 0x40-STORSIZE
 
 	PTR_ADDU	t1, a0			/* end address */
-	.set		reorder
 1:	PTR_ADDIU	a0, 64
 	R10KCBARRIER(0(ra))
 	f_fill64 a0, -64, FILL64RG, .Lfwd_fixup\@, \mode
 	bne		t1, a0, 1b
-	.set		noreorder
 
 .Lmemset_partial\@:
 	R10KCBARRIER(0(ra))
···
 	PTR_SUBU	t1, AT
 	.set		at
 #endif
+	PTR_ADDU	a0, t0			/* dest ptr */
 	jr		t1
-	 PTR_ADDU	a0, t0			/* dest ptr */
 
-	.set		push
-	.set		noreorder
-	.set		nomacro
 	/* ... but first do longs ... */
 	f_fill64 a0, -64, FILL64RG, .Lpartial_fixup\@, \mode
-2:	.set		pop
-	andi		a2, STORMASK		/* At most one long to go */
+2:	andi		a2, STORMASK		/* At most one long to go */
 
+	.set		noreorder
 	beqz		a2, 1f
 #ifdef CONFIG_CPU_HAS_LOAD_STORE_LR
 	 PTR_ADDU	a0, a2			/* What's left */
+	.set		reorder
 	R10KCBARRIER(0(ra))
 #ifdef __MIPSEB__
 	EX(LONG_S_R, a1, -1(a0), .Llast_fixup\@)
···
 #endif
 #else
 	 PTR_SUBU	t0, $0, a2
+	.set		reorder
 	move		a2, zero		/* No remaining longs */
 	PTR_ADDIU	t0, 1
 	STORE_BYTE(0)
···
 #endif
 0:
 #endif
-1:	jr		ra
-	 move		a2, zero
+1:	move		a2, zero
+	jr		ra
 
 .Lsmall_memset\@:
+	PTR_ADDU	t1, a0, a2
 	beqz		a2, 2f
-	 PTR_ADDU	t1, a0, a2
 
 1:	PTR_ADDIU	a0, 1			/* fill bytewise */
 	R10KCBARRIER(0(ra))
+	.set		noreorder
 	bne		t1, a0, 1b
 	 EX(sb, a1, -1(a0), .Lsmall_fixup\@)
+	.set		reorder
 
-2:	jr		ra			/* done */
-	 move		a2, zero
+2:	move		a2, zero
+	jr		ra			/* done */
 	.if __memset == 1
 	END(memset)
 	.set __memset, 0
···
 	 *      a2     =             a2                -              t0                   + 1
 	 */
 	PTR_SUBU	a2, t0
+	PTR_ADDIU	a2, 1
 	jr		ra
-	 PTR_ADDIU	a2, 1
 #endif /* !CONFIG_CPU_HAS_LOAD_STORE_LR */
 
 .Lfirst_fixup\@:
 	/* unset_bytes already in a2 */
 	jr	ra
-	 nop
 
 .Lfwd_fixup\@:
 	/*
···
 	andi		a2, 0x3f
 	LONG_L		t0, THREAD_BUADDR(t0)
 	LONG_ADDU	a2, t1
+	LONG_SUBU	a2, t0
 	jr		ra
-	 LONG_SUBU	a2, t0
 
 .Lpartial_fixup\@:
 	/*
···
 	andi		a2, STORMASK
 	LONG_L		t0, THREAD_BUADDR(t0)
 	LONG_ADDU	a2, a0
+	LONG_SUBU	a2, t0
 	jr		ra
-	 LONG_SUBU	a2, t0
 
 .Llast_fixup\@:
 	/* unset_bytes already in a2 */
 	jr		ra
-	 nop
 
 .Lsmall_fixup\@:
 	/*
 	 * unset_bytes = end_addr - current_addr + 1
 	 *      a2     =    t1    -      a0      + 1
 	 */
-	.set		reorder
 	PTR_SUBU	a2, t1, a0
 	PTR_ADDIU	a2, 1
 	jr		ra
-	.set		noreorder
 
 	.endm
 
···
 
 LEAF(memset)
 EXPORT_SYMBOL(memset)
+	move		v0, a0			/* result */
 	beqz		a1, 1f
-	 move		v0, a0			/* result */
 
 	andi		a1, 0xff		/* spread fillword */
 	LONG_SLL		t1, a1, 8