MIPS: unify cache initialization code

The mips32 & mips64 cache initialization code differs only in that the
mips32 code supports reading the cache size from coprocessor 0 registers
at runtime. Move the more developed mips32 version to a common
arch/mips/lib/cache_init.S & remove the now-redundant mips64 version in
order to reduce duplication. The temporary registers used are shuffled
slightly in order to work for both mips32 & mips64 builds. The RA
register is defined differently to suit mips32 & mips64, but will be
removed by a later commit in the series after further cleanup.

Signed-off-by: Paul Burton <paul.burton@imgtec.com>
Cc: Daniel Schwierzeck <daniel.schwierzeck@gmail.com>
diff --git a/arch/mips/cpu/mips32/Makefile b/arch/mips/cpu/mips32/Makefile
index fa82dd3..19d316a 100644
--- a/arch/mips/cpu/mips32/Makefile
+++ b/arch/mips/cpu/mips32/Makefile
@@ -6,7 +6,6 @@
 #
 
 extra-y	= start.o
-obj-y	= cache.o
-obj-y	+= cpu.o interrupts.o time.o
+obj-y	= cpu.o interrupts.o time.o
 
 obj-$(CONFIG_SOC_AU1X00) += au1x00/
diff --git a/arch/mips/cpu/mips64/Makefile b/arch/mips/cpu/mips64/Makefile
index 899c319..cb4db9c 100644
--- a/arch/mips/cpu/mips64/Makefile
+++ b/arch/mips/cpu/mips64/Makefile
@@ -6,4 +6,4 @@
 #
 
 extra-y	= start.o
-obj-y	= cpu.o interrupts.o time.o cache.o
+obj-y	= cpu.o interrupts.o time.o
diff --git a/arch/mips/cpu/mips64/cache.S b/arch/mips/cpu/mips64/cache.S
deleted file mode 100644
index 36d8688..0000000
--- a/arch/mips/cpu/mips64/cache.S
+++ /dev/null
@@ -1,213 +0,0 @@
-/*
- *  Cache-handling routined for MIPS CPUs
- *
- *  Copyright (c) 2003	Wolfgang Denk <wd@denx.de>
- *
- * SPDX-License-Identifier:	GPL-2.0+
- */
-
-#include <asm-offsets.h>
-#include <config.h>
-#include <asm/asm.h>
-#include <asm/regdef.h>
-#include <asm/mipsregs.h>
-#include <asm/addrspace.h>
-#include <asm/cacheops.h>
-
-#define RA		t9
-
-/*
- * 16kB is the maximum size of instruction and data caches on MIPS 4K,
- * 64kB is on 4KE, 24K, 5K, etc. Set bigger size for convenience.
- *
- * Note that the above size is the maximum size of primary cache. U-Boot
- * doesn't have L2 cache support for now.
- */
-#define MIPS_MAX_CACHE_SIZE	0x10000
-
-#define INDEX_BASE	CKSEG0
-
-	.macro	cache_op op addr
-	.set	push
-	.set	noreorder
-	.set	mips3
-	cache	\op, 0(\addr)
-	.set	pop
-	.endm
-
-	.macro	f_fill64 dst, offset, val
-	LONG_S	\val, (\offset +  0 * LONGSIZE)(\dst)
-	LONG_S	\val, (\offset +  1 * LONGSIZE)(\dst)
-	LONG_S	\val, (\offset +  2 * LONGSIZE)(\dst)
-	LONG_S	\val, (\offset +  3 * LONGSIZE)(\dst)
-	LONG_S	\val, (\offset +  4 * LONGSIZE)(\dst)
-	LONG_S	\val, (\offset +  5 * LONGSIZE)(\dst)
-	LONG_S	\val, (\offset +  6 * LONGSIZE)(\dst)
-	LONG_S	\val, (\offset +  7 * LONGSIZE)(\dst)
-#if LONGSIZE == 4
-	LONG_S	\val, (\offset +  8 * LONGSIZE)(\dst)
-	LONG_S	\val, (\offset +  9 * LONGSIZE)(\dst)
-	LONG_S	\val, (\offset + 10 * LONGSIZE)(\dst)
-	LONG_S	\val, (\offset + 11 * LONGSIZE)(\dst)
-	LONG_S	\val, (\offset + 12 * LONGSIZE)(\dst)
-	LONG_S	\val, (\offset + 13 * LONGSIZE)(\dst)
-	LONG_S	\val, (\offset + 14 * LONGSIZE)(\dst)
-	LONG_S	\val, (\offset + 15 * LONGSIZE)(\dst)
-#endif
-	.endm
-
-/*
- * mips_init_icache(uint PRId, ulong icache_size, unchar icache_linesz)
- */
-LEAF(mips_init_icache)
-	blez		a1, 9f
-	mtc0		zero, CP0_TAGLO
-	/* clear tag to invalidate */
-	PTR_LI		t0, INDEX_BASE
-	PTR_ADDU	t1, t0, a1
-1:	cache_op	INDEX_STORE_TAG_I t0
-	PTR_ADDU	t0, a2
-	bne		t0, t1, 1b
-	/* fill once, so data field parity is correct */
-	PTR_LI		t0, INDEX_BASE
-2:	cache_op	FILL t0
-	PTR_ADDU	t0, a2
-	bne		t0, t1, 2b
-	/* invalidate again - prudent but not strictly neccessary */
-	PTR_LI		t0, INDEX_BASE
-1:	cache_op	INDEX_STORE_TAG_I t0
-	PTR_ADDU	t0, a2
-	bne		t0, t1, 1b
-9:	jr		ra
-	END(mips_init_icache)
-
-/*
- * mips_init_dcache(uint PRId, ulong dcache_size, unchar dcache_linesz)
- */
-LEAF(mips_init_dcache)
-	blez		a1, 9f
-	mtc0		zero, CP0_TAGLO
-	/* clear all tags */
-	PTR_LI		t0, INDEX_BASE
-	PTR_ADDU	t1, t0, a1
-1:	cache_op	INDEX_STORE_TAG_D t0
-	PTR_ADDU	t0, a2
-	bne		t0, t1, 1b
-	/* load from each line (in cached space) */
-	PTR_LI		t0, INDEX_BASE
-2:	LONG_L		zero, 0(t0)
-	PTR_ADDU	t0, a2
-	bne		t0, t1, 2b
-	/* clear all tags */
-	PTR_LI		t0, INDEX_BASE
-1:	cache_op	INDEX_STORE_TAG_D t0
-	PTR_ADDU	t0, a2
-	bne		t0, t1, 1b
-9:	jr		ra
-	END(mips_init_dcache)
-
-/*
- * mips_cache_reset - low level initialisation of the primary caches
- *
- * This routine initialises the primary caches to ensure that they have good
- * parity.  It must be called by the ROM before any cached locations are used
- * to prevent the possibility of data with bad parity being written to memory.
- *
- * To initialise the instruction cache it is essential that a source of data
- * with good parity is available. This routine will initialise an area of
- * memory starting at location zero to be used as a source of parity.
- *
- * RETURNS: N/A
- *
- */
-NESTED(mips_cache_reset, 0, ra)
-	move	RA, ra
-	li	t2, CONFIG_SYS_ICACHE_SIZE
-	li	t3, CONFIG_SYS_DCACHE_SIZE
-	li	t8, CONFIG_SYS_CACHELINE_SIZE
-
-	li	v0, MIPS_MAX_CACHE_SIZE
-
-	/*
-	 * Now clear that much memory starting from zero.
-	 */
-	PTR_LI		a0, CKSEG1
-	PTR_ADDU	a1, a0, v0
-2:	PTR_ADDIU	a0, 64
-	f_fill64	a0, -64, zero
-	bne		a0, a1, 2b
-
-	/*
-	 * The caches are probably in an indeterminate state,
-	 * so we force good parity into them by doing an
-	 * invalidate, load/fill, invalidate for each line.
-	 */
-
-	/*
-	 * Assume bottom of RAM will generate good parity for the cache.
-	 */
-
-	/*
-	 * Initialize the I-cache first,
-	 */
-	move	a1, t2
-	move	a2, t8
-	PTR_LA	v1, mips_init_icache
-	jalr	v1
-
-	/*
-	 * then initialize D-cache.
-	 */
-	move	a1, t3
-	move	a2, t8
-	PTR_LA	v1, mips_init_dcache
-	jalr	v1
-
-	jr	RA
-	END(mips_cache_reset)
-
-/*
- * dcache_status - get cache status
- *
- * RETURNS: 0 - cache disabled; 1 - cache enabled
- *
- */
-LEAF(dcache_status)
-	mfc0	t0, CP0_CONFIG
-	li	t1, CONF_CM_UNCACHED
-	andi	t0, t0, CONF_CM_CMASK
-	move	v0, zero
-	beq	t0, t1, 2f
-	li	v0, 1
-2:	jr	ra
-	END(dcache_status)
-
-/*
- * dcache_disable - disable cache
- *
- * RETURNS: N/A
- *
- */
-LEAF(dcache_disable)
-	mfc0	t0, CP0_CONFIG
-	li	t1, -8
-	and	t0, t0, t1
-	ori	t0, t0, CONF_CM_UNCACHED
-	mtc0	t0, CP0_CONFIG
-	jr	ra
-	END(dcache_disable)
-
-/*
- * dcache_enable - enable cache
- *
- * RETURNS: N/A
- *
- */
-LEAF(dcache_enable)
-	mfc0	t0, CP0_CONFIG
-	ori	t0, CONF_CM_CMASK
-	xori	t0, CONF_CM_CMASK
-	ori	t0, CONF_CM_CACHABLE_NONCOHERENT
-	mtc0	t0, CP0_CONFIG
-	jr	ra
-	END(dcache_enable)
diff --git a/arch/mips/lib/Makefile b/arch/mips/lib/Makefile
index d939ee6..ac536da 100644
--- a/arch/mips/lib/Makefile
+++ b/arch/mips/lib/Makefile
@@ -6,6 +6,7 @@
 #
 
 obj-y	+= cache.o
+obj-y	+= cache_init.o
 obj-y	+= io.o
 
 obj-$(CONFIG_CMD_BOOTM) += bootm.o
diff --git a/arch/mips/cpu/mips32/cache.S b/arch/mips/lib/cache_init.S
similarity index 95%
rename from arch/mips/cpu/mips32/cache.S
rename to arch/mips/lib/cache_init.S
index fb1d84b..6c02bf9 100644
--- a/arch/mips/cpu/mips32/cache.S
+++ b/arch/mips/lib/cache_init.S
@@ -18,7 +18,11 @@
 #define CONFIG_SYS_MIPS_CACHE_MODE CONF_CM_CACHABLE_NONCOHERENT
 #endif
 
-#define RA		t9
+#ifdef CONFIG_64BIT
+# define RA		ta3
+#else
+# define RA		t7
+#endif
 
 #define INDEX_BASE	CKSEG0
 
@@ -117,7 +121,7 @@
 #endif
 
 #ifdef CONFIG_SYS_CACHELINE_SIZE
-	li	t7, CONFIG_SYS_CACHELINE_SIZE
+	li	t9, CONFIG_SYS_CACHELINE_SIZE
 	li	t8, CONFIG_SYS_CACHELINE_SIZE
 #else
 	/* Detect I-cache line size. */
@@ -128,11 +132,11 @@
 	sllv	t8, t6, t8
 
 1:	/* Detect D-cache line size. */
-	srl	t7, t5, MIPS_CONF1_DL_SHIFT
-	andi	t7, t7, (MIPS_CONF1_DL >> MIPS_CONF1_DL_SHIFT)
-	beqz	t7, 1f
+	srl	t9, t5, MIPS_CONF1_DL_SHIFT
+	andi	t9, t9, (MIPS_CONF1_DL >> MIPS_CONF1_DL_SHIFT)
+	beqz	t9, 1f
 	li	t6, 2
-	sllv	t7, t6, t7
+	sllv	t9, t6, t9
 1:
 #endif
 
@@ -168,7 +172,7 @@
 	addi	t6, t6, 1
 	sllv	t4, t4, t6
 1:	/* At this point t4 == I-cache sets. */
-	mul	t3, t4, t7
+	mul	t3, t4, t9
 	srl	t6, t5, MIPS_CONF1_DA_SHIFT
 	andi	t6, t6, (MIPS_CONF1_DA >> MIPS_CONF1_DA_SHIFT)
 	addi	t6, t6, 1
@@ -219,7 +223,7 @@
 	 * then initialize D-cache.
 	 */
 	move	a1, t3
-	move	a2, t7
+	move	a2, t9
 	PTR_LA	v1, mips_init_dcache
 	jalr	v1