/*
 * void __asm_flush_dcache_level(level)
 *
 * clean and invalidate one level cache.
 *
 * x0: cache level
 * x1: 0 flush & invalidate, 1 invalidate only
 * x2~x9: clobbered
 */
.global __asm_flush_dcache_level
__asm_flush_dcache_level:
	lsl	x12, x0, #1
	msr	csselr_el1, x12		/* select cache level */
	isb				/* sync change of cssidr_el1 */
	mrs	x6, ccsidr_el1		/* read the new cssidr_el1 */
	and	x2, x6, #7		/* x2 <- log2(cache line size)-4 */
	add	x2, x2, #4		/* x2 <- log2(cache line size) */
	mov	x3, #0x3ff
	and	x3, x3, x6, lsr #3	/* x3 <- max number of #ways */
	clz	w5, w3			/* bit position of #ways */
	mov	x4, #0x7fff
	and	x4, x4, x6, lsr #13	/* x4 <- max number of #sets */
	/* x12 <- cache level << 1 */
	/* x2 <- line length offset */
	/* x3 <- number of cache ways - 1 */
	/* x4 <- number of cache sets - 1 */
	/* x5 <- bit position of #ways */

loop_set:
	mov	x6, x3			/* x6 <- working copy of #ways */
loop_way:
	lsl	x7, x6, x5
	orr	x9, x12, x7		/* map way and level to cisw value */
	lsl	x7, x4, x2
	orr	x9, x9, x7		/* map set number to cisw value */
	tbz	w1, #0, 1f
	dc	isw, x9
	b	2f
1:	dc	cisw, x9		/* clean & invalidate by set/way */
2:	subs	x6, x6, #1		/* decrement the way */
	b.ge	loop_way
	subs	x4, x4, #1		/* decrement the set */
	b.ge	loop_set

	ret

/*
 * void flush_dcache_all(int invalidate_only)
 *
 * x0: 0 flush & invalidate, 1 invalidate only
 *
 * clean and invalidate all data cache by SET/WAY.
 */
.global __asm_dcache_all
__asm_dcache_all:
	mov	x1, x0
	dsb	sy
	mrs	x10, clidr_el1		/* read clidr_el1 */
	lsr	x11, x10, #24
	and	x11, x11, #0x7		/* x11 <- loc */
	cbz	x11, finished		/* if loc is 0, exit */
	mov	x15, x30
	mov	x0, #0			/* start flush at cache level 0 */
	/* x0  <- cache level */
	/* x10 <- clidr_el1 */
	/* x11 <- loc */
	/* x15 <- return address */

loop_level:
	lsl	x12, x0, #1
	add	x12, x12, x0		/* x0 <- tripled cache level */
	lsr	x12, x10, x12
	and	x12, x12, #7		/* x12 <- cache type */
	cmp	x12, #2
	b.lt	skip			/* skip if no cache or icache */
	bl	__asm_flush_dcache_level	/* x1 = 0 flush, 1 invalidate */
skip:
	add	x0, x0, #1		/* increment cache level */
	cmp	x11, x0
	b.gt	loop_level

	mov	x0, #0
	msr	csselr_el1, x0		/* resotre csselr_el1 */
	dsb	sy
	isb
	mov	x30, x15

finished:
	ret

.global flush_dcache_all
flush_dcache_all:
	mov	x16, x30
	mov	x0, #0
	bl	__asm_dcache_all
	mov	x30, x16
	ret

.global invalidate_dcache_all
invalidate_dcache_all:
	mov	x16, x30
	mov	x0, #0xffff
	bl	__asm_dcache_all
	mov	x30, x16
	ret

/*
 * void flush_dcache_range(start, end)
 *
 * clean & invalidate data cache in the range
 *
 * x0: start address
 * x1: end address
 */
.global flush_dcache_range
flush_dcache_range:
	mrs	x3, ctr_el0
	lsr	x3, x3, #16
	and	x3, x3, #0xf
	mov	x2, #4
	lsl	x2, x2, x3		/* cache line size */

	/* x2 <- minimal cache line size in cache system */
	sub	x3, x2, #1
	bic	x0, x0, x3
1:	dc	civac, x0	/* clean & invalidate data or unified cache */
	add	x0, x0, x2
	cmp	x0, x1
	b.lo	1b
	dsb	sy
	ret

/*
 * void invalidate_icache_all(void)
 *
 * invalidate all icache entries.
 */
.global invalidate_icache_all
invalidate_icache_all:
	ic	ialluis
	isb	sy
	ret

/*
 * void invalidate_tlb_all(void)
 *
 * invalidate all tlb entries.
 */
.global invalidate_tlb_all
invalidate_tlb_all:
	tlbi	alle2
	dsb	sy
	isb
	ret