/* SPDX-License-Identifier: GPL-2.0+ */
/*
 * (C) Copyright 2007 Michal Simek
 * (C) Copyright 2004 Atmark Techno, Inc.
 *
 * Michal  SIMEK <monstr@monstr.eu>
 * Yasushi SHOJI <yashi@atmark-techno.com>
 */

#include <asm-offsets.h>
#include <config.h>

	.text
	.global _start
_start:
	mts	rmsr, r0	/* disable cache */
	mfs	r20, rpc
	addi	r20, r20, -4

	mts	rslr, r0
	mts	rshr, r20

#if defined(CONFIG_SPL_BUILD)
	addi	r1, r0, CONFIG_SPL_STACK_ADDR
#else
	add	r1, r0, r20
#endif

	addi	r1, r1, -4	/* Decrement SP to top of memory */

	/* Call board_init_f_alloc_reserve with the current stack pointer as
	 * parameter. */
	add	r5, r0, r1
	brlid	r15, board_init_f_alloc_reserve
	nop

	/* board_init_f_alloc_reserve returns a pointer to the allocated area
	 * in r3. Set the new stack pointer below this area. */
	add	r1, r0, r3
	mts	rshr, r1
	addi	r1, r1, -4

	/* Call board_init_f_init_reserve with the address returned by
	 * board_init_f_alloc_reserve as parameter. */
	add	r5, r0, r3
	brlid	r15, board_init_f_init_reserve
	nop

#if !defined(CONFIG_SPL_BUILD)
	/* Setup vectors with pre-relocation symbols */
	or	r5, r0, r0
	brlid	r15, __setup_exceptions
	nop
#endif

	/* Flush cache before enable cache */
	addik	r5, r0, 0
	addik	r6, r0, XILINX_DCACHE_BYTE_SIZE
	brlid	r15, flush_cache
	nop

	/* enable instruction and data cache */
	mfs	r12, rmsr
	ori	r12, r12, 0x1a0
	mts	rmsr, r12

clear_bss:
	/* clear BSS segments */
	addi	r5, r0, __bss_start
	addi	r4, r0, __bss_end
	cmp	r6, r5, r4
	beqi	r6, 3f
2:
	swi     r0, r5, 0 /* write zero to loc */
	addi    r5, r5, 4 /* increment to next loc */
	cmp     r6, r5, r4 /* check if we have reach the end */
	bnei    r6, 2b
3:	/* jumping to board_init */
#ifdef CONFIG_DEBUG_UART
	brlid	r15, debug_uart_init
	nop
#endif
#ifndef CONFIG_SPL_BUILD
	or	r5, r0, r0	/* flags - empty */
	bri	board_init_f
#else
	bri	board_init_r
#endif
1:	bri	1b

#ifndef CONFIG_SPL_BUILD
	.text
	.ent	__setup_exceptions
	.align	2
/*
 * Set up reset, interrupt, user exception and hardware exception vectors.
 *
 * Parameters:
 * r5 - relocation offset (zero when setting up vectors before
 *      relocation, and gd->reloc_off when setting up vectors after
 *      relocation)
 *    - the relocation offset is added to the _exception_handler,
 *      _interrupt_handler and _hw_exception_handler symbols to reflect the
 *      post-relocation memory addresses
 *
 * Reserve registers:
 * r10: Stores little/big endian offset for vectors
 * r2: Stores imm opcode
 * r3: Stores brai opcode
 * r4: Stores the vector base address
 */
__setup_exceptions:
	addik	r1, r1, -32
	swi	r2, r1, 4
	swi	r3, r1, 8
	swi	r4, r1, 12
	swi	r6, r1, 16
	swi	r7, r1, 20
	swi	r8, r1, 24
	swi	r10, r1, 28

	/* Find-out if u-boot is running on BIG/LITTLE endian platform
	 * There are some steps which is necessary to keep in mind:
	 * 1. Setup offset value to r6
	 * 2. Store word offset value to address 0x0
	 * 3. Load just byte from address 0x0
	 * 4a) LITTLE endian - r10 contains 0x2 because it is the smallest
	 *     value that's why is on address 0x0
	 * 4b) BIG endian - r10 contains 0x0 because 0x2 offset is on addr 0x3
	 */
	addik	r6, r0, 0x2 /* BIG/LITTLE endian offset */
	sw	r6, r1, r0
	lbu	r10, r1, r0

	/* add opcode instruction for 32bit jump - 2 instruction imm & brai */
	addi	r2, r0, 0xb0000000	/* hex b000 opcode imm */
	addi	r3, r0, 0xb8080000	/* hew b808 opcode brai */

	/* Store the vector base address in r4 */
	addi	r4, r0, CONFIG_XILINX_MICROBLAZE0_VECTOR_BASE_ADDR

	/* reset address */
	swi	r2, r4, 0x0	/* reset address - imm opcode */
	swi	r3, r4, 0x4	/* reset address - brai opcode */

	addik	r6, r0, _start
	/* Intentionally keep reset vector back to origin u-boot location */
	sw	r6, r1, r0
	lhu	r7, r1, r10
	rsubi	r8, r10, 0x2
	sh	r7, r4, r8
	rsubi	r8, r10, 0x6
	sh	r6, r4, r8

#if CONFIG_IS_ENABLED(XILINX_MICROBLAZE0_USR_EXCEP)
	/* user_vector_exception */
	swi	r2, r4, 0x8	/* user vector exception - imm opcode */
	swi	r3, r4, 0xC	/* user vector exception - brai opcode */

	addik	r6, r5, _exception_handler
	sw	r6, r1, r0
	/*
	 * BIG ENDIAN memory map for user exception
	 * 0x8: 0xB000XXXX
	 * 0xC: 0xB808XXXX
	 *
	 * then it is necessary to count address for storing the most significant
	 * 16bits from _exception_handler address and copy it to
	 * 0xa address. Big endian use offset in r10=0 that's why is it just
	 * 0xa address. The same is done for the least significant 16 bits
	 * for 0xe address.
	 *
	 * LITTLE ENDIAN memory map for user exception
	 * 0x8: 0xXXXX00B0
	 * 0xC: 0xXXXX08B8
	 *
	 * Offset is for little endian setup to 0x2. rsubi instruction decrease
	 * address value to ensure that points to proper place which is
	 * 0x8 for the most significant 16 bits and
	 * 0xC for the least significant 16 bits
	 */
	lhu	r7, r1, r10
	rsubi	r8, r10, 0xa
	sh	r7, r4, r8
	rsubi	r8, r10, 0xe
	sh	r6, r4, r8
#endif

	/* interrupt_handler */
	swi	r2, r4, 0x10	/* interrupt - imm opcode */
	swi	r3, r4, 0x14	/* interrupt - brai opcode */

	addik	r6, r5, _interrupt_handler
	sw	r6, r1, r0
	lhu	r7, r1, r10
	rsubi	r8, r10, 0x12
	sh	r7, r4, r8
	rsubi	r8, r10, 0x16
	sh	r6, r4, r8

	/* hardware exception */
	swi	r2, r4, 0x20	/* hardware exception - imm opcode */
	swi	r3, r4, 0x24	/* hardware exception - brai opcode */

	addik	r6, r5, _hw_exception_handler
	sw	r6, r1, r0
	lhu	r7, r1, r10
	rsubi	r8, r10, 0x22
	sh	r7, r4, r8
	rsubi	r8, r10, 0x26
	sh	r6, r4, r8

	lwi	r10, r1, 28
	lwi	r8, r1, 24
	lwi	r7, r1, 20
	lwi	r6, r1, 16
	lwi	r4, r1, 12
	lwi	r3, r1, 8
	lwi	r2, r1, 4
	addik	r1, r1, 32

	rtsd	r15, 8
	or	r0, r0, r0
	.end	__setup_exceptions

/*
 * Read 16bit little endian
 */
	.text
	.global	in16
	.ent	in16
	.align	2
in16:	lhu	r3, r0, r5
	bslli	r4, r3, 8
	bsrli	r3, r3, 8
	andi	r4, r4, 0xffff
	or	r3, r3, r4
	rtsd	r15, 8
	sext16	r3, r3
	.end	in16

/*
 * Write 16bit little endian
 * first parameter(r5) - address, second(r6) - short value
 */
	.text
	.global	out16
	.ent	out16
	.align	2
out16:	bslli	r3, r6, 8
	bsrli	r6, r6, 8
	andi	r3, r3, 0xffff
	or	r3, r3, r6
	sh	r3, r0, r5
	rtsd	r15, 8
	or	r0, r0, r0
	.end	out16

/*
 * Relocate u-boot
 */
	.text
	.global	relocate_code
	.ent	relocate_code
	.align	2
relocate_code:
	/*
	 * r5 - start_addr_sp
	 * r6 - new_gd
	 * r7 - reloc_addr
	 */
	addi	r1, r5, 0 /* Start to use new SP */
	mts	rshr, r1
	addi	r31, r6, 0 /* Start to use new GD */

	/* Relocate text and data - r12 temp value */
	addi	r21, r0, _start
	addi	r22, r0, _end /* Include BSS too */
	addi	r22, r22, -4

	rsub	r6, r21, r22
	or	r5, r0, r0
1:	lw	r12, r21, r5 /* Load u-boot data */
	sw	r12, r7, r5 /* Write zero to loc */
	cmp	r12, r5, r6 /* Check if we have reach the end */
	bneid	r12, 1b
	addi	r5, r5, 4 /* Increment to next loc - relocate code */

	/* R23 points to the base address. */
	rsub	r23, r21, r7 /* keep - this is already here gd->reloc_off */

	/* Setup vectors with post-relocation symbols */
	add	r5, r0, r23 /* load gd->reloc_off to r5 */
	brlid	r15, __setup_exceptions
	nop

	/* Check if GOT exist */
	addik	r21, r23, _got_start
	addik	r22, r23, _got_end
	cmpu	r12, r21, r22
	beqi	r12, 2f /* No GOT table - jump over */

	/* Skip last 3 entries plus 1 because of loop boundary below */
	addik	r22, r22, -0x10

        /* Relocate the GOT. */
3:	lw	r12, r21, r0 /* Load entry */
	addk	r12, r12, r23 /* Add reloc offset */
	sw	r12, r21, r0 /* Save entry back */

	cmpu	r12, r21, r22 /* Check if this cross boundary */
	bneid	r12, 3b
	addik	r21. r21, 4

	/* Flush caches to ensure consistency */
	addik	r5, r0, 0
	addik	r6, r0, XILINX_DCACHE_BYTE_SIZE
	brlid	r15, flush_cache
	nop

2:	addi	r5, r31, 0 /* gd is initialized in board_r.c */
	addi	r6, r0, _start
	addi	r12, r23, board_init_r
	bra	r12 /* Jump to relocated code */

	.end	relocate_code
#endif