/*
 *  Startup Code for MIPS32 CPU-core
 *
 *  Copyright (c) 2003	Wolfgang Denk <wd@denx.de>
 *
 * SPDX-License-Identifier:	GPL-2.0+
 */

#include <asm-offsets.h>
#include <config.h>
#include <asm/regdef.h>
#include <asm/mipsregs.h>

#ifndef CONFIG_SYS_MIPS_CACHE_MODE
#define CONFIG_SYS_MIPS_CACHE_MODE CONF_CM_CACHABLE_NONCOHERENT
#endif

	/*
	 * For the moment disable interrupts, mark the kernel mode and
	 * set ST0_KX so that the CPU does not spit fire when using
	 * 64-bit addresses.
	 */
	.macro	setup_c0_status set clr
	.set	push
	mfc0	t0, CP0_STATUS
	or	t0, ST0_CU0 | \set | 0x1f | \clr
	xor	t0, 0x1f | \clr
	mtc0	t0, CP0_STATUS
	.set	noreorder
	sll	zero, 3				# ehb
	.set	pop
	.endm

	.set noreorder

	.globl _start
	.text
_start:
	/* U-boot entry point */
	b	reset
	 nop

	.org 0x10
#ifdef CONFIG_SYS_XWAY_EBU_BOOTCFG
	/*
	 * Almost all Lantiq XWAY SoC devices have an external bus unit (EBU) to
	 * access external NOR flashes. If the board boots from NOR flash the
	 * internal BootROM does a blind read at address 0xB0000010 to read the
	 * initial configuration for that EBU in order to access the flash
	 * device with correct parameters. This config option is board-specific.
	 */
	.word CONFIG_SYS_XWAY_EBU_BOOTCFG
	.word 0x0
#endif

	.org 0x200
	/* TLB refill, 32 bit task */
1:	b	1b
	 nop

	.org 0x280
	/* XTLB refill, 64 bit task */
1:	b	1b
	 nop

	.org 0x300
	/* Cache error exception */
1:	b	1b
	 nop

	.org 0x380
	/* General exception */
1:	b	1b
	 nop

	.org 0x400
	/* Catch interrupt exceptions */
1:	b	1b
	 nop

	.org 0x480
	/* EJTAG debug exception */
1:	b	1b
	 nop

	.align 4
reset:

	/* Clear watch registers */
	mtc0	zero, CP0_WATCHLO
	mtc0	zero, CP0_WATCHHI

	/* WP(Watch Pending), SW0/1 should be cleared */
	mtc0	zero, CP0_CAUSE

	setup_c0_status 0 0

	/* Init Timer */
	mtc0	zero, CP0_COUNT
	mtc0	zero, CP0_COMPARE

#ifndef CONFIG_SKIP_LOWLEVEL_INIT
	/* CONFIG0 register */
	li	t0, CONF_CM_UNCACHED
	mtc0	t0, CP0_CONFIG
#endif

	/* Initialize $gp */
	bal	1f
	 nop
	.word	_gp
1:
	lw	gp, 0(ra)

#ifndef CONFIG_SKIP_LOWLEVEL_INIT
	/* Initialize any external memory */
	la	t9, lowlevel_init
	jalr	t9
	 nop

	/* Initialize caches... */
	la	t9, mips_cache_reset
	jalr	t9
	 nop

	/* ... and enable them */
	li	t0, CONFIG_SYS_MIPS_CACHE_MODE
	mtc0	t0, CP0_CONFIG
#endif

	/* Set up temporary stack */
	li	sp, CONFIG_SYS_SDRAM_BASE + CONFIG_SYS_INIT_SP_OFFSET

	la	t9, board_init_f
	jr	t9
	 nop

/*
 * void relocate_code (addr_sp, gd, addr_moni)
 *
 * This "function" does not return, instead it continues in RAM
 * after relocating the monitor code.
 *
 * a0 = addr_sp
 * a1 = gd
 * a2 = destination address
 */
	.globl	relocate_code
	.ent	relocate_code
relocate_code:
	move	sp, a0			# set new stack pointer

	move	s0, a1			# save gd in s0
	move	s2, a2			# save destination address in s2

	li	t0, CONFIG_SYS_MONITOR_BASE
	sub	s1, s2, t0		# s1 <-- relocation offset

	la	t3, in_ram
	lw	t2, -12(t3)		# t2 <-- __image_copy_end
	move	t1, a2

	add	gp, s1			# adjust gp

	/*
	 * t0 = source address
	 * t1 = target address
	 * t2 = source end address
	 */
1:
	lw	t3, 0(t0)
	sw	t3, 0(t1)
	addu	t0, 4
	blt	t0, t2, 1b
	 addu	t1, 4

	/* If caches were enabled, we would have to flush them here. */
	sub	a1, t1, s2		# a1 <-- size
	la	t9, flush_cache
	jalr	t9
	 move	a0, s2			# a0 <-- destination address

	/* Jump to where we've relocated ourselves */
	addi	t0, s2, in_ram - _start
	jr	t0
	 nop

	.word	__rel_dyn_end
	.word	__rel_dyn_start
	.word	__image_copy_end
	.word	_GLOBAL_OFFSET_TABLE_
	.word	num_got_entries

in_ram:
	/*
	 * Now we want to update GOT.
	 *
	 * GOT[0] is reserved. GOT[1] is also reserved for the dynamic object
	 * generated by GNU ld. Skip these reserved entries from relocation.
	 */
	lw	t3, -4(t0)		# t3 <-- num_got_entries
	lw	t4, -8(t0)		# t4 <-- _GLOBAL_OFFSET_TABLE_
	add	t4, s1			# t4 now holds relocated _G_O_T_
	addi	t4, t4, 8		# skipping first two entries
	li	t2, 2
1:
	lw	t1, 0(t4)
	beqz	t1, 2f
	 add	t1, s1
	sw	t1, 0(t4)
2:
	addi	t2, 1
	blt	t2, t3, 1b
	 addi	t4, 4

	/* Update dynamic relocations */
	lw	t1, -16(t0)		# t1 <-- __rel_dyn_start
	lw	t2, -20(t0)		# t2 <-- __rel_dyn_end

	b	2f			# skip first reserved entry
	 addi	t1, 8

1:
	lw	t3, -4(t1)		# t3 <-- relocation info

	sub	t3, 3
	bnez	t3, 2f			# skip non R_MIPS_REL32 entries
	 nop

	lw	t3, -8(t1)		# t3 <-- location to fix up in FLASH

	lw	t4, 0(t3)		# t4 <-- original pointer
	add	t4, s1			# t4 <-- adjusted pointer

	add	t3, s1			# t3 <-- location to fix up in RAM
	sw	t4, 0(t3)

2:
	blt	t1, t2, 1b
	 addi	t1, 8			# each rel.dyn entry is 8 bytes

	/*
	 * Clear BSS
	 *
	 * GOT is now relocated. Thus __bss_start and __bss_end can be
	 * accessed directly via $gp.
	 */
	la	t1, __bss_start		# t1 <-- __bss_start
	la	t2, __bss_end		# t2 <-- __bss_end

1:
	sw	zero, 0(t1)
	blt	t1, t2, 1b
	 addi	t1, 4

	move	a0, s0			# a0 <-- gd
	la	t9, board_init_r
	jr	t9
	 move	a1, s2

	.end	relocate_code