/***************************************************************************
 *             __________               __   ___.
 *   Open      \______   \ ____   ____ |  | _\_ |__   _______  ___
 *   Source     |       _//  _ \_/ ___\|  |/ /| __ \ /  _ \  \/  /
 *   Jukebox    |    |   (  <_> )  \___|    < | \_\ (  <_> > <  <
 *   Firmware   |____|_  /\____/ \___  >__|_ \|___  /\____/__/\_ \
 *                     \/            \/     \/    \/            \/
 * $Id$
 *
 * Copyright (C) 2002 by Linus Nielsen Feltzing
 *
 * All files in this archive are subject to the GNU General Public License.
 * See the file COPYING in the source tree root for full license agreement.
 *
 * This software is distributed on an "AS IS" basis, WITHOUT WARRANTY OF ANY
 * KIND, either express or implied.
 *
 ****************************************************************************/
        #include "config.h"
        #include "cpu.h"

//#define BOOTLOADER
        
	.section .init.text
	.global	start
start:
#if defined(ARCHOS_GMINI120)
	;; disable all interrupts
 	clrsr fe
	clrsr ie
	clrsr te
	ld a14, #0x3F0000

	ld r5, 0xA5
	ldb @[a14 + 6], r5	; disable watchdog
	
	ld a11, #(_datacopy)	; where the data section is in the flash
	ld a8, #(_datastart)	; destination

	;; copy data section from flash to ram.
	ld a9, #_datasize
	ld r6, e9
	cmp eq, r6, #0
	brf .data_copy_loop
	cmp eq, r9, #0
	brt .data_copy_end
.data_copy_loop:	
	ldc r2, @a11
	ldw @[a8 + 0], r2
	add a11, #0x2
	add a8, #0x2
	sub r9, #0x2
	sbc r6, #0
	cmp ugt, r6, #0
	brt .data_copy_loop
	cmp ugt, r9, #0
	brt .data_copy_loop
.data_copy_end:

	;; zero out bss
	ld r2, #0
	ld a8, #(_bssstart)	; destination
	ld a9, #_bsssize	
	ld r6, e9

	cmp eq, r6, #0
	brf .bss_init_loop
	cmp eq, r9, #0
	brt .bss_init_end
.bss_init_loop:	
	ldw @[a8 + 0], r2
	add a8, #0x2
	sub r9, #0x2
	sbc r6, #0
	cmp ugt, r6, #0
	brt .bss_init_loop
	cmp ugt, r9, #0
	brt .bss_init_loop
.bss_init_end:

	;; set stack pointer
	ld a15, _stackend

	;; go!
	jsr _main
	
	;; soft reset
	ld a10, #0
	ldc r10, @a10
	jmp a10

	
	.section .vectors, "ax"
irq_handler:

	push r0, r1
	push r2, r3
	push r4, r5
	push r6, r7
	push a8, a9
	push a10, a11
	push a12, a13
	push a14
	ld a13, #0x3f0000
	ldb r0, @[a13 + 0x26]
	add r0, r0
	ld a10, #_interrupt_vector
	ldw a13, @[a10 + r0]
	jsr a13
	pop a14
	pop a13, a12
	pop a11, a10
	pop a9, a8
	pop r7, r6
	pop r5, r4
	pop r3, r2
	pop r1, r0
	ret_irq

#elif defined(IRIVER_H100)

        /* Platform:     iRiver H1xx */
        move.w #0x2700,%sr

        move.l #MBAR+1,%d0
        movec.l %d0,%mbar
        
        move.l #MBAR2+1,%d0
        movec.l %d0,%mbar2

        lea MBAR,%a0
        lea MBAR2,%a1
        
        /* 64K DMA-capable SRAM at 0x10000000
           DMA is enabled and has priority in both banks
           All types of accesses are allowed
           (We might want to restrict that to save power) */
        move.l #0x1000e001,%d0
        movec.l %d0,%rambar1

        /* 32K Non-DMA SRAM at 0x10010000
           All types of accesses are allowed
           (We might want to restrict that to save power) */
        move.l #0x10010001,%d0
        movec.l %d0,%rambar0

        /* Chip select 0 - Flash ROM */
        move.l #0x00000000,%d0 /* CSAR0 - Base = 0x00000000 */
        move.l %d0,(0x080,%a0)
        move.l #0x001f0101,%d0 /* CSMR0 - 2M, All access, write protect */
        move.l %d0,(0x084,%a0)
        move.l #0x00000d80,%d0 /* CSCR0 - 3 wait states, 16 bits, no bursts */
        move.l %d0,(0x088,%a0)
        
        /* Chip select 1 - LCD controller */
        move.l #0xf0000000,%d0 /* CSAR1 - Base = 0xf0000000 */
        move.l %d0,(0x08c,%a0)
        move.l #0x00000075,%d0 /* CSMR1 - 64K, Only data access */
        move.l %d0,(0x090,%a0)
        move.l #0x00002180,%d0 /* CSCR1 - 8 wait states, 16 bits, no bursts */
        move.l %d0,(0x094,%a0)
        
        /* Chip select 2 - ATA controller */
        move.l #0x20000000,%d0 /* CSAR2 - Base = 0x20000000 */
        move.l %d0,(0x098,%a0)
        move.l #0x000f0001,%d0 /* CSMR2 - 64K, Only data access */
        move.l %d0,(0x09c,%a0)
        move.l #0x00000080,%d0 /* CSCR2 - no wait states, 16 bits, no bursts */
        move.l %d0,(0x0a0,%a0) /* NOTE:  I'm not sure about the wait states.
                                  We have to be careful with the access times,
                                  since IORDY isn't connected to the HDD. */
        

#ifdef BOOTLOADER
        /* Set up the DRAM controller. The refresh is based on the 11.2896MHz
           clock (5.6448MHz bus frequency). We haven't yet started the PLL */
        move.l #0x80050000,%d0
        move.l %d0,(0x100,%a0) /* DCR - Synchronous, 80 cycle refresh */

        /* Note: we place the SDRAM on an 0x1000000 (16M) offset because
           the 5249 BGA chip has a fault which disables the use of A24. The
           suggested workaround by FreeScale is to offset the base address by
           half the DRAM size and increase the mask to the double.
           In our case this means that we set the base address 16M ahead and
           use a 64M mask.
        */
        move.l #0x31002520,%d0
        move.l %d0,(0x108,%a0) /* DACR0 - Base 0x31000000, Banks on 23 and up,
                                  CAS latency 1, No refresh yet */
        move.l #0x03fc0001,%d0 /* Size: 64M because of workaround above */
        move.l %d0,(0x10c,%a0) /* DMR0 - 32Mb */

        /* Precharge */
        move.l #0x31002528,%d0
        move.l %d0,(0x108,%a0) /* DACR0[IP] = 1, next access will issue a
                                  Precharge command */
        move.l #0xabcd1234,%d0
        move.l %d0,0x31000000  /* Issue precharge command */

        /* Let it refresh */
        move.l #1000,%d0
.delayloop:
        subq.l #1,%d0
        bne    .delayloop
        
        /* Refresh */
        move.l #0x3100a520,%d0
        move.l %d0,(0x108,%a0) /* Enable refresh */

        /* Mode Register init */
        move.l #0x3100a560,%d0 /* DACR0[IMRS] = 1, next access will set the
                                  Mode Register */
        move.l %d0,(0x108,%a0)

        move.l #0xabcd1234,%d0
        move.l %d0,0x31000800  /* A12=1 means CASL=1 (a0 is not connected) */

        move.l #0x3100a520,%d0 /* Back to normal, the DRAM is now ready */
        move.l %d0,(0x108,%a0)
#endif
#if 1
        lea    _iramcopy,%a2
        lea    _iramstart,%a3
        lea    _iramend,%a4        
.iramloop:
        cmp.l  %a3,%a4
        beq    .iramloopend
        move.w (%a2)+,(%a3)+
        bra    .iramloop
.iramloopend:

#endif
        lea    _edata,%a2
        lea    _end,%a4
        clr.l  %d0
.edataloop:
        cmp.l  %a2,%a4
        beq    .edataloopend
        move.w %d0,(%a2)+
        bra    .edataloop
.edataloopend:

        lea    _datacopy,%a2
        lea    _datastart,%a3
        lea    _dataend,%a4        
.dataloop:
        cmp.l  %a3,%a4
        beq    .dataloopend
        move.w (%a2)+,(%a3)+
        bra    .dataloop
.dataloopend:

        lea    stackend,%sp
        jsr    main
.hoo:
        bra    .hoo

	.section .resetvectors
vectors:
	.long _stackend
	.long start
#else
        /* Platform:     Archos Jukebox
	 * We begin with some tricks. If we have built our code to be loaded
	 * via the standalone GDB stub, we will have out VBR at some other
	 * location than 0x9000000. We must copy the trap vectors for the
	 * GDB stub to our vector table.
	 * If, on the other hand, we are running standalone we will have
	 * the VBR at 0x9000000, and the copy will not do any harm.
	 */
	mov.l	vbr_k,r1	
	mov.l	orig_vbr_k,r2
	
	/* Move the invalid instruction vector (4) */
	mov	#4,r0
	shll2	r0
	mov.l	@(r0,r2),r3
	mov.l	r3,@(r0,r1)
	
	/* Move the invalid slot vector (6) */
	mov	#6,r0
	shll2	r0
	mov.l	@(r0,r2),r3
	mov.l	r3,@(r0,r1)
	
	/* Move the bus error vector (9) */
	mov	#9,r0
	shll2	r0
	mov.l	@(r0,r2),r3
	mov.l	r3,@(r0,r1)
	
	/* Move the DMA bus error vector (10) */
	mov	#10,r0
	shll2	r0
	mov.l	@(r0,r2),r3
	mov.l	r3,@(r0,r1)
	
	/* Move the NMI vector as well (11) */
	mov	#11,r0
	shll2	r0
	mov.l	@(r0,r2),r3
	mov.l	r3,@(r0,r1)
	
	/* Move the breakpoint trap vector (32) */
	mov	#32,r0
	shll2	r0
	mov.l	@(r0,r2),r3
	mov.l	r3,@(r0,r1)
	
	/* Move the IO trap vector (33) */
	mov	#33,r0
	shll2	r0
	mov.l	@(r0,r2),r3
	mov.l	r3,@(r0,r1)
	
	/* Move the serial Rx interrupt vector (105) */
	mov	#105,r0
	shll2	r0
	mov.l	@(r0,r2),r3
	mov.l	r3,@(r0,r1)

	/* Move the single step trap vector (127) */
	mov	#127,r0
	shll2	r0
	mov.l	@(r0,r2),r3
	mov.l	r3,@(r0,r1)
	
	ldc	r1,vbr

	/* Now let's get on with the normal business */
	mov.l	stack_k,r15

	/* zero out bss */
	mov.l	edata_k,r0
	mov.l	end_k,r1
	mov	#0,r2
start_l:
	mov.l	r2,@r0
	add	#4,r0
	cmp/ge	r1,r0
	bf	start_l
	nop

	/* copy the .iram section */
	mov.l	iramcopy_k,r0
	mov.l	iram_k,r1
	mov.l	iramend_k,r2
copy_l:
	mov.l	@r0,r3
	mov.l	r3,@r1
	add	#4,r0
	add	#4,r1
	cmp/ge	r2,r1
	bf	copy_l
	nop

	/* copy the .data section, for rombased execution */
	mov.l	datacopy_k,r0
	mov.l	data_k,r1
	mov.l	dataend_k,r2
        /* Don't copy if src and dest are equal */
        cmp/eq  r0,r1
        bt      nodatacopy
copy_l2:
	mov.l	@r0,r3
	mov.l	r3,@r1
	add	#4,r0
	add	#4,r1
	cmp/ge	r2,r1
	bf	copy_l2
	nop
nodatacopy:
	/* Munge the main thread stack */
	mov.l	stack_k,r2
	mov.l	deadbeef_k,r0
	mov.l	stackbegin_k,r1
munge_loop:
	mov.l	r0,@r1
	add	#4,r1
	cmp/ge	r2,r1
	bf	munge_loop
	nop

	mov     #0,r0
	ldc     r0,gbr
	
	! call the mainline	
	mov.l	main_k,r0
	jsr	@r0
	nop
.hoo:
    bra .hoo

	.align 2
stack_k:
	.long	_stackend	
stackbegin_k:
	.long	_stackbegin
deadbeef_k:
	.long	0xdeadbeef
edata_k:
	.long	_edata
end_k:
	.long	_end
iramcopy_k:
	.long	_iramcopy
iram_k:
	.long	_iramstart
iramend_k:
	.long	_iramend
datacopy_k:
	.long	_datacopy
data_k:
	.long	_datastart
dataend_k:
	.long	_dataend
main_k:
	.long	_main
vbr_k:
	.long	vectors
orig_vbr_k:
	.long 0x9000000

	.section .resetvectors
vectors:
	.long start
	.long _stackend
	.long start
	.long _stackend
#endif