cpu 8086
	org 0x7C00

stack:	equ 0x7C00
framsz: equ 0x300

	jmp init

	; Data on user stack while in kernel (5 words):
	; AX BP IP CS flags

	; Kernel stack:
	; [frame] SP SS
	;      BP-^
	; The start of the frame is available via BP.
	; The frame is followed by a long ptr to the user stack.

	; Main entry point
int20h:	xor ah, ah
int21h:	push bp
	push ax
	; Test if SS=0, if it is, we assume we are already on kernel stack
	mov ax, ss
	test ax, ax
	jz rentry

	; Save user stack
	mov ax, ss
	mov [cs:stack-2], ax
	mov [cs:stack-4], sp

	; Set up kernel stack
	xor ax, ax
	mov ss, ax
	mov sp, (stack-4)

	jmp scall

	; This is like the above, except that it doesnt switch stacks.
	; The 'user' stack data is after (below on stack) the kernel frame.
        ; [frame] SP SS AX BP IP CS flags
	;          \-\--^
rentry:	mov ax, sp
	push ss
	push ax

scall:	sub sp, framsz
	mov bp, sp

	call getax
	call jmptab

	; Restore user stack
	mov sp, [bp+framsz]
	mov ax, [bp+framsz+2]
	mov ss, ax

	pop ax
	pop bp
	iret

	; Lookup address of subfunction
	; Mesh up the stack so we return to subfunction
	; and subfunction later returns to scall
jmptab:	push ax
	push bx
	xor bx, bx
	add bl, ah
	add bl, ah
	mov bx, [cs:bx+sftab]
	mov [bp-4], bx
	pop bx
	ret

	; Get AX from the user stack
getax:	push ds
	push si
	lds si, [bp+framsz]
	lodsw
	pop ds
	pop si
	ret

	; No such subfunction - set carry and exit
sferr:	stc
	ret

	; Subfunction table
sftab:	dw sferr, sferr, sferr, sferr
	dw sferr, sferr, sferr, sferr
	dw sferr, sferr, sferr, sferr
	dw sferr, sferr, sferr, sferr
	; 10
	dw sferr, sferr, sferr, sferr
	dw sferr, sferr, sferr, sferr
	dw sferr, sferr, sferr, sferr
	dw sferr, sferr, sferr, sferr
	; 20
	dw sferr, sferr, sferr, sferr
	dw sferr, setint, sferr, sferr
	dw sferr, sferr, sferr, sferr
	dw sferr, sferr, sferr, sferr
	; 30
	dw sferr, sferr, sferr, sferr
	dw sferr, getint, sferr, sferr
	dw sferr, sferr, sferr, sferr
	dw sferr, sferr, sferr, sferr

	; IN	al	number
	; OUT	bx	al * 4
times4:	push ax
	xor ah, ah
	add al, al
	add al, al
	mov bx, ax
	pop ax
	ret

	; DOS 2+ - GET INTERRUPT VECTOR
	; IN	al	interrupt number
	; OUT	es:bx	current interrupt handler
getint:	push ds
	; DS=0
	xor bx, bx
	mov ds, bx
	; BX=AL*4
	call times4
	les bx, [bx]
	pop ds
	ret

	; DOS 25h: Set interrupt vector
	; IN	al	interrupt number
	;	ds:dx	entry point
setint:	push es
	push bx
	; ES=0
	xor bx, bx
	mov es, bx
	; BX=AL*4
	call times4
	mov [es:bx], dx
	mov ax, ds
	mov [es:bx+2], ds
	pop bx
	pop es
	ret

; ===== end of resident, begin of transient startup code

init:	xor ax, ax
	mov ds, ax
	mov es, ax
	mov ax, 0x1000
	mov ss, ax
	xor sp, sp

	mov al, 0x21
	mov dx, int21h
	call setint

	mov ax, 0x2520
	mov dx, int20h
	int 0x21

	mov ax, 0x3520
	int 0x21

	mov ah, 0x3f
	int 3
	int 0x21
	int 3

main:	push cs
	pop ds
	mov si, msg
	mov ah, 0x0e
	xor bx, bx
loop:	lodsb
	test al, al
	jz end
	int 0x10
	jmp loop
end:	hlt
	jmp end

msg:	db "rdos kernel stub", 0x0A, 0x0D, 0