diff --git a/src/boot.s b/src/boot.s
index ace3604..9ee92cc 100644
--- a/src/boot.s
+++ b/src/boot.s
@@ -13,7 +13,7 @@ mov fs, ax
 
 xor cx, cx
 mov ds, cx
-mov word [0x21*4], 2 ; see kjmp in kernel.s
+mov word [0x21*4], 3 ; see kjmp in kernel.s
 mov word [0x21*4+2], ax
 
 mov ax, 0x200 | NUMSEG
diff --git a/src/int21/0.s b/src/int21/0.s
index ce2c3e1..3aca334 100644
--- a/src/int21/0.s
+++ b/src/int21/0.s
@@ -22,6 +22,10 @@ wrout:
 	; TODO handle ^C and stuff
 	; TODO handle stdout properly
 	push bx
+	cmp dl, 9
+	jnz .print
+	mov dl, ' ' ; "tabs are expanded to blanks"
+.print:
 	mov al, dl
 	mov bl, 0x70
 	call print_character
diff --git a/src/int21/int21.s b/src/int21/int21.s
index 636b405..7867ecb 100644
--- a/src/int21/int21.s
+++ b/src/int21/int21.s
@@ -1,22 +1,27 @@
+; TODO: change to use a stack frame instead of global variables for reentrancy
 int21:
-	push ds
-	push cs
-	pop ds
+	push ax
+	mov ax, ds
+	mov cs:[callerds], ax
+	mov ax, cs
+	mov ds, ax
+	pop ax
 
-	; this probably works
-	mov word [tmp], bx ; cannot use stack for this
+	mov word [tmp], bx
 	movzx bx, ah
 	shl bx, 1
 	add bx, fn
 	mov bx, word [bx]
-	push end21 ; the proper return address
-	; generated return address. should avoid any issues
-	; with things like prefetch or instruction caches
+	push end21
 	push bx
+
 	mov bx, word [tmp]
 	ret
-
-end21:	pop ds
+end21:
+	push ax
+	mov ax, [callerds]
+	mov ds, ax
+	pop ax
 	iret
 
 nul:	stc
@@ -42,6 +47,7 @@ fn:	dw	nul,	rdin_echo,	wrout,	nul,	nul,	nul,	nul,	nul,	nul,	nul,	nul,	nul,	nul,
 	dw	nul,	nul,	nul,	nul,	nul,	nul,	nul,	nul,	nul,	nul,	nul,	nul,	nul,	nul,	nul,	nul	; Ex
 	dw	nul,	nul,	nul,	nul,	nul,	nul,	nul,	nul,	nul,	nul,	nul,	nul,	nul,	nul,	nul,	nul	; Fx
 
+callerds: dw 0
 tmp: dw 0
 retcode: dw 0
 verify: db 0
diff --git a/src/kernel.s b/src/kernel.s
index 448bbc1..b49eae8 100644
--- a/src/kernel.s
+++ b/src/kernel.s
@@ -2,7 +2,7 @@
 
 [map symbols kernel.map]
 
-kjmp: ; MUST be a short jump due to loader config
+kjmp: ; MUST be a jmp due to loader config
 	jmp kernel_entry
 %include "int21/int21.s"
 
@@ -10,6 +10,11 @@ kernel_entry:
 	mov si, hello_string
 	mov bl, 0x70
 	call print_string
+
+	mov ah, 0x2
+	mov dl, '!'
+	int 0x21
+
 	jmp $
 
 hello_string: db "hello world!", 13, 10, '$'
diff --git a/src/vga.s b/src/vga.s
index 46389e3..ddd7d47 100644
--- a/src/vga.s
+++ b/src/vga.s
@@ -52,3 +52,32 @@ print_string:
 	pop es
 	popa
 	ret
+
+; print a u16 to the screen in hex
+; inputs:
+;	BX: value to print
+; outputs:
+;	none
+print_hex:
+	push ax
+	push bx
+	push cx
+	mov cx, 4
+.loop:
+	mov al, bh
+	shr al, 4
+	cmp al, 9
+	jng .num
+	add al, 'A' - 10
+	jmp .eloop
+.num:
+	add al, '0'
+.eloop:
+	call print_character
+	shl bx, 4
+	dec cx
+	jnz .loop
+	pop cx
+	pop bx
+	pop ax
+	ret