Getting a 32bit symbol into a 16bit .word directive

Getting a 32bit symbol into a 16bit .word directive

[MurF]

Chaps,

 I've been using GCC with good results. And I'm slowly finding my way around 
the Protected mode of the IA-32 architechture. I've just started looking into 
interurpts, and want to write my own IDT for a i486. Problem is I can't get 
as to put a 32-bit symbol into a .word directive without getting an error.


 I want to do something like this ....

(A bit of code)
.
.
.
int0:
	mov $0x0,%eax
	mov $0x80,%edx
	out  %al,%dl
loop0:
	jmp loop0
.
.
(A bit of code)
.
.
.globl EXT(idtptr)
EXT(idtptr):
	.word	20*8-1
	.long	idt		
idt:
	//  Offset(15..0) , Segment Selector
	//  Reserved (4..0), '0'(7..5) '0D110'(12..8), DPL(14..13), P(15)
	//  offset(31..16)


	.word	(int0), 0x0010			// Int 0 - Divide Error
	.byte   0x00, 0x0e
	.word	0x0000

.
.
.
.
.

 But when i compile this i get the error 'relocation truncated to fit: 
R_386_16 text' and the rest of the compilation fails. This is due to the line 
'.word (int0),0x0010)'

 If I remove the (int0) and replce with a constant everything works.

 Any ideas?

Re: Getting a 32bit symbol into a 16bit .word directive

[Alan Modra]

On Fri, Jul 18, 2003 at 03:38:01PM +0100, MurF wrote:
> Problem is I can't get 
> as to put a 32-bit symbol into a .word directive without getting an error.
[snip]
> '.word (int0),0x0010)'

You need to link your executable at an address such that "int0" resolves
to a valid 16 bit address.

-- 
Alan Modra
IBM OzLabs - Linux Technology Centre