in the name of zero

lo! the road to elf sorcery is fraught with difficulties and nasty “gotchas”. i don’t know if i’m on the right track, so i’m posting my uncompleted work. and if ever a real wizard passes by, perhaps he could fill me in on some parts i did wrong.

BITS 32
	org	0x08048000
	
; ----------
; elf header
; ----------
ehdr:
	db	0x7f, "ELF", 1, 1, 1
times 9	db	0x0
	dw	2		; e_type
	dw	3		; e_machine
	dd	1		; e_version
	dd	_start		; e_entry
	dd	phdr - $$	; e_phoff
	dd	0		; e_shoff
	dd	0		; e_flags
	dw	ehdrsize	; e_ehsize
	dw	phdrsize	; e_phentsize
	dw	2		; e_phnum
	dw	0		; 3_shentsize
	dw	0		; e_shnum
	dw	0		; e_shstrndx
ehdrsize equ	$ - ehdr
	
; program header macros
%define pt_load	1
%define	p_align	0xf
%define	pf_rx	5
%define	pf_rw	6
	
; --------------------
; program header table
; ---------------------
phdr:
	dd	pt_load		; p_type
	dd	0		; p_offset
	dd	$$		; p_vaddr
	dd	$$		; p_paddr
	dd	s1		; p_filesize
	dd	s1		; p_memsize
	dd	pf_rx		; p_flags
	dd	p_align		; p_align
phdrsize equ	$-phdr
	
phdr1:
	dd	pt_load		; p_type
	dd	dsegoff		; p_offset
	dd	p_vadd1+psize	; p_vaddr
	dd	p_vadd1+psize	; p_paddr
	dd	dseglen		; p_filesize
	dd	dseglen		; p_memsize
	dd	pf_rw		; p_flags
	dd	p_align		; p_align
	
; ----------------------------
; start of segment definitions
; ----------------------------
_start:
	; this will eventually hold sys_exit(0)
	
s1	equ	$-$$
	
dataseg:
	; .data is zero bytes big
	; linux 2.6.12+? requires explicit .data segment declaration
d1	equ	$-dataseg
	
; this part is very  ugly. 
; i don't even know if this is right!!!
; align dataseg p_offset to an 8 byte boundary
%assign liner	(dataseg-$$)
dsegoff	equ	(liner&-8)+8
p_vadd1	equ	$$+(dsegoff)
dseglen	equ	$-dataseg
	
%assign multiplier	(s1/0x1000)+1
assign psize		0x1000*multiplier
	
; --------------------
; section header table
; --------------------

things look good in objdump so far.. i think.

objdump -x elfhead
elfhead:     file format elf32-i386
elfhead
architecture: i386, flags 0x00000102:
EXEC_P, D_PAGED
start address 0x08048074
	
Program Header:
    LOAD off    0x00000000 vaddr 0x08048000 paddr 0x08048000 align 2**12
         filesz 0x00000074 memsz 0x00000074 flags r-x
    LOAD off    0x00000078 vaddr 0x08049078 paddr 0x08049078 align 2**12
         filesz 0x00000000 memsz 0x00000000 flags rw-
	
Sections:
Idx Name          Size      VMA       LMA       File off  Algn
SYMBOL TABLE:
no symbols

contructing the elf header was relatively easier than constructing the program header table thanks to the elf specification. i did look at the elf header example at the whirlwind tutorial from time to time to verify if i was still on the right track of course. objdumps of my former asm programs were very much useful since they provided some sort of blue print-cheat sheet for me.

understanding the difference between “file scope” and “virtal memory scope” was a bit of a stumbling block. again, thanks to my blue print-cheat sheet, i was able to understand those concepts. file scope is offset that is relative to zero. “virtual memory scope” is offset that is relative to the memory address i specified in nasm’s org directive, that being 0x08048000.

anyway, on to the meaty stuff. the code above starts off with an elf header that specifies two (2) program headers (e_phnum) for .text and .data sections respectively. i still haven’t covered the part about “section header table“. reading the elf spec some pages down sure takes a lot toll on your body! i hope to cover more ground tommorrow and post yet another progress report.

for now.. i need steph!