makeelf
makeelf copied to clipboard
v3l0c1r4pt0r
Link program headers to sections
Allow the caller to specify a section to associate a program header to. This will allow the ELF writer to determine a correct file offset for the program header (the file offset of the lowest associated section).
This is the minimal change needed to solve a problem of mine, which was to generate core dump files from scratch. If you'd like some testcase, example usage or fancier implementation I'm open to suggestions of course :) (I also noticed the similar but more fully featured PR that's also open; while that is stewing maybe this is a more lightweight change just to scratch this one itch :) )
Yes, would be great if you provide some test vector for that. I am not developing this project actively, so I guess it would be easier for you than for me at the moment.
Been a while, sorry for the long wait - here's a test :) Let me know what you think.
I also suggest to add function that maps sections to segments to allow changing sizes of sections without segments offsets breaking.
It makes "one allocated section -> one loaded segment" assumption though
## Try to find a mapping from sections to segments to allow to move them around
def map_sections_to_segments(self):
def intersects(a, b):
a1, a2 = a
b1, b2 = b
res = a2 > b1 and b2 > a1
return res
sec: Elf32_Shdr
seg: Elf32_Phdr
for i, sec in enumerate(self.Elf.Shdr_table[1:], 1):
if sec.sh_flags & SHF.SHF_ALLOC == 0:
continue
sec_rg = (sec.sh_addr, sec.sh_addr + sec.sh_size)
for j, seg in enumerate(self.Elf.Phdr_table[1:], 1):
if seg.p_type != PT.PT_LOAD:
continue
seg_rg = (seg.p_vaddr, seg.p_vaddr + seg.p_memsz)
if not intersects(sec_rg, seg_rg):
continue
assert sec_rg[0] >= seg_rg[0] and sec_rg[1] <= seg_rg[1], "Sections cannot span multiple segments!"
seg.sections.append(i)
break
I also suggest to add function that maps sections to segments to allow changing sizes of sections without segments offsets breaking.
May I ask you to add that via a PR of your own? I'm glad I had the time to add test vectors for my change, and I have a feeling @v3l0c1r4pt0r will be (rightfully) asking for test vectors for your function as well :)