| |
| |
| |
|
Page: 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17
Comments:
<0> the things I will probably setup for my ELF loader
<0> for now, will be static program loading.
<0> and expand to support shared objects later.
<1> I need f00d
<2> http://www.youtube.com/watch?v=hDayiW9kfog
<2> good song =)
<3> ELF is dead simple for a dumb loader... 30 lines for static executables
<4> even for relocs
<0> about right?
<0> typedef Elf32_Addrunsigned int;
<0> typedef Elf32_Halfunsigned short;
<0> typedef Elf32_Offunsigned int;
<0> typedef Elf32_Swordint;
<0> typedef Elf32_Wordunsigned int;
<0> ZetItUp, You like Tool?
<0> Im listening to 10,000 days now :D
<0> Im pretty much implementing this straight from the spec.
<0> uysing the same names
<0> lol
<2> Zeii: i dont have any songs with them :(
<2> but i like them :D
<0> :D
<0> They are awesome
<3> Zeii: uh... typedefs are declared the other way around
<0> I own every album they have ever made.
<0> ahh thanks mwk forgot :P
<0> lol
<3> Zeii: https://x-os.homeip.net/~mwk/os/include/elf/
<3> here are lots of definitions taken straight from specs, ready to ude
<0> :) thanks mwk
<0> but, I want to see if I can make a loader.
<0> lol
<2> Zeii: sure you didnt mean Dword?
<2> :D
<2> 23 16:51 | <0> typedef Elf32_Swordint;
<2> :D
<0> nope, Sword
<2> Sword :d
<0> :)
<0> its in the sepc.
<2> kill(Sword);
<0> *spec.
<2> :D
<0> lol
<0> HAhaha
<3> Signed Word that is
<5> ruhroh
<5> http://x-os.homeip.net/x0x0x0x0x0x0x0x0x0/ThisFileMustNotExist
<5> hah
<0> lol
<0> totally from the specs.
<0> #define EI_NIDENT
<0>
<0> typedef struct {
<0> unsigned chare_ident[EI_NIDENT];
<0> Elf32_Half e_type;
<0> Elf32_Halfe_machine;
<0> Elf32_Worde_version;
<0> Elf32_Addre_entry;
<0> Elf32_Offe_phoff;
<0> Elf32_Offe_shoff;
<0> Elf32_Worde_flags;
<0> Elf32_Halfe_ehsize;
<0> Elf32_Halfe_phentsize;
<0> Elf32_Halfe_phnum;
<0> Elf32_Halfe_shentsize;
<0> Elf32_Halfe_shnum;
<0> Elf32_Halfe_shstrndx;
<0> } Elf32_Ehdr;
<3> nn: logs showed this file being accessed A LOT, probably for some exploit... so i decided to create it, just for kicks :)
<5> ahh
<5> :)
<5> interesting language
<0> Mwk, this seem right?
<0> #define ET_NONE
<0> #define ET_REL
<0> #define ET_EXEC
<0> #defineET_DYN
<0> #define ET_CORE
<0> #define ET_LOPROCxFF00
<0> #define ET_HIPROCxFFFF
<0> Just going through the spec now, defining all the strucutres and such
<6> so you're going to paste the whole spec here?
<7> zeii, wtf, i told you use a pastebin
<0> >< jeez, sorry.
<8> It's not so much talking going on anyways.. don't get too harsh to each other .. ;)
<0> :)
<6> code which switches to real mode and back to pmode has to reside within the first megabyte of memory ?
<3> Zeii: yea, in fact, it's in spec, and it agrees with my headers.
<3> shackan_: in first 0x10fff0 bytes of memory, actually...
<6> great, thanks
<0> thanks mw
<0> *mwk
<2> #define DONT_GO_OVER_THIS_BYTE 0x10FFF0
<2> :D
<3> hm
<3> complete SPARC relocation list or play some nethack
<3> that's the question
<2> SPARC
<3> bah... i even managed to find specs for all 42 i386 relocation types, including some listed as 'reserved' in some specs
<0> woah
<3> ELF relocations **** kinda
<3> each architecture has its own relocation types
<3> allocation is totally uncoordinated
<3> on Solaris, someone needed a relocation that takes size of a symbol to implement some stuff in dtrace
<3> since Solaris supports *86 and SPARC, they added new relocations to i386, x86_64 and sparc lists
<0> crazy
<3> the nice part about it is that anyone can use it, even outside of Sun
<3> but... if he wants to use it on, say, PPC, he cannot. cause there's on relocation type ***igned there.
<3> now, some smart*** in GNU invented new optimised TLS model
<3> which makes for great speedup, etc, etc
<3> and is arch-independent
<3> but, this guy only knew FR-V and x86... so only two arches support it
<0> mwk, what to the LOPROC and HIPROC, LORESERVE / HIRESERVE feilds mean to me?
<3> Zeii: LOPROC and HIPROC? depends on your architecture.
<3> LORESERVE/HIRESERVE are range of reserved section indices
<3> which are treated specially by linker and friends
<0> hm
<0> linker as in, compiler/linker or linker as in dynamic linker?
<3> linker and ***embler, in fact
<3> dynamic linker doesn't bother about sections
<0> Ah.
<0> So, it doesn tmean so much to something reading the ELF?
<0> just something building it?
<2> Zeii: http://www.youtube.com/watch?v=yIcDczoDhZ8
<2> <3
<0> lol :)
<3> Zeii: depends what you mean by reading
<3> it doesn't mean anything to program loader reading ELF executable to load it to memory
<0> Ah, I see. :)
<3> in fact, simple executable loader should only bother about ELF header and PT_LOAD program headers
<0> :D Thanks mwk :D
<3> but, it means alot to linker [ld] reading .o files to form the executable/.so
<0> should the Headers be bitpacked?
<3> wtf is bitpacked?
<8> Bitfields, maybe?
<0> no, packed.
<0> __attribute__((packed));
<8> I suppose so.
<0> no space between fields.
<3> no.
<8> No? Oh.
<0> eg: e_type directly after all of e_idents stuff
<3> JimPanic, Zeii: ELF designers were smart and designed them so that there aren't any empty spaces between fields anyway
<3> also, packing structures is almost always dumb idea
<3> the only exceptions being when interfacing to hardware or compressed data formats, like DWARF
<0> thought so, since the spec says it doesnt use any bitfields due to portability reaons. THanks mwk.
<3> also, ferchrissake, ELF specification contains C typedef for all structures. just copy-and-paste it to your header.
<0> I have a question about the section header flag, SHF_MASKPROC
Return to
#osdev
or
Go to some related
logs:
#gentoo
MySQL CREATE FUNCTION Error 1064
#osdev
#suse
#perl
#linux
#php
#linux
#php
#web
|
|