llvm dev - Jan 2016 - [LLD] Question of handling a special relocation in COFF

Hello all,

I am handling a special architecture's relocation in COFF. Such relocation
(IMAGE_REL_RELSEC) is used specifically in static linked executables:

$ llvm-objdump -d -r foo1.obj:

Disassembly of section .text:

00:  mov 04, r2
            00000002:  IMAGE_REL_RELSEC   .text
04: ...

$ llvm-objdump -d -r foo2.obj:

Disassembly of section .text:

00:  mov 04, r2
            00000002:  IMAGE_REL_RELSEC   .text
04: ...

And the expected output will be:
$ lld-link /out:foo.exe foo1.obj foo2.obj

$ llvm-objdump -d foo.exe:

40000000: mov 40000004, r2
40000004: ...
40000008: mov 4000000c, r2
4000000c: ...

PS. 40000004 = Addend (4) + Image Base (40000000) + Instruction RVA (0)
       4000000c = Addend (4) + Image Base (40000000) + Instruction RVA (8)

In Summary, such relocation stores addend that relative to .text section.
Linker is expected to fix the relocation of the instruction by image base
and the instruction RVA. I am wondering whether there is other relocation
like IMAGE_REL_RELSEC therefore I can take a look of it.

Thanks for you reply
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://lists.llvm.org/pipermail/llvm-dev/attachments/20160115/4e145e14/attachment.html>

I don't know if this would answer to your question, but the file you want
to take a look (if you haven't) is
https://github.com/llvm-mirror/lld/blob/master/COFF/Chunks.cpp. All
reocations are implemented in this file.

It sounded like you want to add something like add32(Off, S + ... +
Config->ImageBase) for your relocation. There are a lot of examples in the
file.

You are trying to define a new type of relocation, right? Or, you are
trying to post this to a new architecture?

On Thu, Jan 14, 2016 at 12:08 PM, Yi-Hong Lyu via llvm-dev <
llvm-dev at lists.llvm.org> wrote:
> Hello all,
>
> I am handling a special architecture's relocation in COFF. Such
relocation
> (IMAGE_REL_RELSEC) is used specifically in static linked executables:
>
> $ llvm-objdump -d -r foo1.obj:
>
> Disassembly of section .text:
>
> 00:  mov 04, r2
>             00000002:  IMAGE_REL_RELSEC   .text
> 04: ...
>
> $ llvm-objdump -d -r foo2.obj:
>
> Disassembly of section .text:
>
> 00:  mov 04, r2
>             00000002:  IMAGE_REL_RELSEC   .text
> 04: ...
>
> And the expected output will be:
> $ lld-link /out:foo.exe foo1.obj foo2.obj
>
> $ llvm-objdump -d foo.exe:
>
> 40000000: mov 40000004, r2
> 40000004: ...
> 40000008: mov 4000000c, r2
> 4000000c: ...
>
> PS. 40000004 = Addend (4) + Image Base (40000000) + Instruction RVA (0)
>        4000000c = Addend (4) + Image Base (40000000) + Instruction RVA (8)
>
> In Summary, such relocation stores addend that relative to .text section.
> Linker is expected to fix the relocation of the instruction by image base
> and the instruction RVA. I am wondering whether there is other relocation
> like IMAGE_REL_RELSEC therefore I can take a look of it.
>
> Thanks for you reply
>
> _______________________________________________
> LLVM Developers mailing list
> llvm-dev at lists.llvm.org
> http://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-dev
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://lists.llvm.org/pipermail/llvm-dev/attachments/20160114/f5709060/attachment.html>