You have a blob of some data which you would like to embed into your C program. Perhaps a splash screen, or a special font, firmware for your scsi card, or whatever. The usual approach which I think most people are familiar with is to run something like xxd -i to generate a source file with a large array of hex constants. Or write your own little script for that purpose.

But what if we could just link binaries into our program directly? We can. Mostly. I had no idea this was possible until after reading this list posting which explains the technique. It requires an intermediate object file, but it’s much faster to generate and compile then a textual equivalent file.

I thought this was pretty cool and needed to try it out. I’m going to write a little quine like program that prints its own source.

quine.c:

#include <stdio.h>

int
main(int argc, char **argv)
{
        extern const char _binary_quine_c_start, _binary_quine_c_end;
        const char *start = &_binary_quine_c_start;
        const char *end = &_binary_quine_c_end;

        fwrite(start, end - start, 1, stdout);
}

Now we need to prepare the object. This is where I ran into just a touch of trouble. Over here, ld is ld.lld, the llvm linker and it produces an error message at first.

> ld -r -b binary quine.c -o myself.o
ld: error: target emulation unknown: -m or at least one .o file required

The ld man page is most unhelpful, telling me -m can set an emulation, but that is the only occurrence of the word emulation in the entire page and there is nothing more to be learned. Running ld.bfd just works. However, digging into the source at llvm/tools/lld/ELF/Driver.cpp one can find a list of supported emulations, thankfully not too far from the top of the file.

And now we’re on our way.

> ld -r -b binary quine.c -o myself.o -m elf_amd64
> cc -c quine.c
> cc quine.o myself.o

Running the result and we get the expected output.

> ./a.out                                                                             
#include <stdio.h>

int
main(int argc, char **argv)
{
        extern const char _binary_quine_c_start, _binary_quine_c_end;
        const char *start = &_binary_quine_c_start;
        const char *end = &_binary_quine_c_end;

        fwrite(start, end - start, 1, stdout);
}

Tada.