> Hi,
>
> I'm using msp430-objdump with the IAR disassembler option to generate an
IAR
> assembly file from my linked elf file. I can then re-assembler and
re-link
> it using IAR's tools. However, the .data section is causing me trouble.
If
> I use "-dST" for the objdump, then the .data section is missing, which
means
> that any initialised data gets 0xff (i.e., blank flash) as its contents.
If
> I use "-DST", then the .data section is included (it is disassembled,
which
> looks a bit odd, but the data is correct), but the section address is the
> VMA of the section (0x0200), not the LMA (0xe768, in my case). What I
need
> is a way to get the .data section out of the elf file, preferably along
with
> the rest of the disassembly, but with the origin set to the LMA rather
than
> the VMA. Does anyone know if there is a way to do this?
>
> Many thanks,
>
> David Brown
> Norway.
>
I have now fixed the situation, by making some changes to objdump.c . The
three changes I had to make were making the label use the lma rather than
vma (about line 1936);
objdump_print_IAR_label(abfd, section, sym,section->vma +
addr_offset,
&disasm_info, false);
to
objdump_print_IAR_label(abfd, section, sym,section->lma +
addr_offset,
&disasm_info, false);
I also had to make the -dST flag include the .data section, because -DST was
causing the .vectors section to be strangely generated. This meant adding a
line
&& (section->flags & SEC_DATA) == 0
at line 1827, where the disassemble_data() function determines whether or
not to print the section.
Finally, I had to comment out the lines to add +0x... strings to the labels
in objdump_print_IAR_label() (these get generated because the address being
printed is now the lma, not the vma), since such labels cause errors in the
IAR assembler.
I guess that this should really be made as a patch in some way, but I don't
know how that works, and I think the changes I made should be added in a
neater way (for example, it is not good to have objdump_print_IAR_label()
declared with a parameter "vma" and then pass the lma to it). I also
haven't thought through any possible consequences of these changes, and for
some reason part of the data section turns up in disassembly (ugly, but it
works), and part as a raw DW data dump (much neater). Also, it's 21:45, and
the project has to be working for tomorrow, so I'm going to have to skip the
nicities for now.
mvh.
David
