Re: [PATCH] [BINUTILS] [MICROBLAZE] PIC Data Text Relative

[Michael Eager <eager at eagercon dot com>]

> > >       (microblaze_bfd_write_branch_absolute_value_64): replace relative
> > > branch with absolute in case 'adjust_insn_abs_refs' is true
> > >       (microblaze_bfd_revert_base_reg_value_64): revert base register
> > > from r20 to r0 in case 'adjust_insn_abs_refs' is true
> >
> >
> > These two functions are short and only referenced in
> > microblaze_elf_relocate_section().  It might be clearer to put them
> > inline in this function.
> >
> > I'm unclear what is going on in these functions or why.  Why are you
> > replacing PCREL or TEXTREL reloc with a generic reloc?
> >
> > I'm not comfortable with editing the instructions to replace opcodes or
> > registers.  Shouldn't this be handled in GCC by generating the correct
> > instruction in the first place?  What this means is that the source
> > assembly generated by GCC will be different from that generated by objdump.
>
> In fact, the idea behind 'adjust-insn-abs-refs' option is when making
> static linking
> to a position independent executable with the base program using -R,
> the relocate function
> checks whether the symbol is coming from the external file or not then
> re-adjust the instruction
> to use absolute addressing instead since the reference location in
> memory is fixed and shall be
> excluded from PIC/PIE.
> - For function call: brlid => bralid and accordingly PCREL => Generic Reloc
> - For data reference => replace base register  r0 => r20 and TEXTREL => Generic
>
> I understand it's not the best clean way. I agree with you that it
> would be better to handle it in GCC.
> So, I came up with this solution: Make a variable/function declaration
> attribute in microblaze.c called for example
> 'absolute_reference' which then excludes the symbol reference from PIC
> and generate the corresponding
> instruction. In addition, replace 'adjust-insn-abs-refs' in linker
> with something like 'validate-insn-abs-refs'
> that only checks if symbol references coming from the external file
> are invoked with absolute addressing
> and generate error/warning if not. What do you think ?

I'll have to give this some thought.

Do any other targets support a text relative addressing mode?


> I get this error:
>
> *Hi. This is the qmail-send program at sourceware.org.
> *I'm afraid I wasn't able to deliver your message to the following addresses.
> *This is a permanent error; I've given up. Sorry it didn't work out.
>
> *<gcc-patches@gcc.gnu.org>:
> *Invalid mime type "text/html" detected in message text or
> *attachment.  Please send plain text messages only.
> *See http://sourceware.org/lists.html#sourceware-list-info for more information.
>
> I checked the website, it's stated that only plain text is allowed.

Set your email client to send text only.  If you are using gmail, click
the arrow at the bottom right of the compose window and select "Plain
Text Mode".

--
Michael Eager    eager@eagercon.com
1960 Park Blvd., Palo Alto, CA 94306