RE: reloc against common symbols

["Vineet Sharma, Noida" <vineets at noida dot hcltech dot com>]

Hi,

>It's hard to say without more information.  What is the backtrace when
>coff_XXX_reloc() is called?  There are several possible paths to that
>point, and I'm not sure which you are using.

It is called from 

		bfd_perform_relocation()

which in turn is called from

	bfd_generic_get_relocated_section_contents()

Hope that gives you the idea?



My current output is as follow:
For the code below

	_y:
        .short  _x+0
        .short _xx+0
        .short _xxx+0
        .global _z
        .align  1
	_z:
        .short  1
        .comm _xxx, 4
        .comm _xx,5
        .comm _x,6
	
	
If .bss and .common follows .data and data ends (say at 0x0008)
then the address allocated by linker are wrong(i guess)

Sections:
Idx Name          Size      VMA       LMA       File off  Algn  Flags
  0 .text         00000000  00000000  00000000  00000000  2**2  ALLOC, LOAD,
CODE
  1 .data         00000008  00000000  00000000  000000a8  2**2  CONTENTS,
ALLOC, LOAD, DATA
  2 .bss          00000011  00000008  00000008  00000000  2**2  ALLOC

So the address allocated for
		_x   is 0x000c
		_xx  is 0x0014
		_xxx is 0x0008

Dont you thinks some thing wrong in this _xx being declared first(before _x)
should start at lower address
than that of (_x).

 The allocated symbols should have been set by the generic linker code to
hold the allocated address.
So i guess my code is not interfering with the genric part.


Regards
Vineet








-----Original Message-----
From: Ian Lance Taylor [mailto:ian@wasabisystems.com]
Sent: Tuesday, February 10, 2004 8:15 PM
To: Vineet Sharma, Noida
Cc: binutils@sources.redhat.com
Subject: Re: reloc against common symbols


"Vineet Sharma, Noida" <vineets@noida.hcltech.com> writes:

> 	Actually i have ported binutils to new target (a 16 bit proceessor).
> The problem is with 
> linker(although same holds true for assembler) .Lets talk about linker
> particularly coff_XXX_reloc()
> 
> Now when i need to fix a relocation against a common symbol(in function
> coff_XXX_reloc())
> (Refer example below ".short  _x+28" _x is common symbol)
> I use the follwing method to calculate the value of the final symbol
> value/address in coff_XXX_reloc()
> 
> 	static long 
> 			get_symbol_value (symbol)
> 					     asymbol *symbol;
> 		{
> 			  long relocation = 0;
>  
> 
> 			  if (bfd_is_com_section (symbol->section))
> 				    relocation = 0;
> 			  else
> 			    relocation = symbol->value +
> 				      symbol->section->output_section->vma +
> 				      symbol->section->output_offset;
> 	
> 
> 			  return relocation;
> 		}
> Now for a normal symbol(non-common symbol) "symbol->value" contains the
> offset of the symbol in the section.
> Every thing works fine.And the function returns the final address of the
> symbol.Good..
> 
> 
> Now for the common symbol "symbol->value" does not contain the offset of
the
> symbol in the section, rather it contains the lenght of the common
symbol(if
> its say .comm _x,120) "symbol->value" contians 120(the lenght of the
common
> symbol).Thus i am not able to calculate the final address of the common
> symbol.

When the linker calls coff_XXX_reloc(), presumably via
bfd_generic_get_relocated_section_contents(), there should be no
symbol for which bfd_is_com_section (symbol->section) is true.  All
the symbols should have been allocated into real sections by
lang_common().  The allocated symbols should have been set by the
generic linker code to hold the allocated address.

> >What is the actual problem?
> 
> How to calculate the final address of the common symbol in
coff_XXX_reloc()?
> 
> The above method get_symbol_value() does not work for common symbols.Whats
> wrong?

It's hard to say without more information.  What is the backtrace when
coff_XXX_reloc() is called?  There are several possible paths to that
point, and I'm not sure which you are using.

Ian