Re: question about output section LMA

[Nick Clifton <nickc at redhat dot com>]

Hi miloody,

> when I link the script as I attached, the linker will complain that
> .data section overlaps .rodata.str1.4 and .rodata.cst8 as below:

You do not say which version of the linker you are using, so I will 
assume that it is the latest.

> mips2_fp_le-ld: section .data [00000000800090c0 -> 00000000800093d7]
> overlaps section .rodata.str1.4 [00000000800090c0 -> 0000000080009143]

This happens because you have not specified where any of the .rodata.* 
sections should be placed in your link.xn linker script.  The linker 
chooses to place the .rodata.str1.4 section at the end of the .text 
section (since both of these sections are read-only), but your script 
also places the .data section immediately after the end of the .text 
section.

> but if I mark LMA part in .data section as below:
>   .data :
> /*  AT( _etext_init ) */

By commenting out the AT statement you are allowing the .data section to 
be moved so that the linker can place it after the end of the .rodata.* 
sections, and hence eliminating the overlaps.


> but why linker will tell me sections are overlapped when I use LMA?

Because the linker is telling you that at *load* time, the .data section 
and the .rodata.str1.4 sections are both being loaded into the same 
memory region.  This is not where those sections may be once *exection* 
starts, since that is controlled by the VMA address.  But the LMA 
address specifies where sections are loaded into memory, before they are 
moved to their VMA addresses.  So if two sections are going to be loaded 
into the same region of memory there is going to be a clash.

Cheers
  Nick