Garbage values in LMA of .data section

["Jeffi Edward.J" <j.jeffi@xxxxxxxxxxx>]

Hi,

I have a doubt in using different LMA and VMA for .data, .sdata sections.

I have logically splitted my 256MB RAM in my embedded system into two
regions as follows:

MEMORY
{
  swimage(rwx) : ORIGIN = 0x0, LENGTH = 0x3000000
  rtarea (rwx) : ORIGIN = 0x3000000, LENGTH = 0xD000000
}

SECTIONS
{
   /* swimage region starts */
   .text : {
      *(.init)
      *(.text)
      *(.fini)
      ...
   } > swimage
  __code_end = .;
  
 /* rtarea region starts */
 __runtime_area_start = 0x3000000;

 __data_start = __runtime_area_start;
 .data __runtime_area_start {
   *(.data)
 } > rtarea AT > swimage

 .sdata {
   *(.sdata)
 } > rtarea AT > swimage
 __data_end = .;
 __data_size = __data_end  - __data_start ;

 .sbss {
   *(.sbss)
 } > rtarea

 .bss : {
   *(.bss)
 } > rtarea

 /* allocation of space for heap/stack goes here ... */
 ...
 ...
 ...
}

I do copying of .data from LMA to VMA as follows:

memcpy(__data_start, __code_end, (__data_end - __data_start));

I'm using the following command to specify LMA region of .data.

 .data __runtime_area_start {
   *(.data)
 } > rtarea AT > swimage


After copying, if I refer to a global variable at VMA, it has garbage value.

Using --print-map option to linker, I generated map file of symbols.
I found the LMA of .data section in map file.
VMA of my .data is 0x3000000(as per above skeleton script).
Assume my LMA is 0x2000000 (__code_end).
I calculated the offset of that global variable with respect to LMA region.
I found in my LMA region itself, the global variables are not loaded
properly. i.e the global variable has the same junk value in LMA region.

I found from map file that some additional section are generated in between
.data and .sdata.
.eh_frame, .jcr, .data.rel.local, .data.rel.ro.local, .data.rel.

Are these sections needed? Is my LMA region corrupted my any other section?
On seeing the map file, I found that many additional section
.text._ZN6DArray***, .rela.text._ZN6DArray***, .rela.sdata.__ZTIS***,
.rela.rodata.__ZTIS***.

Apparently there is no overlapping of .data in LMA with other section.

I'm really not able to guess the cause of the problem. I'm struggling for
more than a week.

Please tell me what can cause of the problem and how to rectify.

Thanks and regards,
Jeffi

-- 
View this message in context: http://old.nabble.com/Garbage-values-in-LMA-of-.data-section-tp27140436p27140436.html
Sent from the gcc - Help mailing list archive at Nabble.com.