Advice on linker script creation and verification - compilation

Question

Long story short. I wish to learn how to create a good linker script so that should I change platforms/architectures/vendors, I'm not stuck at ground zero again with not knowing what to do. I'm not concerned with the difficulty of the task, so much as understanding it.
I've started a sort of project, as it were, to create a base or skeleton for programing and developing on STM's 32-bit Cortex-M3 chips. With the help of jsiei97 Beginning with the STM32F103RB (I also have a TI Stellaris LM3S828, but that's another issue), without the need of a licensed IDE. Since I am a student, and most students can't afford such things.
I understand that there's the ODev, and Eclipse Plugins and what not, and have read various blogs, wikis, docs/man pages and most projects provide you with a linker script with little to know explanation as to why and where things have been defined.
I've compiled an arm-none-eabi toolchain for the STM32 but where I get hung up is in the linker script. CodeSourcery also requires one as well. I have a basic concept of how to create them and their syntax after reading the gnu man pages, but I simply haven't a clue where to start with putting in the various extra sections apart from the obvious .text, .bss and .data.
I created a rudimentary version but I get linking errors asking for section definitions and that's where I get stuck. I know how to define them, but knowing if what I'm doing is even close to right is the problem.

Answer 1

I have a simple linker script I reuse regularly across platforms, just change some addresses as needed.
http://github.com/dwelch67/
There are a number of samples many with gcc samples and most of those have linker scripts.
MEMORY
{
rom : ORIGIN = 0x00000000, LENGTH = 0x40000
ram : ORIGIN = 0x10000000, LENGTH = 30K
}
SECTIONS
{
.text : { *(.text*) } > rom
.bss : { *(.bss*) } > ram
}

Answer 2

Here is a working linker script for an STM32F105RB (there are also versions for R8 and RC):
https://github.com/anrope/stm-arp/blob/github/arp.rb.ld (text below)
My top-of-the-head guess is that you wont have to change anything. Maybe the origin of the regions defined in the MEMORY{} statement. Hopefully the comments will be helpful to you.
I used this with a GNU/GCC cross-compiler I rolled myself. After compiling, it's helpful to run nm on your code to make sure sections are being placed at the correct addresses.
Edit:
I pieced this linker script together by using the GNU ld documentation:
http://sourceware.org/binutils/docs/ld/
and by examining the output of a GCC cross-compile with the standard linker script, using nm. I basically identified all the sections that were being output and figured out which ones were actually useful, and where in memory they should go for the STM32F105.
I made notes in the linker script of the purpose of each section.
/*
arp.{r8,rb,rc}.ld :
These linker scripts (one for each memory density of the stm32f105) are used by
the linker to arrange program symbols and sections in memory. This is especially
important for sections like the interrupt vector, which must be placed where the
processor is hard-coded to look for it.
*/
/*stm32f105 dev board linker script*/
/*
OUTPUT_FORMAT() defines the BFD (binary file descriptor) format
OUTPUT_FORMAT(default, big, little)
*/
OUTPUT_FORMAT ("elf32-littlearm", "elf32-bigarm", "elf32-littlearm")
/* ENTRY() defines the symbol at which to begin executing code */
ENTRY(_start)
/* tell ld where to look for archive libraries */
/*SEARCH_DIR("/home/arp/stm/ctc/arm-eabi/lib")*/
/*SEARCH_DIR("/home/arp/stm/ccbuild/method2/install/arm-eabi/lib")*/
SEARCH_DIR("/home/arp/stm32dev-root/usrlol/arm-eabi/lib")
/*
MEMORY{} defines the memory regions of the target device,
and gives them an alias for use later in the linker script.
*/
/* stm32f105rb */
MEMORY
{
ram (rwx) : ORIGIN = 0x20000000, LENGTH = 32k
flash (rx) : ORIGIN = 0x08000000, LENGTH = 128k
option_bytes_rom (rx) : ORIGIN = 0x1FFFF800, LENGTH = 16
}
_sheap = _ebss + 4;
_sstack = _ebss + 4;
/*placed __stack_base__ trying to figure out
global variable overwrite issue
__stack_base__ = _ebss + 4;*/
_eheap = ORIGIN(ram) + LENGTH(ram) - 1;
_estack = ORIGIN(ram) + LENGTH(ram) - 1;
/* SECTIONS{} defines all the ELF sections we want to create */
SECTIONS
{
/*
set . to an initial value (0 here).
. (dot) is the location counter. New sections are placed at the
location pointed to by the location counter, and the location counter
is automatically moved ahead the length of the new section. It is important
to maintain alignment (not handled automatically by the location counter).
*/
. = SEGMENT_START("text-segment", 0);
/*isr_vector contains the interrupt vector.
isr_vector is read only (could be write too?).
isr_vector must appear at start of flash (USR),
address 0x0800 0000*/
.isr_vector :
{
. = ALIGN(4);
_sisr_vector = .;
*(.isr_vector)
_eisr_vector = .;
} >flash
/*text contains executable code.
text is read and execute.*/
.text :
{
. = ALIGN(4);
*(.text)
. = ALIGN(4);
*(.text.*)
} >flash
/*init contains constructor functions
called before entering main. used by crt (?).*/
.init :
{
. = ALIGN(4);
KEEP(*(.init))
} >flash
/*fini contains destructor functions
called after leaving main. used by crt (?).*/
.fini :
{
. = ALIGN(4);
KEEP(*(.fini))
} >flash
/* rodata contains read only data.*/
.rodata :
{
. = ALIGN(4);
*(.rodata)
/* sidata contains the initial values
for variables in the data section.
sidata is read only.*/
. = ALIGN(4);
_sidata = .;
} >flash
/*data contains all initalized variables.
data is read and write.
.data (NOLOAD) : AT(_sidata)*/
.data : AT(_sidata)
{
. = ALIGN(4);
_sdata = .;
*(.data)
_edata = .;
} >ram
/*bss contains unintialized variables.
bss is read and write.
.bss (NOLOAD) :*/
.bss :
{
. = ALIGN(4);
_sbss = .;
__bss_start__ = .;
*(.bss)
. = ALIGN(4);
/*COMMON is a special section containing
uninitialized data.
Example: (declared globally)
int temp; //this will appear in COMMON */
*(COMMON)
_ebss = .;
__bss_end__ = .;
} >ram AT>flash
. = ALIGN(4);
end = .;
/* remove the debugging information from the standard libraries */
DISCARD :
{
libc.a ( * )
libm.a ( * )
libgcc.a ( * )
}
/* Stabs debugging sections. */
.stab 0 : { *(.stab) }
.stabstr 0 : { *(.stabstr) }
.stab.excl 0 : { *(.stab.excl) }
.stab.exclstr 0 : { *(.stab.exclstr) }
.stab.index 0 : { *(.stab.index) }
.stab.indexstr 0 : { *(.stab.indexstr) }
.comment 0 : { *(.comment) }
/* DWARF debug sections.
Symbols in the DWARF debugging sections are relative to the beginning
of the section so we begin them at 0. */
/* DWARF 1 */
.debug 0 : { *(.debug) }
.line 0 : { *(.line) }
/* GNU DWARF 1 extensions */
.debug_srcinfo 0 : { *(.debug_srcinfo) }
.debug_sfnames 0 : { *(.debug_sfnames) }
/* DWARF 1.1 and DWARF 2 */
.debug_aranges 0 : { *(.debug_aranges) }
.debug_pubnames 0 : { *(.debug_pubnames) }
/* DWARF 2 */
.debug_info 0 : { *(.debug_info .gnu.linkonce.wi.*) }
.debug_abbrev 0 : { *(.debug_abbrev) }
.debug_line 0 : { *(.debug_line) }
.debug_frame 0 : { *(.debug_frame) }
.debug_str 0 : { *(.debug_str) }
.debug_loc 0 : { *(.debug_loc) }
.debug_macinfo 0 : { *(.debug_macinfo) }
/* SGI/MIPS DWARF 2 extensions */
.debug_weaknames 0 : { *(.debug_weaknames) }
.debug_funcnames 0 : { *(.debug_funcnames) }
.debug_typenames 0 : { *(.debug_typenames) }
.debug_varnames 0 : { *(.debug_varnames) }
}