Hi,
I need to write a linker (ld) script for our nRF52 application. Essentially I need to combine the linker script that we already use with the scripts that exist in the nRF SDK.
The SDK examples typically use two script files: the project (example) specific, e.g. ble_app_gatts_c_gcc_nrf52.ld and generic nrf_common.ld.
The SDK declares a few custom sections, for example .fs_data which contains an initialised variable that is placed in RAM. The initialisation value is placed in the flash.
What I don't understand is what causes the LMA address of the .fs_data section to be in the flash area.
In the linker file:
.fs_data :
{
PROVIDE(__start_fs_data = .);
KEEP(*(.fs_data))
PROVIDE(__stop_fs_data = .);
} > RAM
and in the generated map file (note "load address 0x00038da4"):
.fs_data 0x200028f8 0x14 load address 0x00038da4 0x200028f8 PROVIDE (__start_fs_data = .) *(.fs_data) .fs_data 0x200028f8 0x14 _build/nrf52840_xxaa/fds.c.o 0x200028f8 m_fs 0x2000290c PROVIDE (__stop_fs_data = .)
This seems to be the effect of the INSERT AFTER command. If I remove the insert command, LMA and VMA of the fs_data section are equal and in the RAM area. But I can not replicate this behaviour in my linker script. Regardless whether INSERT is present or not in my script, I get LMA=VMA and in RAM. Below is a fragment of my resulting map file.
.fs_data 0x20003e28 0x14
0x20003e28 PROVIDE (__start_fs_data = .)
*(.fs_data)
.fs_data 0x20003e28 0x14 build/release/out/c/fds.o
0x20003e28 m_fs
0x20003e3c PROVIDE (__stop_fs_data = .)
