samples/application_development/code_relocation fails to build with coverage on mps2_an385

[dcpleung]

Describe the bug
When building samples/application_development/code_relocation on mps2_an385 with coverage enabled, it complains that

ERROR: section .sram2_data is 6828 bytes in the virtual/runtime address space, but only 6824 in the loaded/XIP section!

Note that the size of section .sram2_data in zephyr.map is 6824 (0x1AA8) bytes:

.sram2_text     0x0000000020200000      0x2d0 load address 0x00000000000000c0
_SRAM2_RODATA_SECTION_NAME
                0x00000000202002d0        0x4 load address 0x0000000000000390
.sram2_data     0x00000000202002d4     0x1aa8 load address 0x0000000000000394
.sram2_bss      0x0000000020201d80       0x98 load address 0x0000000000001e3c

To Reproduce
Steps to reproduce the behavior:

1. mkdir build; cd build
2. cmake -DBOARD=mps2_an385 -DCONFIG_COVERAGE=y ../
3. make
4. See error

Expected behavior
Should build and run without errors.

Impact
Impacting code coverage.

[nashif]

maybe we should just exclude it when coverage is enabled, this is a very "custom" sample

[dcpleung]

Is there a way to exclude it from coverage in the yaml file?

[nashif]

no, just add CONFIG_COVERAGE=n to the project file.

[dcpleung]

From what I can tell, check_link_map.py doesn't take into account the actual size of the section (in this case 0x1AA8), but tries to infer size with addresses. The section can actually fit inside the relocated (loaded/XIP) space. I don't know enough about relocation and the script to fix it correctly.

[AdithyaBaglody]

There are 2 options here.

1. Disable the coverage.
2. Run this on a hardware which has multiple memory regions.
   The 2nd option is not feasible for mps2_an385 because it has only one memory region. This sample creates another memory region inside sram to show the working of code-relocation script. I am not sure how the gcov would behave when we instrument the code in different memory regions. So I suggest we go with first option and disable coverage for this sample.

[wentongwu]

.sram2_data 0x00000000202002d4 0x1aa8 load address 0x0000000000000394
... ...
.data..LPBX1 0x0000000020201d70 0x4 CMakeFiles/offsets.dir/arch/arm/core/offsets/offsets.c.obj
.data..LPBX1 0x0000000020201d74 0x8 libcode_relocation_source_lib.a(code_relocation.c.obj)
*(SORT_BY_ALIGNMENT(.rel.data..LPBX1))
0x0000000020201d7c . = ALIGN (0x4)
0x0000000020201d7c __sram2_data_end = .
0x00000000202002d4 __sram2_data_start = ADDR (.sram2_data)
0x0000000000001aa8 __sram2_data_size = SIZEOF (.sram2_data)
0x0000000000000394 __sram2_data_rom_start = LOADADDR (.sram2_data)

.sram2_bss 0x0000000020201d80 0x98 load address 0x0000000000001e3c

@AdithyaBaglody I have a question: why vma of sram2_bss is 0x0000000020201d80 instead of 0x0000000020201d7c? I think it should be the problem, Thanks a lot!

[AdithyaBaglody]

I actually don't see this error. I just find that the emulator has crashed. In my case I don't see 4byte offset (but still fails) ,so I cant tell what caused this. But if you think this is the issue please go ahead an create a PR fixing it.