Neither of these have had any traction for a while, due to an impasse with the way flash sizes are declared with Pico's bi_decl being incompatible with how we might want to supply (or retrieve from) flash sizes in memmap_mp.ld.
The general idea is to split FLASH into APP and FILESYSTEM so we have an early failure if the two happen to overlap.
Without a linker error to catch this, a MicroPython build will generate a filesystem at runtime and overwrite itself in potentially catastrophic ways.
Making this tooling work requires adding an memmap_mp.ld to each board directory. Since this does not inherit the flash size from mpconfigboard.h it removes the previous "single source of truth" for flash size and... this can easily catch you out.
In all cases the linker variable _flash_app_size should be ACTUAL_FLASH_SIZE - MICROPY_HW_FLASH_STORAGE_BYTES. Eg: For Pico this is 2048 (2MB) - 1408 (1.4MB) = 640k.
This is a little bit of a redux of the issue I raised here: https://github.com/micropython/micropython/issues/8680
And the PR here: https://github.com/micropython/micropython/pull/8761
Neither of these have had any traction for a while, due to an impasse with the way flash sizes are declared with Pico's
bi_decl
being incompatible with how we might want to supply (or retrieve from) flash sizes inmemmap_mp.ld
.The general idea is to split
FLASH
intoAPP
andFILESYSTEM
so we have an early failure if the two happen to overlap.Without a linker error to catch this, a MicroPython build will generate a filesystem at runtime and overwrite itself in potentially catastrophic ways.
Making this tooling work requires adding an
memmap_mp.ld
to each board directory. Since this does not inherit the flash size frommpconfigboard.h
it removes the previous "single source of truth" for flash size and... this can easily catch you out.In all cases the linker variable
_flash_app_size
should beACTUAL_FLASH_SIZE - MICROPY_HW_FLASH_STORAGE_BYTES
. Eg: For Pico this is2048 (2MB) - 1408 (1.4MB) = 640k
.