There was only one genuine use of this macro, all other places were
using it only as a necessary compensation. While this was fine as long as
it was the first meg of flash which was mapped, it became incorrect and
quite dangerous whenever this assumption did not hold (such as when
running from the second slot in an OTA scenario).
The flash API now uses actual addresses, not translated/mapped
addresses, and the users of this API have been adjusted accordingly.
This makes the flash API work correctly regardless of what flash mapping
is in use.
The old macro is still available under the new name
INTERNAL_FLASH_MAPPED_ADDRESS, and this is used to detect flash writes
where the source is mapped flash (and thus has to be bounced), and to
adjust the _flash_used_end linker symbol when used with
flassh_find_sector() by the filesystem code. The latter usage is not
OTA-proof, but in an OTA scenario the filesystem needs a fixed location
anyway and thus would not use this code path.
Also made the cache on/off configurable via user_config.h. Uncached writes
are not a very good idea, but for read-only deployments a further ~0.5k RAM
can be gained by disabling the cache.
Tweaked the file.read() workhorse to read large chunks at a time rather
than use getc(), to compensate for potential unavailability of cache.
Arnim Läuger
Johny Mattsson
Johny Mattsson
funshine
HuangRui