The const_ram_storage! and ram_storage! macros used to have trait and
result arguments to set the storage trait to implement and the result
type to use. As this defaulted to relative paths, it made the result of
the macro dependent on the context and could cause confusion. At the
same time, there is no real use case for substituting these types.
This patch removes the arguments for good and just uses
$crate::driver::Storage and $crate::io::Result instead.
Previously, we reported the lookahead buffer size in bytes but
littlefs2-sys expects the lookahead buffer size as a multiple of 8
bytes. This could lead to a buffer overflow causing filesystem
corruption. This patch fixes the reported lookahead buffer size.
Note that Storage::LOOKAHEAD_WORDS_SIZE allows users to set invalid
values (as it is measured in 4 bytes, not in 8 bytes). Invalid values
that were previously accepted because of the wrong buffer size
calculation can now be rejected by littlefs2-sys.
This is a combination of two previous patches:
https://github.com/trussed-dev/littlefs2/pull/19https://github.com/Nitrokey/littlefs2/pull/1
Fixes: https://github.com/trussed-dev/littlefs2/issues/16
In real usage, `Storage` will be implemented by a class wrapping
a device's flash peripheral. We may very well need access to this
peripheral outside of littlefs (e.g., to change startup bytes or
any other flash read/write operation).
One possible approach is to "release" the flash peripheral, drop
the Storage, and later recreate it again with the same peripheral.
`ram_storage!` now generates a `Ram`, which is not dropped and
plays the role of the flash peripheral and its backing flash.
Instead, `RamStorage` now stores a `&mut Ram`. The tests intend to
demonstrate that we can reconstruct `RamStorage` and continue to
work with a possibly even mounted filesystem.
The "weird gotcha" is that `assert_eq!(metadata.len(), n)` panics
on an `unwrap_err`, while `assert!(metadata.len() == n)` works.
Robin Krahl
Sosthène Guédonandsosthene-nitrokey
Nicolas StalderandNicolas Stalder