Right now we calculate checksum on the fly as we stream the file to a partition.
This is cheap; not much extra overhead (some CPU time).
However, for increased robustness it is valuable to write first then read it back and calculate the checksum while reading it back instead.
This is important for devices with unreliable storage and/or safety-critical operations, that are OK with the tradeoff of more resource overhead, as well as when debugging storage issues.
- A Mender client configuration option to turn on/off checksum verification by reading it back from storage