eds/store: handle corrupted storage

[walldiss]

Implementation ideas

Underlying storage of eds.Store could be corrupted at some point and we need to handle such situations. Possible roots of corruption could be:

• bug in CAR write operation
• bug in eds.Store that lead to storing corrupted date
• Hardware (SSD) failure
• 3rd party manipulation on CAR file

We can identify such situation by running simple validation of data integrity by hashing:

• CAR header could be validated by calculating dataHash.
• Shares could be validated by verification with inclusion proofs.

We need to figure out how possible handling of such situations could look like. Some possible options could include:

• recover data by data reconstruction
• overwrite corrupted data with data requested from remote peer
• alerting to node operator to handle it manually