Here I have explained the redo and undo concepts. Below are the observations

1. Datafiles may contain uncommitted data (buffers dirtied by uncommitted transaction and checkpoint takes place)
2. Committed data may not be there in data files (Checkpointing not done after commit)
3. Redo/undo for committed changes will always be there in redo logs (Lgwr writes on commit).
4. redo/undo for Uncommitted changes may be there in redo logs(Lgwr writes every 3 seconds)
5. In addition to redo logs, undo information
a) may be present in the buffer cache (checkpoint has not taken place after the change)
b) will be written to undo tablespace on checkpoint