On Tue, Oct 9, 2012 at 9:00 AM, Andy Isaacson <a...@hexapodia.org> wrote: > That's expected, because you're still going through the translation > layer.
Ok, I now see the faulty assumption in my experiment — blocks are added to the mapping only after being erased. > If you had instead hooked up a microcontroller to the data pins > of the flash chips on the circuitboard, you'd find that much of the data > you'd written is still on the flash chips; it's simply in pages that > aren't referenced by the current FTL state. That's not necessarily true, it depends on the algorithm used for wear leveling — see [1, §2.3.1]. [1] “Algorithms and data structures for flash memories”, http://dx.doi.org/10.1145/1089733.1089735. -- Maxim Kammerer Liberté Linux: http://dee.su/liberte -- Unsubscribe, change to digest, or change password at: https://mailman.stanford.edu/mailman/listinfo/liberationtech