On Fri, Jun 02, 2017 at 11:04:06AM +0200, Boris Brezillon wrote: > BTW, MTD_NO_ERASE is not the only problem we have with UBI or JFFS2. > Are we guaranteed that an erase operation fills an eraseblock with > ones? Don't we have mem technologies that are filling them with zeros? > Note that mtdram is artificially setting the mem-region to 0xff in its > dummy erase operation, so maybe it's a implicit rule that ->_erase() is > supposed to fill eraseblocks with 0xff.
I've wondered about the general assumption. But mtdram isn't really a good example, because it clearly calls itself a "test mtd device". So it makes sense it would emulate common MTDs (i.e., flash memory). Brian
