Shel, Re-reading my posts in this thread, I realize they are somewhat misleading and also just plain wrong in at least one statement. I hope to be more accurate and clear in this post, at least as much as my understanding of the systems involved may allow.
FAT 12/16/32 file systems were originally designed for magnetic disks. As such, the control structures, which are the boot record, the File Allocation Table (there's actually two) and the root directory are always written in that order, and in the same physical place on the first sectors of a volume. In FAT32, the root directory is actually treated like any ordinary data structure, but even then it typically follows the FATs. The data area fills the rest of the volume. In a flash device, repeated writing to a particular address increases the risk that that area will fail. However, in flash, the media can only be in the state of erased or non-erased. Only erased blocks can have data bits written to them; you can't actually twiddle individual bits or bytes. So when a change is requested for commitment to an existing cluster at the File System level, the Flash Translation Layer (which is below the FS) actually looks for a free (erased) block and writes the data there, which is then marked as dirty/ allocated (unless that block fails, which then gets marked as bad/ unusable and it seeks another free block). The old block is marked as deleted. The FTL virtually maps memory addresses to what the FS expects, preserving the illusion to the FS that data has been modified when actually it has been written to a new location. Blocks are organized into erase units, which are erased (written to all zeroes or ones depending on the device) as a group when needed, which happens when the card eventually runs out of free blocks. In some cases, the lowest virtual address are actually created and modified in system memory on card insertion due to the frequency of changes made by FAT. In summary, every time any "change" is made to data stored in flash media, it is actually written as new data to a new location on the device. Obviously, this includes control structures. As I understand it, more advanced wear-leveling systems in things such as ATA flash drives as well as wear-level aware file systems track the usage patterns of addresses at the media level to maximize the potential lifespan of the media. This is beyond the scope of cameras and thumb drives. Michael On Dec 4, 2006, at 11:29 AM, Shel Belinkoff wrote: > You lost me on that one. Care to explain? > > Shel > > > >> [Original Message] >> From: Godfrey DiGiorgi > >> they also reposition the directory >> tables as time goes on if the >> media is used a lot, again to preserve >> and extend the media's lifespan. > > > > -- > PDML Pentax-Discuss Mail List > [email protected] > http://pdml.net/mailman/listinfo/pdml_pdml.net -- PDML Pentax-Discuss Mail List [email protected] http://pdml.net/mailman/listinfo/pdml_pdml.net
