Re: nand flash
"Toru Nishimura" <[email protected]> Fri, 5 Oct 2007 11:00:12 +0900
| Newsgroups | gmane.os.netbsd.devel.embedded |
|---|---|
| Message-ID | <000301c806f3$7a63ce80$1718a8c0@tpad3> |
David Young asked; >> but the decision was made weighing "the rootfs is >> seldom written beyond configuration mods by administrators." Software ECC >> is always calc'ed and verified, block write is counted and recorded inside >> NAND spare fields. > > How do you handle the bad blocks that develop as the device "wears" ? The answer is "by manual intervention," just like good-old-days Winchester style hard disk drive bad sector management practice. I wrote a cmdline tool to manage bad block list and relocate block contents. In principle we neglect the whole NAND woe/buzz to happen due to the above mentioned rational. "this tools may cause hazardous results to hurt you, trained service personel only, or don't use this." If I have chose MMC/SD/CF media to go, I would build vnd on a large file on the FAT filesys and forget how good/bad to gauge badblock and wear-leveling schemes inside. Toru Nishimura/ALKYL Technology