Re: Doubts about File Systems on FIASCO on ARM (Cortex-A15)
Maria Soler <[email protected]>
| Newsgroups | gmane.comp.micro-kernel.l4.devel |
|---|---|
| Message-ID | <CALxX4v--AW=ytSGGFGcRKh+=B2w8KiNDUU+KD_fLkeDNCj5oOQ@mail.gmail.com> |
Hi again, Thanks to all for your answers. After further investigation I don't seem to actually need disk access/a file system by now, so I'll leave this thread of work for the moment. Thanks again, María. On Fri, Jun 20, 2014 at 12:02 PM, Martin Schröder <[email protected]> wrote: > Am 19.06.2014 18:06, schrieb Jorge Ventura: > >> This is a good direction. I forgot that, in reality we need two drivers, >> one for a block device and another one for the filesystem itself. >> > > The two drivers are not sufficient if the underlying hardware is some sort > of bare flash devices (NOR, NAND) and you use a traditional file system. The > number of erase cycles of an individual block is limited. So writes of > frequently written blocks (directory, file allocation table) need to be > evenly spread over the device to avoid a premature death of vital blocks. > This is the so called wear levelling which is an additional driver in > between the block device driver and the file system driver. Or you use a > file systems which incorporates wear levelling (JFFS2, UBIFS). > > The two drivers are sufficient only if the underlying hardware is an > intelligent device like a hard disk, SSD, SD card or USB Stick. Such devices > have their own controller performing wear levelling itself. > > > Martin > > > _______________________________________________ > l4-hackers mailing list > [email protected] > http://os.inf.tu-dresden.de/mailman/listinfo/l4-hackers _______________________________________________ l4-hackers mailing list [email protected] http://os.inf.tu-dresden.de/mailman/listinfo/l4-hackers