asanochkin at Lnxw.COM
Fri Dec 7 08:11:11 EST 2001
> > Yes, I've looked through Jörn Engel's code. Our modifications are less
> > significant and are based on other goals in mind.
> > Primary goal was to allow for creating Flash partitions that consist
> > non-consequent sectors. For instance, the user may wish to chose
> > 2 for storing some binary data and use sectors 1,3,4..... as a single
> > JFFS partition.
> > Another goal was a possibility of flexible (re-)configration of the
> > partitions, without having to change any C code. Supported methods
> > build-time (via a CONFIG_... define), boot-time (via a kernel command
> > line), and run-time (via an mtdchar IOCTL command). With all the three
> > methods, the user uses a string notation to specify desired layout of
> > partitions. For instance, "0,3-8:1,2:9-25" means three partitions,
> > sectors 0 and 3 to 8 belonging to the first partition, 1 and 2 to the
> > second, and 9 to 25 to the third.
> We already have partitioning that is independent of the kernels normal
> partitioning. Now you send in the equivalent of the logical volume
> manager. The mtd code is getting there. :)
> What are your reasons to use mtdchar ioctl()s for runtime
> repartitioning? This should be pretty independent of the char device,
> (caching) block device or whatever the future might still bring. Why
> not a /proc interface?
One reason was that the user will most likely have mtdchar in the kernel
anyway, if he wants to use a tool for erasing Flash, but he may want
not to include /proc in the kernel for some reasons. So, we found it
more logical to have an IOCTL command for the repartitioning.
More information about the linux-mtd