Re: [PATCH 2.6.24] block2mtd: removing a device and typo fixes

["Jörn Engel" <joern@logfs.org>]

On Wed, 20 February 2008 17:30:53 +0000, Stephane Chazelas wrote:
> > 
> > Actually, there is /dev/mtd<x>.  Enable CONFIG_MTD_CHAR.
> 
> Yes, my point ;). "block2mtd" creates a "mtd" out of a block
> device and "mtdchar" and "mtdblock" create the "char" and
> "block" devices out of the "mtd". This is a different concept
> from "loop". With "loop", you make a block device out of a file,
> but you do the ioctl on the target loop block device itself.
> With block2mtd, you can't do that.

The fact is that there is _no_ interface to manage mtd.  Part of the
reason is this tradition to have seperate modules every couple of lines
of code.  In other subsystems, CONFIG_MTD_CHAR would not exist and the
code always be compiled in.  Simply to get _some_ interface to handle
the device that is always there.

Without going much deeper into that discussion, I consider it acceptable
to depend on CONFIG_MTD_CHAR for things like device removal.  If you
want device removal and explicitly don't want the extra code from
mtdchar.c, send a patch or go find 500 bytes to save elsewhere. ;)

With that out of the way, the question remains which interface we should
have.  Independently of mimicking loop, I would like to have a generic
interface for all mtd, not just block2mtd.  Whether it is "echo 1 >
/sys/.../mtd/mtdN/remove" or an ioctl(), I don't care much about.

> Actually, that's what I use block2mtd for, in combination with
> "loop" to mount jffs2 filesystem images (always wondered if
> there wasn't a simpler way, BTW (other than mtdram))

Logfs can use a block device directly, which still leaves loop.
Enhancing block2mtd to work with files shouldn't be too hard.  If anyone
want a fun project to hack on...

Jörn

-- 
Data dominates. If you've chosen the right data structures and organized
things well, the algorithms will almost always be self-evident. Data
structures, not algorithms, are central to programming.
-- Rob Pike