MTD partitioning rewrite

MTD partitioning rewrite

[Jörn Engel]

Moin!

I have currently started to do a rewrite of the MTD partitioning and
the project might even grow into a complete rewrite of the central MTD
code. We will see, how much is needed...

The work will break some current code and should _not_ be included in
the cvs for a while. I will maintain my own codebase for a while and
feed the changes to David when they mature.

Here are some quick pointers at my design goals:
- An MTD device will be the equivalent of a hard drive. There will be
several, named mtda, mtdb, mtdc,...
- Each MTD device can be partitioned into mtda0 (whole device), mtda1,
mtda2,...
- The ro-devices don't seem to be necessary. If noone has strong
objections, I will just discard those.
- The partitioning should stay out of the mapping drivers as much as
possible. The mapping driver only supplies a pointer to the partition
table parser. When the device is registered, the partitioning will
automatically be read.
- The physmap driver, being the most general one, should be
copied/changed to map several devices.
- Big goal: When given a new piece of hardware, I only want to change
the kernel configuration, compile, flash and be done with it. Any
piece of hardware that needs source code tweaks is a proof that we are
not done yet.

If anyone has questions, comments, code sniplets, ideas, anything, you
can contact me privately on through the list. I will only try to do
the best, _I_ can do and hope, David approves it.

Jörn

-- 
Often, the most striking and innovative solutions come 
from realizing that your concept of the problem was wrong.

RE: MTD partitioning rewrite

[Jonas Holmberg]

> Here are some quick pointers at my design goals:
> - An MTD device will be the equivalent of a hard drive. There will be
> several, named mtda, mtdb, mtdc,...
> - Each MTD device can be partitioned into mtda0 (whole device), mtda1,
> mtda2,...

Would it still be possible to make a logical partition that extends over
several flash chips (e.g. one big partition of two chips)?

/Jonas

Re: MTD partitioning rewrite

[David Woodhouse]

jonas.holmberg@axis.com said:
>  Would it still be possible to make a logical partition that extends
> over several flash chips (e.g. one big partition of two chips)?

As possible as it is now - the chip probe code has to detect them all in 
one go and make a single MTD device out of them. Otherwise, you have to 
hack up your own wrapper anyway. 

--
dwmw2

Re: MTD partitioning rewrite

[Jörn Engel]

> Here are some quick pointers at my design goals:
> [...]

I forgot one:
- Devices may be assigned numbers. Device 0 will become mtda if that
one is still free or the first one that is free. Device 1 will become
mtdb,...

Jörn

-- 
Open your arms to change, but don't let go
of your values.

Re: MTD partitioning rewrite

[Jörn Engel]

Moin!

> I have currently started to do a rewrite of the MTD partitioning and
> the project might even grow into a complete rewrite of the central MTD
> code. We will see, how much is needed...

The most central work appears to be finished. This is still unusable
for any practical purpose, but maybe I am lucky and find another
developer who has a look at the code and can show me some bug, give a
helpful hint or the like.
So if any of you want to try something really new and really broken,
download the tarball from
http://wh.fh-wedel.de/~joern/software/kernel/mtd-0.0.2.tar.gz

To install it, simply unpack it into a linux source tree, recompile
and - pray.

I will try to provide new versions nearly daily and might have a
running one sometime next week.

Jörn

-- 
Open your arms to change, but don't let go
of your values.

Re: MTD partitioning rewrite

[Jörn Engel]

Hi!

> To install it, simply unpack it into a linux source tree, recompile
> and - pray.

http://wh.fh-wedel.de/~joern/software/kernel/mtd-0.0.4.tar.gz

New version, less FIXME's less TODO's. It is now running on my
hardware with the rw block driver. Next will be a new physmap driver,
then char, then block_ro.

Here is some output, if anyone is interested.

root:~ > ls -l /dev/root 
lr-xr-xr-x   1 root     root           12 Jan  1 00:00 /dev/root ->
mtdblock/0/2
root:~ > ls -l /dev/mtdblock/  
total 0
drwxr-xr-x   1 root     root            0 Jan  1 00:00 0
root:~ > ls -l /dev/mtdblock/0/
total 0
brw-rw-rw-   1 root     root      31,   0 Jan  1 00:00 0
brw-rw-rw-   1 root     root      31,   1 Jan  1 00:00 1
brw-rw-rw-   1 root     root      31,   2 Jan  1 00:00 2
brw-rw-rw-   1 root     root      31,   3 Jan  1 00:00 3
brw-rw-rw-   1 root     root      31,   4 Jan  1 00:00 4
brw-rw-rw-   1 root     root      31,   5 Jan  1 00:00 5

Jörn

-- 
A defeated army first battles and then seeks victory.
	Sun Tzu

Re: MTD partitioning rewrite

[Jörn Engel]

Hi!

> New version, less FIXME's less TODO's. It is now running on my
> hardware with the rw block driver. Next will be a new physmap
> driver,
> then char, then block_ro.

http://wh.fh-wedel.de/~joern/software/kernel/mtd-0.0.6.tar.gz

Two people have downloaded the last version, but noone told me that I
forgot to include mtdblock.c. Well...

Now the char driver and the procfs file are working, too. The physmap
driver *might* be working with multiple devices, it *is* working with
only one. Since this might be a problem with either my code or my
hardware, this is postponed for a while.

I haven't done any tests without devfs support, yet. That should work,
but I haven't even proven it correct.

Jörn

-- 
Beware of bugs in the above code; I have only proved it correct,
but not tried it.
  -- Donald Knuth