infradead.org vs kernel.org sources

infradead.org vs kernel.org sources

[Kent Borg]

Hello, a newbie here, and I thought I was making good progress
(debugging why I couldn't talk to our flash at all), and then I got
confused over sources.

I am working on an embedded PPC 405 controller.  We started with 2.4.2
sources from mvista.com, which seem quite close to kernel.org.  And
the main MTD web page <http://www.linux-mtd.infradead.org/> says "Very
occasionally, I make snapshot releases. Now that the MTD code is in
the 2.4 kernel, it's become even rarer. ", which I took to mean that
the kernel.org sources were up to date.

But now I had occasion to look at the stuff I grabbed from the
infradead cvs repository.  It looks rather different.

For example, mtd/drivers/mtd in the infradead sources seems to have
files I find in linux/drivers/mtd in kernel.org sources, but the
infradead sources are more elaborate and have sub-directories.

I guess my next move is to drop that infradead sub-tree into the
linux/drivers/mtd directory and see what happens.  (I think we need to
directly link, not be a loadable module, so the sources need to be
merged.)

Is this what I want?


Thanks,

-kb, the Kent who hopes to get JFFS working on top of MTD.

RE: infradead.org vs kernel.org sources

[Vipin Malik]

>But now I had occasion to look at the stuff I grabbed from the
>infradead cvs repository.  It looks rather different.

AFAIK latest stuff is in the CVS repo. IF you are comparing with the stuff
in the release kernels (is that what you meant by kernel.org?) then you
should be working off the CVS.

>I guess my next move is to drop that infradead sub-tree into the
>linux/drivers/mtd directory and see what happens.  (I think we need to
>directly link, not be a loadable module, so the sources need to be
>merged.)

There is no diff in doing modules vs "merged in" as far as how you link in
the CVS code into the kernel. You can do both (rather either).

>Is this what I want?

Sounds like it. Is it? :)

>-kb, the Kent who hopes to get JFFS working on top of MTD.
Why? Why not JFFS2? It's more stable and full featured (and not to mention
next gen from JFFS). As long as you are using 2.4.x as JFFS2 does not yet
work with 2.2.x

Vipin

Re: infradead.org vs kernel.org sources

[Kent Borg]

On Tue, Jul 31, 2001 at 05:05:33PM -0500, Vipin Malik wrote:
> AFAIK latest stuff is in the CVS repo. IF you are comparing with the stuff
> in the release kernels (is that what you meant by kernel.org?) then you
> should be working off the CVS.

Cool.  

> >-kb, the Kent who hopes to get JFFS working on top of MTD.
> Why? Why not JFFS2? It's more stable and full featured (and not to mention
> next gen from JFFS). As long as you are using 2.4.x as JFFS2 does not yet
> work with 2.2.x

Oops.  I was being sloppy and using "JFFS" as a kinda generic term.
Certainly I expect to go for JFFS2.


Thanks a bunch,

-kb, the Kent who has hacked at flash before, but who is new at these
specifics.

Building MTD with JFFS2

[Kent Borg]

On Tue, Jul 31, 2001 at 05:54:42PM -0400, I wrote:
> Hello, a newbie here, and I thought I was making good progress
> (debugging why I couldn't talk to our flash at all), and then I got
> confused over sources.

And now I am getting closer.

Following e-mailed advice of Abraham vd Merwe <abraham@2d3d.co.za> I
used the infradead.org sources.  He sent me a script that copied the
right cvs-ed files into the right places in the regular 2.4 kernel
tree, but that left me with no configuration mention of JFFS2 and a
compile error in the old JFFS directory.

Resorting to a little RTFM I looked at the mtd-jffs-HOWTO and followed
its approach ("sh patchin.sh /usr/src/linux"), and got the same
results: no mention of JFFS2 in the configuring and the same compile
error in linux/fs/jffs (inode-v23.c, line 93, no member named
s_maxbytes).

Looking at the mtd/patches/Configure.help I see mention of
CONFIG_JFFS2_FS and CONFIG_JFFS2_FS_DEBUG, I tried manually defining
them in autoconf.h as 1 and 3 respectively, and in linux/fs/Makefile I
changed the instances of jffs and JFFS to jffs2 and JFFS2 respectively.

This made my compile happy, but, though the make dep seems to have
visited the linux/fs/jffs directory, the make itself didn't.

Hmmm.  Ideas on how to get out of the newbie mud?


In the mean time, because we don't have CFI chips, I think I will look
into whether I need to write my own mtd driver.  (AMD and Intel chips,
but old ones.)


Thanks,

-kb, the Kent who is glad there are those who have been here before
him.

Re: Building MTD with JFFS2

[David Woodhouse]

kentborg@borg.org said:
>  This made my compile happy, but, though the make dep seems to have
> visited the linux/fs/jffs directory, the make itself didn't.

> Hmmm.  Ideas on how to get out of the newbie mud? 

Look at the 2.4-ac patches. They have all the necessary glue you need to 
add the fs/jffs2 directory to the build.


kentborg@borg.org said:
>  In the mean time, because we don't have CFI chips, I think I will
> look into whether I need to write my own mtd driver.  (AMD and Intel
> chips, but old ones.) 

We have drivers for them, but just can't necessarily recognise them. Rather 
than doing a whole new driver, you need to just write a probe routine which 
passes control to the cfi_cmdset_000x drivers like cfi_probe.c does.

The actual commands are the same - CFI only defines how you probe the chips
to identify them.

--
dwmw2