Re: How do I add yaffs file system to mainline?

[Charles Manning <manningc2@actrix.gen.nz>]

On Friday 29 October 2010 07:08:45 Greg KH wrote:
> On Fri, Oct 29, 2010 at 06:55:39AM +1300, Charles Manning wrote:
> > On Friday 29 October 2010 06:44:32 Greg KH wrote:
> > > On Thu, Oct 28, 2010 at 10:26:41AM -0700, Randy Dunlap wrote:
> > > > On Fri, 29 Oct 2010 04:55:02 +1300 Charles Manning wrote:
> > > > > YAFFS has been used for many years as a third-party patch-in.
> > > > >
> > > > > I have recently been through the exercise of changing all the
> > > > > symbols to be more kernel friendly with the intention of mainlining
> > > > > into the linux tree.
> > > > >
> > > > > The code is in git at
> > > > > http://github.com/cdhmanning/linux-yaffs-integration/
> > > >
> > > > It's difficult to review & comment on a git tree.
> > > > We prefer patches via email for review.
> > > >
> > > > > Thanks to CELF and Google for sponsoring the effort so far.
> > > > >
> > > > > What still needs to be done to mainline this?
> > > > > Who do I need to approach?
> > > >
> > > > Either ask Stephen Rothwell to add the git tree to the linux-next
> > > > daily tree or ask Greg KH to add it to the drivers/staging/ area.
> > >
> > > I'd be glad to add the code to the staging tree, but to do so, do you
> > > have a list of things that are left to do in order to get it properly
> > > merged to the "real" portion of the kernel?
> >
> > We're getting into a Catch-22 discussion here... Until I fully understand
> > what is needed, I can't really say what needs to be done :-).
> >
> > At this stage the code works, is integrated into, and compiles cleanly
> > against Linus' 2.6 git. My git is based on Linus' 2.6 as of yesterday.
> >
> > I have done symbol cleaning changing the old yaffs names of the form
> > yaffs_ScanBackwards() to yaffs_scan_backward() etc.
> >
> > What I really need is someone to look at what's there and tell me if
> > there are still style issues etc that need changing.
> >
> > I know there is a style guide etc, but when we're talking about
> > processing 15k loc then some of those rules might be slightly bendable.
>
> Not really.
>
> Have you run it through sparse and scripts/checkpatch.pl?  If not,
> please do so and resolve all of the issues that they bring up.
>
> Then break up the filesystem into reviewable patches and send them to
> the linux-fsdevel list for review.
>
> Again, read Documentation/SubmittingPatches for the details on how to
> properly do this, it is described in very good detail :)
>
> > > And if so, what is preventing you from doing those tasks right now to
> > > get the code into the .38 kernel merge?
> >
> > My biggest problem is not fully understanding the process.
>
> The process is very well documented, have you read the documentation?
> If so, what part is lacking?

Thanks to all for the input. I'll be re-reading the docs and running the 
various scripts etc. That will probably keep me busy for a bit.

The one thing I'm still blurry on is whether it is OK to send a whole fs of 
15kloc as a single patch to all. SubmitPatch seems to be written from the 
perspective of writing smaller patches.I guess checkpatch.pl will tell 
me :-).

-- Charles