Oh great, not yet-another-kernel-tree, just what the world needs...

Yes, this is an announcement of a new kernel tree, linux-staging.  It is
a quilt series of patches that can be found at:
	git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/staging.git


In a long and meandering thread with some of the other kernel developers
a week or so ago, it came up that there is no single place for companies
and developers to put their code for testing while it gets cleaned up
for submission into the kernel tree.  All of the different subsystems
have trees, but they generally only want code that is about to go into
this release, or the next one.  For stuff that is farther off, there is
no place to go.

So, here's the tree for it.  From the README:

PURPOSE

The linux-staging tree was created to hold drivers and filesystems and
other semi-major additions to the Linux kernel that are not ready to be
merged at this point in time.  It is here for companies and authors to
get a wider range of testing, and to allow for other members of the
community to help with the development of these features for the
eventual inclusion into the main kernel tree.

This tree will be included in the daily linux-next builds, and will get
testing by all users of that tree.

The rules of what can be included here is as follows:
	- the code must be released under a Linux kernel-compatible
	  license
	- the goal of the developers must be to merge this code into the
	  main kernel tree in the near future, but not for the next
	  kernel release.
	- the code must build properly on the x86 platform
	- this is not a tree for bugfixes or rewrites of existing kernel
	  code, this should be for new features, drivers, and
	  filesystems.
	- the patches included must detail exactly what is needed to be
	  completed in order for them to be included into the main
	  kernel tree.
	- there must be some email address associated with the patch
	  that can be used for bug reporting and questions about
	  cleanups and ...
[ message continues ]

Here is a list of the current patches in the linux-staging tree, and
their current status.

 - 17 patches
 - 192 files changed, 131073 insertions(+), 651 deletions(-)

---------
Patch: aectc: add the aectc driver
Contact: Brandon Philips <brandon@ifup.org>

Status:
  Being reworked to use the UIO interface instead of char driver
  Part of the linuxdriverproject.org development effort.

---------
Patch: DVB: add firesat driver
Contact: Manu Abraham <manu@linuxtv.org> and Greg KH <greg@kroah.com>

Status:
  Currently waiting for more DVB tuner development to finish off the
  remaining features.  Will be submitted to the DVB developers for
  inclusion in the kernel tree soon, as waiting for that effort might be
  a while.
  Part of the linuxdriverproject.org development effort.

---------
Patch: Framebuffer: add the VIA framebuffer driver
Contact: Benjamin Pan <BenjaminPan@viatech.com> and Greg Kroah-Hartman <greg@kroah.com>

Status:
  many TODOs still left to finish (see patch for details)
  Part of the linuxdriverproject.org development effort.

---------
Patch: Input: add appleir driver
Contact: Greg Kroah-Hartman <greg@kroah.com>

Status:
  looks like this is not needed at all, but still need to determine the
  proper HID quirk.
  Part of the linuxdriverproject.org development effort.

---------
Patch: me4000: add pci data collection driver
Patch: me4000 firmware file
Contact: Greg Kroah-Hartman <greg@kroah.com>

Status:
  many TODOs still left to finish (see patch for details)
  Part of the linuxdriverproject.org development effort.

---------
Patch: NET: add Alacritech slicoss driver
Contact: Greg Kroah-Hartman <greg@kroah.com>

Status:
  many TODOs still left to finish (see patch for details)
  Part of the linuxdriverproject.org development effort.

---------
Patch: NET: add et131x driver
Contact: Greg Kroah-Hartman <greg@kroah.com>

Status:
  many TODOs still left to finish (see patch for details)
  Part of the linuxdriverproject.org ...
[ message continues ]

Does this mean that the nature of linux-next is changing?  I thought
the whole point of linux-next was only to have what would be pushed to
Linus in the near future, so we could check for patch compatibility
issues.  For that reason, for example, I don't push the unstable set
of patches in the ext4 tree to linux-next, since they aren't ready for
merging yet in their current form.  

But if linux-staging is going to be pushed to linux-next, doesn't that
violate the ground rules of Linux-next?  Or are we allowing in this
case because these are filesystems and/or device drivers that don't
exist at all in the mainline tree yet?

						- Ted

--

[ message continues ]

I'm asking for the rule to be bent for this tree, not that the whole
nature of linux-next is changing.

-staging is for only whole new drivers/filesystems, not changes/features
to existing code that is not yet ready for merging.  The main reason
these drivers are not in mainline is usually:
	- coding style issues
	- sparse cleanups needed
	- ioctl 32/64 cleanups
	- locking review
	- direct access to hardware through memory pointers (only works
	  on x86)

If you look at what I currently have, there's nothing earth-shattering
there, but there is stuff that users can use to get hardware to work
that currently is not supported on kernel.org kernels at all.

It would be nice if distros also pick it up if they want to support
these devices and give me some feedback.  There are 2 big network
drivers in there that support a wide range of devices that some people
would like to see working :)

thanks,

greg k-h
--

[ message continues ]

hi,


That's great. I started working on the test and measurement class
driver. Do you want me to send preliminary patches or just the final
one ?

-- 
Best Regards,

Felipe Balbi
felipebalbi@users.sourceforge.net
--

[ message continues ]

I'll take whatever you have, prelim patches is nice if you are at a
stopping point and want to make sure I get them applied :)

thanks,

greg k-h
--

[ message continues ]

Hi.

Would you consider including TuxOnIce in it?

I do still want to get it merged and would appreciate feedback.

Regards,

Nigel

--

[ message continues ]

Is the patch "stand-alone", only adding new code in discrete chunks like
a new driver or filesystem would?

If not, I don't think it is relevant.  Odds are you want to be your own
series of patches, like we discussed years ago, right?

thanks,

greg k-h
--

[ message continues ]

Hi Greg.


The patch I distribute now does have a few parts to it that could be
separated into distinct patches (cryptoapi LZF support, fuse freezer
support), but the bulk of it is TuxOnIce itself, which just adds new
files and inserts the hooks necessary to share the lowlevel code with
[u]swsusp. I think, therefore, it would akin to adding a new driver or

I don't think I do want to have my own series of patches, because
TuxOnIce doesn't remove or rework swsusp or uswsusp, but sits along side
them. I'm not trying to mutate swsusp into TuxOnIce, because that would
require a complete rework of swsusp from the ground up (TuxOnIce does
everything but the atomic copy/restore and associated prep/cleanup
differently).

Regards,

Nigel

--

[ message continues ]

Like always, you need to divide your changes up into logical chunks in
order to get them approved and reviewed.  For such a core functionality
like suspend, this is extra important.

I do not think that -staging is proper for this kind of feature at this
point in time.

thanks,

greg k-h
--

[ message continues ]

Hi again.


Just another thought - I should have said that I don't really know what
patches to add a new driver or filesystem look like. I can only imagine
a filesystem going in all-in-one-patch, so I'd probably appreciate an
example/template to work off, as might others.

Regards,

Nigel

--

[ message continues ]

If you look in the -staging tree, there are 17 "template/example"
patches in there to go off of :)

Basically, one patch per feature.  It sounds like the tuxonice is much
too big for this kind of thing.

Just respin your patches and post them to lkml for review.  You should
be able to set up your own tree for inclusion in linux-next if they are
ready to go in.

thanks,

greg k-h
--

[ message continues ]

Greg,

The OSD initiator (see
http://git.open-osd.org/gitweb.cgi?p=open-osd.git;a=summary)
looks in principle like a great fit for linux-staging.

What would be the mechanics of including it?

Benny


--

[ message continues ]

The patches are not yet there, Benny. They are currently 
out-of-tree here:
http://git.open-osd.org/gitweb.cgi?p=osd-lib.git;a=summary

It will take me until end of next week to separate them into

I want to send them to linux-scsi-ml and also To: Greg KH
Requesting to be included in "linux-staging tree". Is that

<snip>

--

[ message continues ]

Hi,


just don't forget to add copyright notes in all files :-)--
Best Regards,

Felipe Balbi
felipebalbi@users.sourceforge.net
--

[ message continues ]

If they are to go into linux-scsi, why would you need/want them in the
-staging tree as well?

thanks,

greg k-h
--

[ message continues ]

We're jumping the guns here a bit, but ...

The code is pretty stable and robust as far as the protocol and performance
is concerned, surly once I in-tree them, divide them into patches, prettify
comments, add file-headers, and checkpatch them.

But the bigger implications are not yet clear, and will need advise from
the list, which could take time. Mainly in regard to upper ULD and tests.
Currently there are 4 users for this code:
- pNFS-over-objects layout driver
- pNFS-over-objects Simple Server implementation (spNFS)
- OSDVFS - Virtual psuedo file system to access and debug OSD luns from user mode.
          (This is not the OSDFS from IBM which is a general FS over objects,
           but a direct representation of the OSD Lun to user mode)
- Testing

The first 3 are their own ULD and do not need a proper SCSI-ULD. With some
changes to sg.c they can manage with what we have now. The later is just
for debugging.

On the other hand a true OSD-ULD that exports /dev/osdx char and/or block
devices, has merits and future directions of it's own. And will eliminate
the need for changes in sg.c

So the mechanics are pretty much there but the direction is not clear, which
will govern the folders/exports/dependencies. But I'm not sure -staging tree
will help in any of that. I do have my git.open-osd.org exports and can manage

Thanks
Boaz
--

[ message continues ]

Why isn't that just going directly into the linux-scsi tree when ready?

It doesn't look like it has coding style issues, does it?  What is

Just send me a patch :)

thanks,

greg k-h
--

[ message continues ]

Would the linux-staging tree be an appropriate place to merge a new
architecture? Or would that be too large a change and should go via
its own tree?
--

[ message continues ]

That is probably too big to go into -staging and should deserve its own
tree based on the size of patches that have gone into creating a new
architecture in the past.

Do you have an example of one that is currently not included in the main
kernel tree right now?

thanks,

greg k-h
--

[ message continues ]

I have a tree that is currently about half way to public consumption,
although I don't think there's any point pushing it until we have a
--

[ message continues ]

Yeah, that would help as well :)

thanks,

greg k-h
--

[ message continues ]

This is great. To let this be useful for wireless we'll need
wireless-testing.git merged as we rely on it for the latest and

We have a few drivers which are not yet ready for
wireless-testing.git. Airgo is one.

Anyway, good stuff. Let me know what you think about letting this work
for wireless too.

  Luis
--

[ message continues ]

I don't think directly pulling wireless-testing is a good idea of
-staging for process-related reasons.  However I think we can arrange
to send something for the new tree.  at76, airgo, and mrv8k are all
decent candidates.

Greg, how does the -staging tree get managed?  Is it periodically
rebuilt (like -next)?  Or would I need to send you pull requests?

John
-- 
John W. Linville
linville@tuxdriver.com
--

[ message continues ]

On Fri, Jun 13, 2008 at 7:33 AM, John W. Linville

Its just quit patches.

  Luis
--

[ message continues ]

Just send me patches, it's a quilt tree.

thanks,

greg k-h
--

[ message continues ]

Hm, why?  Is it really needed for you to have this pulled in?  What is

If you have individual drivers, I'd be glad to take them into the tree
if you think it will help out.  Just send me a patch that can build
properly.

thanks,

greg k-h
--

[ message continues ]

Hmm, how do I tap into that developer pool? :-)

Kohjinsha wifi driver is in state where it was ported to mac80211, and
it still works with minimum features, but it needs dead code removal
and rewriting the rest. That code is really very ugly.

I can do the handholding; I can probably do the rewrite, too (its
10000LoC), but it will take time...

							Pavel
-- 
(english) http://www.livejournal.com/~pavelmachek
(cesky, pictures) http://atrey.karlin.mff.cuni.cz/~pavel/picture/horses/blog.html
--

[ message continues ]

If the patch at least builds, send it to me for inclusion in -staging if
you want to get some exposure for it.

thanks,

greg k-h
--

[ message continues ]

Hi Greg,

Is there a separate mailing-list for staging?

-- 
Thanks and Regards,
Jike

--

[ message continues ]

Nope, lkml is fine for now, just be sure to cc: me so I notice it.

thanks,

greg k-h
--

[ message continues ]