Re: Handling of suspend/hibernation patches

[Andrew Morton <akpm@linux-foundation.org>]

On Thu, 3 Jul 2008 22:47:32 +0200
"Rafael J. Wysocki" <rjw@sisk.pl> wrote:

> Hi,
> 
> Recently, we've had some problems with the handling of suspend/hibernation
> patches, because they tend to touch multiple subsystems at a time.  As a
> result, it usually is not clear which tree they should be included in and at
> the moment there are suspend/hibernation patches in the PCI, ACPI, x86
> trees, as well as in -mm.  Of course the resulting dependencies between those
> trees are a pain to Stephen and their maintainers.
> 
> After the last Kernel Summit we tried to create a branch in the ACPI tree for
> suspend/hibernation patches (thanks Len!) and that had worked quite well until
> we started to work on the core device power management.  This coincided with
> the creation of linux-next and collecting suspend/hibernation patches in the
> ACPI tree became inconvenient.
> 
> Now, we could go back to the old way of handling suspend/hibernation patches,
> which was to merge them through -mm, but the disadvantage of this would be that
> the patches wouldn't go through linux-next.  However, I'd like
> suspend/hibernation patches to be included in linux-next, so that they can get
> as much testing as possible.

I need to get butt into gear and get most-of-mm into linux-next.  I'll
be picking that up when 2.6.27-rc1 is done.

So we can just keep this tree in -mm (and hence linux-next) if you like.

> For this reason I'd like to create a quilt-based series of suspend/hibernation
> (PM sleep) patches, that Stephen will be able to pull from, with the following
> rules:
> * the patch series will be based on linux-next and will be rebased as often
>   as necessary
> * the patch series will be periodically (most probably once a week) posted
>   to linux-pm and linux-acpi; additionally the maintainers of the subsystems
>   affected by the patches will be sent CCs of the messages with them, so that
>   they can pick those patches up
> * patches taken into the other subsystem trees will be dropped from the
>   patchset and the resulting dependencies will be reflected by the annotations
>   in the 'series' file
> * the patches that are still in the series when the merge window opens, but
>   are accepted by the appropriate maintainers, will be put into a git branch
>   and pushed to Linus towards the end of the merge window.
> 
> If you agree to this plan, I'd like to start by creating the patch series out
> of the suspend/hibernation patches that are in -mm at the moment and one
> additional patch that was accepted by Len, but not added to the ACPI tree
> because of some dependencies with the other subsystem trees.