mbox

[GIT,PULL,3/4] ARM: mvebu: boards for v3.9

Message ID 1359733919.c3D01c3.22262@triton
State New
Headers show

Pull-request

git://git.infradead.org/users/jcooper/linux.git tags/boards_for_v3.9

Message

Jason Cooper Feb. 1, 2013, 3:51 p.m. UTC
The following changes since commit 949db153b6466c6f7cad5a427ecea94985927311:

  Linux 3.8-rc5

are available in the git repository at:

  git://git.infradead.org/users/jcooper/linux.git tags/boards_for_v3.9

for you to fetch changes up to c7064904895f69b2e33117b450a05746f75abf3a:

  ARM: kirkwood: convert Guruplug Server Plus to use the device tree

----------------------------------------------------------------
mvebu boards for v3.9
     - Guruplug Server Plus DT board
     - RD-A370-A1 board
     - mvebu improved SMP support in interrupt controller
     - update defconfigs
     - split legacy and DT setup for dove
     - remove some redundant clock aliases
    
    depends on:
     - tags/mvebu_fixes_for_v3.8-rc6
     - tags/cleanup_for_v3.9_round2
     - tags/drivers_for_v3.9
     - mmc/mmc-next up to:
        d293875 mmc: mvsdio: add pinctrl integration
    
    conflicts:
     - arm-soc/for-next
        - arch/arm/mach-dove/common.c
           * remove all DT code (it's now in .../board-dt.c)
    
    build fixes (not merge conflict)
     - arm-soc/for-next
        - arch/arm/mach-dove/board-dt.c
           - .timer no longer used
           * s/.timer = &dove_timer/.init_time = dove_timer_init/ (not valid sed)
----------------------------------------------------------------
Andrew Lunn (2):
      ARM: Kirkwood: Remove redundent USB clock alias
      ARM: Kirkwood: Remove redundent SDIO clock alias

Florian Fainelli (1):
      arm: mvebu: add DTS file for Marvell RD-A370-A1 board

Gregory CLEMENT (2):
      arm: mvebu: Update defconfig with Marvell RTC support
      arm: mvebu: Improve the SMP support of the interrupt controller

Jason Cooper (4):
      Merge tag 'tags/mvebu_fixes_for_v3.8-rc6' into mvebu/boards
      Merge tag 'tags/cleanup_for_v3.9_round2' into mvebu/boards
      Merge tag 'tags/drivers_for_v3.9' into mvebu/boards
      Merge commit 'd293875' into mvebu/boards

Olof Johansson (1):
      ARM: dove: update dove_defconfig with a few useful options

Sebastian Hesselbarth (1):
      ARM: Dove: split legacy and DT setup

Thomas Petazzoni (4):
      arm: mvebu: enable SDIO support in mvebu_defconfig
      arm: mvebu: enable mwifiex driver in mvebu_defconfig
      arm: mvebu: enable btmrvl driver in mvebu_defconfig
      arm: mvebu: add LEDs support to defconfig file

Willy Tarreau (1):
      ARM: kirkwood: convert Guruplug Server Plus to use the device tree

 .../bindings/mmc/brcm,bcm2835-sdhci.txt         |  18 ++
 .../devicetree/bindings/mmc/orion-sdio.txt      |  17 ++
 arch/arm/boot/dts/Makefile                      |   2 +
 arch/arm/boot/dts/armada-370-rd.dts             |  61 ++++
 arch/arm/boot/dts/armada-xp.dtsi                |   2 +-
 arch/arm/boot/dts/dove.dtsi                     |   2 +
 .../boot/dts/kirkwood-guruplug-server-plus.dts  |  94 +++++++
 arch/arm/configs/dove_defconfig                 |  28 +-
 arch/arm/configs/kirkwood_defconfig             |   1 +
 arch/arm/configs/mvebu_defconfig                |  16 ++
 arch/arm/mach-dove/Kconfig                      |   5 +
 arch/arm/mach-dove/Makefile                     |   4 +-
 arch/arm/mach-dove/board-dt.c                   | 102 +++++++
 arch/arm/mach-dove/common.c                     | 114 +-------
 arch/arm/mach-kirkwood/Kconfig                  |   7 +
 arch/arm/mach-kirkwood/Makefile                 |   2 +-
 arch/arm/mach-kirkwood/board-dt.c               |  15 +-
 arch/arm/mach-kirkwood/board-guruplug.c         |  39 +++
 arch/arm/mach-kirkwood/board-ib62x0.c           |   1 -
 arch/arm/mach-kirkwood/board-mplcec4.c          |   1 -
 arch/arm/mach-kirkwood/board-nsa310.c           |   8 +-
 arch/arm/mach-kirkwood/common.c                 |  23 ++
 arch/arm/mach-kirkwood/common.h                 |   6 +
 arch/arm/mach-kirkwood/dockstar-setup.c         |   1 -
 arch/arm/mach-kirkwood/include/mach/kirkwood.h  |   3 +-
 arch/arm/mach-kirkwood/pcie.c                   |  10 +-
 arch/arm/mach-mvebu/irq-armada-370-xp.c         |  72 +++++
 arch/arm/plat-orion/mpp.c                       |   2 +-
 drivers/cpuidle/Kconfig                         |   6 +
 drivers/cpuidle/Makefile                        |   1 +
 .../cpuidle/cpuidle-kirkwood.c                  |  45 ++-
 drivers/mmc/card/block.c                        |  30 +-
 drivers/mmc/card/queue.c                        |  32 ++-
 drivers/mmc/card/queue.h                        |   3 +
 drivers/mmc/core/bus.c                          |   1 +
 drivers/mmc/core/core.c                         | 121 +++++++-
 drivers/mmc/core/core.h                         |   1 +
 drivers/mmc/core/sdio.c                         |  33 ++-
 drivers/mmc/core/slot-gpio.c                    |  57 +++-
 drivers/mmc/host/Kconfig                        |  11 +
 drivers/mmc/host/Makefile                       |   1 +
 drivers/mmc/host/mvsdio.c                       | 131 +++++----
 drivers/mmc/host/sdhci-bcm2835.c                | 227 +++++++++++++++
 drivers/mmc/host/sdhci-esdhc-imx.c              |  59 +---
 drivers/mmc/host/sdhci-pxav3.c                  |  12 +-
 drivers/mmc/host/sdhci.c                        | 117 +++++---
 drivers/mmc/host/sh_mmcif.c                     | 280 +++++++++++--------
 drivers/mmc/host/tmio_mmc_pio.c                 |   8 -
 drivers/rtc/Kconfig                             |   2 +-
 include/linux/mmc/card.h                        |  12 +
 include/linux/mmc/core.h                        |   3 +-
 include/linux/mmc/host.h                        |  25 ++
 52 files changed, 1381 insertions(+), 493 deletions(-)
 create mode 100644 Documentation/devicetree/bindings/mmc/brcm,bcm2835-sdhci.txt
 create mode 100644 Documentation/devicetree/bindings/mmc/orion-sdio.txt
 create mode 100644 arch/arm/boot/dts/armada-370-rd.dts
 create mode 100644 arch/arm/boot/dts/kirkwood-guruplug-server-plus.dts
 create mode 100644 arch/arm/mach-dove/board-dt.c
 create mode 100644 arch/arm/mach-kirkwood/board-guruplug.c
 rename arch/arm/mach-kirkwood/cpuidle.c => drivers/cpuidle/cpuidle-kirkwood.c (61%)
 create mode 100644 drivers/mmc/host/sdhci-bcm2835.c

Comments

Olof Johansson Feb. 5, 2013, 7:10 a.m. UTC | #1
Hi,

On Fri, Feb 01, 2013 at 03:51:59PM +0000, Jason Cooper wrote:
> The following changes since commit 949db153b6466c6f7cad5a427ecea94985927311:
> 
>   Linux 3.8-rc5
> 
> are available in the git repository at:
> 
>   git://git.infradead.org/users/jcooper/linux.git tags/boards_for_v3.9
> 
> for you to fetch changes up to c7064904895f69b2e33117b450a05746f75abf3a:
> 
>   ARM: kirkwood: convert Guruplug Server Plus to use the device tree
> 
> ----------------------------------------------------------------
> mvebu boards for v3.9
>      - Guruplug Server Plus DT board
>      - RD-A370-A1 board
>      - mvebu improved SMP support in interrupt controller
>      - update defconfigs
>      - split legacy and DT setup for dove
>      - remove some redundant clock aliases
>     
>     depends on:
>      - tags/mvebu_fixes_for_v3.8-rc6
>      - tags/cleanup_for_v3.9_round2
>      - tags/drivers_for_v3.9
>      - mmc/mmc-next up to:
>         d293875 mmc: mvsdio: add pinctrl integration

Hm. What do you need the mmc branch for? Is it just the removal of the
redundant SDIO clock alias?

If so, I'd say let's keep that as a cleanup until 3.10 instead of bringing this
in as a dependency. If it's something more substantial than that, then let me
know and we'll still pull this in.

Also, it's really hard to tell where code came from when you just merge in
a hash like below:

> Jason Cooper (4):
>       Merge tag 'tags/mvebu_fixes_for_v3.8-rc6' into mvebu/boards
>       Merge tag 'tags/cleanup_for_v3.9_round2' into mvebu/boards
>       Merge tag 'tags/drivers_for_v3.9' into mvebu/boards
>       Merge commit 'd293875' into mvebu/boards

That's the mmc-next branch at that commit, but it'd be nice to have a
more descriptive summary in the shortlog.


-Olof
Jason Cooper Feb. 5, 2013, 11:21 a.m. UTC | #2
On Mon, Feb 04, 2013 at 11:10:10PM -0800, Olof Johansson wrote:
> Hi,
> 
> On Fri, Feb 01, 2013 at 03:51:59PM +0000, Jason Cooper wrote:
> > The following changes since commit 949db153b6466c6f7cad5a427ecea94985927311:
> > 
> >   Linux 3.8-rc5
> > 
> > are available in the git repository at:
> > 
> >   git://git.infradead.org/users/jcooper/linux.git tags/boards_for_v3.9
> > 
> > for you to fetch changes up to c7064904895f69b2e33117b450a05746f75abf3a:
> > 
> >   ARM: kirkwood: convert Guruplug Server Plus to use the device tree
> > 
> > ----------------------------------------------------------------
> > mvebu boards for v3.9
> >      - Guruplug Server Plus DT board
> >      - RD-A370-A1 board
> >      - mvebu improved SMP support in interrupt controller
> >      - update defconfigs
> >      - split legacy and DT setup for dove
> >      - remove some redundant clock aliases
> >     
> >     depends on:
> >      - tags/mvebu_fixes_for_v3.8-rc6
> >      - tags/cleanup_for_v3.9_round2
> >      - tags/drivers_for_v3.9
> >      - mmc/mmc-next up to:
> >         d293875 mmc: mvsdio: add pinctrl integration
> 
> Hm. What do you need the mmc branch for? Is it just the removal of the
> redundant SDIO clock alias?

No, it's the four patches of Thomas' to the mvsdio driver:

d293875 mmc: mvsdio: add pinctrl integration
182ce21 mmc: mvsdio: implement a Device Tree binding
aa3738e mmc: mvsdio: use slot-gpio for card detect gpio
9d8b28e5 mmc: mvsdio: use slot-gpio infrastructure for write protect gpio

which mvebu/boards needs before we can activate defconfig options.

> Also, it's really hard to tell where code came from when you just merge in
> a hash like below:
> 
> > Jason Cooper (4):
> >       Merge tag 'tags/mvebu_fixes_for_v3.8-rc6' into mvebu/boards
> >       Merge tag 'tags/cleanup_for_v3.9_round2' into mvebu/boards
> >       Merge tag 'tags/drivers_for_v3.9' into mvebu/boards
> >       Merge commit 'd293875' into mvebu/boards
> 
> That's the mmc-next branch at that commit, but it'd be nice to have a
> more descriptive summary in the shortlog.

Agreed, I started doing this after the first time I pulled in pinctrl as
a dependency.  LinusW recommended pulling just up to the commit needed
so that if he did have to do a rebase, the closer it was to the tip of
his tree, the fewer people it would affect.

Would this be better?

"Merge mmc/mmc-next up to 'd293875' into mvebu/boards"

thx,

Jason.
Olof Johansson Feb. 5, 2013, 6:23 p.m. UTC | #3
On Tue, Feb 05, 2013 at 06:21:04AM -0500, Jason Cooper wrote:
> On Mon, Feb 04, 2013 at 11:10:10PM -0800, Olof Johansson wrote:
> > Hi,
> > 
> > On Fri, Feb 01, 2013 at 03:51:59PM +0000, Jason Cooper wrote:
> > > The following changes since commit 949db153b6466c6f7cad5a427ecea94985927311:
> > > 
> > >   Linux 3.8-rc5
> > > 
> > > are available in the git repository at:
> > > 
> > >   git://git.infradead.org/users/jcooper/linux.git tags/boards_for_v3.9
> > > 
> > > for you to fetch changes up to c7064904895f69b2e33117b450a05746f75abf3a:
> > > 
> > >   ARM: kirkwood: convert Guruplug Server Plus to use the device tree
> > > 
> > > ----------------------------------------------------------------
> > > mvebu boards for v3.9
> > >      - Guruplug Server Plus DT board
> > >      - RD-A370-A1 board
> > >      - mvebu improved SMP support in interrupt controller
> > >      - update defconfigs
> > >      - split legacy and DT setup for dove
> > >      - remove some redundant clock aliases
> > >     
> > >     depends on:
> > >      - tags/mvebu_fixes_for_v3.8-rc6
> > >      - tags/cleanup_for_v3.9_round2
> > >      - tags/drivers_for_v3.9
> > >      - mmc/mmc-next up to:
> > >         d293875 mmc: mvsdio: add pinctrl integration
> > 
> > Hm. What do you need the mmc branch for? Is it just the removal of the
> > redundant SDIO clock alias?
> 
> No, it's the four patches of Thomas' to the mvsdio driver:
> 
> d293875 mmc: mvsdio: add pinctrl integration
> 182ce21 mmc: mvsdio: implement a Device Tree binding
> aa3738e mmc: mvsdio: use slot-gpio for card detect gpio
> 9d8b28e5 mmc: mvsdio: use slot-gpio infrastructure for write protect gpio
> 
> which mvebu/boards needs before we can activate defconfig options.

Yeah, those are the patches you need to bring in. What I was wondering is why
you need to bring them in? If it's just for defconfig updates and a couple of
tiny cleanups, then I would say it's not worth the overhead of adding the
dependency branch to do those -- defconfig updates can go in anyway and the
cleanups can happen in the next release.

> > Also, it's really hard to tell where code came from when you just merge in
> > a hash like below:
> > 
> > > Jason Cooper (4):
> > >       Merge tag 'tags/mvebu_fixes_for_v3.8-rc6' into mvebu/boards
> > >       Merge tag 'tags/cleanup_for_v3.9_round2' into mvebu/boards
> > >       Merge tag 'tags/drivers_for_v3.9' into mvebu/boards
> > >       Merge commit 'd293875' into mvebu/boards
> > 
> > That's the mmc-next branch at that commit, but it'd be nice to have a
> > more descriptive summary in the shortlog.
> 
> Agreed, I started doing this after the first time I pulled in pinctrl as
> a dependency.  LinusW recommended pulling just up to the commit needed
> so that if he did have to do a rebase, the closer it was to the tip of
> his tree, the fewer people it would affect.
> 
> Would this be better?
> 
> "Merge mmc/mmc-next up to 'd293875' into mvebu/boards"

Yeah, that would have been better (but see above).


-Olof
Jason Cooper Feb. 5, 2013, 6:30 p.m. UTC | #4
On Tue, Feb 05, 2013 at 10:23:33AM -0800, Olof Johansson wrote:
> On Tue, Feb 05, 2013 at 06:21:04AM -0500, Jason Cooper wrote:
> > On Mon, Feb 04, 2013 at 11:10:10PM -0800, Olof Johansson wrote:
> > > Hi,
> > > 
> > > On Fri, Feb 01, 2013 at 03:51:59PM +0000, Jason Cooper wrote:
> > > > The following changes since commit 949db153b6466c6f7cad5a427ecea94985927311:
> > > > 
> > > >   Linux 3.8-rc5
> > > > 
> > > > are available in the git repository at:
> > > > 
> > > >   git://git.infradead.org/users/jcooper/linux.git tags/boards_for_v3.9
> > > > 
> > > > for you to fetch changes up to c7064904895f69b2e33117b450a05746f75abf3a:
> > > > 
> > > >   ARM: kirkwood: convert Guruplug Server Plus to use the device tree
> > > > 
> > > > ----------------------------------------------------------------
> > > > mvebu boards for v3.9
> > > >      - Guruplug Server Plus DT board
> > > >      - RD-A370-A1 board
> > > >      - mvebu improved SMP support in interrupt controller
> > > >      - update defconfigs
> > > >      - split legacy and DT setup for dove
> > > >      - remove some redundant clock aliases
> > > >     
> > > >     depends on:
> > > >      - tags/mvebu_fixes_for_v3.8-rc6
> > > >      - tags/cleanup_for_v3.9_round2
> > > >      - tags/drivers_for_v3.9
> > > >      - mmc/mmc-next up to:
> > > >         d293875 mmc: mvsdio: add pinctrl integration
> > > 
> > > Hm. What do you need the mmc branch for? Is it just the removal of the
> > > redundant SDIO clock alias?
> > 
> > No, it's the four patches of Thomas' to the mvsdio driver:
> > 
> > d293875 mmc: mvsdio: add pinctrl integration
> > 182ce21 mmc: mvsdio: implement a Device Tree binding
> > aa3738e mmc: mvsdio: use slot-gpio for card detect gpio
> > 9d8b28e5 mmc: mvsdio: use slot-gpio infrastructure for write protect gpio
> > 
> > which mvebu/boards needs before we can activate defconfig options.
> 
> Yeah, those are the patches you need to bring in. What I was wondering is why
> you need to bring them in? If it's just for defconfig updates and a couple of
> tiny cleanups, then I would say it's not worth the overhead of adding the
> dependency branch to do those -- defconfig updates can go in anyway and the
> cleanups can happen in the next release.

Well, I brought them in to mvebu/boards because of defconfig, but I
need them in mvebu/dt (which depends on mvebu/boards) because we remove
the legacy mvsdio init calls, and add the mvsdio dt blocks to the dts
files.

Would you prefer I put all the defconfig changes into /dt since it's at
the end of the dependency chain?

> > > Also, it's really hard to tell where code came from when you just merge in
> > > a hash like below:
> > > 
> > > > Jason Cooper (4):
> > > >       Merge tag 'tags/mvebu_fixes_for_v3.8-rc6' into mvebu/boards
> > > >       Merge tag 'tags/cleanup_for_v3.9_round2' into mvebu/boards
> > > >       Merge tag 'tags/drivers_for_v3.9' into mvebu/boards
> > > >       Merge commit 'd293875' into mvebu/boards
> > > 
> > > That's the mmc-next branch at that commit, but it'd be nice to have a
> > > more descriptive summary in the shortlog.
> > 
> > Agreed, I started doing this after the first time I pulled in pinctrl as
> > a dependency.  LinusW recommended pulling just up to the commit needed
> > so that if he did have to do a rebase, the closer it was to the tip of
> > his tree, the fewer people it would affect.
> > 
> > Would this be better?
> > 
> > "Merge mmc/mmc-next up to 'd293875' into mvebu/boards"
> 
> Yeah, that would have been better (but see above).

Ok, I'll do that in the future.

thx,

Jason.
Olof Johansson Feb. 5, 2013, 11:26 p.m. UTC | #5
On Tue, Feb 05, 2013 at 01:30:50PM -0500, Jason Cooper wrote:
> On Tue, Feb 05, 2013 at 10:23:33AM -0800, Olof Johansson wrote:
> > On Tue, Feb 05, 2013 at 06:21:04AM -0500, Jason Cooper wrote:
> > > On Mon, Feb 04, 2013 at 11:10:10PM -0800, Olof Johansson wrote:
> > > > Hi,
> > > > 
> > > > On Fri, Feb 01, 2013 at 03:51:59PM +0000, Jason Cooper wrote:
> > > > > The following changes since commit 949db153b6466c6f7cad5a427ecea94985927311:
> > > > > 
> > > > >   Linux 3.8-rc5
> > > > > 
> > > > > are available in the git repository at:
> > > > > 
> > > > >   git://git.infradead.org/users/jcooper/linux.git tags/boards_for_v3.9
> > > > > 
> > > > > for you to fetch changes up to c7064904895f69b2e33117b450a05746f75abf3a:
> > > > > 
> > > > >   ARM: kirkwood: convert Guruplug Server Plus to use the device tree
> > > > > 
> > > > > ----------------------------------------------------------------
> > > > > mvebu boards for v3.9
> > > > >      - Guruplug Server Plus DT board
> > > > >      - RD-A370-A1 board
> > > > >      - mvebu improved SMP support in interrupt controller
> > > > >      - update defconfigs
> > > > >      - split legacy and DT setup for dove
> > > > >      - remove some redundant clock aliases
> > > > >     
> > > > >     depends on:
> > > > >      - tags/mvebu_fixes_for_v3.8-rc6
> > > > >      - tags/cleanup_for_v3.9_round2
> > > > >      - tags/drivers_for_v3.9
> > > > >      - mmc/mmc-next up to:
> > > > >         d293875 mmc: mvsdio: add pinctrl integration
> > > > 
> > > > Hm. What do you need the mmc branch for? Is it just the removal of the
> > > > redundant SDIO clock alias?
> > > 
> > > No, it's the four patches of Thomas' to the mvsdio driver:
> > > 
> > > d293875 mmc: mvsdio: add pinctrl integration
> > > 182ce21 mmc: mvsdio: implement a Device Tree binding
> > > aa3738e mmc: mvsdio: use slot-gpio for card detect gpio
> > > 9d8b28e5 mmc: mvsdio: use slot-gpio infrastructure for write protect gpio
> > > 
> > > which mvebu/boards needs before we can activate defconfig options.
> > 
> > Yeah, those are the patches you need to bring in. What I was wondering is why
> > you need to bring them in? If it's just for defconfig updates and a couple of
> > tiny cleanups, then I would say it's not worth the overhead of adding the
> > dependency branch to do those -- defconfig updates can go in anyway and the
> > cleanups can happen in the next release.
> 
> Well, I brought them in to mvebu/boards because of defconfig, but I
> need them in mvebu/dt (which depends on mvebu/boards) because we remove
> the legacy mvsdio init calls, and add the mvsdio dt blocks to the dts
> files.
> 
> Would you prefer I put all the defconfig changes into /dt since it's at
> the end of the dependency chain?

The defconfig changes can go in either branch. They end up selecting
new options that will be silently turned off unless the driver change
is also in the same tree, but that's not a big deal since things will
eventually converge.

What makes sense is to hold off the patches that adds dts entries and
removes the platform device registrations to the next release to avoid
this dependency.


-Olof
Jason Cooper Feb. 6, 2013, 1:20 p.m. UTC | #6
On Tue, Feb 05, 2013 at 03:26:04PM -0800, Olof Johansson wrote:
> What makes sense is to hold off the patches that adds dts entries and
> removes the platform device registrations to the next release to avoid
> this dependency.

Hmm, this makes life a *lot* simpler, but it also really slows things
down.  I'll go ahead and do this from now on.

Would you like me to redo /boards and /dt in light of this, or just do
it from here forward?

thx,

Jason.
Olof Johansson Feb. 10, 2013, 12:31 a.m. UTC | #7
On Wed, Feb 06, 2013 at 08:20:15AM -0500, Jason Cooper wrote:
> On Tue, Feb 05, 2013 at 03:26:04PM -0800, Olof Johansson wrote:
> > What makes sense is to hold off the patches that adds dts entries and
> > removes the platform device registrations to the next release to avoid
> > this dependency.
> 
> Hmm, this makes life a *lot* simpler, but it also really slows things
> down.  I'll go ahead and do this from now on.
> 
> Would you like me to redo /boards and /dt in light of this, or just do
> it from here forward?

It doesn't necessarily slow things down -- it increases latency though. I.e.
you'll end up building a pipeline over time.

I agree that we're not entirely clear on this, and we're also going back and
forth a bit on how much of this we push back on and not. I don't think we've
found quite a good balance of this yet. Hm.


-Olof
Jason Cooper Feb. 10, 2013, 9:05 p.m. UTC | #8
On Sat, Feb 09, 2013 at 04:31:13PM -0800, Olof Johansson wrote:
> On Wed, Feb 06, 2013 at 08:20:15AM -0500, Jason Cooper wrote:
> > On Tue, Feb 05, 2013 at 03:26:04PM -0800, Olof Johansson wrote:
> > > What makes sense is to hold off the patches that adds dts entries and
> > > removes the platform device registrations to the next release to avoid
> > > this dependency.
> > 
> > Hmm, this makes life a *lot* simpler, but it also really slows things
> > down.  I'll go ahead and do this from now on.
> > 
> > Would you like me to redo /boards and /dt in light of this, or just do
> > it from here forward?
> 
> It doesn't necessarily slow things down -- it increases latency though. I.e.
> you'll end up building a pipeline over time.

Yes, that's not a problem.

> I agree that we're not entirely clear on this, and we're also going back and
> forth a bit on how much of this we push back on and not. I don't think we've
> found quite a good balance of this yet. Hm.

Well, it appears to me the crux of it is: How hard is it to merge and
re-merge branches with dependencies on driver trees?  If it's a PITA,
then let's do as you've suggested.  A sane, but slower process will
incur fewer mistakes.

Also, we can sit down an go ever this (with a few beers) at ELC.  ;-)

I'll redo as you've suggested for this go-around, and we can hash it out
in a few weeks.

thx,

Jason.