diff mbox

[1/6] video: tegra: Add nvhost driver

Message ID 1353577684-7896-2-git-send-email-tbergstrom@nvidia.com
State Not Applicable, archived
Headers show

Commit Message

Terje Bergstrom Nov. 22, 2012, 9:47 a.m. UTC
Add nvhost, the driver for host1x and 2D, which is a client device
for host1x.

Change-Id: Id93a28491dc2d54049e0adce4ad287c5985b2bca
Signed-off-by: Terje Bergstrom <tbergstrom@nvidia.com>
Signed-off-by: Arto Merilainen <amerilainen@nvidia.com>
---
 drivers/video/Kconfig                              |    2 +
 drivers/video/Makefile                             |    2 +
 drivers/video/tegra/host/Kconfig                   |   20 +
 drivers/video/tegra/host/Makefile                  |   21 +
 drivers/video/tegra/host/bus_client.c              |  101 ++++
 drivers/video/tegra/host/bus_client.h              |   32 ++
 drivers/video/tegra/host/chip_support.c            |   68 +++
 drivers/video/tegra/host/chip_support.h            |  179 +++++++
 drivers/video/tegra/host/debug.c                   |  255 ++++++++++
 drivers/video/tegra/host/debug.h                   |   50 ++
 drivers/video/tegra/host/dev.c                     |  104 ++++
 drivers/video/tegra/host/dev.h                     |   33 ++
 drivers/video/tegra/host/dmabuf.c                  |  151 ++++++
 drivers/video/tegra/host/dmabuf.h                  |   51 ++
 drivers/video/tegra/host/host1x/Makefile           |    6 +
 drivers/video/tegra/host/host1x/host1x.c           |  320 ++++++++++++
 drivers/video/tegra/host/host1x/host1x.h           |   97 ++++
 .../video/tegra/host/host1x/host1x01_hardware.h    |  157 ++++++
 drivers/video/tegra/host/host1x/host1x_cdma.c      |  488 ++++++++++++++++++
 drivers/video/tegra/host/host1x/host1x_cdma.h      |   39 ++
 drivers/video/tegra/host/host1x/host1x_channel.c   |  157 ++++++
 drivers/video/tegra/host/host1x/host1x_debug.c     |  405 +++++++++++++++
 drivers/video/tegra/host/host1x/host1x_intr.c      |  263 ++++++++++
 drivers/video/tegra/host/host1x/host1x_syncpt.c    |  170 +++++++
 .../video/tegra/host/host1x/hw_host1x01_channel.h  |  182 +++++++
 drivers/video/tegra/host/host1x/hw_host1x01_sync.h |  398 +++++++++++++++
 .../video/tegra/host/host1x/hw_host1x01_uclass.h   |  474 +++++++++++++++++
 drivers/video/tegra/host/nvhost_acm.c              |  532 ++++++++++++++++++++
 drivers/video/tegra/host/nvhost_acm.h              |   49 ++
 drivers/video/tegra/host/nvhost_cdma.c             |  473 +++++++++++++++++
 drivers/video/tegra/host/nvhost_cdma.h             |  116 +++++
 drivers/video/tegra/host/nvhost_channel.c          |  129 +++++
 drivers/video/tegra/host/nvhost_channel.h          |   65 +++
 drivers/video/tegra/host/nvhost_intr.c             |  391 ++++++++++++++
 drivers/video/tegra/host/nvhost_intr.h             |  110 ++++
 drivers/video/tegra/host/nvhost_job.c              |  398 +++++++++++++++
 drivers/video/tegra/host/nvhost_memmgr.c           |  252 ++++++++++
 drivers/video/tegra/host/nvhost_memmgr.h           |   66 +++
 drivers/video/tegra/host/nvhost_syncpt.c           |  453 +++++++++++++++++
 drivers/video/tegra/host/nvhost_syncpt.h           |  148 ++++++
 drivers/video/tegra/host/t20/Makefile              |    6 +
 drivers/video/tegra/host/t20/t20.c                 |   78 +++
 drivers/video/tegra/host/t20/t20.h                 |   29 ++
 drivers/video/tegra/host/t30/Makefile              |    6 +
 drivers/video/tegra/host/t30/t30.c                 |   80 +++
 drivers/video/tegra/host/t30/t30.h                 |   29 ++
 include/linux/nvhost.h                             |  302 +++++++++++
 include/trace/events/nvhost.h                      |  249 +++++++++
 48 files changed, 8186 insertions(+)
 create mode 100644 drivers/video/tegra/host/Kconfig
 create mode 100644 drivers/video/tegra/host/Makefile
 create mode 100644 drivers/video/tegra/host/bus_client.c
 create mode 100644 drivers/video/tegra/host/bus_client.h
 create mode 100644 drivers/video/tegra/host/chip_support.c
 create mode 100644 drivers/video/tegra/host/chip_support.h
 create mode 100644 drivers/video/tegra/host/debug.c
 create mode 100644 drivers/video/tegra/host/debug.h
 create mode 100644 drivers/video/tegra/host/dev.c
 create mode 100644 drivers/video/tegra/host/dev.h
 create mode 100644 drivers/video/tegra/host/dmabuf.c
 create mode 100644 drivers/video/tegra/host/dmabuf.h
 create mode 100644 drivers/video/tegra/host/host1x/Makefile
 create mode 100644 drivers/video/tegra/host/host1x/host1x.c
 create mode 100644 drivers/video/tegra/host/host1x/host1x.h
 create mode 100644 drivers/video/tegra/host/host1x/host1x01_hardware.h
 create mode 100644 drivers/video/tegra/host/host1x/host1x_cdma.c
 create mode 100644 drivers/video/tegra/host/host1x/host1x_cdma.h
 create mode 100644 drivers/video/tegra/host/host1x/host1x_channel.c
 create mode 100644 drivers/video/tegra/host/host1x/host1x_debug.c
 create mode 100644 drivers/video/tegra/host/host1x/host1x_intr.c
 create mode 100644 drivers/video/tegra/host/host1x/host1x_syncpt.c
 create mode 100644 drivers/video/tegra/host/host1x/hw_host1x01_channel.h
 create mode 100644 drivers/video/tegra/host/host1x/hw_host1x01_sync.h
 create mode 100644 drivers/video/tegra/host/host1x/hw_host1x01_uclass.h
 create mode 100644 drivers/video/tegra/host/nvhost_acm.c
 create mode 100644 drivers/video/tegra/host/nvhost_acm.h
 create mode 100644 drivers/video/tegra/host/nvhost_cdma.c
 create mode 100644 drivers/video/tegra/host/nvhost_cdma.h
 create mode 100644 drivers/video/tegra/host/nvhost_channel.c
 create mode 100644 drivers/video/tegra/host/nvhost_channel.h
 create mode 100644 drivers/video/tegra/host/nvhost_intr.c
 create mode 100644 drivers/video/tegra/host/nvhost_intr.h
 create mode 100644 drivers/video/tegra/host/nvhost_job.c
 create mode 100644 drivers/video/tegra/host/nvhost_memmgr.c
 create mode 100644 drivers/video/tegra/host/nvhost_memmgr.h
 create mode 100644 drivers/video/tegra/host/nvhost_syncpt.c
 create mode 100644 drivers/video/tegra/host/nvhost_syncpt.h
 create mode 100644 drivers/video/tegra/host/t20/Makefile
 create mode 100644 drivers/video/tegra/host/t20/t20.c
 create mode 100644 drivers/video/tegra/host/t20/t20.h
 create mode 100644 drivers/video/tegra/host/t30/Makefile
 create mode 100644 drivers/video/tegra/host/t30/t30.c
 create mode 100644 drivers/video/tegra/host/t30/t30.h
 create mode 100644 include/linux/nvhost.h
 create mode 100644 include/trace/events/nvhost.h

Comments

Thierry Reding Nov. 23, 2012, 11:38 p.m. UTC | #1
On Thu, Nov 22, 2012 at 11:47:59AM +0200, Terje Bergstrom wrote:
> Add nvhost, the driver for host1x and 2D, which is a client device
> for host1x.
> 
> Change-Id: Id93a28491dc2d54049e0adce4ad287c5985b2bca
> Signed-off-by: Terje Bergstrom <tbergstrom@nvidia.com>
> Signed-off-by: Arto Merilainen <amerilainen@nvidia.com>
> ---
>  drivers/video/Kconfig                              |    2 +
>  drivers/video/Makefile                             |    2 +
>  drivers/video/tegra/host/Kconfig                   |   20 +
>  drivers/video/tegra/host/Makefile                  |   21 +
>  drivers/video/tegra/host/bus_client.c              |  101 ++++
>  drivers/video/tegra/host/bus_client.h              |   32 ++
>  drivers/video/tegra/host/chip_support.c            |   68 +++
>  drivers/video/tegra/host/chip_support.h            |  179 +++++++
>  drivers/video/tegra/host/debug.c                   |  255 ++++++++++
>  drivers/video/tegra/host/debug.h                   |   50 ++
>  drivers/video/tegra/host/dev.c                     |  104 ++++
>  drivers/video/tegra/host/dev.h                     |   33 ++
>  drivers/video/tegra/host/dmabuf.c                  |  151 ++++++
>  drivers/video/tegra/host/dmabuf.h                  |   51 ++
>  drivers/video/tegra/host/host1x/Makefile           |    6 +
>  drivers/video/tegra/host/host1x/host1x.c           |  320 ++++++++++++
>  drivers/video/tegra/host/host1x/host1x.h           |   97 ++++
>  .../video/tegra/host/host1x/host1x01_hardware.h    |  157 ++++++
>  drivers/video/tegra/host/host1x/host1x_cdma.c      |  488 ++++++++++++++++++
>  drivers/video/tegra/host/host1x/host1x_cdma.h      |   39 ++
>  drivers/video/tegra/host/host1x/host1x_channel.c   |  157 ++++++
>  drivers/video/tegra/host/host1x/host1x_debug.c     |  405 +++++++++++++++
>  drivers/video/tegra/host/host1x/host1x_intr.c      |  263 ++++++++++
>  drivers/video/tegra/host/host1x/host1x_syncpt.c    |  170 +++++++
>  .../video/tegra/host/host1x/hw_host1x01_channel.h  |  182 +++++++
>  drivers/video/tegra/host/host1x/hw_host1x01_sync.h |  398 +++++++++++++++
>  .../video/tegra/host/host1x/hw_host1x01_uclass.h   |  474 +++++++++++++++++
>  drivers/video/tegra/host/nvhost_acm.c              |  532 ++++++++++++++++++++
>  drivers/video/tegra/host/nvhost_acm.h              |   49 ++
>  drivers/video/tegra/host/nvhost_cdma.c             |  473 +++++++++++++++++
>  drivers/video/tegra/host/nvhost_cdma.h             |  116 +++++
>  drivers/video/tegra/host/nvhost_channel.c          |  129 +++++
>  drivers/video/tegra/host/nvhost_channel.h          |   65 +++
>  drivers/video/tegra/host/nvhost_intr.c             |  391 ++++++++++++++
>  drivers/video/tegra/host/nvhost_intr.h             |  110 ++++
>  drivers/video/tegra/host/nvhost_job.c              |  398 +++++++++++++++
>  drivers/video/tegra/host/nvhost_memmgr.c           |  252 ++++++++++
>  drivers/video/tegra/host/nvhost_memmgr.h           |   66 +++
>  drivers/video/tegra/host/nvhost_syncpt.c           |  453 +++++++++++++++++
>  drivers/video/tegra/host/nvhost_syncpt.h           |  148 ++++++
>  drivers/video/tegra/host/t20/Makefile              |    6 +
>  drivers/video/tegra/host/t20/t20.c                 |   78 +++
>  drivers/video/tegra/host/t20/t20.h                 |   29 ++
>  drivers/video/tegra/host/t30/Makefile              |    6 +
>  drivers/video/tegra/host/t30/t30.c                 |   80 +++
>  drivers/video/tegra/host/t30/t30.h                 |   29 ++
>  include/linux/nvhost.h                             |  302 +++++++++++
>  include/trace/events/nvhost.h                      |  249 +++++++++
>  48 files changed, 8186 insertions(+)
>  create mode 100644 drivers/video/tegra/host/Kconfig
>  create mode 100644 drivers/video/tegra/host/Makefile
>  create mode 100644 drivers/video/tegra/host/bus_client.c
>  create mode 100644 drivers/video/tegra/host/bus_client.h
>  create mode 100644 drivers/video/tegra/host/chip_support.c
>  create mode 100644 drivers/video/tegra/host/chip_support.h
>  create mode 100644 drivers/video/tegra/host/debug.c
>  create mode 100644 drivers/video/tegra/host/debug.h
>  create mode 100644 drivers/video/tegra/host/dev.c
>  create mode 100644 drivers/video/tegra/host/dev.h
>  create mode 100644 drivers/video/tegra/host/dmabuf.c
>  create mode 100644 drivers/video/tegra/host/dmabuf.h
>  create mode 100644 drivers/video/tegra/host/host1x/Makefile
>  create mode 100644 drivers/video/tegra/host/host1x/host1x.c
>  create mode 100644 drivers/video/tegra/host/host1x/host1x.h
>  create mode 100644 drivers/video/tegra/host/host1x/host1x01_hardware.h
>  create mode 100644 drivers/video/tegra/host/host1x/host1x_cdma.c
>  create mode 100644 drivers/video/tegra/host/host1x/host1x_cdma.h
>  create mode 100644 drivers/video/tegra/host/host1x/host1x_channel.c
>  create mode 100644 drivers/video/tegra/host/host1x/host1x_debug.c
>  create mode 100644 drivers/video/tegra/host/host1x/host1x_intr.c
>  create mode 100644 drivers/video/tegra/host/host1x/host1x_syncpt.c
>  create mode 100644 drivers/video/tegra/host/host1x/hw_host1x01_channel.h
>  create mode 100644 drivers/video/tegra/host/host1x/hw_host1x01_sync.h
>  create mode 100644 drivers/video/tegra/host/host1x/hw_host1x01_uclass.h
>  create mode 100644 drivers/video/tegra/host/nvhost_acm.c
>  create mode 100644 drivers/video/tegra/host/nvhost_acm.h
>  create mode 100644 drivers/video/tegra/host/nvhost_cdma.c
>  create mode 100644 drivers/video/tegra/host/nvhost_cdma.h
>  create mode 100644 drivers/video/tegra/host/nvhost_channel.c
>  create mode 100644 drivers/video/tegra/host/nvhost_channel.h
>  create mode 100644 drivers/video/tegra/host/nvhost_intr.c
>  create mode 100644 drivers/video/tegra/host/nvhost_intr.h
>  create mode 100644 drivers/video/tegra/host/nvhost_job.c
>  create mode 100644 drivers/video/tegra/host/nvhost_memmgr.c
>  create mode 100644 drivers/video/tegra/host/nvhost_memmgr.h
>  create mode 100644 drivers/video/tegra/host/nvhost_syncpt.c
>  create mode 100644 drivers/video/tegra/host/nvhost_syncpt.h
>  create mode 100644 drivers/video/tegra/host/t20/Makefile
>  create mode 100644 drivers/video/tegra/host/t20/t20.c
>  create mode 100644 drivers/video/tegra/host/t20/t20.h
>  create mode 100644 drivers/video/tegra/host/t30/Makefile
>  create mode 100644 drivers/video/tegra/host/t30/t30.c
>  create mode 100644 drivers/video/tegra/host/t30/t30.h
>  create mode 100644 include/linux/nvhost.h
>  create mode 100644 include/trace/events/nvhost.h

Okay, there's a huge amount of code here. From a quick look this seems
like the whole nvhost implementation as found in the L4T kernel. It'll
obviously take quite a bit of time to go through all of it.

I would really like this to be phased in in more manageable chunks. For
instance syncpoints could be added on top of the existing host1x
infrastructure (the display controllers can make use of it for VBLANK
reporting, right?), followed by channel support for command submission.

While VBLANK reporting is already rudimentarily supported using the
display controllers' VBLANK interrupts, adding syncpt based support
should be easy to do.

Last but not least, I'd really appreciate it if we could settle on one
name for this component. One place calls it graphics host, another one
refers to it as host1x and now we have an nvhost driver for it. The TRM
isn't very consistent either, unfortunately, but a number of chapters
refer to it as host1x and the clock that drives the block is also named
host1x. Those are pretty much the reasons why I chose to call the DT
node and the driver host1x and I would like to suggest we stay with it
to keep things less confusing.

Thierry
Terje Bergstrom Nov. 24, 2012, 7 a.m. UTC | #2
On 24.11.2012 01:38, Thierry Reding wrote:
> Okay, there's a huge amount of code here. From a quick look this seems
> like the whole nvhost implementation as found in the L4T kernel. It'll
> obviously take quite a bit of time to go through all of it.

Thanks Thierry for your comments.

It's actually about a third of of the downstream nvhost. I left out
support for all client modules, ioctl APIs, hardware context, nvmap
(obviously) and various other features.

But yes, the amount of code is still significant and I feel the pain,
too. I'm attempting to keep downstream and upstream nvhost as similar to
each other as possible so that we can port changes easily between the
two trees. In downstream kernel nvhost manages all its clients (except
DC only partially), so it's kept as its own driver. I'm hoping we'd end
up using tegradrm in downstream, and keeping nvhost as separate entity
is playing a role there.

If it helps, I could write up an essay on the structure of nvhost and
why it works that way. That should help in understanding some of the
design decisions in code.

> I would really like this to be phased in in more manageable chunks. For
> instance syncpoints could be added on top of the existing host1x
> infrastructure (the display controllers can make use of it for VBLANK
> reporting, right?), followed by channel support for command submission.
>
> While VBLANK reporting is already rudimentarily supported using the
> display controllers' VBLANK interrupts, adding syncpt based support
> should be easy to do.

Yep, vblank is a sync point register.

I attempted to tune the code down to only having syncpt support, but it
pulled in still enough (interrupts, syncpts, power management) that the
patch went down about 40% in size. I thought that as the staging didn't
bring really small parts, I deleted that attempt.

I could redo that if it helps. I guess removing dynamic power management
would be the first step, and channels the second. In review process,
they're anyway in separate files so I'm not sure about the benefits, though.

I'm not convinced about moving the code to live inside tegradrm. There's
really not that much host1x infrastructure in tegradrm's host1x.c that
we could leverage except binding device and driver and enabling and
disabling host1x clocks. Most of code in current host1x.c is dedicated
to managing the drm sub-clients, which anyway logically belong to drm.c.

To implement sync points in tegradrm, we'd need to re-implement syncpt
and interrupt support in tegradrm. That's a significant part of code,
and we already have code for that, so I'd like to leverage it as far as
possible.

> Last but not least, I'd really appreciate it if we could settle on one
> name for this component. One place calls it graphics host, another one
> refers to it as host1x and now we have an nvhost driver for it. The TRM
> isn't very consistent either, unfortunately, but a number of chapters
> refer to it as host1x and the clock that drives the block is also named
> host1x. Those are pretty much the reasons why I chose to call the DT
> node and the driver host1x and I would like to suggest we stay with it
> to keep things less confusing.

Graphics host and host1x are names for the hardware block. Graphics host
is an older name, host1x is the proper name.

nvhost is the driver for host1x and its client modules in downstream
kernel. The naming gets confusing in upstream kernel because of shared
responsibility of client modules. tegradrm's 2D driver is just glue
between tegradrm and nvhost.

host1x as the name of the driver wouldn't cover the fact that it's also
managing client modules.

I'll still try to figure out a way to clear up the naming, and
suggestions are welcome. :-)

Terje
--
To unsubscribe from this list: send the line "unsubscribe linux-tegra" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Thierry Reding Nov. 24, 2012, 7:04 p.m. UTC | #3
On Sat, Nov 24, 2012 at 09:00:20AM +0200, Terje Bergström wrote:
> On 24.11.2012 01:38, Thierry Reding wrote:
> > Okay, there's a huge amount of code here. From a quick look this seems
> > like the whole nvhost implementation as found in the L4T kernel. It'll
> > obviously take quite a bit of time to go through all of it.
> 
> Thanks Thierry for your comments.
> 
> It's actually about a third of of the downstream nvhost. I left out
> support for all client modules, ioctl APIs, hardware context, nvmap
> (obviously) and various other features.
> 
> But yes, the amount of code is still significant and I feel the pain,
> too. I'm attempting to keep downstream and upstream nvhost as similar to
> each other as possible so that we can port changes easily between the
> two trees. In downstream kernel nvhost manages all its clients (except
> DC only partially), so it's kept as its own driver. I'm hoping we'd end
> up using tegradrm in downstream, and keeping nvhost as separate entity
> is playing a role there.

The usual way to do this is to adapt downstream kernels to upstream
changes, not the other way around.

> If it helps, I could write up an essay on the structure of nvhost and
> why it works that way. That should help in understanding some of the
> design decisions in code.

Some more documentation about host1x would certainly be very welcome. If
you write something up, maybe you can push to get it included in the TRM
as well. The host1x chapter in the TRM is a bit sketchy and something
more like a programming guide would be really useful.

Funny enough, the chapter about the 2D engine says that it is merely
there to document the registers and explicitly not as a programming
guide because it is expected that NVIDIA supplied drivers will always be
used. I had a good laugh when I read that. =)

Still I guess it's better than nothing.

> > I would really like this to be phased in in more manageable chunks. For
> > instance syncpoints could be added on top of the existing host1x
> > infrastructure (the display controllers can make use of it for VBLANK
> > reporting, right?), followed by channel support for command submission.
> >
> > While VBLANK reporting is already rudimentarily supported using the
> > display controllers' VBLANK interrupts, adding syncpt based support
> > should be easy to do.
> 
> Yep, vblank is a sync point register.
> 
> I attempted to tune the code down to only having syncpt support, but it
> pulled in still enough (interrupts, syncpts, power management) that the
> patch went down about 40% in size. I thought that as the staging didn't
> bring really small parts, I deleted that attempt.
> 
> I could redo that if it helps. I guess removing dynamic power management
> would be the first step, and channels the second. In review process,
> they're anyway in separate files so I'm not sure about the benefits, though.

The benefit is that smaller changes that add a single particular feature
are a lot easier to review and verify.

> I'm not convinced about moving the code to live inside tegradrm. There's
> really not that much host1x infrastructure in tegradrm's host1x.c that
> we could leverage except binding device and driver and enabling and
> disabling host1x clocks.

The reason there isn't more infrastructure is that I explicitly wanted
to start small. Just enough to get the display to work. I find that to
be a very good method because it allows the code to evolve in small and
incremental steps.

> Most of code in current host1x.c is dedicated to managing the drm sub-
> clients, which anyway logically belong to drm.c.

In fact the code in host1x.c is supposed to be managing host1x clients,
which is exactly what nvhost does. It isn't doing much else for sole
reason that it doesn't have to at this point.

> To implement sync points in tegradrm, we'd need to re-implement syncpt
> and interrupt support in tegradrm. That's a significant part of code,
> and we already have code for that, so I'd like to leverage it as far as
> possible.

I'm all for reusing your code. My concern is that the structure of this
is quite different from what can be found in the L4T downstream kernels
and therefore some rethinking may be required.

> > Last but not least, I'd really appreciate it if we could settle on one
> > name for this component. One place calls it graphics host, another one
> > refers to it as host1x and now we have an nvhost driver for it. The TRM
> > isn't very consistent either, unfortunately, but a number of chapters
> > refer to it as host1x and the clock that drives the block is also named
> > host1x. Those are pretty much the reasons why I chose to call the DT
> > node and the driver host1x and I would like to suggest we stay with it
> > to keep things less confusing.
> 
> Graphics host and host1x are names for the hardware block. Graphics host
> is an older name, host1x is the proper name.
> 
> nvhost is the driver for host1x and its client modules in downstream
> kernel. The naming gets confusing in upstream kernel because of shared
> responsibility of client modules. tegradrm's 2D driver is just glue
> between tegradrm and nvhost.

But that's precisely my point. If the hardware is called host1x, why not
call the driver host1x as well. We do the same for all other hardware
blocks. The I2C controller driver is named i2c-tegra just as the RTC
driver is named rtc-tegra. Why should the driver for host1x not be named
host1x?

> host1x as the name of the driver wouldn't cover the fact that it's also
> managing client modules.

That's like saying an I2C controller driver shouldn't be named I2C
because it addresses slaves. Furthermore, just naming it nvhost doesn't
change anything about that. How does nvhost make it any clearer that it
manages client modules compared to host1x?

> I'll still try to figure out a way to clear up the naming, and
> suggestions are welcome. :-)

host1x hardware and host1x clients are controlled by the "host1x"
driver. Can it be any clearer than that?

Thierry
Terje Bergstrom Nov. 24, 2012, 8:02 p.m. UTC | #4
I'll reply to the rest later, but here some initial comments.

On 24.11.2012 21:04, Thierry Reding wrote:
> The usual way to do this is to adapt downstream kernels to upstream
> changes, not the other way around.

Yep, we're doing both ways - leveraging downstream and upstream code.

> Some more documentation about host1x would certainly be very welcome. If
> you write something up, maybe you can push to get it included in the TRM
> as well. The host1x chapter in the TRM is a bit sketchy and something
> more like a programming guide would be really useful.

Got it. I hope user space code would help here, as it would be a
definite explanation of how the thing is supposed to be used.

> Funny enough, the chapter about the 2D engine says that it is merely
> there to document the registers and explicitly not as a programming
> guide because it is expected that NVIDIA supplied drivers will always be
> used. I had a good laugh when I read that. =)

Ouch. :-)

> The benefit is that smaller changes that add a single particular feature
> are a lot easier to review and verify.

Yep, I'm working on this. I think I found a way.

> But that's precisely my point. If the hardware is called host1x, why not
> call the driver host1x as well. We do the same for all other hardware
> blocks. The I2C controller driver is named i2c-tegra just as the RTC
> driver is named rtc-tegra. Why should the driver for host1x not be named
> host1x?

Ok, I got it.

Terje
--
To unsubscribe from this list: send the line "unsubscribe linux-tegra" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Terje Bergstrom Nov. 26, 2012, 8:21 a.m. UTC | #5
On 24.11.2012 21:04, Thierry Reding wrote:
>>> I would really like this to be phased in in more manageable chunks. For
>>> instance syncpoints could be added on top of the existing host1x
>>> infrastructure (the display controllers can make use of it for VBLANK
>>> reporting, right?), followed by channel support for command submission.

I staged nvhost into 4 patches that each add some functionality:
 * Add driver with API to read and increment sync point values
 * Add syncpt wait and interrupt infrastructure
 * Add channel and client support and related infrastructure (jobs, memmgr)
 * Add debug support

I hope that helps in reviews.

> I'm all for reusing your code. My concern is that the structure of this
> is quite different from what can be found in the L4T downstream kernels
> and therefore some rethinking may be required.

As long as I can keep nvhost as a separate entity, I can port changes
from upstream to downstream and vice versa and there's no architectural
reason why we couldn't gradually move in downstream kernel to using
tegradrm.

If we merge host1x management code to tegradrm, there's no logical way
of taking tegradrm into use in downstream.

Terje
--
To unsubscribe from this list: send the line "unsubscribe linux-tegra" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Thierry Reding Nov. 26, 2012, 9:25 a.m. UTC | #6
On Mon, Nov 26, 2012 at 10:21:36AM +0200, Terje Bergström wrote:
> On 24.11.2012 21:04, Thierry Reding wrote:
> >>> I would really like this to be phased in in more manageable chunks. For
> >>> instance syncpoints could be added on top of the existing host1x
> >>> infrastructure (the display controllers can make use of it for VBLANK
> >>> reporting, right?), followed by channel support for command submission.
> 
> I staged nvhost into 4 patches that each add some functionality:
>  * Add driver with API to read and increment sync point values
>  * Add syncpt wait and interrupt infrastructure
>  * Add channel and client support and related infrastructure (jobs, memmgr)
>  * Add debug support
> 
> I hope that helps in reviews.

Sounds better.

> > I'm all for reusing your code. My concern is that the structure of this
> > is quite different from what can be found in the L4T downstream kernels
> > and therefore some rethinking may be required.
> 
> As long as I can keep nvhost as a separate entity, I can port changes
> from upstream to downstream and vice versa and there's no architectural
> reason why we couldn't gradually move in downstream kernel to using
> tegradrm.
> 
> If we merge host1x management code to tegradrm, there's no logical way
> of taking tegradrm into use in downstream.

I think I understand what you're saying. While I agree that it might be
better to move host1x out of tegra-drm in the long term, in particular
because other frameworks like V4L2 will start to use it eventually, I
have some reservations about whether it makes sense to do it right away.

Furthermore as I understand it, most of the work required to use
tegra-drm downstream would involve changes to the userspace components,
wouldn't it? None of that would be dependent on where in the kernel the
host1x driver resides.

Thierry
Terje Bergstrom Nov. 26, 2012, 9:33 a.m. UTC | #7
On 26.11.2012 11:25, Thierry Reding wrote:
> I think I understand what you're saying. While I agree that it might be
> better to move host1x out of tegra-drm in the long term, in particular
> because other frameworks like V4L2 will start to use it eventually, I
> have some reservations about whether it makes sense to do it right away.
> 
> Furthermore as I understand it, most of the work required to use
> tegra-drm downstream would involve changes to the userspace components,
> wouldn't it? None of that would be dependent on where in the kernel the
> host1x driver resides.

If we start using tegradrm in downstream and it takes over host1x, we'd
lose ability to operate 3D, camera and multimedia hw. In current form we
expose enough APIs to add support for them in separate drivers.

Terje
--
To unsubscribe from this list: send the line "unsubscribe linux-tegra" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Terje Bergstrom Nov. 26, 2012, 1:42 p.m. UTC | #8
On 26.11.2012 11:25, Thierry Reding wrote:
> On Mon, Nov 26, 2012 at 10:21:36AM +0200, Terje Bergström wrote:
>> I staged nvhost into 4 patches that each add some functionality:
>>  * Add driver with API to read and increment sync point values
>>  * Add syncpt wait and interrupt infrastructure
>>  * Add channel and client support and related infrastructure (jobs, memmgr)
>>  * Add debug support
>>
>> I hope that helps in reviews.
> Sounds better.

Done now, and patches sent. Let me know if you believe the patches
should still be divided.

Best regards,
Terje
--
To unsubscribe from this list: send the line "unsubscribe linux-tegra" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
diff mbox

Patch

diff --git a/drivers/video/Kconfig b/drivers/video/Kconfig
index fb9a14e..94c861b 100644
--- a/drivers/video/Kconfig
+++ b/drivers/video/Kconfig
@@ -2463,4 +2463,6 @@  config FB_SH_MOBILE_MERAM
 	  Up to 4 memory channels can be configured, allowing 4 RGB or
 	  2 YCbCr framebuffers to be configured.
 
+source "drivers/video/tegra/host/Kconfig"
+
 endmenu
diff --git a/drivers/video/Makefile b/drivers/video/Makefile
index b936b00..aae33a1 100644
--- a/drivers/video/Makefile
+++ b/drivers/video/Makefile
@@ -17,6 +17,8 @@  obj-y				  += backlight/
 
 obj-$(CONFIG_EXYNOS_VIDEO)     += exynos/
 
+obj-$(CONFIG_TEGRA_GRHOST)     += tegra/host/
+
 obj-$(CONFIG_FB_CFB_FILLRECT)  += cfbfillrect.o
 obj-$(CONFIG_FB_CFB_COPYAREA)  += cfbcopyarea.o
 obj-$(CONFIG_FB_CFB_IMAGEBLIT) += cfbimgblt.o
diff --git a/drivers/video/tegra/host/Kconfig b/drivers/video/tegra/host/Kconfig
new file mode 100644
index 0000000..2954bea
--- /dev/null
+++ b/drivers/video/tegra/host/Kconfig
@@ -0,0 +1,20 @@ 
+config TEGRA_GRHOST
+	tristate "Tegra graphics host driver"
+	help
+	  Driver for the Tegra graphics host hardware.
+
+config TEGRA_GRHOST_USE_DMABUF
+	depends on TEGRA_GRHOST
+	bool "Support dmabuf buffers"
+	default y
+	select DMA_SHARED_BUFFER
+	help
+	  Support dmabuf buffers.
+
+config TEGRA_GRHOST_DEFAULT_TIMEOUT
+	depends on TEGRA_GRHOST
+	int "Default timeout for submits"
+	default 10000
+	help
+	  Default timeout for jobs in milliseconds. Set to zero for no timeout.
+
diff --git a/drivers/video/tegra/host/Makefile b/drivers/video/tegra/host/Makefile
new file mode 100644
index 0000000..bd12f98
--- /dev/null
+++ b/drivers/video/tegra/host/Makefile
@@ -0,0 +1,21 @@ 
+ccflags-y = -Idrivers/video/tegra/host -Iarch/arm/mach-tegra
+
+nvhost-objs = \
+	nvhost_acm.o \
+	nvhost_syncpt.o \
+	nvhost_cdma.o \
+	nvhost_intr.o \
+	nvhost_channel.o \
+	nvhost_job.o \
+	dev.o \
+	debug.o \
+	bus_client.o \
+	chip_support.o \
+	nvhost_memmgr.o
+
+obj-$(CONFIG_TEGRA_GRHOST) += host1x/
+obj-$(CONFIG_TEGRA_GRHOST) += t20/
+obj-$(CONFIG_TEGRA_GRHOST) += t30/
+obj-$(CONFIG_TEGRA_GRHOST) += nvhost.o
+
+obj-$(CONFIG_TEGRA_GRHOST_USE_DMABUF) += dmabuf.o
diff --git a/drivers/video/tegra/host/bus_client.c b/drivers/video/tegra/host/bus_client.c
new file mode 100644
index 0000000..886d3fe
--- /dev/null
+++ b/drivers/video/tegra/host/bus_client.c
@@ -0,0 +1,101 @@ 
+/*
+ * drivers/video/tegra/host/bus_client.c
+ *
+ * Tegra Graphics Host Client Module
+ *
+ * Copyright (c) 2010-2012, NVIDIA Corporation.
+ *
+ * This program is free software; you can redistribute it and/or modify it
+ * under the terms and conditions of the GNU General Public License,
+ * version 2, as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
+ * more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program.  If not, see <http://www.gnu.org/licenses/>.
+ */
+
+#include <linux/module.h>
+#include <linux/slab.h>
+#include <linux/string.h>
+#include <linux/spinlock.h>
+#include <linux/fs.h>
+#include <linux/cdev.h>
+#include <linux/uaccess.h>
+#include <linux/file.h>
+#include <linux/clk.h>
+#include <linux/hrtimer.h>
+#include <linux/export.h>
+
+#include <trace/events/nvhost.h>
+
+#include <linux/io.h>
+#include <linux/string.h>
+
+#include <linux/nvhost.h>
+
+#include "debug.h"
+#include "bus_client.h"
+#include "dev.h"
+#include "nvhost_memmgr.h"
+#include "chip_support.h"
+#include "nvhost_acm.h"
+
+#include "nvhost_channel.h"
+
+int nvhost_client_device_init(struct platform_device *dev)
+{
+	int err;
+	struct nvhost_master *nvhost_master = nvhost_get_host(dev);
+	struct nvhost_channel *ch;
+	struct nvhost_device_data *pdata = platform_get_drvdata(dev);
+
+	ch = nvhost_alloc_channel(dev);
+	if (ch == NULL)
+		return -ENODEV;
+
+	/* store the pointer to this device for channel */
+	ch->dev = dev;
+
+	err = nvhost_channel_init(ch, nvhost_master, pdata->index);
+	if (err)
+		goto fail;
+
+	err = nvhost_module_init(dev);
+	if (err)
+		goto fail;
+
+	err = nvhost_device_list_add(dev);
+	if (err)
+		goto fail;
+
+	nvhost_device_debug_init(dev);
+
+	dev_info(&dev->dev, "initialized\n");
+
+	return 0;
+
+fail:
+	/* Add clean-up */
+	nvhost_free_channel(ch);
+	return err;
+}
+EXPORT_SYMBOL(nvhost_client_device_init);
+
+int nvhost_client_device_suspend(struct platform_device *dev)
+{
+	int ret = 0;
+	struct nvhost_device_data *pdata = platform_get_drvdata(dev);
+
+	ret = nvhost_channel_suspend(pdata->channel);
+	if (ret)
+		return ret;
+
+	dev_info(&dev->dev, "suspend status: %d\n", ret);
+
+	return ret;
+}
+EXPORT_SYMBOL(nvhost_client_device_suspend);
diff --git a/drivers/video/tegra/host/bus_client.h b/drivers/video/tegra/host/bus_client.h
new file mode 100644
index 0000000..2b56213
--- /dev/null
+++ b/drivers/video/tegra/host/bus_client.h
@@ -0,0 +1,32 @@ 
+/*
+ * drivers/video/tegra/host/bus_client.h
+ *
+ * Tegra Graphics Host client
+ *
+ * Copyright (c) 2010-2012, NVIDIA Corporation.
+ *
+ * This program is free software; you can redistribute it and/or modify it
+ * under the terms and conditions of the GNU General Public License,
+ * version 2, as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
+ * more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program.  If not, see <http://www.gnu.org/licenses/>.
+ */
+
+#ifndef __NVHOST_BUS_CLIENT_H
+#define __NVHOST_BUS_CLIENT_H
+
+#include <linux/types.h>
+
+struct platform_device;
+
+int nvhost_client_device_init(struct platform_device *dev);
+
+int nvhost_client_device_suspend(struct platform_device *dev);
+
+#endif
diff --git a/drivers/video/tegra/host/chip_support.c b/drivers/video/tegra/host/chip_support.c
new file mode 100644
index 0000000..fca15a6
--- /dev/null
+++ b/drivers/video/tegra/host/chip_support.c
@@ -0,0 +1,68 @@ 
+/*
+ * drivers/video/tegra/host/chip_support.c
+ *
+ * Tegra Graphics Host Chip support module
+ *
+ * Copyright (c) 2012, NVIDIA Corporation.
+ *
+ * This program is free software; you can redistribute it and/or modify it
+ * under the terms and conditions of the GNU General Public License,
+ * version 2, as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
+ * more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program.  If not, see <http://www.gnu.org/licenses/>.
+ */
+
+#include <linux/errno.h>
+#include <linux/types.h>
+#include <linux/bug.h>
+#include <linux/slab.h>
+
+#include "chip_support.h"
+#include "t20/t20.h"
+#include "t30/t30.h"
+
+#include "fuse.h"
+
+struct nvhost_chip_support *nvhost_chip_ops;
+
+struct nvhost_chip_support *nvhost_get_chip_ops(void)
+{
+	return nvhost_chip_ops;
+}
+
+int nvhost_init_chip_support(struct nvhost_master *host)
+{
+	int err = 0;
+
+	if (nvhost_chip_ops == NULL) {
+		nvhost_chip_ops = kzalloc(sizeof(*nvhost_chip_ops), GFP_KERNEL);
+		if (nvhost_chip_ops == NULL) {
+			pr_err("%s: Cannot allocate nvhost_chip_support\n",
+				__func__);
+			return 0;
+		}
+	}
+
+	switch (tegra_chip_id) {
+	case TEGRA20:
+		nvhost_chip_ops->soc_name = "tegra2x";
+		err = nvhost_init_t20_support(host, nvhost_chip_ops);
+		break;
+
+	case TEGRA30:
+		nvhost_chip_ops->soc_name = "tegra3x";
+		err = nvhost_init_t30_support(host, nvhost_chip_ops);
+		break;
+
+	default:
+		err = -ENODEV;
+	}
+
+	return err;
+}
diff --git a/drivers/video/tegra/host/chip_support.h b/drivers/video/tegra/host/chip_support.h
new file mode 100644
index 0000000..ce2ac22
--- /dev/null
+++ b/drivers/video/tegra/host/chip_support.h
@@ -0,0 +1,179 @@ 
+/*
+ * drivers/video/tegra/host/chip_support.h
+ *
+ * Tegra Graphics Host Chip Support
+ *
+ * Copyright (c) 2011-2012, NVIDIA Corporation.
+ *
+ * This program is free software; you can redistribute it and/or modify it
+ * under the terms and conditions of the GNU General Public License,
+ * version 2, as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
+ * more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program.  If not, see <http://www.gnu.org/licenses/>.
+ */
+#ifndef _NVHOST_CHIP_SUPPORT_H_
+#define _NVHOST_CHIP_SUPPORT_H_
+
+#include <linux/types.h>
+
+struct output;
+
+struct nvhost_master;
+struct nvhost_intr;
+struct nvhost_syncpt;
+struct nvhost_userctx_timeout;
+struct nvhost_channel;
+struct nvhost_cdma;
+struct nvhost_job;
+struct push_buffer;
+struct nvhost_syncpt;
+struct dentry;
+struct nvhost_job;
+struct nvhost_job_unpin_data;
+struct nvhost_intr_syncpt;
+struct mem_handle;
+struct mem_mgr;
+struct platform_device;
+
+struct nvhost_channel_ops {
+	const char *soc_name;
+	int (*init)(struct nvhost_channel *,
+		    struct nvhost_master *,
+		    int chid);
+	int (*submit)(struct nvhost_job *job);
+};
+
+struct nvhost_cdma_ops {
+	void (*start)(struct nvhost_cdma *);
+	void (*stop)(struct nvhost_cdma *);
+	void (*kick)(struct  nvhost_cdma *);
+	int (*timeout_init)(struct nvhost_cdma *,
+			    u32 syncpt_id);
+	void (*timeout_destroy)(struct nvhost_cdma *);
+	void (*timeout_teardown_begin)(struct nvhost_cdma *);
+	void (*timeout_teardown_end)(struct nvhost_cdma *,
+				     u32 getptr);
+	void (*timeout_cpu_incr)(struct nvhost_cdma *,
+				 u32 getptr,
+				 u32 syncpt_incrs,
+				 u32 syncval,
+				 u32 nr_slots);
+};
+
+struct nvhost_pushbuffer_ops {
+	void (*reset)(struct push_buffer *);
+	int (*init)(struct push_buffer *);
+	void (*destroy)(struct push_buffer *);
+	void (*push_to)(struct push_buffer *,
+			struct mem_mgr *, struct mem_handle *,
+			u32 op1, u32 op2);
+	void (*pop_from)(struct push_buffer *,
+			 unsigned int slots);
+	u32 (*space)(struct push_buffer *);
+	u32 (*putptr)(struct push_buffer *);
+};
+
+struct nvhost_debug_ops {
+	void (*debug_init)(struct dentry *de);
+	void (*show_channel_cdma)(struct nvhost_master *,
+				  struct nvhost_channel *,
+				  struct output *,
+				  int chid);
+	void (*show_channel_fifo)(struct nvhost_master *,
+				  struct nvhost_channel *,
+				  struct output *,
+				  int chid);
+	void (*show_mlocks)(struct nvhost_master *m,
+			    struct output *o);
+
+};
+
+struct nvhost_syncpt_ops {
+	void (*reset)(struct nvhost_syncpt *, u32 id);
+	void (*reset_wait_base)(struct nvhost_syncpt *, u32 id);
+	void (*read_wait_base)(struct nvhost_syncpt *, u32 id);
+	u32 (*update_min)(struct nvhost_syncpt *, u32 id);
+	void (*cpu_incr)(struct nvhost_syncpt *, u32 id);
+	int (*patch_wait)(struct nvhost_syncpt *sp,
+			void *patch_addr);
+	void (*debug)(struct nvhost_syncpt *);
+	const char * (*name)(struct nvhost_syncpt *, u32 id);
+};
+
+struct nvhost_intr_ops {
+	void (*init_host_sync)(struct nvhost_intr *);
+	void (*set_host_clocks_per_usec)(
+		struct nvhost_intr *, u32 clocks);
+	void (*set_syncpt_threshold)(
+		struct nvhost_intr *, u32 id, u32 thresh);
+	void (*enable_syncpt_intr)(struct nvhost_intr *, u32 id);
+	void (*disable_syncpt_intr)(struct nvhost_intr *, u32 id);
+	void (*disable_all_syncpt_intrs)(struct nvhost_intr *);
+	int  (*request_host_general_irq)(struct nvhost_intr *);
+	void (*free_host_general_irq)(struct nvhost_intr *);
+	int (*free_syncpt_irq)(struct nvhost_intr *);
+};
+
+struct nvhost_dev_ops {
+	struct nvhost_channel *(*alloc_nvhost_channel)(
+			struct platform_device *dev);
+	void (*free_nvhost_channel)(struct nvhost_channel *ch);
+};
+
+struct nvhost_mem_ops {
+	struct mem_mgr *(*alloc_mgr)(void);
+	void (*put_mgr)(struct mem_mgr *);
+	struct mem_mgr *(*get_mgr)(struct mem_mgr *);
+	struct mem_mgr *(*get_mgr_file)(int fd);
+	struct mem_handle *(*alloc)(struct mem_mgr *,
+			size_t size, size_t align,
+			int flags);
+	struct mem_handle *(*get)(struct mem_mgr *,
+			u32 id, struct platform_device *);
+	void (*put)(struct mem_mgr *, struct mem_handle *);
+	struct sg_table *(*pin)(struct mem_mgr *, struct mem_handle *);
+	void (*unpin)(struct mem_mgr *, struct mem_handle *, struct sg_table *);
+	void *(*mmap)(struct mem_handle *);
+	void (*munmap)(struct mem_handle *, void *);
+	void *(*kmap)(struct mem_handle *, unsigned int);
+	void (*kunmap)(struct mem_handle *, unsigned int, void *);
+	int (*pin_array_ids)(struct mem_mgr *,
+			struct platform_device *,
+			long unsigned *,
+			dma_addr_t *,
+			u32,
+			struct nvhost_job_unpin_data *);
+};
+
+struct nvhost_chip_support {
+	const char *soc_name;
+	struct nvhost_channel_ops channel;
+	struct nvhost_cdma_ops cdma;
+	struct nvhost_pushbuffer_ops push_buffer;
+	struct nvhost_debug_ops debug;
+	struct nvhost_syncpt_ops syncpt;
+	struct nvhost_intr_ops intr;
+	struct nvhost_dev_ops nvhost_dev;
+	struct nvhost_mem_ops mem;
+};
+
+struct nvhost_chip_support *nvhost_get_chip_ops(void);
+
+#define host_device_op()	(nvhost_get_chip_ops()->nvhost_dev)
+#define channel_cdma_op()	(nvhost_get_chip_ops()->cdma)
+#define channel_op()		(nvhost_get_chip_ops()->channel)
+#define syncpt_op()		(nvhost_get_chip_ops()->syncpt)
+#define intr_op()		(nvhost_get_chip_ops()->intr)
+#define cdma_op()		(nvhost_get_chip_ops()->cdma)
+#define cdma_pb_op()		(nvhost_get_chip_ops()->push_buffer)
+#define mem_op()		(nvhost_get_chip_ops()->mem)
+
+int nvhost_init_chip_support(struct nvhost_master *host);
+
+#endif /* _NVHOST_CHIP_SUPPORT_H_ */
diff --git a/drivers/video/tegra/host/debug.c b/drivers/video/tegra/host/debug.c
new file mode 100644
index 0000000..4f912f2
--- /dev/null
+++ b/drivers/video/tegra/host/debug.c
@@ -0,0 +1,255 @@ 
+/*
+ * drivers/video/tegra/host/debug.c
+ *
+ * Copyright (C) 2010 Google, Inc.
+ * Author: Erik Gilling <konkers@android.com>
+ *
+ * Copyright (C) 2011-2012 NVIDIA Corporation
+ *
+ * This software is licensed under the terms of the GNU General Public
+ * License version 2, as published by the Free Software Foundation, and
+ * may be copied, distributed, and modified under those terms.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ *
+ */
+
+#include <linux/debugfs.h>
+#include <linux/seq_file.h>
+#include <linux/uaccess.h>
+
+#include <linux/io.h>
+
+#include "dev.h"
+#include "debug.h"
+#include "nvhost_acm.h"
+#include "nvhost_channel.h"
+#include "chip_support.h"
+
+pid_t nvhost_debug_null_kickoff_pid;
+unsigned int nvhost_debug_trace_cmdbuf;
+
+pid_t nvhost_debug_force_timeout_pid;
+u32 nvhost_debug_force_timeout_val;
+u32 nvhost_debug_force_timeout_channel;
+
+void nvhost_debug_output(struct output *o, const char *fmt, ...)
+{
+	va_list args;
+	int len;
+
+	va_start(args, fmt);
+	len = vsnprintf(o->buf, sizeof(o->buf), fmt, args);
+	va_end(args);
+	o->fn(o->ctx, o->buf, len);
+}
+
+static int show_channels(struct platform_device *pdev, void *data)
+{
+	struct nvhost_channel *ch;
+	struct output *o = data;
+	struct nvhost_master *m;
+	struct nvhost_device_data *pdata;
+
+	if (pdev == NULL)
+		return 0;
+
+	pdata = platform_get_drvdata(pdev);
+	m = nvhost_get_host(pdev);
+	ch = pdata->channel;
+	if (ch) {
+		mutex_lock(&ch->reflock);
+		if (ch->refcount) {
+			mutex_lock(&ch->cdma.lock);
+			nvhost_get_chip_ops()->debug.show_channel_fifo(
+				m, ch, o, pdata->index);
+			nvhost_get_chip_ops()->debug.show_channel_cdma(
+				m, ch, o, pdata->index);
+			mutex_unlock(&ch->cdma.lock);
+		}
+		mutex_unlock(&ch->reflock);
+	}
+
+	return 0;
+}
+
+static void show_syncpts(struct nvhost_master *m, struct output *o)
+{
+	int i;
+	nvhost_debug_output(o, "---- syncpts ----\n");
+	for (i = 0; i < nvhost_syncpt_nb_pts(&m->syncpt); i++) {
+		u32 max = nvhost_syncpt_read_max(&m->syncpt, i);
+		u32 min = nvhost_syncpt_update_min(&m->syncpt, i);
+		if (!min && !max)
+			continue;
+		nvhost_debug_output(o, "id %d (%s) min %d max %d\n",
+			i, nvhost_get_chip_ops()->syncpt.name(&m->syncpt, i),
+			min, max);
+	}
+
+	for (i = 0; i < nvhost_syncpt_nb_bases(&m->syncpt); i++) {
+		u32 base_val;
+		base_val = nvhost_syncpt_read_wait_base(&m->syncpt, i);
+		if (base_val)
+			nvhost_debug_output(o, "waitbase id %d val %d\n",
+					i, base_val);
+	}
+
+	nvhost_debug_output(o, "\n");
+}
+
+static void show_all(struct nvhost_master *m, struct output *o)
+{
+	nvhost_module_busy(m->dev);
+
+	nvhost_get_chip_ops()->debug.show_mlocks(m, o);
+	show_syncpts(m, o);
+	nvhost_debug_output(o, "---- channels ----\n");
+	nvhost_device_list_for_all(o, show_channels);
+
+	nvhost_module_idle(m->dev);
+}
+
+#ifdef CONFIG_DEBUG_FS
+static int show_channels_no_fifo(struct platform_device *pdev, void *data)
+{
+	struct nvhost_channel *ch;
+	struct output *o = data;
+	struct nvhost_master *m;
+	struct nvhost_device_data *pdata;
+
+	if (pdev == NULL)
+		return 0;
+
+	pdata = platform_get_drvdata(pdev);
+	m = nvhost_get_host(pdev);
+	ch = pdata->channel;
+	if (ch) {
+		mutex_lock(&ch->reflock);
+		if (ch->refcount) {
+			mutex_lock(&ch->cdma.lock);
+			nvhost_get_chip_ops()->debug.show_channel_cdma(m,
+					ch, o, pdata->index);
+			mutex_unlock(&ch->cdma.lock);
+		}
+		mutex_unlock(&ch->reflock);
+	}
+
+	return 0;
+}
+
+static void show_all_no_fifo(struct nvhost_master *m, struct output *o)
+{
+	nvhost_module_busy(m->dev);
+
+	nvhost_get_chip_ops()->debug.show_mlocks(m, o);
+	show_syncpts(m, o);
+	nvhost_debug_output(o, "---- channels ----\n");
+	nvhost_device_list_for_all(o, show_channels_no_fifo);
+
+	nvhost_module_idle(m->dev);
+}
+
+static int nvhost_debug_show_all(struct seq_file *s, void *unused)
+{
+	struct output o = {
+		.fn = write_to_seqfile,
+		.ctx = s
+	};
+	show_all(s->private, &o);
+	return 0;
+}
+
+static int nvhost_debug_show(struct seq_file *s, void *unused)
+{
+	struct output o = {
+		.fn = write_to_seqfile,
+		.ctx = s
+	};
+	show_all_no_fifo(s->private, &o);
+	return 0;
+}
+
+static int nvhost_debug_open_all(struct inode *inode, struct file *file)
+{
+	return single_open(file, nvhost_debug_show_all, inode->i_private);
+}
+
+static const struct file_operations nvhost_debug_all_fops = {
+	.open		= nvhost_debug_open_all,
+	.read		= seq_read,
+	.llseek		= seq_lseek,
+	.release	= single_release,
+};
+
+static int nvhost_debug_open(struct inode *inode, struct file *file)
+{
+	return single_open(file, nvhost_debug_show, inode->i_private);
+}
+
+static const struct file_operations nvhost_debug_fops = {
+	.open		= nvhost_debug_open,
+	.read		= seq_read,
+	.llseek		= seq_lseek,
+	.release	= single_release,
+};
+
+void nvhost_device_debug_init(struct platform_device *dev)
+{
+	struct dentry *de = NULL;
+	struct nvhost_device_data *pdata = platform_get_drvdata(dev);
+
+	de = debugfs_create_dir(dev->name, de);
+
+	pdata->debugfs = de;
+}
+
+void nvhost_debug_init(struct nvhost_master *master)
+{
+	struct nvhost_device_data *pdata;
+	struct dentry *de = debugfs_create_dir("tegra_host", NULL);
+
+	if (!de)
+		return;
+
+	pdata = platform_get_drvdata(master->dev);
+
+	/* Store the created entry */
+	pdata->debugfs = de;
+
+	debugfs_create_file("status", S_IRUGO, de,
+			master, &nvhost_debug_fops);
+	debugfs_create_file("status_all", S_IRUGO, de,
+			master, &nvhost_debug_all_fops);
+
+	debugfs_create_u32("null_kickoff_pid", S_IRUGO|S_IWUSR, de,
+			&nvhost_debug_null_kickoff_pid);
+	debugfs_create_u32("trace_cmdbuf", S_IRUGO|S_IWUSR, de,
+			&nvhost_debug_trace_cmdbuf);
+
+	if (nvhost_get_chip_ops()->debug.debug_init)
+		nvhost_get_chip_ops()->debug.debug_init(de);
+
+	debugfs_create_u32("force_timeout_pid", S_IRUGO|S_IWUSR, de,
+			&nvhost_debug_force_timeout_pid);
+	debugfs_create_u32("force_timeout_val", S_IRUGO|S_IWUSR, de,
+			&nvhost_debug_force_timeout_val);
+	debugfs_create_u32("force_timeout_channel", S_IRUGO|S_IWUSR, de,
+			&nvhost_debug_force_timeout_channel);
+}
+#else
+void nvhost_debug_init(struct nvhost_master *master)
+{
+}
+#endif
+
+void nvhost_debug_dump(struct nvhost_master *master)
+{
+	struct output o = {
+		.fn = write_to_printk
+	};
+	show_all(master, &o);
+}
diff --git a/drivers/video/tegra/host/debug.h b/drivers/video/tegra/host/debug.h
new file mode 100644
index 0000000..0d1110a
--- /dev/null
+++ b/drivers/video/tegra/host/debug.h
@@ -0,0 +1,50 @@ 
+/*
+ * drivers/video/tegra/host/debug.h
+ *
+ * Tegra Graphics Host Debug
+ *
+ * Copyright (c) 2011-2012 NVIDIA Corporation.
+ *
+ * This program is free software; you can redistribute it and/or modify it
+ * under the terms and conditions of the GNU General Public License,
+ * version 2, as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
+ * more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program.  If not, see <http://www.gnu.org/licenses/>.
+ */
+#ifndef __NVHOST_DEBUG_H
+#define __NVHOST_DEBUG_H
+
+#include <linux/debugfs.h>
+#include <linux/seq_file.h>
+
+struct output {
+	void (*fn)(void *ctx, const char *str, size_t len);
+	void *ctx;
+	char buf[256];
+};
+
+static inline void write_to_seqfile(void *ctx, const char *str, size_t len)
+{
+	seq_write((struct seq_file *)ctx, str, len);
+}
+
+static inline void write_to_printk(void *ctx, const char *str, size_t len)
+{
+	pr_info("%s", str);
+}
+
+void nvhost_debug_output(struct output *o, const char *fmt, ...);
+
+extern pid_t nvhost_debug_null_kickoff_pid;
+extern pid_t nvhost_debug_force_timeout_pid;
+extern u32 nvhost_debug_force_timeout_val;
+extern u32 nvhost_debug_force_timeout_channel;
+extern unsigned int nvhost_debug_trace_cmdbuf;
+
+#endif /*__NVHOST_DEBUG_H */
diff --git a/drivers/video/tegra/host/dev.c b/drivers/video/tegra/host/dev.c
new file mode 100644
index 0000000..4743039
--- /dev/null
+++ b/drivers/video/tegra/host/dev.c
@@ -0,0 +1,104 @@ 
+/*
+ * drivers/video/tegra/host/dev.c
+ *
+ * Tegra Graphics Host Driver Entrypoint
+ *
+ * Copyright (c) 2010-2012, NVIDIA Corporation.
+ *
+ * This program is free software; you can redistribute it and/or modify it
+ * under the terms and conditions of the GNU General Public License,
+ * version 2, as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
+ * more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program.  If not, see <http://www.gnu.org/licenses/>.
+ */
+
+#define CREATE_TRACE_POINTS
+#include <trace/events/nvhost.h>
+
+#include <linux/module.h>
+#include <linux/io.h>
+#include <linux/nvhost.h>
+#include <linux/list.h>
+#include <linux/slab.h>
+
+/* host1x device list is used in 2 places:
+ * 1. In ioctl(NVHOST_IOCTL_CTRL_MODULE_REGRDWR) of host1x device
+ * 2. debug-fs dump of host1x and client device
+ *    as well as channel state */
+struct nvhost_device_list {
+	struct list_head list;
+	struct platform_device *pdev;
+};
+
+/* HEAD for the host1x device list */
+static struct nvhost_device_list ndev_head;
+
+/* Constructor for the host1x device list */
+void nvhost_device_list_init(void)
+{
+	INIT_LIST_HEAD(&ndev_head.list);
+}
+
+/* Adds a device to tail of host1x device list */
+int nvhost_device_list_add(struct platform_device *pdev)
+{
+	struct nvhost_device_list *list;
+
+	list = kzalloc(sizeof(struct nvhost_device_list), GFP_KERNEL);
+	if (!list)
+		return -ENOMEM;
+
+	list->pdev = pdev;
+	list_add_tail(&list->list, &ndev_head.list);
+
+	return 0;
+}
+
+/* Iterator function for host1x device list
+ * It takes a fptr as an argument and calls that function for each
+ * device in the list */
+void nvhost_device_list_for_all(void *data,
+	int (*fptr)(struct platform_device *pdev, void *fdata))
+{
+	struct list_head *pos;
+	struct nvhost_device_list *nlist;
+	int ret;
+
+	list_for_each(pos, &ndev_head.list) {
+		nlist = list_entry(pos, struct nvhost_device_list, list);
+		if (nlist && nlist->pdev && fptr) {
+			ret = fptr(nlist->pdev, data);
+			if (ret) {
+				pr_info("%s: iterator error\n", __func__);
+				break;
+			}
+		}
+	}
+}
+
+/* Removes a device from the host1x device list */
+void nvhost_device_list_remove(struct platform_device *pdev)
+{
+	struct list_head *pos;
+	struct nvhost_device_list *nlist;
+
+	list_for_each(pos, &ndev_head.list) {
+		nlist = list_entry(pos, struct nvhost_device_list, list);
+		if (nlist && nlist->pdev == pdev) {
+			list_del(&nlist->list);
+			kfree(nlist);
+		}
+	}
+}
+
+MODULE_AUTHOR("Terje Bergstrom <tbergstrom@nvidia.com>");
+MODULE_DESCRIPTION("Graphics host driver for Tegra products");
+MODULE_VERSION("1.0");
+MODULE_LICENSE("GPL");
+MODULE_ALIAS("platform-nvhost");
diff --git a/drivers/video/tegra/host/dev.h b/drivers/video/tegra/host/dev.h
new file mode 100644
index 0000000..12dfda5
--- /dev/null
+++ b/drivers/video/tegra/host/dev.h
@@ -0,0 +1,33 @@ 
+/*
+ * drivers/video/tegra/host/dev.h
+ *
+ * Copyright (c) 2012, NVIDIA Corporation.
+ *
+ * This program is free software; you can redistribute it and/or modify it
+ * under the terms and conditions of the GNU General Public License,
+ * version 2, as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
+ * more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program.  If not, see <http://www.gnu.org/licenses/>.
+ */
+
+#ifndef NVHOST_DEV_H
+#define NVHOST_DEV_H
+
+#include "host1x/host1x.h"
+
+struct platform_device;
+
+void nvhost_device_list_init(void);
+int nvhost_device_list_add(struct platform_device *pdev);
+void nvhost_device_list_for_all(void *data,
+	int (*fptr)(struct platform_device *pdev, void *fdata));
+struct platform_device *nvhost_device_list_match_by_id(u32 id);
+void nvhost_device_list_remove(struct platform_device *pdev);
+
+#endif
diff --git a/drivers/video/tegra/host/dmabuf.c b/drivers/video/tegra/host/dmabuf.c
new file mode 100644
index 0000000..0ae9d78
--- /dev/null
+++ b/drivers/video/tegra/host/dmabuf.c
@@ -0,0 +1,151 @@ 
+/*
+ * drivers/video/tegra/host/dmabuf.c
+ *
+ * Tegra Graphics Host DMA-BUF support
+ *
+ * Copyright (c) 2012, NVIDIA Corporation.
+ *
+ * This program is free software; you can redistribute it and/or modify it
+ * under the terms and conditions of the GNU General Public License,
+ * version 2, as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
+ * more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program.  If not, see <http://www.gnu.org/licenses/>.
+ */
+
+#include <linux/dma-buf.h>
+#include <linux/nvhost.h>
+#include "chip_support.h"
+#include "nvhost_memmgr.h"
+
+static inline struct dma_buf_attachment *to_dmabuf_att(struct mem_handle *h)
+{
+	return (struct dma_buf_attachment *)(((u32)h) & ~0x3);
+}
+
+static inline struct dma_buf *to_dmabuf(struct mem_handle *h)
+{
+	return to_dmabuf_att(h)->dmabuf;
+}
+
+static inline int to_dmabuf_fd(u32 id)
+{
+	return nvhost_memmgr_id(id) >> 2;
+}
+struct mem_handle *nvhost_dmabuf_alloc(size_t size, size_t align, int flags)
+{
+	/* TODO: Add allocation via DMA Mapping API */
+	return NULL;
+}
+
+void nvhost_dmabuf_put(struct mem_handle *handle)
+{
+	struct dma_buf_attachment *attach = to_dmabuf_att(handle);
+	struct dma_buf *dmabuf = attach->dmabuf;
+	dma_buf_detach(dmabuf, attach);
+	dma_buf_put(dmabuf);
+}
+
+struct sg_table *nvhost_dmabuf_pin(struct mem_handle *handle)
+{
+	return dma_buf_map_attachment(to_dmabuf_att(handle),
+				DMA_BIDIRECTIONAL);
+}
+
+void nvhost_dmabuf_unpin(struct mem_handle *handle, struct sg_table *sgt)
+{
+	dma_buf_unmap_attachment(to_dmabuf_att(handle), sgt, DMA_BIDIRECTIONAL);
+}
+
+
+void *nvhost_dmabuf_mmap(struct mem_handle *handle)
+{
+	return dma_buf_vmap(to_dmabuf(handle));
+}
+
+void nvhost_dmabuf_munmap(struct mem_handle *handle, void *addr)
+{
+	dma_buf_vunmap(to_dmabuf(handle), addr);
+}
+
+void *nvhost_dmabuf_kmap(struct mem_handle *handle, unsigned int pagenum)
+{
+	return dma_buf_kmap(to_dmabuf(handle), pagenum);
+}
+
+void nvhost_dmabuf_kunmap(struct mem_handle *handle, unsigned int pagenum,
+		void *addr)
+{
+	dma_buf_kunmap(to_dmabuf(handle), pagenum, addr);
+}
+
+struct mem_handle *nvhost_dmabuf_get(u32 id, struct platform_device *dev)
+{
+	struct mem_handle *h;
+	struct dma_buf *buf;
+
+	buf = dma_buf_get(to_dmabuf_fd(id));
+	if (IS_ERR_OR_NULL(buf))
+		return (struct mem_handle *)buf;
+	else {
+		h = (struct mem_handle *)dma_buf_attach(buf, &dev->dev);
+		if (IS_ERR_OR_NULL(h))
+			dma_buf_put(buf);
+	}
+
+	return (struct mem_handle *) ((u32)h | mem_mgr_type_dmabuf);
+}
+
+int nvhost_dmabuf_pin_array_ids(struct platform_device *dev,
+		long unsigned *ids,
+		long unsigned id_type_mask,
+		long unsigned id_type,
+		u32 count,
+		struct nvhost_job_unpin_data *unpin_data,
+		dma_addr_t *phys_addr) {
+	int i;
+	int pin_count = 0;
+	int err;
+
+	for (i = 0; i < count; i++) {
+		struct mem_handle *handle;
+		struct sg_table *sgt;
+
+		if ((ids[i] & id_type_mask) != id_type)
+			continue;
+
+		handle = nvhost_dmabuf_get(ids[i], dev);
+
+		if (IS_ERR(handle)) {
+			err = PTR_ERR(handle);
+			goto fail;
+		}
+
+		sgt = nvhost_dmabuf_pin(handle);
+		if (IS_ERR_OR_NULL(sgt)) {
+			nvhost_dmabuf_put(handle);
+			err = PTR_ERR(sgt);
+			goto fail;
+		}
+
+		phys_addr[i] = sg_dma_address(sgt->sgl);
+
+		unpin_data[pin_count].h = handle;
+		unpin_data[pin_count].mem = sgt;
+		pin_count++;
+	}
+	return pin_count;
+fail:
+	while (pin_count) {
+		pin_count--;
+		nvhost_dmabuf_unpin(unpin_data[pin_count].h,
+				unpin_data[pin_count].mem);
+		nvhost_dmabuf_put(unpin_data[pin_count].h);
+	}
+	return err;
+}
diff --git a/drivers/video/tegra/host/dmabuf.h b/drivers/video/tegra/host/dmabuf.h
new file mode 100644
index 0000000..fc26477
--- /dev/null
+++ b/drivers/video/tegra/host/dmabuf.h
@@ -0,0 +1,51 @@ 
+/*
+ * drivers/video/tegra/host/dmabuf.h
+ *
+ * Tegra Graphics Host dmabuf memory manager
+ *
+ * Copyright (c) 2012, NVIDIA Corporation.
+ *
+ * This program is free software; you can redistribute it and/or modify it
+ * under the terms and conditions of the GNU General Public License,
+ * version 2, as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
+ * more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program.  If not, see <http://www.gnu.org/licenses/>.
+ */
+
+#ifndef __NVHOST_DMABUF_H
+#define __NVHOST_DMABUF_H
+
+#include "nvhost_memmgr.h"
+
+struct nvhost_chip_support;
+struct platform_device;
+
+struct mem_mgr *nvhost_dmabuf_alloc_mgr(void);
+void nvhost_dmabuf_put_mgr(struct mem_mgr *mgr);
+struct mem_mgr *nvhost_dmabuf_get_mgr(struct mem_mgr *mgr);
+struct mem_mgr *nvhost_dmabuf_get_mgr_file(int fd);
+struct mem_handle *nvhost_dmabuf_alloc(struct mem_mgr *mgr,
+		size_t size, size_t align, int flags);
+void nvhost_dmabuf_put(struct mem_handle *handle);
+struct sg_table *nvhost_dmabuf_pin(struct mem_handle *handle);
+void nvhost_dmabuf_unpin(struct mem_handle *handle, struct sg_table *sgt);
+void *nvhost_dmabuf_mmap(struct mem_handle *handle);
+void nvhost_dmabuf_munmap(struct mem_handle *handle, void *addr);
+void *nvhost_dmabuf_kmap(struct mem_handle *handle, unsigned int pagenum);
+void nvhost_dmabuf_kunmap(struct mem_handle *handle, unsigned int pagenum,
+		void *addr);
+int nvhost_dmabuf_get(u32 id, struct platform_device *dev);
+int nvhost_dmabuf_pin_array_ids(struct platform_device *dev,
+		long unsigned *ids,
+		long unsigned id_type_mask,
+		long unsigned id_type,
+		u32 count,
+		struct nvhost_job_unpin_data *unpin_data,
+		dma_addr_t *phys_addr);
+#endif
diff --git a/drivers/video/tegra/host/host1x/Makefile b/drivers/video/tegra/host/host1x/Makefile
new file mode 100644
index 0000000..516e16f
--- /dev/null
+++ b/drivers/video/tegra/host/host1x/Makefile
@@ -0,0 +1,6 @@ 
+ccflags-y = -Idrivers/video/tegra/host
+
+nvhost-host1x-objs  = \
+	host1x.o
+
+obj-$(CONFIG_TEGRA_GRHOST) += nvhost-host1x.o
diff --git a/drivers/video/tegra/host/host1x/host1x.c b/drivers/video/tegra/host/host1x/host1x.c
new file mode 100644
index 0000000..db4389d
--- /dev/null
+++ b/drivers/video/tegra/host/host1x/host1x.c
@@ -0,0 +1,320 @@ 
+/*
+ * drivers/video/tegra/host/dev.c
+ *
+ * Tegra Graphics Host Driver Entrypoint
+ *
+ * Copyright (c) 2010-2012, NVIDIA Corporation.
+ *
+ * This program is free software; you can redistribute it and/or modify it
+ * under the terms and conditions of the GNU General Public License,
+ * version 2, as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
+ * more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program.  If not, see <http://www.gnu.org/licenses/>.
+ */
+
+#include <linux/slab.h>
+#include <linux/fs.h>
+#include <linux/cdev.h>
+#include <linux/uaccess.h>
+#include <linux/file.h>
+#include <linux/clk.h>
+#include <linux/hrtimer.h>
+#include <linux/module.h>
+#include <linux/of.h>
+
+#include "dev.h"
+#include <trace/events/nvhost.h>
+
+#include <linux/nvhost.h>
+
+#include "debug.h"
+#include "bus_client.h"
+#include "nvhost_acm.h"
+#include "nvhost_channel.h"
+#include "chip_support.h"
+
+#define DRIVER_NAME		"tegra-host1x"
+
+struct nvhost_master *nvhost;
+
+/* public sync point API */
+u32 host1x_syncpt_incr_max(u32 id, u32 incrs)
+{
+	struct nvhost_syncpt *sp = &nvhost->syncpt;
+	return nvhost_syncpt_incr_max(sp, id, incrs);
+}
+EXPORT_SYMBOL(host1x_syncpt_incr_max);
+
+void host1x_syncpt_incr(u32 id)
+{
+	struct nvhost_syncpt *sp = &nvhost->syncpt;
+	nvhost_syncpt_incr(sp, id);
+}
+EXPORT_SYMBOL(host1x_syncpt_incr);
+
+u32 host1x_syncpt_read(u32 id)
+{
+	struct nvhost_syncpt *sp = &nvhost->syncpt;
+	return nvhost_syncpt_read(sp, id);
+}
+EXPORT_SYMBOL(host1x_syncpt_read);
+
+int host1x_syncpt_wait(u32 id, u32 thresh,
+	u32 timeout, u32 *value)
+{
+	struct nvhost_syncpt *sp = &nvhost->syncpt;
+	return nvhost_syncpt_wait_timeout(sp, id, thresh, timeout, value);
+}
+EXPORT_SYMBOL(host1x_syncpt_wait);
+
+struct nvhost_ctrl_userctx {
+	struct nvhost_master *dev;
+	u32 *mod_locks;
+};
+
+static void power_on_host(struct platform_device *dev)
+{
+	struct nvhost_master *host = nvhost_get_private_data(dev);
+
+	nvhost_syncpt_reset(&host->syncpt);
+}
+
+static int power_off_host(struct platform_device *dev)
+{
+	struct nvhost_master *host = nvhost_get_private_data(dev);
+
+	nvhost_syncpt_save(&host->syncpt);
+	return 0;
+}
+
+static void clock_on_host(struct platform_device *dev)
+{
+	struct nvhost_device_data *pdata = platform_get_drvdata(dev);
+	struct nvhost_master *host = nvhost_get_private_data(dev);
+	nvhost_intr_start(&host->intr, clk_get_rate(pdata->clk[0]));
+}
+
+static int clock_off_host(struct platform_device *dev)
+{
+	struct nvhost_master *host = nvhost_get_private_data(dev);
+	nvhost_intr_stop(&host->intr);
+	return 0;
+}
+
+static int __devinit nvhost_user_init(struct nvhost_master *host)
+{
+	int err = 0;
+
+	host->nvhost_class = class_create(THIS_MODULE, IFACE_NAME);
+	if (IS_ERR(host->nvhost_class)) {
+		err = PTR_ERR(host->nvhost_class);
+		dev_err(&host->dev->dev, "failed to create class\n");
+	}
+
+	return err;
+}
+
+struct nvhost_channel *nvhost_alloc_channel(struct platform_device *dev)
+{
+	return host_device_op().alloc_nvhost_channel(dev);
+}
+
+void nvhost_free_channel(struct nvhost_channel *ch)
+{
+	host_device_op().free_nvhost_channel(ch);
+}
+
+static void nvhost_free_resources(struct nvhost_master *host)
+{
+	kfree(host->intr.syncpt);
+	host->intr.syncpt = 0;
+}
+
+static int __devinit nvhost_alloc_resources(struct nvhost_master *host)
+{
+	int err;
+
+	err = nvhost_init_chip_support(host);
+	if (err)
+		return err;
+
+	host->intr.syncpt = devm_kzalloc(&host->dev->dev,
+			sizeof(struct nvhost_intr_syncpt) *
+			nvhost_syncpt_nb_pts(&host->syncpt),
+			GFP_KERNEL);
+
+	if (!host->intr.syncpt) {
+		/* frees happen in the support removal phase */
+		return -ENOMEM;
+	}
+
+	return 0;
+}
+
+static int __devinit nvhost_probe(struct platform_device *dev)
+{
+	struct nvhost_master *host;
+	struct resource *regs, *intr0, *intr1;
+	int i, err;
+	struct nvhost_device_data *pdata =
+		(struct nvhost_device_data *)dev->dev.platform_data;
+
+	regs = platform_get_resource(dev, IORESOURCE_MEM, 0);
+	intr0 = platform_get_resource(dev, IORESOURCE_IRQ, 0);
+	intr1 = platform_get_resource(dev, IORESOURCE_IRQ, 1);
+
+	if (!regs || !intr0 || !intr1) {
+		dev_err(&dev->dev, "missing required platform resources\n");
+		return -ENXIO;
+	}
+
+	host = devm_kzalloc(&dev->dev, sizeof(*host), GFP_KERNEL);
+	if (!host)
+		return -ENOMEM;
+
+	nvhost = host;
+
+	host->dev = dev;
+
+	/* Copy host1x parameters. The private_data gets replaced
+	 * by nvhost_master later */
+	memcpy(&host->info, pdata->private_data,
+			sizeof(struct host1x_device_info));
+
+	pdata->finalize_poweron = power_on_host;
+	pdata->prepare_poweroff = power_off_host;
+	pdata->prepare_clockoff = clock_off_host;
+	pdata->finalize_clockon = clock_on_host;
+
+	pdata->pdev = dev;
+
+	/* set common host1x device data */
+	platform_set_drvdata(dev, pdata);
+
+	/* set private host1x device data */
+	nvhost_set_private_data(dev, host);
+
+	host->aperture = devm_request_and_ioremap(&dev->dev, regs);
+	if (!host->aperture) {
+		dev_err(&dev->dev, "failed to remap host registers\n");
+		err = -ENXIO;
+		goto fail;
+	}
+
+	err = nvhost_alloc_resources(host);
+	if (err) {
+		dev_err(&dev->dev, "failed to init chip support\n");
+		goto fail;
+	}
+
+	host->memmgr = mem_op().alloc_mgr();
+	if (!host->memmgr) {
+		dev_err(&dev->dev, "unable to create memory client\n");
+		err = -EIO;
+		goto fail;
+	}
+
+	err = nvhost_syncpt_init(dev, &host->syncpt);
+	if (err)
+		goto fail;
+
+	err = nvhost_intr_init(&host->intr, intr1->start, intr0->start);
+	if (err)
+		goto fail;
+
+	err = nvhost_user_init(host);
+	if (err)
+		goto fail;
+
+	err = nvhost_module_init(dev);
+	if (err)
+		goto fail;
+
+	for (i = 0; i < pdata->num_clks; i++)
+		clk_prepare_enable(pdata->clk[i]);
+	nvhost_syncpt_reset(&host->syncpt);
+	for (i = 0; i < pdata->num_clks; i++)
+		clk_disable_unprepare(pdata->clk[i]);
+
+	nvhost_device_list_init();
+	err = nvhost_device_list_add(dev);
+	if (err)
+		goto fail;
+
+	nvhost_debug_init(host);
+
+	dev_info(&dev->dev, "initialized\n");
+
+	return 0;
+
+fail:
+	nvhost_free_resources(host);
+	if (host->memmgr)
+		mem_op().put_mgr(host->memmgr);
+	kfree(host);
+	return err;
+}
+
+static int __exit nvhost_remove(struct platform_device *dev)
+{
+	struct nvhost_master *host = nvhost_get_private_data(dev);
+	nvhost_intr_deinit(&host->intr);
+	nvhost_syncpt_deinit(&host->syncpt);
+	nvhost_free_resources(host);
+	return 0;
+}
+
+static int nvhost_suspend(struct platform_device *dev, pm_message_t state)
+{
+	struct nvhost_master *host = nvhost_get_private_data(dev);
+	int ret = 0;
+
+	ret = nvhost_module_suspend(host->dev);
+	dev_info(&dev->dev, "suspend status: %d\n", ret);
+
+	return ret;
+}
+
+static int nvhost_resume(struct platform_device *dev)
+{
+	dev_info(&dev->dev, "resuming\n");
+	return 0;
+}
+
+static struct of_device_id host1x_match[] __devinitdata = {
+	{ .compatible = "nvidia,tegra20-host1x", },
+	{ .compatible = "nvidia,tegra30-host1x", },
+	{ },
+};
+
+static struct platform_driver platform_driver = {
+	.probe = nvhost_probe,
+	.remove = __exit_p(nvhost_remove),
+	.suspend = nvhost_suspend,
+	.resume = nvhost_resume,
+	.driver = {
+		.owner = THIS_MODULE,
+		.name = DRIVER_NAME,
+		.of_match_table = of_match_ptr(host1x_match),
+	},
+};
+
+static int __init nvhost_mod_init(void)
+{
+	return platform_driver_register(&platform_driver);
+}
+
+static void __exit nvhost_mod_exit(void)
+{
+	platform_driver_unregister(&platform_driver);
+}
+
+module_init(nvhost_mod_init);
+module_exit(nvhost_mod_exit);
+
diff --git a/drivers/video/tegra/host/host1x/host1x.h b/drivers/video/tegra/host/host1x/host1x.h
new file mode 100644
index 0000000..46ecdf4
--- /dev/null
+++ b/drivers/video/tegra/host/host1x/host1x.h
@@ -0,0 +1,97 @@ 
+/*
+ * drivers/video/tegra/host/host1x/host1x.h
+ *
+ * Tegra Graphics Host Driver Entrypoint
+ *
+ * Copyright (c) 2010-2012, NVIDIA Corporation.
+ *
+ * This program is free software; you can redistribute it and/or modify it
+ * under the terms and conditions of the GNU General Public License,
+ * version 2, as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
+ * more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program.  If not, see <http://www.gnu.org/licenses/>.
+ */
+
+#ifndef __NVHOST_HOST1X_H
+#define __NVHOST_HOST1X_H
+
+#include <linux/cdev.h>
+#include <linux/nvhost.h>
+
+#include "nvhost_syncpt.h"
+#include "nvhost_intr.h"
+
+#define TRACE_MAX_LENGTH	128U
+#define IFACE_NAME		"nvhost"
+
+struct nvhost_channel;
+struct mem_mgr;
+
+struct nvhost_master {
+	void __iomem *aperture;
+	void __iomem *sync_aperture;
+	struct class *nvhost_class;
+	struct cdev cdev;
+	struct device *ctrl;
+	struct nvhost_syncpt syncpt;
+	struct mem_mgr *memmgr;
+	struct nvhost_intr intr;
+	struct platform_device *dev;
+	atomic_t clientid;
+	struct host1x_device_info info;
+};
+
+extern struct nvhost_master *nvhost;
+
+void nvhost_debug_init(struct nvhost_master *master);
+void nvhost_device_debug_init(struct platform_device *dev);
+void nvhost_debug_dump(struct nvhost_master *master);
+
+struct nvhost_channel *nvhost_alloc_channel(struct platform_device *dev);
+void nvhost_free_channel(struct nvhost_channel *ch);
+
+extern pid_t nvhost_debug_null_kickoff_pid;
+
+static inline void *nvhost_get_private_data(struct platform_device *_dev)
+{
+	struct nvhost_device_data *pdata =
+		(struct nvhost_device_data *)platform_get_drvdata(_dev);
+	WARN_ON(!pdata);
+	return (pdata && pdata->private_data) ? pdata->private_data : NULL;
+}
+
+static inline void nvhost_set_private_data(struct platform_device *_dev,
+	void *priv_data)
+{
+	struct nvhost_device_data *pdata =
+		(struct nvhost_device_data *)platform_get_drvdata(_dev);
+	WARN_ON(!pdata);
+	if (pdata)
+		pdata->private_data = priv_data;
+}
+
+static inline
+struct nvhost_master *nvhost_get_host(struct platform_device *_dev)
+{
+	struct platform_device *pdev;
+
+	if (_dev->dev.parent) {
+		pdev = to_platform_device(_dev->dev.parent);
+		return nvhost_get_private_data(pdev);
+	} else
+		return nvhost_get_private_data(_dev);
+}
+
+static inline
+struct platform_device *nvhost_get_parent(struct platform_device *_dev)
+{
+	return _dev->dev.parent ? to_platform_device(_dev->dev.parent) : NULL;
+}
+
+#endif
diff --git a/drivers/video/tegra/host/host1x/host1x01_hardware.h b/drivers/video/tegra/host/host1x/host1x01_hardware.h
new file mode 100644
index 0000000..75468de
--- /dev/null
+++ b/drivers/video/tegra/host/host1x/host1x01_hardware.h
@@ -0,0 +1,157 @@ 
+/*
+ * drivers/video/tegra/host/host1x/host1x01_hardware.h
+ *
+ * Tegra Graphics Host Register Offsets for T20/T30
+ *
+ * Copyright (c) 2010-2012 NVIDIA Corporation.
+ *
+ * This program is free software; you can redistribute it and/or modify it
+ * under the terms and conditions of the GNU General Public License,
+ * version 2, as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
+ * more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program.  If not, see <http://www.gnu.org/licenses/>.
+ */
+
+#ifndef __NVHOST_HOST1X01_HARDWARE_H
+#define __NVHOST_HOST1X01_HARDWARE_H
+
+#include <linux/types.h>
+#include <linux/bitops.h>
+#include "hw_host1x01_channel.h"
+#include "hw_host1x01_sync.h"
+#include "hw_host1x01_uclass.h"
+
+/* channel registers */
+#define NV_HOST1X_CHANNEL_MAP_SIZE_BYTES 16384
+#define NV_HOST1X_SYNC_MLOCK_NUM 16
+
+/* sync registers */
+#define HOST1X_CHANNEL_SYNC_REG_BASE   0x3000
+#define NV_HOST1X_NB_MLOCKS 16
+
+static inline u32 nvhost_class_host_wait_syncpt(
+	unsigned indx, unsigned threshold)
+{
+	return host1x_uclass_wait_syncpt_indx_f(indx)
+		| host1x_uclass_wait_syncpt_thresh_f(threshold);
+}
+
+static inline u32 nvhost_class_host_load_syncpt_base(
+	unsigned indx, unsigned threshold)
+{
+	return host1x_uclass_load_syncpt_base_base_indx_f(indx)
+		| host1x_uclass_load_syncpt_base_value_f(threshold);
+}
+
+static inline u32 nvhost_class_host_wait_syncpt_base(
+	unsigned indx, unsigned base_indx, unsigned offset)
+{
+	return host1x_uclass_wait_syncpt_base_indx_f(indx)
+		| host1x_uclass_wait_syncpt_base_base_indx_f(base_indx)
+		| host1x_uclass_wait_syncpt_base_offset_f(offset);
+}
+
+static inline u32 nvhost_class_host_incr_syncpt_base(
+	unsigned base_indx, unsigned offset)
+{
+	return host1x_uclass_incr_syncpt_base_base_indx_f(base_indx)
+		| host1x_uclass_incr_syncpt_base_offset_f(offset);
+}
+
+static inline u32 nvhost_class_host_incr_syncpt(
+	unsigned cond, unsigned indx)
+{
+	return host1x_uclass_incr_syncpt_cond_f(cond)
+		| host1x_uclass_incr_syncpt_indx_f(indx);
+}
+
+static inline u32 nvhost_class_host_indoff_reg_write(
+	unsigned mod_id, unsigned offset, bool auto_inc)
+{
+	u32 v = host1x_uclass_indoff_indbe_f(0xf)
+		| host1x_uclass_indoff_indmodid_f(mod_id)
+		| host1x_uclass_indoff_indroffset_f(offset);
+	if (auto_inc)
+		v |= host1x_uclass_indoff_autoinc_f(1);
+	return v;
+}
+
+static inline u32 nvhost_class_host_indoff_reg_read(
+	unsigned mod_id, unsigned offset, bool auto_inc)
+{
+	u32 v = host1x_uclass_indoff_indmodid_f(mod_id)
+		| host1x_uclass_indoff_indroffset_f(offset)
+		| host1x_uclass_indoff_rwn_read_v();
+	if (auto_inc)
+		v |= host1x_uclass_indoff_autoinc_f(1);
+	return v;
+}
+
+
+/* cdma opcodes */
+static inline u32 nvhost_opcode_setclass(
+	unsigned class_id, unsigned offset, unsigned mask)
+{
+	return (0 << 28) | (offset << 16) | (class_id << 6) | mask;
+}
+
+static inline u32 nvhost_opcode_incr(unsigned offset, unsigned count)
+{
+	return (1 << 28) | (offset << 16) | count;
+}
+
+static inline u32 nvhost_opcode_nonincr(unsigned offset, unsigned count)
+{
+	return (2 << 28) | (offset << 16) | count;
+}
+
+static inline u32 nvhost_opcode_mask(unsigned offset, unsigned mask)
+{
+	return (3 << 28) | (offset << 16) | mask;
+}
+
+static inline u32 nvhost_opcode_imm(unsigned offset, unsigned value)
+{
+	return (4 << 28) | (offset << 16) | value;
+}
+
+static inline u32 nvhost_opcode_imm_incr_syncpt(unsigned cond, unsigned indx)
+{
+	return nvhost_opcode_imm(host1x_uclass_incr_syncpt_r(),
+		nvhost_class_host_incr_syncpt(cond, indx));
+}
+
+static inline u32 nvhost_opcode_restart(unsigned address)
+{
+	return (5 << 28) | (address >> 4);
+}
+
+static inline u32 nvhost_opcode_gather(unsigned count)
+{
+	return (6 << 28) | count;
+}
+
+static inline u32 nvhost_opcode_gather_nonincr(unsigned offset,	unsigned count)
+{
+	return (6 << 28) | (offset << 16) | BIT(15) | count;
+}
+
+static inline u32 nvhost_opcode_gather_incr(unsigned offset, unsigned count)
+{
+	return (6 << 28) | (offset << 16) | BIT(15) | BIT(14) | count;
+}
+
+#define NVHOST_OPCODE_NOOP nvhost_opcode_nonincr(0, 0)
+
+static inline u32 nvhost_mask2(unsigned x, unsigned y)
+{
+	return 1 | (1 << (y - x));
+}
+
+#endif
diff --git a/drivers/video/tegra/host/host1x/host1x_cdma.c b/drivers/video/tegra/host/host1x/host1x_cdma.c
new file mode 100644
index 0000000..224b721
--- /dev/null
+++ b/drivers/video/tegra/host/host1x/host1x_cdma.c
@@ -0,0 +1,488 @@ 
+/*
+ * drivers/video/tegra/host/host1x/host1x_cdma.c
+ *
+ * Tegra Graphics Host Command DMA
+ *
+ * Copyright (c) 2010-2012, NVIDIA Corporation.
+ *
+ * This program is free software; you can redistribute it and/or modify it
+ * under the terms and conditions of the GNU General Public License,
+ * version 2, as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
+ * more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program.  If not, see <http://www.gnu.org/licenses/>.
+ */
+
+#include <linux/slab.h>
+#include <linux/scatterlist.h>
+#include <linux/dma-mapping.h>
+#include "nvhost_acm.h"
+#include "nvhost_cdma.h"
+#include "nvhost_channel.h"
+#include "dev.h"
+#include "chip_support.h"
+#include "nvhost_memmgr.h"
+
+#include "host1x_cdma.h"
+
+static inline u32 host1x_channel_dmactrl(int stop, int get_rst, int init_get)
+{
+	return host1x_channel_dmactrl_dmastop_f(stop)
+		| host1x_channel_dmactrl_dmagetrst_f(get_rst)
+		| host1x_channel_dmactrl_dmainitget_f(init_get);
+}
+
+static void cdma_timeout_handler(struct work_struct *work);
+
+/*
+ * push_buffer
+ *
+ * The push buffer is a circular array of words to be fetched by command DMA.
+ * Note that it works slightly differently to the sync queue; fence == cur
+ * means that the push buffer is full, not empty.
+ */
+
+
+/**
+ * Reset to empty push buffer
+ */
+static void push_buffer_reset(struct push_buffer *pb)
+{
+	pb->fence = PUSH_BUFFER_SIZE - 8;
+	pb->cur = 0;
+}
+
+/**
+ * Init push buffer resources
+ */
+static void push_buffer_destroy(struct push_buffer *pb);
+static int push_buffer_init(struct push_buffer *pb)
+{
+	struct nvhost_cdma *cdma = pb_to_cdma(pb);
+	struct nvhost_master *master = cdma_to_dev(cdma);
+	pb->mapped = NULL;
+	pb->phys = 0;
+	pb->client_handle = NULL;
+
+	cdma_pb_op().reset(pb);
+
+	/* allocate and map pushbuffer memory */
+	pb->mapped = dma_alloc_writecombine(&master->dev->dev,
+			PUSH_BUFFER_SIZE + 4, &pb->phys, GFP_KERNEL);
+	if (IS_ERR_OR_NULL(pb->mapped)) {
+		pb->mapped = NULL;
+		goto fail;
+	}
+
+	/* memory for storing mem client and handles for each opcode pair */
+	pb->client_handle = kzalloc(NVHOST_GATHER_QUEUE_SIZE *
+				sizeof(struct mem_mgr_handle),
+			GFP_KERNEL);
+	if (!pb->client_handle)
+		goto fail;
+
+	/* put the restart at the end of pushbuffer memory */
+	*(pb->mapped + (PUSH_BUFFER_SIZE >> 2)) =
+		nvhost_opcode_restart(pb->phys);
+
+	return 0;
+
+fail:
+	push_buffer_destroy(pb);
+	return -ENOMEM;
+}
+
+/**
+ * Clean up push buffer resources
+ */
+static void push_buffer_destroy(struct push_buffer *pb)
+{
+	struct nvhost_cdma *cdma = pb_to_cdma(pb);
+	struct nvhost_master *master = cdma_to_dev(cdma);
+
+	if (pb->phys != 0)
+		dma_free_writecombine(&master->dev->dev,
+				PUSH_BUFFER_SIZE + 4,
+				pb->mapped, pb->phys);
+
+	kfree(pb->client_handle);
+
+	pb->mapped = NULL;
+	pb->phys = 0;
+	pb->client_handle = 0;
+}
+
+/**
+ * Push two words to the push buffer
+ * Caller must ensure push buffer is not full
+ */
+static void push_buffer_push_to(struct push_buffer *pb,
+		struct mem_mgr *client, struct mem_handle *handle,
+		u32 op1, u32 op2)
+{
+	u32 cur = pb->cur;
+	u32 *p = (u32 *)((u32)pb->mapped + cur);
+	u32 cur_mem = (cur/8) & (NVHOST_GATHER_QUEUE_SIZE - 1);
+	WARN_ON(cur == pb->fence);
+	*(p++) = op1;
+	*(p++) = op2;
+	pb->client_handle[cur_mem].client = client;
+	pb->client_handle[cur_mem].handle = handle;
+	pb->cur = (cur + 8) & (PUSH_BUFFER_SIZE - 1);
+}
+
+/**
+ * Pop a number of two word slots from the push buffer
+ * Caller must ensure push buffer is not empty
+ */
+static void push_buffer_pop_from(struct push_buffer *pb,
+		unsigned int slots)
+{
+	/* Clear the mem references for old items from pb */
+	unsigned int i;
+	u32 fence_mem = pb->fence/8;
+	for (i = 0; i < slots; i++) {
+		int cur_fence_mem = (fence_mem+i)
+				& (NVHOST_GATHER_QUEUE_SIZE - 1);
+		struct mem_mgr_handle *h = &pb->client_handle[cur_fence_mem];
+		h->client = NULL;
+		h->handle = NULL;
+	}
+	/* Advance the next write position */
+	pb->fence = (pb->fence + slots * 8) & (PUSH_BUFFER_SIZE - 1);
+}
+
+/**
+ * Return the number of two word slots free in the push buffer
+ */
+static u32 push_buffer_space(struct push_buffer *pb)
+{
+	return ((pb->fence - pb->cur) & (PUSH_BUFFER_SIZE - 1)) / 8;
+}
+
+static u32 push_buffer_putptr(struct push_buffer *pb)
+{
+	return pb->phys + pb->cur;
+}
+
+/*
+ * The syncpt incr buffer is filled with methods to increment syncpts, which
+ * is later GATHER-ed into the mainline PB. It's used when a timed out context
+ * is interleaved with other work, so needs to inline the syncpt increments
+ * to maintain the count (but otherwise does no work).
+ */
+
+/**
+ * Init timeout resources
+ */
+static int cdma_timeout_init(struct nvhost_cdma *cdma,
+				 u32 syncpt_id)
+{
+	if (syncpt_id == NVSYNCPT_INVALID)
+		return -EINVAL;
+
+	INIT_DELAYED_WORK(&cdma->timeout.wq, cdma_timeout_handler);
+	cdma->timeout.initialized = true;
+
+	return 0;
+}
+
+/**
+ * Clean up timeout resources
+ */
+static void cdma_timeout_destroy(struct nvhost_cdma *cdma)
+{
+	if (cdma->timeout.initialized)
+		cancel_delayed_work(&cdma->timeout.wq);
+	cdma->timeout.initialized = false;
+}
+
+/**
+ * Increment timedout buffer's syncpt via CPU.
+ */
+static void cdma_timeout_cpu_incr(struct nvhost_cdma *cdma, u32 getptr,
+				u32 syncpt_incrs, u32 syncval, u32 nr_slots)
+{
+	struct nvhost_master *dev = cdma_to_dev(cdma);
+	struct push_buffer *pb = &cdma->push_buffer;
+	u32 i, getidx;
+
+	for (i = 0; i < syncpt_incrs; i++)
+		nvhost_syncpt_cpu_incr(&dev->syncpt, cdma->timeout.syncpt_id);
+
+	/* after CPU incr, ensure shadow is up to date */
+	nvhost_syncpt_update_min(&dev->syncpt, cdma->timeout.syncpt_id);
+
+	/* NOP all the PB slots */
+	getidx = getptr - pb->phys;
+	while (nr_slots--) {
+		u32 *p = (u32 *)((u32)pb->mapped + getidx);
+		*(p++) = NVHOST_OPCODE_NOOP;
+		*(p++) = NVHOST_OPCODE_NOOP;
+		dev_dbg(&dev->dev->dev, "%s: NOP at 0x%x\n",
+			__func__, pb->phys + getidx);
+		getidx = (getidx + 8) & (PUSH_BUFFER_SIZE - 1);
+	}
+	wmb();
+}
+
+/**
+ * Start channel DMA
+ */
+static void cdma_start(struct nvhost_cdma *cdma)
+{
+	void __iomem *chan_regs = cdma_to_channel(cdma)->aperture;
+
+	if (cdma->running)
+		return;
+
+	cdma->last_put = cdma_pb_op().putptr(&cdma->push_buffer);
+
+	writel(host1x_channel_dmactrl(true, false, false),
+		chan_regs + host1x_channel_dmactrl_r());
+
+	/* set base, put, end pointer (all of memory) */
+	writel(0, chan_regs + host1x_channel_dmastart_r());
+	writel(cdma->last_put, chan_regs + host1x_channel_dmaput_r());
+	writel(0xFFFFFFFF, chan_regs + host1x_channel_dmaend_r());
+
+	/* reset GET */
+	writel(host1x_channel_dmactrl(true, true, true),
+		chan_regs + host1x_channel_dmactrl_r());
+
+	/* start the command DMA */
+	writel(host1x_channel_dmactrl(false, false, false),
+		chan_regs + host1x_channel_dmactrl_r());
+
+	cdma->running = true;
+}
+
+/**
+ * Similar to cdma_start(), but rather than starting from an idle
+ * state (where DMA GET is set to DMA PUT), on a timeout we restore
+ * DMA GET from an explicit value (so DMA may again be pending).
+ */
+static void cdma_timeout_restart(struct nvhost_cdma *cdma, u32 getptr)
+{
+	struct nvhost_master *dev = cdma_to_dev(cdma);
+	void __iomem *chan_regs = cdma_to_channel(cdma)->aperture;
+
+	if (cdma->running)
+		return;
+
+	cdma->last_put = cdma_pb_op().putptr(&cdma->push_buffer);
+
+	writel(host1x_channel_dmactrl(true, false, false),
+		chan_regs + host1x_channel_dmactrl_r());
+
+	/* set base, end pointer (all of memory) */
+	writel(0, chan_regs + host1x_channel_dmastart_r());
+	writel(0xFFFFFFFF, chan_regs + host1x_channel_dmaend_r());
+
+	/* set GET, by loading the value in PUT (then reset GET) */
+	writel(getptr, chan_regs + host1x_channel_dmaput_r());
+	writel(host1x_channel_dmactrl(true, true, true),
+		chan_regs + host1x_channel_dmactrl_r());
+
+	dev_dbg(&dev->dev->dev,
+		"%s: DMA GET 0x%x, PUT HW 0x%x / shadow 0x%x\n",
+		__func__,
+		readl(chan_regs + host1x_channel_dmaget_r()),
+		readl(chan_regs + host1x_channel_dmaput_r()),
+		cdma->last_put);
+
+	/* deassert GET reset and set PUT */
+	writel(host1x_channel_dmactrl(true, false, false),
+		chan_regs + host1x_channel_dmactrl_r());
+	writel(cdma->last_put, chan_regs + host1x_channel_dmaput_r());
+
+	/* start the command DMA */
+	writel(host1x_channel_dmactrl(false, false, false),
+		chan_regs + host1x_channel_dmactrl_r());
+
+	cdma->running = true;
+}
+
+/**
+ * Kick channel DMA into action by writing its PUT offset (if it has changed)
+ */
+static void cdma_kick(struct nvhost_cdma *cdma)
+{
+	u32 put;
+
+	put = cdma_pb_op().putptr(&cdma->push_buffer);
+
+	if (put != cdma->last_put) {
+		void __iomem *chan_regs = cdma_to_channel(cdma)->aperture;
+		writel(put, chan_regs + host1x_channel_dmaput_r());
+		cdma->last_put = put;
+	}
+}
+
+static void cdma_stop(struct nvhost_cdma *cdma)
+{
+	void __iomem *chan_regs = cdma_to_channel(cdma)->aperture;
+
+	mutex_lock(&cdma->lock);
+	if (cdma->running) {
+		nvhost_cdma_wait_locked(cdma, CDMA_EVENT_SYNC_QUEUE_EMPTY);
+		writel(host1x_channel_dmactrl(true, false, false),
+			chan_regs + host1x_channel_dmactrl_r());
+		cdma->running = false;
+	}
+	mutex_unlock(&cdma->lock);
+}
+
+/**
+ * Stops both channel's command processor and CDMA immediately.
+ * Also, tears down the channel and resets corresponding module.
+ */
+static void cdma_timeout_teardown_begin(struct nvhost_cdma *cdma)
+{
+	struct nvhost_master *dev = cdma_to_dev(cdma);
+	struct nvhost_channel *ch = cdma_to_channel(cdma);
+	u32 cmdproc_stop;
+
+	if (cdma->torndown && !cdma->running) {
+		dev_warn(&dev->dev->dev, "Already torn down\n");
+		return;
+	}
+
+	dev_dbg(&dev->dev->dev,
+		"begin channel teardown (channel id %d)\n", ch->chid);
+
+	cmdproc_stop = readl(dev->sync_aperture + host1x_sync_cmdproc_stop_r());
+	cmdproc_stop |= BIT(ch->chid);
+	writel(cmdproc_stop, dev->sync_aperture + host1x_sync_cmdproc_stop_r());
+
+	dev_dbg(&dev->dev->dev,
+		"%s: DMA GET 0x%x, PUT HW 0x%x / shadow 0x%x\n",
+		__func__,
+		readl(ch->aperture + host1x_channel_dmaget_r()),
+		readl(ch->aperture + host1x_channel_dmaput_r()),
+		cdma->last_put);
+
+	writel(host1x_channel_dmactrl(true, false, false),
+		ch->aperture + host1x_channel_dmactrl_r());
+
+	writel(BIT(ch->chid), dev->sync_aperture + host1x_sync_ch_teardown_r());
+
+	cdma->running = false;
+	cdma->torndown = true;
+}
+
+static void cdma_timeout_teardown_end(struct nvhost_cdma *cdma, u32 getptr)
+{
+	struct nvhost_master *dev = cdma_to_dev(cdma);
+	struct nvhost_channel *ch = cdma_to_channel(cdma);
+	u32 cmdproc_stop;
+
+	dev_dbg(&dev->dev->dev,
+		"end channel teardown (id %d, DMAGET restart = 0x%x)\n",
+		ch->chid, getptr);
+
+	cmdproc_stop = readl(dev->sync_aperture + host1x_sync_cmdproc_stop_r());
+	cmdproc_stop &= ~(BIT(ch->chid));
+	writel(cmdproc_stop, dev->sync_aperture + host1x_sync_cmdproc_stop_r());
+
+	cdma->torndown = false;
+	cdma_timeout_restart(cdma, getptr);
+}
+
+/**
+ * If this timeout fires, it indicates the current sync_queue entry has
+ * exceeded its TTL and the userctx should be timed out and remaining
+ * submits already issued cleaned up (future submits return an error).
+ */
+static void cdma_timeout_handler(struct work_struct *work)
+{
+	struct nvhost_cdma *cdma;
+	struct nvhost_master *dev;
+	struct nvhost_syncpt *sp;
+	struct nvhost_channel *ch;
+
+	u32 syncpt_val;
+
+	u32 prev_cmdproc, cmdproc_stop;
+
+	cdma = container_of(to_delayed_work(work), struct nvhost_cdma,
+			    timeout.wq);
+	dev = cdma_to_dev(cdma);
+	sp = &dev->syncpt;
+	ch = cdma_to_channel(cdma);
+
+	nvhost_debug_dump(cdma_to_dev(cdma));
+
+	mutex_lock(&cdma->lock);
+
+	if (!cdma->timeout.clientid) {
+		dev_dbg(&dev->dev->dev,
+			 "cdma_timeout: expired, but has no clientid\n");
+		mutex_unlock(&cdma->lock);
+		return;
+	}
+
+	/* stop processing to get a clean snapshot */
+	prev_cmdproc = readl(dev->sync_aperture + host1x_sync_cmdproc_stop_r());
+	cmdproc_stop = prev_cmdproc | BIT(ch->chid);
+	writel(cmdproc_stop, dev->sync_aperture + host1x_sync_cmdproc_stop_r());
+
+	dev_dbg(&dev->dev->dev, "cdma_timeout: cmdproc was 0x%x is 0x%x\n",
+		prev_cmdproc, cmdproc_stop);
+
+	syncpt_val = nvhost_syncpt_update_min(&dev->syncpt,
+			cdma->timeout.syncpt_id);
+
+	/* has buffer actually completed? */
+	if ((s32)(syncpt_val - cdma->timeout.syncpt_val) >= 0) {
+		dev_dbg(&dev->dev->dev,
+			 "cdma_timeout: expired, but buffer had completed\n");
+		/* restore */
+		cmdproc_stop = prev_cmdproc & ~(BIT(ch->chid));
+		writel(cmdproc_stop,
+			dev->sync_aperture + host1x_sync_cmdproc_stop_r());
+		mutex_unlock(&cdma->lock);
+		return;
+	}
+
+	dev_warn(&dev->dev->dev,
+		"%s: timeout: %d (%s), HW thresh %d, done %d\n",
+		__func__,
+		cdma->timeout.syncpt_id,
+		syncpt_op().name(sp, cdma->timeout.syncpt_id),
+		syncpt_val, cdma->timeout.syncpt_val);
+
+	/* stop HW, resetting channel/module */
+	cdma_op().timeout_teardown_begin(cdma);
+
+	nvhost_cdma_update_sync_queue(cdma, sp, ch->dev);
+	mutex_unlock(&cdma->lock);
+}
+
+static const struct nvhost_cdma_ops host1x_cdma_ops = {
+	.start = cdma_start,
+	.stop = cdma_stop,
+	.kick = cdma_kick,
+
+	.timeout_init = cdma_timeout_init,
+	.timeout_destroy = cdma_timeout_destroy,
+	.timeout_teardown_begin = cdma_timeout_teardown_begin,
+	.timeout_teardown_end = cdma_timeout_teardown_end,
+	.timeout_cpu_incr = cdma_timeout_cpu_incr,
+};
+
+static const struct nvhost_pushbuffer_ops host1x_pushbuffer_ops = {
+	.reset = push_buffer_reset,
+	.init = push_buffer_init,
+	.destroy = push_buffer_destroy,
+	.push_to = push_buffer_push_to,
+	.pop_from = push_buffer_pop_from,
+	.space = push_buffer_space,
+	.putptr = push_buffer_putptr,
+};
+
diff --git a/drivers/video/tegra/host/host1x/host1x_cdma.h b/drivers/video/tegra/host/host1x/host1x_cdma.h
new file mode 100644
index 0000000..94bfc09
--- /dev/null
+++ b/drivers/video/tegra/host/host1x/host1x_cdma.h
@@ -0,0 +1,39 @@ 
+/*
+ * drivers/video/tegra/host/host1x/host1x_cdma.h
+ *
+ * Tegra Graphics Host Command DMA
+ *
+ * Copyright (c) 2011-2012, NVIDIA Corporation.
+ *
+ * This program is free software; you can redistribute it and/or modify it
+ * under the terms and conditions of the GNU General Public License,
+ * version 2, as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
+ * more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program.  If not, see <http://www.gnu.org/licenses/>.
+ */
+
+#ifndef __NVHOST_HOST1X_HOST1X_CDMA_H
+#define __NVHOST_HOST1X_HOST1X_CDMA_H
+
+/* Size of the sync queue. If it is too small, we won't be able to queue up
+ * many command buffers. If it is too large, we waste memory. */
+#define NVHOST_SYNC_QUEUE_SIZE 512
+
+/* Number of gathers we allow to be queued up per channel. Must be a
+ * power of two. Currently sized such that pushbuffer is 4KB (512*8B). */
+#define NVHOST_GATHER_QUEUE_SIZE 512
+
+/* 8 bytes per slot. (This number does not include the final RESTART.) */
+#define PUSH_BUFFER_SIZE (NVHOST_GATHER_QUEUE_SIZE * 8)
+
+/* 4K page containing GATHERed methods to increment channel syncpts
+ * and replaces the original timed out contexts GATHER slots */
+#define SYNCPT_INCR_BUFFER_SIZE_WORDS   (4096 / sizeof(u32))
+
+#endif
diff --git a/drivers/video/tegra/host/host1x/host1x_channel.c b/drivers/video/tegra/host/host1x/host1x_channel.c
new file mode 100644
index 0000000..d39fe60
--- /dev/null
+++ b/drivers/video/tegra/host/host1x/host1x_channel.c
@@ -0,0 +1,157 @@ 
+/*
+ * drivers/video/tegra/host/host1x/channel_host1x.c
+ *
+ * Tegra Graphics Host Channel
+ *
+ * Copyright (c) 2010-2012, NVIDIA Corporation.
+ *
+ * This program is free software; you can redistribute it and/or modify it
+ * under the terms and conditions of the GNU General Public License,
+ * version 2, as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
+ * more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program.  If not, see <http://www.gnu.org/licenses/>.
+ */
+
+#include <linux/nvhost.h>
+#include "nvhost_channel.h"
+#include "dev.h"
+#include "nvhost_acm.h"
+#include <trace/events/nvhost.h>
+#include <linux/slab.h>
+#include "nvhost_intr.h"
+
+#define NV_FIFO_READ_TIMEOUT 200000
+
+static void submit_gathers(struct nvhost_job *job)
+{
+	/* push user gathers */
+	int i;
+	for (i = 0 ; i < job->num_gathers; i++) {
+		struct nvhost_job_gather *g = &job->gathers[i];
+		u32 op1 = nvhost_opcode_gather(g->words);
+		u32 op2 = g->mem_base + g->offset;
+		nvhost_cdma_push_gather(&job->ch->cdma,
+				job->memmgr,
+				job->gathers[i].ref,
+				job->gathers[i].offset,
+				op1, op2);
+	}
+}
+
+static int host1x_channel_submit(struct nvhost_job *job)
+{
+	struct nvhost_channel *ch = job->ch;
+	struct nvhost_syncpt *sp = &nvhost_get_host(job->ch->dev)->syncpt;
+	u32 user_syncpt_incrs = job->syncpt_incrs;
+	u32 prev_max = 0;
+	u32 syncval;
+	int err;
+	void *completed_waiter = NULL;
+	struct nvhost_device_data *pdata = platform_get_drvdata(ch->dev);
+
+	/* Turn on the client module and host1x */
+	nvhost_module_busy(ch->dev);
+
+	/* before error checks, return current max */
+	prev_max = job->syncpt_end =
+		nvhost_syncpt_read_max(sp, job->syncpt_id);
+
+	/* get submit lock */
+	err = mutex_lock_interruptible(&ch->submitlock);
+	if (err) {
+		nvhost_module_idle(ch->dev);
+		goto error;
+	}
+
+	completed_waiter = nvhost_intr_alloc_waiter();
+	if (!completed_waiter) {
+		nvhost_module_idle(ch->dev);
+		mutex_unlock(&ch->submitlock);
+		err = -ENOMEM;
+		goto error;
+	}
+
+	/* begin a CDMA submit */
+	err = nvhost_cdma_begin(&ch->cdma, job);
+	if (err) {
+		mutex_unlock(&ch->submitlock);
+		nvhost_module_idle(ch->dev);
+		goto error;
+	}
+
+	if (pdata->serialize) {
+		/* Force serialization by inserting a host wait for the
+		 * previous job to finish before this one can commence. */
+		nvhost_cdma_push(&ch->cdma,
+				nvhost_opcode_setclass(NV_HOST1X_CLASS_ID,
+					host1x_uclass_wait_syncpt_r(),
+					1),
+				nvhost_class_host_wait_syncpt(job->syncpt_id,
+					nvhost_syncpt_read_max(sp,
+						job->syncpt_id)));
+	}
+
+	syncval = nvhost_syncpt_incr_max(sp,
+			job->syncpt_id, user_syncpt_incrs);
+
+	job->syncpt_end = syncval;
+
+	/* add a setclass for modules that require it */
+	if (pdata->class)
+		nvhost_cdma_push(&ch->cdma,
+			nvhost_opcode_setclass(pdata->class, 0, 0),
+			NVHOST_OPCODE_NOOP);
+
+	submit_gathers(job);
+
+	/* end CDMA submit & stash pinned hMems into sync queue */
+	nvhost_cdma_end(&ch->cdma, job);
+
+	trace_nvhost_channel_submitted(ch->dev->name,
+			prev_max, syncval);
+
+	/* schedule a submit complete interrupt */
+	err = nvhost_intr_add_action(&nvhost_get_host(ch->dev)->intr,
+			job->syncpt_id, syncval,
+			NVHOST_INTR_ACTION_SUBMIT_COMPLETE, ch,
+			completed_waiter,
+			NULL);
+	completed_waiter = NULL;
+	WARN(err, "Failed to set submit complete interrupt");
+
+	mutex_unlock(&ch->submitlock);
+
+	return 0;
+
+error:
+	kfree(completed_waiter);
+	return err;
+}
+
+static inline void __iomem *host1x_channel_aperture(void __iomem *p, int ndx)
+{
+	p += ndx * NV_HOST1X_CHANNEL_MAP_SIZE_BYTES;
+	return p;
+}
+
+static int host1x_channel_init(struct nvhost_channel *ch,
+	struct nvhost_master *dev, int index)
+{
+	ch->chid = index;
+	mutex_init(&ch->reflock);
+	mutex_init(&ch->submitlock);
+
+	ch->aperture = host1x_channel_aperture(dev->aperture, index);
+	return 0;
+}
+
+static const struct nvhost_channel_ops host1x_channel_ops = {
+	.init = host1x_channel_init,
+	.submit = host1x_channel_submit,
+};
diff --git a/drivers/video/tegra/host/host1x/host1x_debug.c b/drivers/video/tegra/host/host1x/host1x_debug.c
new file mode 100644
index 0000000..ce6074e
--- /dev/null
+++ b/drivers/video/tegra/host/host1x/host1x_debug.c
@@ -0,0 +1,405 @@ 
+/*
+ * drivers/video/tegra/host/host1x/host1x_debug.c
+ *
+ * Copyright (C) 2010 Google, Inc.
+ * Author: Erik Gilling <konkers@android.com>
+ *
+ * Copyright (C) 2011 NVIDIA Corporation
+ *
+ * This software is licensed under the terms of the GNU General Public
+ * License version 2, as published by the Free Software Foundation, and
+ * may be copied, distributed, and modified under those terms.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ *
+ */
+
+#include <linux/debugfs.h>
+#include <linux/seq_file.h>
+#include <linux/mm.h>
+#include <linux/scatterlist.h>
+
+#include <linux/io.h>
+
+#include "dev.h"
+#include "debug.h"
+#include "nvhost_cdma.h"
+#include "nvhost_channel.h"
+#include "chip_support.h"
+#include "nvhost_memmgr.h"
+
+#define NVHOST_DEBUG_MAX_PAGE_OFFSET 102400
+
+enum {
+	NVHOST_DBG_STATE_CMD = 0,
+	NVHOST_DBG_STATE_DATA = 1,
+	NVHOST_DBG_STATE_GATHER = 2
+};
+
+static int show_channel_command(struct output *o, u32 addr, u32 val, int *count)
+{
+	unsigned mask;
+	unsigned subop;
+
+	switch (val >> 28) {
+	case 0x0:
+		mask = val & 0x3f;
+		if (mask) {
+			nvhost_debug_output(o,
+				"SETCL(class=%03x, offset=%03x, mask=%02x, [",
+				val >> 6 & 0x3ff, val >> 16 & 0xfff, mask);
+			*count = hweight8(mask);
+			return NVHOST_DBG_STATE_DATA;
+		} else {
+			nvhost_debug_output(o, "SETCL(class=%03x)\n",
+				val >> 6 & 0x3ff);
+			return NVHOST_DBG_STATE_CMD;
+		}
+
+	case 0x1:
+		nvhost_debug_output(o, "INCR(offset=%03x, [",
+			val >> 16 & 0xfff);
+		*count = val & 0xffff;
+		return NVHOST_DBG_STATE_DATA;
+
+	case 0x2:
+		nvhost_debug_output(o, "NONINCR(offset=%03x, [",
+			val >> 16 & 0xfff);
+		*count = val & 0xffff;
+		return NVHOST_DBG_STATE_DATA;
+
+	case 0x3:
+		mask = val & 0xffff;
+		nvhost_debug_output(o, "MASK(offset=%03x, mask=%03x, [",
+			   val >> 16 & 0xfff, mask);
+		*count = hweight16(mask);
+		return NVHOST_DBG_STATE_DATA;
+
+	case 0x4:
+		nvhost_debug_output(o, "IMM(offset=%03x, data=%03x)\n",
+			   val >> 16 & 0xfff, val & 0xffff);
+		return NVHOST_DBG_STATE_CMD;
+
+	case 0x5:
+		nvhost_debug_output(o, "RESTART(offset=%08x)\n", val << 4);
+		return NVHOST_DBG_STATE_CMD;
+
+	case 0x6:
+		nvhost_debug_output(o,
+			"GATHER(offset=%03x, insert=%d, type=%d, count=%04x, addr=[",
+			val >> 16 & 0xfff, val >> 15 & 0x1, val >> 14 & 0x1,
+			val & 0x3fff);
+		*count = val & 0x3fff; /* TODO: insert */
+		return NVHOST_DBG_STATE_GATHER;
+
+	case 0xe:
+		subop = val >> 24 & 0xf;
+		if (subop == 0)
+			nvhost_debug_output(o, "ACQUIRE_MLOCK(index=%d)\n",
+				val & 0xff);
+		else if (subop == 1)
+			nvhost_debug_output(o, "RELEASE_MLOCK(index=%d)\n",
+				val & 0xff);
+		else
+			nvhost_debug_output(o, "EXTEND_UNKNOWN(%08x)\n", val);
+		return NVHOST_DBG_STATE_CMD;
+
+	default:
+		return NVHOST_DBG_STATE_CMD;
+	}
+}
+
+static void show_channel_gather(struct output *o, u32 addr,
+		phys_addr_t phys_addr, u32 words, struct nvhost_cdma *cdma);
+
+static void show_channel_word(struct output *o, int *state, int *count,
+		u32 addr, u32 val, struct nvhost_cdma *cdma)
+{
+	static int start_count, dont_print;
+
+	switch (*state) {
+	case NVHOST_DBG_STATE_CMD:
+		if (addr)
+			nvhost_debug_output(o, "%08x: %08x:", addr, val);
+		else
+			nvhost_debug_output(o, "%08x:", val);
+
+		*state = show_channel_command(o, addr, val, count);
+		dont_print = 0;
+		start_count = *count;
+		if (*state == NVHOST_DBG_STATE_DATA && *count == 0) {
+			*state = NVHOST_DBG_STATE_CMD;
+			nvhost_debug_output(o, "])\n");
+		}
+		break;
+
+	case NVHOST_DBG_STATE_DATA:
+		(*count)--;
+		if (start_count - *count < 64)
+			nvhost_debug_output(o, "%08x%s",
+				val, *count > 0 ? ", " : "])\n");
+		else if (!dont_print && (*count > 0)) {
+			nvhost_debug_output(o, "[truncated; %d more words]\n",
+				*count);
+			dont_print = 1;
+		}
+		if (*count == 0)
+			*state = NVHOST_DBG_STATE_CMD;
+		break;
+
+	case NVHOST_DBG_STATE_GATHER:
+		*state = NVHOST_DBG_STATE_CMD;
+		nvhost_debug_output(o, "%08x]):\n", val);
+		if (cdma) {
+			show_channel_gather(o, addr, val,
+					*count, cdma);
+		}
+		break;
+	}
+}
+
+static void do_show_channel_gather(struct output *o,
+		phys_addr_t phys_addr,
+		u32 words, struct nvhost_cdma *cdma,
+		phys_addr_t pin_addr, u32 *map_addr)
+{
+	/* Map dmaget cursor to corresponding mem handle */
+	u32 offset;
+	int state, count, i;
+
+	offset = phys_addr - pin_addr;
+	/*
+	 * Sometimes we're given different hardware address to the same
+	 * page - in these cases the offset will get an invalid number and
+	 * we just have to bail out.
+	 */
+	if (offset > NVHOST_DEBUG_MAX_PAGE_OFFSET) {
+		nvhost_debug_output(o, "[address mismatch]\n");
+	} else {
+		/* GATHER buffer starts always with commands */
+		state = NVHOST_DBG_STATE_CMD;
+		for (i = 0; i < words; i++)
+			show_channel_word(o, &state, &count,
+					phys_addr + i * 4,
+					*(map_addr + offset/4 + i),
+					cdma);
+	}
+}
+
+static void show_channel_gather(struct output *o, u32 addr,
+		phys_addr_t phys_addr,
+		u32 words, struct nvhost_cdma *cdma)
+{
+	/* Map dmaget cursor to corresponding mem handle */
+	struct push_buffer *pb = &cdma->push_buffer;
+	u32 cur = addr - pb->phys;
+	struct mem_mgr_handle *mem = &pb->client_handle[cur/8];
+	u32 *map_addr, offset;
+	struct sg_table *sgt;
+
+	if (!mem || !mem->handle || !mem->client) {
+		nvhost_debug_output(o, "[already deallocated]\n");
+		return;
+	}
+
+	map_addr = mem_op().mmap(mem->handle);
+	if (!map_addr) {
+		nvhost_debug_output(o, "[could not mmap]\n");
+		return;
+	}
+
+	/* Get base address from mem */
+	sgt = mem_op().pin(mem->client, mem->handle);
+	if (IS_ERR(sgt)) {
+		nvhost_debug_output(o, "[couldn't pin]\n");
+		mem_op().munmap(mem->handle, map_addr);
+		return;
+	}
+
+	offset = phys_addr - sg_dma_address(sgt->sgl);
+	do_show_channel_gather(o, phys_addr, words, cdma,
+			sg_dma_address(sgt->sgl), map_addr);
+	mem_op().unpin(mem->client, mem->handle, sgt);
+	mem_op().munmap(mem->handle, map_addr);
+}
+
+static void show_channel_gathers(struct output *o, struct nvhost_cdma *cdma)
+{
+	struct nvhost_job *job;
+
+	list_for_each_entry(job, &cdma->sync_queue, list) {
+		int i;
+		nvhost_debug_output(o, "\n%p: JOB, syncpt_id=%d, syncpt_val=%d,"
+				" first_get=%08x, timeout=%d"
+				" num_slots=%d, num_handles=%d\n",
+				job,
+				job->syncpt_id,
+				job->syncpt_end,
+				job->first_get,
+				job->timeout,
+				job->num_slots,
+				job->num_unpins);
+
+		for (i = 0; i < job->num_gathers; i++) {
+			struct nvhost_job_gather *g = &job->gathers[i];
+			u32 *mapped = mem_op().mmap(g->ref);
+			if (!mapped) {
+				nvhost_debug_output(o, "[could not mmap]\n");
+				continue;
+			}
+
+			nvhost_debug_output(o,
+				"    GATHER at %08x+%04x, %d words\n",
+				g->mem_base, g->offset, g->words);
+
+			do_show_channel_gather(o, g->mem_base + g->offset,
+					g->words, cdma, g->mem_base, mapped);
+			mem_op().munmap(g->ref, mapped);
+		}
+	}
+}
+
+static void host1x_debug_show_channel_cdma(struct nvhost_master *m,
+	struct nvhost_channel *ch, struct output *o, int chid)
+{
+	struct nvhost_channel *channel = ch;
+	struct nvhost_cdma *cdma = &channel->cdma;
+	u32 dmaput, dmaget, dmactrl;
+	u32 cbstat, cbread;
+	u32 val, base, baseval;
+	struct nvhost_device_data *pdata = platform_get_drvdata(channel->dev);
+
+	dmaput = readl(channel->aperture + host1x_channel_dmaput_r());
+	dmaget = readl(channel->aperture + host1x_channel_dmaget_r());
+	dmactrl = readl(channel->aperture + host1x_channel_dmactrl_r());
+	cbread = readl(m->sync_aperture + host1x_sync_cbread0_r() + 4 * chid);
+	cbstat = readl(m->sync_aperture + host1x_sync_cbstat_0_r() + 4 * chid);
+
+	nvhost_debug_output(o, "%d-%s (%d): ", chid,
+			    channel->dev->name,
+			    pdata->refcount);
+
+	if (host1x_channel_dmactrl_dmastop_v(dmactrl)
+		|| !channel->cdma.push_buffer.mapped) {
+		nvhost_debug_output(o, "inactive\n\n");
+		return;
+	}
+
+	switch (cbstat) {
+	case 0x00010008:
+		nvhost_debug_output(o, "waiting on syncpt %d val %d\n",
+			cbread >> 24, cbread & 0xffffff);
+		break;
+
+	case 0x00010009:
+		base = (cbread >> 16) & 0xff;
+		baseval = readl(m->sync_aperture +
+				host1x_sync_syncpt_base_0_r() + 4 * base);
+		val = cbread & 0xffff;
+		nvhost_debug_output(o, "waiting on syncpt %d val %d "
+			  "(base %d = %d; offset = %d)\n",
+			cbread >> 24, baseval + val,
+			base, baseval, val);
+		break;
+
+	default:
+		nvhost_debug_output(o,
+				"active class %02x, offset %04x, val %08x\n",
+				host1x_sync_cbstat_0_cbclass0_v(cbstat),
+				host1x_sync_cbstat_0_cboffset0_v(cbstat),
+				cbread);
+		break;
+	}
+
+	nvhost_debug_output(o, "DMAPUT %08x, DMAGET %08x, DMACTL %08x\n",
+		dmaput, dmaget, dmactrl);
+	nvhost_debug_output(o, "CBREAD %08x, CBSTAT %08x\n", cbread, cbstat);
+
+	show_channel_gathers(o, cdma);
+	nvhost_debug_output(o, "\n");
+}
+
+static void host1x_debug_show_channel_fifo(struct nvhost_master *m,
+	struct nvhost_channel *ch, struct output *o, int chid)
+{
+	u32 val, rd_ptr, wr_ptr, start, end;
+	struct nvhost_channel *channel = ch;
+	int state, count;
+
+	nvhost_debug_output(o, "%d: fifo:\n", chid);
+
+	val = readl(channel->aperture + host1x_channel_fifostat_r());
+	nvhost_debug_output(o, "FIFOSTAT %08x\n", val);
+	if (host1x_channel_fifostat_cfempty_v(val)) {
+		nvhost_debug_output(o, "[empty]\n");
+		return;
+	}
+
+	writel(0x0, m->sync_aperture + host1x_sync_cfpeek_ctrl_r());
+	writel(host1x_sync_cfpeek_ctrl_cfpeek_ena_f(1)
+			| host1x_sync_cfpeek_ctrl_cfpeek_channr_f(chid),
+		m->sync_aperture + host1x_sync_cfpeek_ctrl_r());
+
+	val = readl(m->sync_aperture + host1x_sync_cfpeek_ptrs_r());
+	rd_ptr = host1x_sync_cfpeek_ptrs_cf_rd_ptr_v(val);
+	wr_ptr = host1x_sync_cfpeek_ptrs_cf_wr_ptr_v(val);
+
+	val = readl(m->sync_aperture + host1x_sync_cf0_setup_r() + 4 * chid);
+	start = host1x_sync_cf0_setup_cf0_base_v(val);
+	end = host1x_sync_cf0_setup_cf0_limit_v(val);
+
+	state = NVHOST_DBG_STATE_CMD;
+
+	do {
+		writel(0x0, m->sync_aperture + host1x_sync_cfpeek_ctrl_r());
+		writel(host1x_sync_cfpeek_ctrl_cfpeek_ena_f(1)
+				| host1x_sync_cfpeek_ctrl_cfpeek_channr_f(chid)
+				| host1x_sync_cfpeek_ctrl_cfpeek_addr_f(rd_ptr),
+			m->sync_aperture + host1x_sync_cfpeek_ctrl_r());
+		val = readl(m->sync_aperture + host1x_sync_cfpeek_read_r());
+
+		show_channel_word(o, &state, &count, 0, val, NULL);
+
+		if (rd_ptr == end)
+			rd_ptr = start;
+		else
+			rd_ptr++;
+	} while (rd_ptr != wr_ptr);
+
+	if (state == NVHOST_DBG_STATE_DATA)
+		nvhost_debug_output(o, ", ...])\n");
+	nvhost_debug_output(o, "\n");
+
+	writel(0x0, m->sync_aperture + host1x_sync_cfpeek_ctrl_r());
+}
+
+static void host1x_debug_show_mlocks(struct nvhost_master *m, struct output *o)
+{
+	u32 __iomem *mlo_regs = m->sync_aperture +
+		host1x_sync_mlock_owner_0_r();
+	int i;
+
+	nvhost_debug_output(o, "---- mlocks ----\n");
+	for (i = 0; i < NV_HOST1X_NB_MLOCKS; i++) {
+		u32 owner = readl(mlo_regs + i);
+		if (host1x_sync_mlock_owner_0_mlock_ch_owns_0_v(owner))
+			nvhost_debug_output(o, "%d: locked by channel %d\n",
+				i,
+				host1x_sync_mlock_owner_0_mlock_owner_chid_0_f(
+					owner));
+		else if (host1x_sync_mlock_owner_0_mlock_cpu_owns_0_v(owner))
+			nvhost_debug_output(o, "%d: locked by cpu\n", i);
+		else
+			nvhost_debug_output(o, "%d: unlocked\n", i);
+	}
+	nvhost_debug_output(o, "\n");
+}
+
+static const struct nvhost_debug_ops host1x_debug_ops = {
+	.show_channel_cdma = host1x_debug_show_channel_cdma,
+	.show_channel_fifo = host1x_debug_show_channel_fifo,
+	.show_mlocks = host1x_debug_show_mlocks,
+};
diff --git a/drivers/video/tegra/host/host1x/host1x_intr.c b/drivers/video/tegra/host/host1x/host1x_intr.c
new file mode 100644
index 0000000..bf816bb
--- /dev/null
+++ b/drivers/video/tegra/host/host1x/host1x_intr.c
@@ -0,0 +1,263 @@ 
+/*
+ * drivers/video/tegra/host/host1x/host1x_intr.c
+ *
+ * Tegra Graphics Host Interrupt Management
+ *
+ * Copyright (C) 2010 Google, Inc.
+ * Copyright (c) 2010-2012, NVIDIA Corporation.
+ *
+ * This program is free software; you can redistribute it and/or modify it
+ * under the terms and conditions of the GNU General Public License,
+ * version 2, as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
+ * more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program.  If not, see <http://www.gnu.org/licenses/>.
+ */
+
+#include <linux/interrupt.h>
+#include <linux/irq.h>
+#include <linux/io.h>
+#include <asm/mach/irq.h>
+
+#include "nvhost_intr.h"
+#include "dev.h"
+
+/* Spacing between sync registers */
+#define REGISTER_STRIDE 4
+
+static void host1x_intr_syncpt_thresh_isr(struct nvhost_intr_syncpt *syncpt);
+
+static void syncpt_thresh_cascade_fn(struct work_struct *work)
+{
+	struct nvhost_intr_syncpt *sp =
+		container_of(work, struct nvhost_intr_syncpt, work);
+	nvhost_syncpt_thresh_fn(sp->irq, sp);
+}
+
+static irqreturn_t syncpt_thresh_cascade_isr(int irq, void *dev_id)
+{
+	struct nvhost_master *dev = dev_id;
+	void __iomem *sync_regs = dev->sync_aperture;
+	struct nvhost_intr *intr = &dev->intr;
+	unsigned long reg;
+	int i, id;
+
+	for (i = 0; i < dev->info.nb_pts / BITS_PER_LONG; i++) {
+		reg = readl(sync_regs +
+				host1x_sync_syncpt_thresh_cpu0_int_status_r() +
+				i * REGISTER_STRIDE);
+		for_each_set_bit(id, &reg, BITS_PER_LONG) {
+			struct nvhost_intr_syncpt *sp =
+				intr->syncpt + (i * BITS_PER_LONG + id);
+			host1x_intr_syncpt_thresh_isr(sp);
+			queue_work(intr->wq, &sp->work);
+		}
+	}
+
+	return IRQ_HANDLED;
+}
+
+static void host1x_intr_init_host_sync(struct nvhost_intr *intr)
+{
+	struct nvhost_master *dev = intr_to_dev(intr);
+	void __iomem *sync_regs = dev->sync_aperture;
+	int i, err, irq;
+
+	writel(0xffffffffUL,
+		sync_regs + host1x_sync_syncpt_thresh_int_disable_r());
+	writel(0xffffffffUL,
+		sync_regs + host1x_sync_syncpt_thresh_cpu0_int_status_r());
+
+	for (i = 0; i < dev->info.nb_pts; i++)
+		INIT_WORK(&intr->syncpt[i].work, syncpt_thresh_cascade_fn);
+
+	irq = platform_get_irq(dev->dev, 0);
+	WARN_ON(IS_ERR_VALUE(irq));
+	err = devm_request_irq(&dev->dev->dev, irq,
+				syncpt_thresh_cascade_isr,
+				IRQF_SHARED, "host_syncpt", dev);
+	WARN_ON(IS_ERR_VALUE(err));
+
+	/* disable the ip_busy_timeout. this prevents write drops, etc.
+	 * there's no real way to recover from a hung client anyway.
+	 */
+	writel(0, sync_regs + host1x_sync_ip_busy_timeout_r());
+
+	/* increase the auto-ack timout to the maximum value. 2d will hang
+	 * otherwise on Tegra2.
+	 */
+	writel(0xff, sync_regs + host1x_sync_ctxsw_timeout_cfg_r());
+}
+
+static void host1x_intr_set_host_clocks_per_usec(struct nvhost_intr *intr,
+		u32 cpm)
+{
+	struct nvhost_master *dev = intr_to_dev(intr);
+	void __iomem *sync_regs = dev->sync_aperture;
+	/* write microsecond clock register */
+	writel(cpm, sync_regs + host1x_sync_usec_clk_r());
+}
+
+static void host1x_intr_set_syncpt_threshold(struct nvhost_intr *intr,
+	u32 id, u32 thresh)
+{
+	struct nvhost_master *dev = intr_to_dev(intr);
+	void __iomem *sync_regs = dev->sync_aperture;
+	writel(thresh, sync_regs +
+		(host1x_sync_syncpt_int_thresh_0_r() + id * REGISTER_STRIDE));
+}
+
+static void host1x_intr_enable_syncpt_intr(struct nvhost_intr *intr, u32 id)
+{
+	struct nvhost_master *dev = intr_to_dev(intr);
+	void __iomem *sync_regs = dev->sync_aperture;
+
+	writel(BIT_MASK(id), sync_regs +
+			host1x_sync_syncpt_thresh_int_enable_cpu0_r() +
+			BIT_WORD(id) * REGISTER_STRIDE);
+}
+
+static void host1x_intr_disable_syncpt_intr(struct nvhost_intr *intr, u32 id)
+{
+	struct nvhost_master *dev = intr_to_dev(intr);
+	void __iomem *sync_regs = dev->sync_aperture;
+
+	writel(BIT_MASK(id), sync_regs +
+			host1x_sync_syncpt_thresh_int_disable_r() +
+			BIT_WORD(id) * REGISTER_STRIDE);
+
+	writel(BIT_MASK(id), sync_regs +
+		host1x_sync_syncpt_thresh_cpu0_int_status_r() +
+		BIT_WORD(id) * REGISTER_STRIDE);
+}
+
+static void host1x_intr_disable_all_syncpt_intrs(struct nvhost_intr *intr)
+{
+	struct nvhost_master *dev = intr_to_dev(intr);
+	void __iomem *sync_regs = dev->sync_aperture;
+	u32 reg;
+
+	for (reg = 0; reg <= BIT_WORD(dev->info.nb_pts) * REGISTER_STRIDE;
+			reg += REGISTER_STRIDE) {
+		writel(0xffffffffu, sync_regs +
+				host1x_sync_syncpt_thresh_int_disable_r() +
+				reg);
+
+		writel(0xffffffffu, sync_regs +
+			host1x_sync_syncpt_thresh_cpu0_int_status_r() + reg);
+	}
+}
+
+/**
+ * Sync point threshold interrupt service function
+ * Handles sync point threshold triggers, in interrupt context
+ */
+static void host1x_intr_syncpt_thresh_isr(struct nvhost_intr_syncpt *syncpt)
+{
+	unsigned int id = syncpt->id;
+	struct nvhost_intr *intr = intr_syncpt_to_intr(syncpt);
+
+	void __iomem *sync_regs = intr_to_dev(intr)->sync_aperture;
+
+	u32 reg = BIT_WORD(id) * REGISTER_STRIDE;
+
+	writel(BIT_MASK(id), sync_regs +
+		host1x_sync_syncpt_thresh_int_disable_r() + reg);
+	writel(BIT_MASK(id), sync_regs +
+		host1x_sync_syncpt_thresh_cpu0_int_status_r() + reg);
+}
+
+/**
+ * Host general interrupt service function
+ * Handles read / write failures
+ */
+static irqreturn_t host1x_intr_host1x_isr(int irq, void *dev_id)
+{
+	struct nvhost_intr *intr = dev_id;
+	void __iomem *sync_regs = intr_to_dev(intr)->sync_aperture;
+	u32 stat;
+	u32 ext_stat;
+	u32 addr;
+
+	stat = readl(sync_regs + host1x_sync_hintstatus_r());
+	ext_stat = readl(sync_regs + host1x_sync_hintstatus_ext_r());
+
+	if (host1x_sync_hintstatus_ext_ip_read_int_v(ext_stat)) {
+		addr = readl(sync_regs + host1x_sync_ip_read_timeout_addr_r());
+		pr_err("Host read timeout at address %x\n", addr);
+	}
+
+	if (host1x_sync_hintstatus_ext_ip_write_int_v(ext_stat)) {
+		addr = readl(sync_regs + host1x_sync_ip_write_timeout_addr_r());
+		pr_err("Host write timeout at address %x\n", addr);
+	}
+
+	writel(ext_stat, sync_regs + host1x_sync_hintstatus_ext_r());
+	writel(stat, sync_regs + host1x_sync_hintstatus_r());
+
+	return IRQ_HANDLED;
+}
+static int host1x_intr_request_host_general_irq(struct nvhost_intr *intr)
+{
+	void __iomem *sync_regs = intr_to_dev(intr)->sync_aperture;
+	int err;
+
+	/* master disable for general (not syncpt) host interrupts */
+	writel(0, sync_regs + host1x_sync_intmask_r());
+
+	/* clear status & extstatus */
+	writel(0xfffffffful, sync_regs + host1x_sync_hintstatus_ext_r());
+	writel(0xfffffffful, sync_regs + host1x_sync_hintstatus_r());
+
+	err = request_irq(intr->host_general_irq, host1x_intr_host1x_isr, 0,
+			"host_status", intr);
+	if (err)
+		return err;
+
+	/* enable extra interrupt sources IP_READ_INT and IP_WRITE_INT */
+	writel(BIT(30) | BIT(31), sync_regs + host1x_sync_hintmask_ext_r());
+
+	/* enable extra interrupt sources */
+	writel(BIT(12) | BIT(31), sync_regs + host1x_sync_hintmask_r());
+
+	/* enable host module interrupt to CPU0 */
+	writel(BIT(0), sync_regs + host1x_sync_intc0mask_r());
+
+	/* master enable for general (not syncpt) host interrupts */
+	writel(BIT(0), sync_regs + host1x_sync_intmask_r());
+
+	return err;
+}
+
+static void host1x_intr_free_host_general_irq(struct nvhost_intr *intr)
+{
+	void __iomem *sync_regs = intr_to_dev(intr)->sync_aperture;
+
+	/* master disable for general (not syncpt) host interrupts */
+	writel(0, sync_regs + host1x_sync_intmask_r());
+
+	free_irq(intr->host_general_irq, intr);
+}
+
+static int host1x_free_syncpt_irq(struct nvhost_intr *intr)
+{
+	flush_workqueue(intr->wq);
+	return 0;
+}
+
+static const struct nvhost_intr_ops host1x_intr_ops = {
+	.init_host_sync = host1x_intr_init_host_sync,
+	.set_host_clocks_per_usec = host1x_intr_set_host_clocks_per_usec,
+	.set_syncpt_threshold = host1x_intr_set_syncpt_threshold,
+	.enable_syncpt_intr = host1x_intr_enable_syncpt_intr,
+	.disable_syncpt_intr = host1x_intr_disable_syncpt_intr,
+	.disable_all_syncpt_intrs = host1x_intr_disable_all_syncpt_intrs,
+	.request_host_general_irq = host1x_intr_request_host_general_irq,
+	.free_host_general_irq = host1x_intr_free_host_general_irq,
+	.free_syncpt_irq = host1x_free_syncpt_irq,
+};
diff --git a/drivers/video/tegra/host/host1x/host1x_syncpt.c b/drivers/video/tegra/host/host1x/host1x_syncpt.c
new file mode 100644
index 0000000..ba0080a
--- /dev/null
+++ b/drivers/video/tegra/host/host1x/host1x_syncpt.c
@@ -0,0 +1,170 @@ 
+/*
+ * drivers/video/tegra/host/host1x/host1x_syncpt.c
+ *
+ * Tegra Graphics Host Syncpoints for HOST1X
+ *
+ * Copyright (c) 2010-2012, NVIDIA Corporation.
+ *
+ * This program is free software; you can redistribute it and/or modify it
+ * under the terms and conditions of the GNU General Public License,
+ * version 2, as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
+ * more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program.  If not, see <http://www.gnu.org/licenses/>.
+ */
+
+#include <linux/io.h>
+#include <trace/events/nvhost.h>
+#include "nvhost_syncpt.h"
+#include "nvhost_acm.h"
+#include "host1x.h"
+#include "chip_support.h"
+
+/**
+ * Write the current syncpoint value back to hw.
+ */
+static void host1x_syncpt_reset(struct nvhost_syncpt *sp, u32 id)
+{
+	struct nvhost_master *dev = syncpt_to_dev(sp);
+	int min = nvhost_syncpt_read_min(sp, id);
+	writel(min, dev->sync_aperture + (host1x_sync_syncpt_0_r() + id * 4));
+}
+
+/**
+ * Write the current waitbase value back to hw.
+ */
+static void host1x_syncpt_reset_wait_base(struct nvhost_syncpt *sp, u32 id)
+{
+	struct nvhost_master *dev = syncpt_to_dev(sp);
+	writel(sp->base_val[id],
+		dev->sync_aperture + (host1x_sync_syncpt_base_0_r() + id * 4));
+}
+
+/**
+ * Read waitbase value from hw.
+ */
+static void host1x_syncpt_read_wait_base(struct nvhost_syncpt *sp, u32 id)
+{
+	struct nvhost_master *dev = syncpt_to_dev(sp);
+	sp->base_val[id] = readl(dev->sync_aperture +
+				(host1x_sync_syncpt_base_0_r() + id * 4));
+}
+
+/**
+ * Updates the last value read from hardware.
+ * (was nvhost_syncpt_update_min)
+ */
+static u32 host1x_syncpt_update_min(struct nvhost_syncpt *sp, u32 id)
+{
+	struct nvhost_master *dev = syncpt_to_dev(sp);
+	void __iomem *sync_regs = dev->sync_aperture;
+	u32 old, live;
+
+	do {
+		old = nvhost_syncpt_read_min(sp, id);
+		live = readl(sync_regs + (host1x_sync_syncpt_0_r() + id * 4));
+	} while ((u32)atomic_cmpxchg(&sp->min_val[id], old, live) != old);
+
+	if (!nvhost_syncpt_check_max(sp, id, live))
+		dev_err(&syncpt_to_dev(sp)->dev->dev,
+				"%s failed: id=%u, min=%d, max=%d\n",
+				__func__,
+				id,
+				nvhost_syncpt_read_min(sp, id),
+				nvhost_syncpt_read_max(sp, id));
+
+	return live;
+}
+
+/**
+ * Write a cpu syncpoint increment to the hardware, without touching
+ * the cache. Caller is responsible for host being powered.
+ */
+static void host1x_syncpt_cpu_incr(struct nvhost_syncpt *sp, u32 id)
+{
+	struct nvhost_master *dev = syncpt_to_dev(sp);
+	u32 reg_offset = id / 32;
+
+	if (!nvhost_module_powered(dev->dev)) {
+		dev_err(&syncpt_to_dev(sp)->dev->dev,
+			"Trying to access host1x when it's off");
+		return;
+	}
+
+	if (!nvhost_syncpt_client_managed(sp, id)
+			&& nvhost_syncpt_min_eq_max(sp, id)) {
+		dev_err(&syncpt_to_dev(sp)->dev->dev,
+			"Trying to increment syncpoint id %d beyond max\n",
+			id);
+		nvhost_debug_dump(syncpt_to_dev(sp));
+		return;
+	}
+	writel(BIT_MASK(id), dev->sync_aperture +
+			host1x_sync_syncpt_cpu_incr_r() + reg_offset * 4);
+	wmb();
+}
+
+/* remove a wait pointed to by patch_addr */
+static int host1x_syncpt_patch_wait(struct nvhost_syncpt *sp,
+		void *patch_addr)
+{
+	u32 override = nvhost_class_host_wait_syncpt(
+			NVSYNCPT_GRAPHICS_HOST, 0);
+	__raw_writel(override, patch_addr);
+	return 0;
+}
+
+
+static const char *host1x_syncpt_name(struct nvhost_syncpt *sp, u32 id)
+{
+	struct host1x_device_info *info = &syncpt_to_dev(sp)->info;
+	const char *name = NULL;
+
+	if (id < info->nb_pts)
+		name = info->syncpt_names[id];
+
+	return name ? name : "";
+}
+
+static void host1x_syncpt_debug(struct nvhost_syncpt *sp)
+{
+	u32 i;
+	for (i = 0; i < nvhost_syncpt_nb_pts(sp); i++) {
+		u32 max = nvhost_syncpt_read_max(sp, i);
+		u32 min = nvhost_syncpt_update_min(sp, i);
+		if (!max && !min)
+			continue;
+		dev_info(&syncpt_to_dev(sp)->dev->dev,
+			"id %d (%s) min %d max %d\n",
+			i, syncpt_op().name(sp, i),
+			min, max);
+
+	}
+
+	for (i = 0; i < nvhost_syncpt_nb_bases(sp); i++) {
+		u32 base_val;
+		host1x_syncpt_read_wait_base(sp, i);
+		base_val = sp->base_val[i];
+		if (base_val)
+			dev_info(&syncpt_to_dev(sp)->dev->dev,
+					"waitbase id %d val %d\n",
+					i, base_val);
+
+	}
+}
+
+static const struct nvhost_syncpt_ops host1x_syncpt_ops = {
+	.reset = host1x_syncpt_reset,
+	.reset_wait_base = host1x_syncpt_reset_wait_base,
+	.read_wait_base = host1x_syncpt_read_wait_base,
+	.update_min = host1x_syncpt_update_min,
+	.cpu_incr = host1x_syncpt_cpu_incr,
+	.patch_wait = host1x_syncpt_patch_wait,
+	.debug = host1x_syncpt_debug,
+	.name = host1x_syncpt_name,
+};
diff --git a/drivers/video/tegra/host/host1x/hw_host1x01_channel.h b/drivers/video/tegra/host/host1x/hw_host1x01_channel.h
new file mode 100644
index 0000000..ca2f9a0
--- /dev/null
+++ b/drivers/video/tegra/host/host1x/hw_host1x01_channel.h
@@ -0,0 +1,182 @@ 
+/*
+ * drivers/video/tegra/host/host1x/hw_host1x_channel_host1x.h
+ *
+ * Copyright (c) 2012, NVIDIA Corporation.
+ *
+ * This program is free software; you can redistribute it and/or modify it
+ * under the terms and conditions of the GNU General Public License,
+ * version 2, as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
+ * more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program.  If not, see <http://www.gnu.org/licenses/>.
+ *
+ */
+
+ /*
+  * Function naming determines intended use:
+  *
+  *     <x>_r(void) : Returns the offset for register <x>.
+  *
+  *     <x>_w(void) : Returns the word offset for word (4 byte) element <x>.
+  *
+  *     <x>_<y>_s(void) : Returns size of field <y> of register <x> in bits.
+  *
+  *     <x>_<y>_f(u32 v) : Returns a value based on 'v' which has been shifted
+  *         and masked to place it at field <y> of register <x>.  This value
+  *         can be |'d with others to produce a full register value for
+  *         register <x>.
+  *
+  *     <x>_<y>_m(void) : Returns a mask for field <y> of register <x>.  This
+  *         value can be ~'d and then &'d to clear the value of field <y> for
+  *         register <x>.
+  *
+  *     <x>_<y>_<z>_f(void) : Returns the constant value <z> after being shifted
+  *         to place it at field <y> of register <x>.  This value can be |'d
+  *         with others to produce a full register value for <x>.
+  *
+  *     <x>_<y>_v(u32 r) : Returns the value of field <y> from a full register
+  *         <x> value 'r' after being shifted to place its LSB at bit 0.
+  *         This value is suitable for direct comparison with other unshifted
+  *         values appropriate for use in field <y> of register <x>.
+  *
+  *     <x>_<y>_<z>_v(void) : Returns the constant value for <z> defined for
+  *         field <y> of register <x>.  This value is suitable for direct
+  *         comparison with unshifted values appropriate for use in field <y>
+  *         of register <x>.
+  */
+
+#ifndef __hw_host1x_channel_host1x_h__
+#define __hw_host1x_channel_host1x_h__
+/*This file is autogenerated.  Do not edit. */
+
+static inline u32 host1x_channel_fifostat_r(void)
+{
+	return 0x0;
+}
+static inline u32 host1x_channel_fifostat_cfempty_s(void)
+{
+	return 1;
+}
+static inline u32 host1x_channel_fifostat_cfempty_f(u32 v)
+{
+	return (v & 0x1) << 10;
+}
+static inline u32 host1x_channel_fifostat_cfempty_m(void)
+{
+	return 0x1 << 10;
+}
+static inline u32 host1x_channel_fifostat_cfempty_v(u32 r)
+{
+	return (r >> 10) & 0x1;
+}
+static inline u32 host1x_channel_fifostat_cfempty_notempty_v(void)
+{
+	return 0;
+}
+static inline u32 host1x_channel_fifostat_cfempty_empty_v(void)
+{
+	return 1;
+}
+static inline u32 host1x_channel_fifostat_outfentries_s(void)
+{
+	return 5;
+}
+static inline u32 host1x_channel_fifostat_outfentries_f(u32 v)
+{
+	return (v & 0x1f) << 24;
+}
+static inline u32 host1x_channel_fifostat_outfentries_m(void)
+{
+	return 0x1f << 24;
+}
+static inline u32 host1x_channel_fifostat_outfentries_v(u32 r)
+{
+	return (r >> 24) & 0x1f;
+}
+static inline u32 host1x_channel_inddata_r(void)
+{
+	return 0xc;
+}
+static inline u32 host1x_channel_dmastart_r(void)
+{
+	return 0x14;
+}
+static inline u32 host1x_channel_dmaput_r(void)
+{
+	return 0x18;
+}
+static inline u32 host1x_channel_dmaget_r(void)
+{
+	return 0x1c;
+}
+static inline u32 host1x_channel_dmaend_r(void)
+{
+	return 0x20;
+}
+static inline u32 host1x_channel_dmactrl_r(void)
+{
+	return 0x24;
+}
+static inline u32 host1x_channel_dmactrl_dmastop_s(void)
+{
+	return 1;
+}
+static inline u32 host1x_channel_dmactrl_dmastop_f(u32 v)
+{
+	return (v & 0x1) << 0;
+}
+static inline u32 host1x_channel_dmactrl_dmastop_m(void)
+{
+	return 0x1 << 0;
+}
+static inline u32 host1x_channel_dmactrl_dmastop_v(u32 r)
+{
+	return (r >> 0) & 0x1;
+}
+static inline u32 host1x_channel_dmactrl_dmastop_run_v(void)
+{
+	return 0;
+}
+static inline u32 host1x_channel_dmactrl_dmastop_stop_v(void)
+{
+	return 1;
+}
+static inline u32 host1x_channel_dmactrl_dmagetrst_s(void)
+{
+	return 1;
+}
+static inline u32 host1x_channel_dmactrl_dmagetrst_f(u32 v)
+{
+	return (v & 0x1) << 1;
+}
+static inline u32 host1x_channel_dmactrl_dmagetrst_m(void)
+{
+	return 0x1 << 1;
+}
+static inline u32 host1x_channel_dmactrl_dmagetrst_v(u32 r)
+{
+	return (r >> 1) & 0x1;
+}
+static inline u32 host1x_channel_dmactrl_dmainitget_s(void)
+{
+	return 1;
+}
+static inline u32 host1x_channel_dmactrl_dmainitget_f(u32 v)
+{
+	return (v & 0x1) << 2;
+}
+static inline u32 host1x_channel_dmactrl_dmainitget_m(void)
+{
+	return 0x1 << 2;
+}
+static inline u32 host1x_channel_dmactrl_dmainitget_v(u32 r)
+{
+	return (r >> 2) & 0x1;
+}
+
+#endif /* __hw_host1x_channel_host1x_h__ */
diff --git a/drivers/video/tegra/host/host1x/hw_host1x01_sync.h b/drivers/video/tegra/host/host1x/hw_host1x01_sync.h
new file mode 100644
index 0000000..67f0cbf
--- /dev/null
+++ b/drivers/video/tegra/host/host1x/hw_host1x01_sync.h
@@ -0,0 +1,398 @@ 
+/*
+ * drivers/video/tegra/host/host1x/hw_host1x_sync_host1x.h
+ *
+ * Copyright (c) 2012, NVIDIA Corporation.
+ *
+ * This program is free software; you can redistribute it and/or modify it
+ * under the terms and conditions of the GNU General Public License,
+ * version 2, as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
+ * more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program.  If not, see <http://www.gnu.org/licenses/>.
+ *
+ */
+
+ /*
+  * Function naming determines intended use:
+  *
+  *     <x>_r(void) : Returns the offset for register <x>.
+  *
+  *     <x>_w(void) : Returns the word offset for word (4 byte) element <x>.
+  *
+  *     <x>_<y>_s(void) : Returns size of field <y> of register <x> in bits.
+  *
+  *     <x>_<y>_f(u32 v) : Returns a value based on 'v' which has been shifted
+  *         and masked to place it at field <y> of register <x>.  This value
+  *         can be |'d with others to produce a full register value for
+  *         register <x>.
+  *
+  *     <x>_<y>_m(void) : Returns a mask for field <y> of register <x>.  This
+  *         value can be ~'d and then &'d to clear the value of field <y> for
+  *         register <x>.
+  *
+  *     <x>_<y>_<z>_f(void) : Returns the constant value <z> after being shifted
+  *         to place it at field <y> of register <x>.  This value can be |'d
+  *         with others to produce a full register value for <x>.
+  *
+  *     <x>_<y>_v(u32 r) : Returns the value of field <y> from a full register
+  *         <x> value 'r' after being shifted to place its LSB at bit 0.
+  *         This value is suitable for direct comparison with other unshifted
+  *         values appropriate for use in field <y> of register <x>.
+  *
+  *     <x>_<y>_<z>_v(void) : Returns the constant value for <z> defined for
+  *         field <y> of register <x>.  This value is suitable for direct
+  *         comparison with unshifted values appropriate for use in field <y>
+  *         of register <x>.
+  */
+
+#ifndef __hw_host1x_sync_host1x_h__
+#define __hw_host1x_sync_host1x_h__
+/*This file is autogenerated.  Do not edit. */
+
+static inline u32 host1x_sync_intmask_r(void)
+{
+	return 0x4;
+}
+static inline u32 host1x_sync_intc0mask_r(void)
+{
+	return 0x8;
+}
+static inline u32 host1x_sync_hintstatus_r(void)
+{
+	return 0x20;
+}
+static inline u32 host1x_sync_hintmask_r(void)
+{
+	return 0x24;
+}
+static inline u32 host1x_sync_hintstatus_ext_r(void)
+{
+	return 0x28;
+}
+static inline u32 host1x_sync_hintstatus_ext_ip_read_int_s(void)
+{
+	return 1;
+}
+static inline u32 host1x_sync_hintstatus_ext_ip_read_int_f(u32 v)
+{
+	return (v & 0x1) << 30;
+}
+static inline u32 host1x_sync_hintstatus_ext_ip_read_int_m(void)
+{
+	return 0x1 << 30;
+}
+static inline u32 host1x_sync_hintstatus_ext_ip_read_int_v(u32 r)
+{
+	return (r >> 30) & 0x1;
+}
+static inline u32 host1x_sync_hintstatus_ext_ip_write_int_s(void)
+{
+	return 1;
+}
+static inline u32 host1x_sync_hintstatus_ext_ip_write_int_f(u32 v)
+{
+	return (v & 0x1) << 31;
+}
+static inline u32 host1x_sync_hintstatus_ext_ip_write_int_m(void)
+{
+	return 0x1 << 31;
+}
+static inline u32 host1x_sync_hintstatus_ext_ip_write_int_v(u32 r)
+{
+	return (r >> 31) & 0x1;
+}
+static inline u32 host1x_sync_hintmask_ext_r(void)
+{
+	return 0x2c;
+}
+static inline u32 host1x_sync_syncpt_thresh_cpu0_int_status_r(void)
+{
+	return 0x40;
+}
+static inline u32 host1x_sync_syncpt_thresh_cpu1_int_status_r(void)
+{
+	return 0x48;
+}
+static inline u32 host1x_sync_syncpt_thresh_int_disable_r(void)
+{
+	return 0x60;
+}
+static inline u32 host1x_sync_syncpt_thresh_int_enable_cpu0_r(void)
+{
+	return 0x68;
+}
+static inline u32 host1x_sync_cf0_setup_r(void)
+{
+	return 0x80;
+}
+static inline u32 host1x_sync_cf0_setup_cf0_base_s(void)
+{
+	return 9;
+}
+static inline u32 host1x_sync_cf0_setup_cf0_base_f(u32 v)
+{
+	return (v & 0x1ff) << 0;
+}
+static inline u32 host1x_sync_cf0_setup_cf0_base_m(void)
+{
+	return 0x1ff << 0;
+}
+static inline u32 host1x_sync_cf0_setup_cf0_base_v(u32 r)
+{
+	return (r >> 0) & 0x1ff;
+}
+static inline u32 host1x_sync_cf0_setup_cf0_limit_s(void)
+{
+	return 9;
+}
+static inline u32 host1x_sync_cf0_setup_cf0_limit_f(u32 v)
+{
+	return (v & 0x1ff) << 16;
+}
+static inline u32 host1x_sync_cf0_setup_cf0_limit_m(void)
+{
+	return 0x1ff << 16;
+}
+static inline u32 host1x_sync_cf0_setup_cf0_limit_v(u32 r)
+{
+	return (r >> 16) & 0x1ff;
+}
+static inline u32 host1x_sync_cmdproc_stop_r(void)
+{
+	return 0xac;
+}
+static inline u32 host1x_sync_ch_teardown_r(void)
+{
+	return 0xb0;
+}
+static inline u32 host1x_sync_usec_clk_r(void)
+{
+	return 0x1a4;
+}
+static inline u32 host1x_sync_ctxsw_timeout_cfg_r(void)
+{
+	return 0x1a8;
+}
+static inline u32 host1x_sync_ip_busy_timeout_r(void)
+{
+	return 0x1bc;
+}
+static inline u32 host1x_sync_ip_read_timeout_addr_r(void)
+{
+	return 0x1c0;
+}
+static inline u32 host1x_sync_ip_write_timeout_addr_r(void)
+{
+	return 0x1c4;
+}
+static inline u32 host1x_sync_mlock_0_r(void)
+{
+	return 0x2c0;
+}
+static inline u32 host1x_sync_mlock_owner_0_r(void)
+{
+	return 0x340;
+}
+static inline u32 host1x_sync_mlock_owner_0_mlock_owner_chid_0_s(void)
+{
+	return 4;
+}
+static inline u32 host1x_sync_mlock_owner_0_mlock_owner_chid_0_f(u32 v)
+{
+	return (v & 0xf) << 8;
+}
+static inline u32 host1x_sync_mlock_owner_0_mlock_owner_chid_0_m(void)
+{
+	return 0xf << 8;
+}
+static inline u32 host1x_sync_mlock_owner_0_mlock_owner_chid_0_v(u32 r)
+{
+	return (r >> 8) & 0xf;
+}
+static inline u32 host1x_sync_mlock_owner_0_mlock_cpu_owns_0_s(void)
+{
+	return 1;
+}
+static inline u32 host1x_sync_mlock_owner_0_mlock_cpu_owns_0_f(u32 v)
+{
+	return (v & 0x1) << 1;
+}
+static inline u32 host1x_sync_mlock_owner_0_mlock_cpu_owns_0_m(void)
+{
+	return 0x1 << 1;
+}
+static inline u32 host1x_sync_mlock_owner_0_mlock_cpu_owns_0_v(u32 r)
+{
+	return (r >> 1) & 0x1;
+}
+static inline u32 host1x_sync_mlock_owner_0_mlock_ch_owns_0_s(void)
+{
+	return 1;
+}
+static inline u32 host1x_sync_mlock_owner_0_mlock_ch_owns_0_f(u32 v)
+{
+	return (v & 0x1) << 0;
+}
+static inline u32 host1x_sync_mlock_owner_0_mlock_ch_owns_0_m(void)
+{
+	return 0x1 << 0;
+}
+static inline u32 host1x_sync_mlock_owner_0_mlock_ch_owns_0_v(u32 r)
+{
+	return (r >> 0) & 0x1;
+}
+static inline u32 host1x_sync_syncpt_0_r(void)
+{
+	return 0x400;
+}
+static inline u32 host1x_sync_syncpt_int_thresh_0_r(void)
+{
+	return 0x500;
+}
+static inline u32 host1x_sync_syncpt_base_0_r(void)
+{
+	return 0x600;
+}
+static inline u32 host1x_sync_syncpt_cpu_incr_r(void)
+{
+	return 0x700;
+}
+static inline u32 host1x_sync_cbread0_r(void)
+{
+	return 0x720;
+}
+static inline u32 host1x_sync_cfpeek_ctrl_r(void)
+{
+	return 0x74c;
+}
+static inline u32 host1x_sync_cfpeek_ctrl_cfpeek_addr_s(void)
+{
+	return 9;
+}
+static inline u32 host1x_sync_cfpeek_ctrl_cfpeek_addr_f(u32 v)
+{
+	return (v & 0x1ff) << 0;
+}
+static inline u32 host1x_sync_cfpeek_ctrl_cfpeek_addr_m(void)
+{
+	return 0x1ff << 0;
+}
+static inline u32 host1x_sync_cfpeek_ctrl_cfpeek_addr_v(u32 r)
+{
+	return (r >> 0) & 0x1ff;
+}
+static inline u32 host1x_sync_cfpeek_ctrl_cfpeek_channr_s(void)
+{
+	return 3;
+}
+static inline u32 host1x_sync_cfpeek_ctrl_cfpeek_channr_f(u32 v)
+{
+	return (v & 0x7) << 16;
+}
+static inline u32 host1x_sync_cfpeek_ctrl_cfpeek_channr_m(void)
+{
+	return 0x7 << 16;
+}
+static inline u32 host1x_sync_cfpeek_ctrl_cfpeek_channr_v(u32 r)
+{
+	return (r >> 16) & 0x7;
+}
+static inline u32 host1x_sync_cfpeek_ctrl_cfpeek_ena_s(void)
+{
+	return 1;
+}
+static inline u32 host1x_sync_cfpeek_ctrl_cfpeek_ena_f(u32 v)
+{
+	return (v & 0x1) << 31;
+}
+static inline u32 host1x_sync_cfpeek_ctrl_cfpeek_ena_m(void)
+{
+	return 0x1 << 31;
+}
+static inline u32 host1x_sync_cfpeek_ctrl_cfpeek_ena_v(u32 r)
+{
+	return (r >> 31) & 0x1;
+}
+static inline u32 host1x_sync_cfpeek_read_r(void)
+{
+	return 0x750;
+}
+static inline u32 host1x_sync_cfpeek_ptrs_r(void)
+{
+	return 0x754;
+}
+static inline u32 host1x_sync_cfpeek_ptrs_cf_rd_ptr_s(void)
+{
+	return 9;
+}
+static inline u32 host1x_sync_cfpeek_ptrs_cf_rd_ptr_f(u32 v)
+{
+	return (v & 0x1ff) << 0;
+}
+static inline u32 host1x_sync_cfpeek_ptrs_cf_rd_ptr_m(void)
+{
+	return 0x1ff << 0;
+}
+static inline u32 host1x_sync_cfpeek_ptrs_cf_rd_ptr_v(u32 r)
+{
+	return (r >> 0) & 0x1ff;
+}
+static inline u32 host1x_sync_cfpeek_ptrs_cf_wr_ptr_s(void)
+{
+	return 9;
+}
+static inline u32 host1x_sync_cfpeek_ptrs_cf_wr_ptr_f(u32 v)
+{
+	return (v & 0x1ff) << 16;
+}
+static inline u32 host1x_sync_cfpeek_ptrs_cf_wr_ptr_m(void)
+{
+	return 0x1ff << 16;
+}
+static inline u32 host1x_sync_cfpeek_ptrs_cf_wr_ptr_v(u32 r)
+{
+	return (r >> 16) & 0x1ff;
+}
+static inline u32 host1x_sync_cbstat_0_r(void)
+{
+	return 0x758;
+}
+static inline u32 host1x_sync_cbstat_0_cboffset0_s(void)
+{
+	return 16;
+}
+static inline u32 host1x_sync_cbstat_0_cboffset0_f(u32 v)
+{
+	return (v & 0xffff) << 0;
+}
+static inline u32 host1x_sync_cbstat_0_cboffset0_m(void)
+{
+	return 0xffff << 0;
+}
+static inline u32 host1x_sync_cbstat_0_cboffset0_v(u32 r)
+{
+	return (r >> 0) & 0xffff;
+}
+static inline u32 host1x_sync_cbstat_0_cbclass0_s(void)
+{
+	return 10;
+}
+static inline u32 host1x_sync_cbstat_0_cbclass0_f(u32 v)
+{
+	return (v & 0x3ff) << 16;
+}
+static inline u32 host1x_sync_cbstat_0_cbclass0_m(void)
+{
+	return 0x3ff << 16;
+}
+static inline u32 host1x_sync_cbstat_0_cbclass0_v(u32 r)
+{
+	return (r >> 16) & 0x3ff;
+}
+
+#endif /* __hw_host1x_sync_host1x_h__ */
diff --git a/drivers/video/tegra/host/host1x/hw_host1x01_uclass.h b/drivers/video/tegra/host/host1x/hw_host1x01_uclass.h
new file mode 100644
index 0000000..ed6e4b7
--- /dev/null
+++ b/drivers/video/tegra/host/host1x/hw_host1x01_uclass.h
@@ -0,0 +1,474 @@ 
+/*
+ * drivers/video/tegra/host/host1x/hw_host1x_uclass_host1x.h
+ *
+ * Copyright (c) 2012, NVIDIA Corporation.
+ *
+ * This program is free software; you can redistribute it and/or modify it
+ * under the terms and conditions of the GNU General Public License,
+ * version 2, as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
+ * more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program.  If not, see <http://www.gnu.org/licenses/>.
+ *
+ */
+
+ /*
+  * Function naming determines intended use:
+  *
+  *     <x>_r(void) : Returns the offset for register <x>.
+  *
+  *     <x>_w(void) : Returns the word offset for word (4 byte) element <x>.
+  *
+  *     <x>_<y>_s(void) : Returns size of field <y> of register <x> in bits.
+  *
+  *     <x>_<y>_f(u32 v) : Returns a value based on 'v' which has been shifted
+  *         and masked to place it at field <y> of register <x>.  This value
+  *         can be |'d with others to produce a full register value for
+  *         register <x>.
+  *
+  *     <x>_<y>_m(void) : Returns a mask for field <y> of register <x>.  This
+  *         value can be ~'d and then &'d to clear the value of field <y> for
+  *         register <x>.
+  *
+  *     <x>_<y>_<z>_f(void) : Returns the constant value <z> after being shifted
+  *         to place it at field <y> of register <x>.  This value can be |'d
+  *         with others to produce a full register value for <x>.
+  *
+  *     <x>_<y>_v(u32 r) : Returns the value of field <y> from a full register
+  *         <x> value 'r' after being shifted to place its LSB at bit 0.
+  *         This value is suitable for direct comparison with other unshifted
+  *         values appropriate for use in field <y> of register <x>.
+  *
+  *     <x>_<y>_<z>_v(void) : Returns the constant value for <z> defined for
+  *         field <y> of register <x>.  This value is suitable for direct
+  *         comparison with unshifted values appropriate for use in field <y>
+  *         of register <x>.
+  */
+
+#ifndef __hw_host1x_uclass_host1x_h__
+#define __hw_host1x_uclass_host1x_h__
+/*This file is autogenerated.  Do not edit. */
+
+static inline u32 host1x_uclass_incr_syncpt_r(void)
+{
+	return 0x0;
+}
+static inline u32 host1x_uclass_incr_syncpt_cond_s(void)
+{
+	return 8;
+}
+static inline u32 host1x_uclass_incr_syncpt_cond_f(u32 v)
+{
+	return (v & 0xff) << 8;
+}
+static inline u32 host1x_uclass_incr_syncpt_cond_m(void)
+{
+	return 0xff << 8;
+}
+static inline u32 host1x_uclass_incr_syncpt_cond_v(u32 r)
+{
+	return (r >> 8) & 0xff;
+}
+static inline u32 host1x_uclass_incr_syncpt_cond_immediate_v(void)
+{
+	return 0;
+}
+static inline u32 host1x_uclass_incr_syncpt_cond_op_done_v(void)
+{
+	return 1;
+}
+static inline u32 host1x_uclass_incr_syncpt_cond_rd_done_v(void)
+{
+	return 2;
+}
+static inline u32 host1x_uclass_incr_syncpt_cond_reg_wr_safe_v(void)
+{
+	return 3;
+}
+static inline u32 host1x_uclass_incr_syncpt_indx_s(void)
+{
+	return 8;
+}
+static inline u32 host1x_uclass_incr_syncpt_indx_f(u32 v)
+{
+	return (v & 0xff) << 0;
+}
+static inline u32 host1x_uclass_incr_syncpt_indx_m(void)
+{
+	return 0xff << 0;
+}
+static inline u32 host1x_uclass_incr_syncpt_indx_v(u32 r)
+{
+	return (r >> 0) & 0xff;
+}
+static inline u32 host1x_uclass_wait_syncpt_r(void)
+{
+	return 0x8;
+}
+static inline u32 host1x_uclass_wait_syncpt_indx_s(void)
+{
+	return 8;
+}
+static inline u32 host1x_uclass_wait_syncpt_indx_f(u32 v)
+{
+	return (v & 0xff) << 24;
+}
+static inline u32 host1x_uclass_wait_syncpt_indx_m(void)
+{
+	return 0xff << 24;
+}
+static inline u32 host1x_uclass_wait_syncpt_indx_v(u32 r)
+{
+	return (r >> 24) & 0xff;
+}
+static inline u32 host1x_uclass_wait_syncpt_thresh_s(void)
+{
+	return 24;
+}
+static inline u32 host1x_uclass_wait_syncpt_thresh_f(u32 v)
+{
+	return (v & 0xffffff) << 0;
+}
+static inline u32 host1x_uclass_wait_syncpt_thresh_m(void)
+{
+	return 0xffffff << 0;
+}
+static inline u32 host1x_uclass_wait_syncpt_thresh_v(u32 r)
+{
+	return (r >> 0) & 0xffffff;
+}
+static inline u32 host1x_uclass_wait_syncpt_base_r(void)
+{
+	return 0x9;
+}
+static inline u32 host1x_uclass_wait_syncpt_base_indx_s(void)
+{
+	return 8;
+}
+static inline u32 host1x_uclass_wait_syncpt_base_indx_f(u32 v)
+{
+	return (v & 0xff) << 24;
+}
+static inline u32 host1x_uclass_wait_syncpt_base_indx_m(void)
+{
+	return 0xff << 24;
+}
+static inline u32 host1x_uclass_wait_syncpt_base_indx_v(u32 r)
+{
+	return (r >> 24) & 0xff;
+}
+static inline u32 host1x_uclass_wait_syncpt_base_base_indx_s(void)
+{
+	return 8;
+}
+static inline u32 host1x_uclass_wait_syncpt_base_base_indx_f(u32 v)
+{
+	return (v & 0xff) << 16;
+}
+static inline u32 host1x_uclass_wait_syncpt_base_base_indx_m(void)
+{
+	return 0xff << 16;
+}
+static inline u32 host1x_uclass_wait_syncpt_base_base_indx_v(u32 r)
+{
+	return (r >> 16) & 0xff;
+}
+static inline u32 host1x_uclass_wait_syncpt_base_offset_s(void)
+{
+	return 16;
+}
+static inline u32 host1x_uclass_wait_syncpt_base_offset_f(u32 v)
+{
+	return (v & 0xffff) << 0;
+}
+static inline u32 host1x_uclass_wait_syncpt_base_offset_m(void)
+{
+	return 0xffff << 0;
+}
+static inline u32 host1x_uclass_wait_syncpt_base_offset_v(u32 r)
+{
+	return (r >> 0) & 0xffff;
+}
+static inline u32 host1x_uclass_load_syncpt_base_r(void)
+{
+	return 0xb;
+}
+static inline u32 host1x_uclass_load_syncpt_base_base_indx_s(void)
+{
+	return 8;
+}
+static inline u32 host1x_uclass_load_syncpt_base_base_indx_f(u32 v)
+{
+	return (v & 0xff) << 24;
+}
+static inline u32 host1x_uclass_load_syncpt_base_base_indx_m(void)
+{
+	return 0xff << 24;
+}
+static inline u32 host1x_uclass_load_syncpt_base_base_indx_v(u32 r)
+{
+	return (r >> 24) & 0xff;
+}
+static inline u32 host1x_uclass_load_syncpt_base_value_s(void)
+{
+	return 24;
+}
+static inline u32 host1x_uclass_load_syncpt_base_value_f(u32 v)
+{
+	return (v & 0xffffff) << 0;
+}
+static inline u32 host1x_uclass_load_syncpt_base_value_m(void)
+{
+	return 0xffffff << 0;
+}
+static inline u32 host1x_uclass_load_syncpt_base_value_v(u32 r)
+{
+	return (r >> 0) & 0xffffff;
+}
+static inline u32 host1x_uclass_incr_syncpt_base_r(void)
+{
+	return 0xc;
+}
+static inline u32 host1x_uclass_incr_syncpt_base_base_indx_s(void)
+{
+	return 8;
+}
+static inline u32 host1x_uclass_incr_syncpt_base_base_indx_f(u32 v)
+{
+	return (v & 0xff) << 24;
+}
+static inline u32 host1x_uclass_incr_syncpt_base_base_indx_m(void)
+{
+	return 0xff << 24;
+}
+static inline u32 host1x_uclass_incr_syncpt_base_base_indx_v(u32 r)
+{
+	return (r >> 24) & 0xff;
+}
+static inline u32 host1x_uclass_incr_syncpt_base_offset_s(void)
+{
+	return 24;
+}
+static inline u32 host1x_uclass_incr_syncpt_base_offset_f(u32 v)
+{
+	return (v & 0xffffff) << 0;
+}
+static inline u32 host1x_uclass_incr_syncpt_base_offset_m(void)
+{
+	return 0xffffff << 0;
+}
+static inline u32 host1x_uclass_incr_syncpt_base_offset_v(u32 r)
+{
+	return (r >> 0) & 0xffffff;
+}
+static inline u32 host1x_uclass_indoff_r(void)
+{
+	return 0x2d;
+}
+static inline u32 host1x_uclass_indoff_indbe_s(void)
+{
+	return 4;
+}
+static inline u32 host1x_uclass_indoff_indbe_f(u32 v)
+{
+	return (v & 0xf) << 28;
+}
+static inline u32 host1x_uclass_indoff_indbe_m(void)
+{
+	return 0xf << 28;
+}
+static inline u32 host1x_uclass_indoff_indbe_v(u32 r)
+{
+	return (r >> 28) & 0xf;
+}
+static inline u32 host1x_uclass_indoff_autoinc_s(void)
+{
+	return 1;
+}
+static inline u32 host1x_uclass_indoff_autoinc_f(u32 v)
+{
+	return (v & 0x1) << 27;
+}
+static inline u32 host1x_uclass_indoff_autoinc_m(void)
+{
+	return 0x1 << 27;
+}
+static inline u32 host1x_uclass_indoff_autoinc_v(u32 r)
+{
+	return (r >> 27) & 0x1;
+}
+static inline u32 host1x_uclass_indoff_spool_s(void)
+{
+	return 1;
+}
+static inline u32 host1x_uclass_indoff_spool_f(u32 v)
+{
+	return (v & 0x1) << 26;
+}
+static inline u32 host1x_uclass_indoff_spool_m(void)
+{
+	return 0x1 << 26;
+}
+static inline u32 host1x_uclass_indoff_spool_v(u32 r)
+{
+	return (r >> 26) & 0x1;
+}
+static inline u32 host1x_uclass_indoff_indoffset_s(void)
+{
+	return 24;
+}
+static inline u32 host1x_uclass_indoff_indoffset_f(u32 v)
+{
+	return (v & 0xffffff) << 2;
+}
+static inline u32 host1x_uclass_indoff_indoffset_m(void)
+{
+	return 0xffffff << 2;
+}
+static inline u32 host1x_uclass_indoff_indoffset_v(u32 r)
+{
+	return (r >> 2) & 0xffffff;
+}
+static inline u32 host1x_uclass_indoff_indmodid_s(void)
+{
+	return 8;
+}
+static inline u32 host1x_uclass_indoff_indmodid_f(u32 v)
+{
+	return (v & 0xff) << 18;
+}
+static inline u32 host1x_uclass_indoff_indmodid_m(void)
+{
+	return 0xff << 18;
+}
+static inline u32 host1x_uclass_indoff_indmodid_v(u32 r)
+{
+	return (r >> 18) & 0xff;
+}
+static inline u32 host1x_uclass_indoff_indmodid_host1x_v(void)
+{
+	return 0;
+}
+static inline u32 host1x_uclass_indoff_indmodid_mpe_v(void)
+{
+	return 1;
+}
+static inline u32 host1x_uclass_indoff_indmodid_vi_v(void)
+{
+	return 2;
+}
+static inline u32 host1x_uclass_indoff_indmodid_epp_v(void)
+{
+	return 3;
+}
+static inline u32 host1x_uclass_indoff_indmodid_isp_v(void)
+{
+	return 4;
+}
+static inline u32 host1x_uclass_indoff_indmodid_gr2d_v(void)
+{
+	return 5;
+}
+static inline u32 host1x_uclass_indoff_indmodid_gr3d_v(void)
+{
+	return 6;
+}
+static inline u32 host1x_uclass_indoff_indmodid_display_v(void)
+{
+	return 8;
+}
+static inline u32 host1x_uclass_indoff_indmodid_tvo_v(void)
+{
+	return 11;
+}
+static inline u32 host1x_uclass_indoff_indmodid_displayb_v(void)
+{
+	return 9;
+}
+static inline u32 host1x_uclass_indoff_indmodid_dsi_v(void)
+{
+	return 12;
+}
+static inline u32 host1x_uclass_indoff_indmodid_hdmi_v(void)
+{
+	return 10;
+}
+static inline u32 host1x_uclass_indoff_indmodid_dsib_v(void)
+{
+	return 16;
+}
+static inline u32 host1x_uclass_indoff_indroffset_s(void)
+{
+	return 16;
+}
+static inline u32 host1x_uclass_indoff_indroffset_f(u32 v)
+{
+	return (v & 0xffff) << 2;
+}
+static inline u32 host1x_uclass_indoff_indroffset_m(void)
+{
+	return 0xffff << 2;
+}
+static inline u32 host1x_uclass_indoff_indroffset_v(u32 r)
+{
+	return (r >> 2) & 0xffff;
+}
+static inline u32 host1x_uclass_indoff_acctype_s(void)
+{
+	return 1;
+}
+static inline u32 host1x_uclass_indoff_acctype_f(u32 v)
+{
+	return (v & 0x1) << 1;
+}
+static inline u32 host1x_uclass_indoff_acctype_m(void)
+{
+	return 0x1 << 1;
+}
+static inline u32 host1x_uclass_indoff_acctype_v(u32 r)
+{
+	return (r >> 1) & 0x1;
+}
+static inline u32 host1x_uclass_indoff_acctype_reg_v(void)
+{
+	return 0;
+}
+static inline u32 host1x_uclass_indoff_acctype_fb_v(void)
+{
+	return 1;
+}
+static inline u32 host1x_uclass_indoff_rwn_s(void)
+{
+	return 1;
+}
+static inline u32 host1x_uclass_indoff_rwn_f(u32 v)
+{
+	return (v & 0x1) << 0;
+}
+static inline u32 host1x_uclass_indoff_rwn_m(void)
+{
+	return 0x1 << 0;
+}
+static inline u32 host1x_uclass_indoff_rwn_v(u32 r)
+{
+	return (r >> 0) & 0x1;
+}
+static inline u32 host1x_uclass_indoff_rwn_write_v(void)
+{
+	return 0;
+}
+static inline u32 host1x_uclass_indoff_rwn_read_v(void)
+{
+	return 1;
+}
+static inline u32 host1x_uclass_inddata_r(void)
+{
+	return 0x2e;
+}
+
+#endif /* __hw_host1x_uclass_host1x_h__ */
diff --git a/drivers/video/tegra/host/nvhost_acm.c b/drivers/video/tegra/host/nvhost_acm.c
new file mode 100644
index 0000000..6da816a
--- /dev/null
+++ b/drivers/video/tegra/host/nvhost_acm.c
@@ -0,0 +1,532 @@ 
+/*
+ * drivers/video/tegra/host/nvhost_acm.c
+ *
+ * Tegra Graphics Host Automatic Clock Management
+ *
+ * Copyright (c) 2010-2012, NVIDIA Corporation. All rights reserved.
+ *
+ * This program is free software; you can redistribute it and/or modify it
+ * under the terms and conditions of the GNU General Public License,
+ * version 2, as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
+ * more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program.  If not, see <http://www.gnu.org/licenses/>.
+ */
+
+#include <linux/module.h>
+#include <linux/slab.h>
+#include <linux/stat.h>
+#include <linux/string.h>
+#include <linux/sched.h>
+#include <linux/clk.h>
+#include <linux/err.h>
+#include <linux/device.h>
+#include <linux/delay.h>
+#include <linux/platform_device.h>
+#include <trace/events/nvhost.h>
+
+#include <mach/powergate.h>
+#include <mach/clk.h>
+
+#include "nvhost_acm.h"
+#include "dev.h"
+
+#define ACM_SUSPEND_WAIT_FOR_IDLE_TIMEOUT	(2 * HZ)
+#define POWERGATE_DELAY				10
+#define MAX_DEVID_LENGTH			16
+
+static void do_powergate_locked(int id)
+{
+	if (id != -1 && tegra_powergate_is_powered(id))
+		tegra_powergate_power_off(id);
+}
+
+static void do_unpowergate_locked(int id)
+{
+	if (id != -1)
+		tegra_powergate_power_on(id);
+}
+
+static void to_state_clockgated_locked(struct platform_device *dev)
+{
+	struct nvhost_device_data *pdata = platform_get_drvdata(dev);
+
+	if (pdata->powerstate == NVHOST_POWER_STATE_RUNNING) {
+		int i, err;
+		if (pdata->prepare_clockoff) {
+			err = pdata->prepare_clockoff(dev);
+			if (err) {
+				dev_err(&dev->dev, "error clock gating");
+				return;
+			}
+		}
+
+		for (i = 0; i < pdata->num_clks; i++)
+			clk_disable_unprepare(pdata->clk[i]);
+		if (dev->dev.parent)
+			nvhost_module_idle(to_platform_device(dev->dev.parent));
+	} else if (pdata->powerstate == NVHOST_POWER_STATE_POWERGATED
+			&& pdata->can_powergate) {
+		do_unpowergate_locked(pdata->powergate_ids[0]);
+		do_unpowergate_locked(pdata->powergate_ids[1]);
+	}
+	pdata->powerstate = NVHOST_POWER_STATE_CLOCKGATED;
+}
+
+static void to_state_running_locked(struct platform_device *dev)
+{
+	struct nvhost_device_data *pdata = platform_get_drvdata(dev);
+	int prev_state = pdata->powerstate;
+
+	if (pdata->powerstate == NVHOST_POWER_STATE_POWERGATED)
+		to_state_clockgated_locked(dev);
+
+	if (pdata->powerstate == NVHOST_POWER_STATE_CLOCKGATED) {
+		int i;
+
+		if (dev->dev.parent)
+			nvhost_module_busy(to_platform_device(dev->dev.parent));
+
+		for (i = 0; i < pdata->num_clks; i++) {
+			int err = clk_prepare_enable(pdata->clk[i]);
+			if (err) {
+				dev_err(&dev->dev, "Cannot turn on clock %s",
+					pdata->clocks[i].name);
+				return;
+			}
+		}
+
+		if (pdata->finalize_clockon)
+			pdata->finalize_clockon(dev);
+
+		/* Invoke callback after power un-gating. This is used for
+		 * restoring context. */
+		if (prev_state == NVHOST_POWER_STATE_POWERGATED
+				&& pdata->finalize_poweron)
+			pdata->finalize_poweron(dev);
+	}
+	pdata->powerstate = NVHOST_POWER_STATE_RUNNING;
+}
+
+/* This gets called from powergate_handler() and from module suspend.
+ * Module suspend is done for all modules, runtime power gating only
+ * for modules with can_powergate set.
+ */
+static int to_state_powergated_locked(struct platform_device *dev)
+{
+	struct nvhost_device_data *pdata = platform_get_drvdata(dev);
+	int err = 0;
+
+	if (pdata->prepare_poweroff &&
+		pdata->powerstate != NVHOST_POWER_STATE_POWERGATED) {
+		/* Clock needs to be on in prepare_poweroff */
+		to_state_running_locked(dev);
+		err = pdata->prepare_poweroff(dev);
+		if (err)
+			return err;
+	}
+
+	if (pdata->powerstate == NVHOST_POWER_STATE_RUNNING)
+		to_state_clockgated_locked(dev);
+
+	if (pdata->can_powergate) {
+		do_powergate_locked(pdata->powergate_ids[0]);
+		do_powergate_locked(pdata->powergate_ids[1]);
+	}
+
+	pdata->powerstate = NVHOST_POWER_STATE_POWERGATED;
+	return 0;
+}
+
+static void schedule_powergating_locked(struct platform_device *dev)
+{
+	struct nvhost_device_data *pdata = platform_get_drvdata(dev);
+	if (pdata->can_powergate)
+		schedule_delayed_work(&pdata->powerstate_down,
+				msecs_to_jiffies(pdata->powergate_delay));
+}
+
+static void schedule_clockgating_locked(struct platform_device *dev)
+{
+	struct nvhost_device_data *pdata = platform_get_drvdata(dev);
+	schedule_delayed_work(&pdata->powerstate_down,
+			msecs_to_jiffies(pdata->clockgate_delay));
+}
+
+void nvhost_module_busy(struct platform_device *dev)
+{
+	struct nvhost_device_data *pdata = platform_get_drvdata(dev);
+
+	if (pdata->busy)
+		pdata->busy(dev);
+
+	mutex_lock(&pdata->lock);
+	cancel_delayed_work(&pdata->powerstate_down);
+
+	pdata->refcount++;
+	if (pdata->refcount > 0 && !nvhost_module_powered(dev))
+		to_state_running_locked(dev);
+	mutex_unlock(&pdata->lock);
+}
+
+static void powerstate_down_handler(struct work_struct *work)
+{
+	struct platform_device *dev;
+	struct nvhost_device_data *pdata;
+
+	pdata = container_of(to_delayed_work(work),
+			struct nvhost_device_data,
+			powerstate_down);
+
+	dev = pdata->pdev;
+
+	mutex_lock(&pdata->lock);
+	if (pdata->refcount == 0) {
+		switch (pdata->powerstate) {
+		case NVHOST_POWER_STATE_RUNNING:
+			to_state_clockgated_locked(dev);
+			schedule_powergating_locked(dev);
+			break;
+		case NVHOST_POWER_STATE_CLOCKGATED:
+			if (to_state_powergated_locked(dev))
+				schedule_powergating_locked(dev);
+			break;
+		default:
+			break;
+		}
+	}
+	mutex_unlock(&pdata->lock);
+}
+
+void nvhost_module_idle_mult(struct platform_device *dev, int refs)
+{
+	struct nvhost_device_data *pdata = platform_get_drvdata(dev);
+	bool kick = false;
+
+	mutex_lock(&pdata->lock);
+	pdata->refcount -= refs;
+	if (pdata->refcount == 0) {
+		if (nvhost_module_powered(dev))
+			schedule_clockgating_locked(dev);
+		kick = true;
+	}
+	mutex_unlock(&pdata->lock);
+
+	if (kick) {
+		wake_up(&pdata->idle_wq);
+
+		if (pdata->idle)
+			pdata->idle(dev);
+	}
+}
+
+static ssize_t refcount_show(struct kobject *kobj,
+	struct kobj_attribute *attr, char *buf)
+{
+	int ret;
+	struct nvhost_device_power_attr *power_attribute =
+		container_of(attr, struct nvhost_device_power_attr,
+			power_attr[NVHOST_POWER_SYSFS_ATTRIB_REFCOUNT]);
+	struct platform_device *dev = power_attribute->ndev;
+	struct nvhost_device_data *pdata = platform_get_drvdata(dev);
+
+	mutex_lock(&pdata->lock);
+	ret = sprintf(buf, "%d\n", pdata->refcount);
+	mutex_unlock(&pdata->lock);
+
+	return ret;
+}
+
+static ssize_t powergate_delay_store(struct kobject *kobj,
+	struct kobj_attribute *attr, const char *buf, size_t count)
+{
+	int powergate_delay = 0, ret = 0;
+	struct nvhost_device_power_attr *power_attribute =
+		container_of(attr, struct nvhost_device_power_attr,
+			power_attr[NVHOST_POWER_SYSFS_ATTRIB_POWERGATE_DELAY]);
+	struct platform_device *dev = power_attribute->ndev;
+	struct nvhost_device_data *pdata = platform_get_drvdata(dev);
+
+	if (!pdata->can_powergate) {
+		dev_info(&dev->dev, "does not support power-gating\n");
+		return count;
+	}
+
+	mutex_lock(&pdata->lock);
+	ret = sscanf(buf, "%d", &powergate_delay);
+	if (ret == 1 && powergate_delay >= 0)
+		pdata->powergate_delay = powergate_delay;
+	else
+		dev_err(&dev->dev, "Invalid powergate delay\n");
+	mutex_unlock(&pdata->lock);
+
+	return count;
+}
+
+static ssize_t powergate_delay_show(struct kobject *kobj,
+	struct kobj_attribute *attr, char *buf)
+{
+	int ret;
+	struct nvhost_device_power_attr *power_attribute =
+		container_of(attr, struct nvhost_device_power_attr,
+			power_attr[NVHOST_POWER_SYSFS_ATTRIB_POWERGATE_DELAY]);
+	struct platform_device *dev = power_attribute->ndev;
+	struct nvhost_device_data *pdata = platform_get_drvdata(dev);
+
+	mutex_lock(&pdata->lock);
+	ret = sprintf(buf, "%d\n", pdata->powergate_delay);
+	mutex_unlock(&pdata->lock);
+
+	return ret;
+}
+
+static ssize_t clockgate_delay_store(struct kobject *kobj,
+	struct kobj_attribute *attr, const char *buf, size_t count)
+{
+	int clockgate_delay = 0, ret = 0;
+	struct nvhost_device_power_attr *power_attribute =
+		container_of(attr, struct nvhost_device_power_attr,
+			power_attr[NVHOST_POWER_SYSFS_ATTRIB_CLOCKGATE_DELAY]);
+	struct platform_device *dev = power_attribute->ndev;
+	struct nvhost_device_data *pdata = platform_get_drvdata(dev);
+
+	mutex_lock(&pdata->lock);
+	ret = sscanf(buf, "%d", &clockgate_delay);
+	if (ret == 1 && clockgate_delay >= 0)
+		pdata->clockgate_delay = clockgate_delay;
+	else
+		dev_err(&dev->dev, "Invalid clockgate delay\n");
+	mutex_unlock(&pdata->lock);
+
+	return count;
+}
+
+static ssize_t clockgate_delay_show(struct kobject *kobj,
+	struct kobj_attribute *attr, char *buf)
+{
+	int ret;
+	struct nvhost_device_power_attr *power_attribute =
+		container_of(attr, struct nvhost_device_power_attr,
+			power_attr[NVHOST_POWER_SYSFS_ATTRIB_CLOCKGATE_DELAY]);
+	struct platform_device *dev = power_attribute->ndev;
+	struct nvhost_device_data *pdata = platform_get_drvdata(dev);
+
+	mutex_lock(&pdata->lock);
+	ret = sprintf(buf, "%d\n", pdata->clockgate_delay);
+	mutex_unlock(&pdata->lock);
+
+	return ret;
+}
+
+int nvhost_module_init(struct platform_device *dev)
+{
+	int i = 0, err = 0;
+	struct kobj_attribute *attr = NULL;
+	struct nvhost_device_data *pdata = platform_get_drvdata(dev);
+
+	/* initialize clocks to known state */
+	while (pdata->clocks[i].name && i < NVHOST_MODULE_MAX_CLOCKS) {
+		char devname[MAX_DEVID_LENGTH];
+		long rate = pdata->clocks[i].default_rate;
+		struct clk *c;
+
+		snprintf(devname, MAX_DEVID_LENGTH,
+				"tegra_%s", dev_name(&dev->dev));
+		c = devm_clk_get(&dev->dev, pdata->clocks[i].name);
+		if (IS_ERR_OR_NULL(c)) {
+			dev_err(&dev->dev, "Cannot get clock %s\n",
+					pdata->clocks[i].name);
+			return -ENODEV;
+		}
+
+		rate = clk_round_rate(c, rate);
+		clk_prepare_enable(c);
+		clk_set_rate(c, rate);
+		clk_disable_unprepare(c);
+		pdata->clk[i] = c;
+		i++;
+	}
+	pdata->num_clks = i;
+
+	mutex_init(&pdata->lock);
+	init_waitqueue_head(&pdata->idle_wq);
+	INIT_DELAYED_WORK(&pdata->powerstate_down, powerstate_down_handler);
+
+	/* power gate units that we can power gate */
+	if (pdata->can_powergate) {
+		do_powergate_locked(pdata->powergate_ids[0]);
+		do_powergate_locked(pdata->powergate_ids[1]);
+		pdata->powerstate = NVHOST_POWER_STATE_POWERGATED;
+	} else {
+		do_unpowergate_locked(pdata->powergate_ids[0]);
+		do_unpowergate_locked(pdata->powergate_ids[1]);
+		pdata->powerstate = NVHOST_POWER_STATE_CLOCKGATED;
+	}
+
+	/* Init the power sysfs attributes for this device */
+	pdata->power_attrib = devm_kzalloc(&dev->dev,
+			sizeof(struct nvhost_device_power_attr),
+		GFP_KERNEL);
+	if (!pdata->power_attrib) {
+		dev_err(&dev->dev, "Unable to allocate sysfs attributes\n");
+		return -ENOMEM;
+	}
+	pdata->power_attrib->ndev = dev;
+
+	pdata->power_kobj = kobject_create_and_add("acm", &dev->dev.kobj);
+	if (!pdata->power_kobj) {
+		dev_err(&dev->dev, "Could not add dir 'power'\n");
+		err = -EIO;
+		goto fail_attrib_alloc;
+	}
+
+	attr = &pdata->power_attrib->power_attr[NVHOST_POWER_SYSFS_ATTRIB_CLOCKGATE_DELAY];
+	attr->attr.name = "clockgate_delay";
+	attr->attr.mode = S_IWUSR | S_IRUGO;
+	attr->show = clockgate_delay_show;
+	attr->store = clockgate_delay_store;
+	if (sysfs_create_file(pdata->power_kobj, &attr->attr)) {
+		dev_err(&dev->dev, "Could not create sysfs attribute clockgate_delay\n");
+		err = -EIO;
+		goto fail_clockdelay;
+	}
+
+	attr = &pdata->power_attrib->power_attr[NVHOST_POWER_SYSFS_ATTRIB_POWERGATE_DELAY];
+	attr->attr.name = "powergate_delay";
+	attr->attr.mode = S_IWUSR | S_IRUGO;
+	attr->show = powergate_delay_show;
+	attr->store = powergate_delay_store;
+	if (sysfs_create_file(pdata->power_kobj, &attr->attr)) {
+		dev_err(&dev->dev, "Could not create sysfs attribute powergate_delay\n");
+		err = -EIO;
+		goto fail_powergatedelay;
+	}
+
+	attr = &pdata->power_attrib->power_attr[NVHOST_POWER_SYSFS_ATTRIB_REFCOUNT];
+	attr->attr.name = "refcount";
+	attr->attr.mode = S_IRUGO;
+	attr->show = refcount_show;
+	if (sysfs_create_file(pdata->power_kobj, &attr->attr)) {
+		dev_err(&dev->dev, "Could not create sysfs attribute refcount\n");
+		err = -EIO;
+		goto fail_refcount;
+	}
+
+	return 0;
+
+fail_refcount:
+	attr = &pdata->power_attrib->power_attr[NVHOST_POWER_SYSFS_ATTRIB_POWERGATE_DELAY];
+	sysfs_remove_file(pdata->power_kobj, &attr->attr);
+
+fail_powergatedelay:
+	attr = &pdata->power_attrib->power_attr[NVHOST_POWER_SYSFS_ATTRIB_CLOCKGATE_DELAY];
+	sysfs_remove_file(pdata->power_kobj, &attr->attr);
+
+fail_clockdelay:
+	kobject_put(pdata->power_kobj);
+
+fail_attrib_alloc:
+	kfree(pdata->power_attrib);
+
+	return err;
+}
+
+static int is_module_idle(struct platform_device *dev)
+{
+	int count;
+	struct nvhost_device_data *pdata = platform_get_drvdata(dev);
+
+	mutex_lock(&pdata->lock);
+	count = pdata->refcount;
+	mutex_unlock(&pdata->lock);
+
+	return (count == 0);
+}
+
+int nvhost_module_suspend(struct platform_device *dev)
+{
+	int ret;
+	struct nvhost_device_data *pdata = platform_get_drvdata(dev);
+
+	ret = wait_event_timeout(pdata->idle_wq, is_module_idle(dev),
+			ACM_SUSPEND_WAIT_FOR_IDLE_TIMEOUT);
+	if (ret == 0) {
+		dev_info(&dev->dev, "%s prevented suspend\n",
+				dev_name(&dev->dev));
+		return -EBUSY;
+	}
+
+	mutex_lock(&pdata->lock);
+	cancel_delayed_work(&pdata->powerstate_down);
+	to_state_powergated_locked(dev);
+	mutex_unlock(&pdata->lock);
+
+	if (pdata->suspend_ndev)
+		pdata->suspend_ndev(dev);
+
+	return 0;
+}
+
+void nvhost_module_deinit(struct platform_device *dev)
+{
+	int i;
+	struct nvhost_device_data *pdata = platform_get_drvdata(dev);
+
+	if (pdata->deinit)
+		pdata->deinit(dev);
+
+	nvhost_module_suspend(dev);
+	for (i = 0; i < pdata->num_clks; i++)
+		clk_put(pdata->clk[i]);
+	pdata->powerstate = NVHOST_POWER_STATE_DEINIT;
+}
+
+/* public host1x power management APIs */
+bool nvhost_module_powered_ext(struct platform_device *dev)
+{
+	bool ret = 0;
+
+	/* get the parent */
+	if (dev->dev.parent) {
+		struct platform_device *pdev;
+		pdev = to_platform_device(dev->dev.parent);
+
+		ret = nvhost_module_powered(pdev);
+	} else {
+		dev_warn(&dev->dev, "Cannot return power state, no parent\n");
+	}
+
+	return ret;
+}
+
+void nvhost_module_busy_ext(struct platform_device *dev)
+{
+	/* get the parent */
+	if (dev->dev.parent) {
+		struct platform_device *pdev;
+		pdev = to_platform_device(dev->dev.parent);
+
+		nvhost_module_busy(pdev);
+	} else {
+		dev_warn(&dev->dev, "Cannot turn on, no parent\n");
+	}
+}
+EXPORT_SYMBOL(nvhost_module_busy_ext);
+
+void nvhost_module_idle_ext(struct platform_device *dev)
+{
+	/* get the parent */
+	if (dev->dev.parent) {
+		struct platform_device *pdev;
+		pdev = to_platform_device(dev->dev.parent);
+
+		nvhost_module_idle(pdev);
+	} else {
+		dev_warn(&dev->dev, "Cannot idle, no parent\n");
+	}
+}
diff --git a/drivers/video/tegra/host/nvhost_acm.h b/drivers/video/tegra/host/nvhost_acm.h
new file mode 100644
index 0000000..65ab3f0
--- /dev/null
+++ b/drivers/video/tegra/host/nvhost_acm.h
@@ -0,0 +1,49 @@ 
+/*
+ * drivers/video/tegra/host/nvhost_acm.h
+ *
+ * Tegra Graphics Host Automatic Clock Management
+ *
+ * Copyright (c) 2010-2012, NVIDIA Corporation.
+ *
+ * This program is free software; you can redistribute it and/or modify it
+ * under the terms and conditions of the GNU General Public License,
+ * version 2, as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
+ * more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program.  If not, see <http://www.gnu.org/licenses/>.
+ */
+
+#ifndef __NVHOST_ACM_H
+#define __NVHOST_ACM_H
+
+#include <linux/workqueue.h>
+#include <linux/wait.h>
+#include <linux/mutex.h>
+#include <linux/clk.h>
+#include <linux/nvhost.h>
+
+/* Sets clocks and powergating state for a module */
+int nvhost_module_init(struct platform_device *ndev);
+void nvhost_module_deinit(struct platform_device *dev);
+int nvhost_module_suspend(struct platform_device *dev);
+
+void nvhost_module_busy(struct platform_device *dev);
+void nvhost_module_idle_mult(struct platform_device *dev, int refs);
+
+static inline bool nvhost_module_powered(struct platform_device *dev)
+{
+	struct nvhost_device_data *pdata = platform_get_drvdata(dev);
+	return pdata->powerstate == NVHOST_POWER_STATE_RUNNING;
+}
+
+static inline void nvhost_module_idle(struct platform_device *dev)
+{
+	nvhost_module_idle_mult(dev, 1);
+}
+
+#endif
diff --git a/drivers/video/tegra/host/nvhost_cdma.c b/drivers/video/tegra/host/nvhost_cdma.c
new file mode 100644
index 0000000..71f0343
--- /dev/null
+++ b/drivers/video/tegra/host/nvhost_cdma.c
@@ -0,0 +1,473 @@ 
+/*
+ * drivers/video/tegra/host/nvhost_cdma.c
+ *
+ * Tegra Graphics Host Command DMA
+ *
+ * Copyright (c) 2010-2012, NVIDIA Corporation.
+ *
+ * This program is free software; you can redistribute it and/or modify it
+ * under the terms and conditions of the GNU General Public License,
+ * version 2, as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
+ * more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program.  If not, see <http://www.gnu.org/licenses/>.
+ */
+
+#include "nvhost_cdma.h"
+#include "nvhost_channel.h"
+#include "dev.h"
+#include "debug.h"
+#include "nvhost_memmgr.h"
+#include "chip_support.h"
+#include <asm/cacheflush.h>
+
+#include <linux/slab.h>
+#include <linux/kfifo.h>
+#include <trace/events/nvhost.h>
+#include <linux/interrupt.h>
+
+/*
+ * TODO:
+ *   stats
+ *     - for figuring out what to optimize further
+ *   resizable push buffer
+ *     - some channels hardly need any, some channels (3d) could use more
+ */
+
+/**
+ * Add an entry to the sync queue.
+ */
+static void add_to_sync_queue(struct nvhost_cdma *cdma,
+			      struct nvhost_job *job,
+			      u32 nr_slots,
+			      u32 first_get)
+{
+	if (job->syncpt_id == NVSYNCPT_INVALID) {
+		dev_warn(&job->ch->dev->dev, "%s: Invalid syncpt\n",
+				__func__);
+		return;
+	}
+
+	job->first_get = first_get;
+	job->num_slots = nr_slots;
+	nvhost_job_get(job);
+	list_add_tail(&job->list, &cdma->sync_queue);
+}
+
+/**
+ * Return the status of the cdma's sync queue or push buffer for the given event
+ *  - sq empty: returns 1 for empty, 0 for not empty (as in "1 empty queue" :-)
+ *  - pb space: returns the number of free slots in the channel's push buffer
+ * Must be called with the cdma lock held.
+ */
+static unsigned int cdma_status_locked(struct nvhost_cdma *cdma,
+		enum cdma_event event)
+{
+	switch (event) {
+	case CDMA_EVENT_SYNC_QUEUE_EMPTY:
+		return list_empty(&cdma->sync_queue) ? 1 : 0;
+	case CDMA_EVENT_PUSH_BUFFER_SPACE: {
+		struct push_buffer *pb = &cdma->push_buffer;
+		return cdma_pb_op().space(pb);
+	}
+	default:
+		return 0;
+	}
+}
+
+/**
+ * Sleep (if necessary) until the requested event happens
+ *   - CDMA_EVENT_SYNC_QUEUE_EMPTY : sync queue is completely empty.
+ *     - Returns 1
+ *   - CDMA_EVENT_PUSH_BUFFER_SPACE : there is space in the push buffer
+ *     - Return the amount of space (> 0)
+ * Must be called with the cdma lock held.
+ */
+unsigned int nvhost_cdma_wait_locked(struct nvhost_cdma *cdma,
+		enum cdma_event event)
+{
+	for (;;) {
+		unsigned int space = cdma_status_locked(cdma, event);
+		if (space)
+			return space;
+
+		trace_nvhost_wait_cdma(cdma_to_channel(cdma)->dev->name,
+				event);
+
+		/* If somebody has managed to already start waiting, yield */
+		if (cdma->event != CDMA_EVENT_NONE) {
+			mutex_unlock(&cdma->lock);
+			schedule();
+			mutex_lock(&cdma->lock);
+			continue;
+		}
+		cdma->event = event;
+
+		mutex_unlock(&cdma->lock);
+		down(&cdma->sem);
+		mutex_lock(&cdma->lock);
+	}
+	return 0;
+}
+
+/**
+ * Start timer for a buffer submition that has completed yet.
+ * Must be called with the cdma lock held.
+ */
+static void cdma_start_timer_locked(struct nvhost_cdma *cdma,
+		struct nvhost_job *job)
+{
+	if (cdma->timeout.clientid) {
+		/* timer already started */
+		return;
+	}
+
+	cdma->timeout.clientid = job->clientid;
+	cdma->timeout.syncpt_id = job->syncpt_id;
+	cdma->timeout.syncpt_val = job->syncpt_end;
+	cdma->timeout.start_ktime = ktime_get();
+
+	schedule_delayed_work(&cdma->timeout.wq,
+			msecs_to_jiffies(job->timeout));
+}
+
+/**
+ * Stop timer when a buffer submition completes.
+ * Must be called with the cdma lock held.
+ */
+static void stop_cdma_timer_locked(struct nvhost_cdma *cdma)
+{
+	cancel_delayed_work(&cdma->timeout.wq);
+	cdma->timeout.clientid = 0;
+}
+
+/**
+ * For all sync queue entries that have already finished according to the
+ * current sync point registers:
+ *  - unpin & unref their mems
+ *  - pop their push buffer slots
+ *  - remove them from the sync queue
+ * This is normally called from the host code's worker thread, but can be
+ * called manually if necessary.
+ * Must be called with the cdma lock held.
+ */
+static void update_cdma_locked(struct nvhost_cdma *cdma)
+{
+	bool signal = false;
+	struct nvhost_master *dev = cdma_to_dev(cdma);
+	struct nvhost_syncpt *sp = &dev->syncpt;
+	struct nvhost_job *job, *n;
+
+	/* If CDMA is stopped, queue is cleared and we can return */
+	if (!cdma->running)
+		return;
+
+	/*
+	 * Walk the sync queue, reading the sync point registers as necessary,
+	 * to consume as many sync queue entries as possible without blocking
+	 */
+	list_for_each_entry_safe(job, n, &cdma->sync_queue, list) {
+		/* Check whether this syncpt has completed, and bail if not */
+		if (!nvhost_syncpt_is_expired(sp,
+				job->syncpt_id, job->syncpt_end)) {
+			/* Start timer on next pending syncpt */
+			if (job->timeout)
+				cdma_start_timer_locked(cdma, job);
+			break;
+		}
+
+		/* Cancel timeout, when a buffer completes */
+		if (cdma->timeout.clientid)
+			stop_cdma_timer_locked(cdma);
+
+		/* Unpin the memory */
+		nvhost_job_unpin(job);
+
+		/* Pop push buffer slots */
+		if (job->num_slots) {
+			struct push_buffer *pb = &cdma->push_buffer;
+			cdma_pb_op().pop_from(pb, job->num_slots);
+			if (cdma->event == CDMA_EVENT_PUSH_BUFFER_SPACE)
+				signal = true;
+		}
+
+		list_del(&job->list);
+		nvhost_job_put(job);
+	}
+
+	if (list_empty(&cdma->sync_queue) &&
+				cdma->event == CDMA_EVENT_SYNC_QUEUE_EMPTY)
+			signal = true;
+
+	/* Wake up CdmaWait() if the requested event happened */
+	if (signal) {
+		cdma->event = CDMA_EVENT_NONE;
+		up(&cdma->sem);
+	}
+}
+
+void nvhost_cdma_update_sync_queue(struct nvhost_cdma *cdma,
+		struct nvhost_syncpt *syncpt, struct platform_device *dev)
+{
+	u32 get_restart;
+	u32 syncpt_incrs;
+	struct nvhost_job *job = NULL;
+	u32 syncpt_val;
+
+	syncpt_val = nvhost_syncpt_update_min(syncpt, cdma->timeout.syncpt_id);
+
+	dev_dbg(&dev->dev,
+		"%s: starting cleanup (thresh %d)\n",
+		__func__, syncpt_val);
+
+	/*
+	 * Move the sync_queue read pointer to the first entry that hasn't
+	 * completed based on the current HW syncpt value. It's likely there
+	 * won't be any (i.e. we're still at the head), but covers the case
+	 * where a syncpt incr happens just prior/during the teardown.
+	 */
+
+	dev_dbg(&dev->dev,
+		"%s: skip completed buffers still in sync_queue\n",
+		__func__);
+
+	list_for_each_entry(job, &cdma->sync_queue, list) {
+		if (syncpt_val < job->syncpt_end)
+			break;
+
+		nvhost_job_dump(&dev->dev, job);
+	}
+
+	/*
+	 * Walk the sync_queue, first incrementing with the CPU syncpts that
+	 * are partially executed (the first buffer) or fully skipped while
+	 * still in the current context (slots are also NOP-ed).
+	 *
+	 * At the point contexts are interleaved, syncpt increments must be
+	 * done inline with the pushbuffer from a GATHER buffer to maintain
+	 * the order (slots are modified to be a GATHER of syncpt incrs).
+	 *
+	 * Note: save in get_restart the location where the timed out buffer
+	 * started in the PB, so we can start the refetch from there (with the
+	 * modified NOP-ed PB slots). This lets things appear to have completed
+	 * properly for this buffer and resources are freed.
+	 */
+
+	dev_dbg(&dev->dev,
+		"%s: perform CPU incr on pending same ctx buffers\n",
+		__func__);
+
+	get_restart = cdma->last_put;
+	if (!list_empty(&cdma->sync_queue))
+		get_restart = job->first_get;
+
+	/* do CPU increments as long as this context continues */
+	list_for_each_entry_from(job, &cdma->sync_queue, list) {
+		/* different context, gets us out of this loop */
+		if (job->clientid != cdma->timeout.clientid)
+			break;
+
+		/* won't need a timeout when replayed */
+		job->timeout = 0;
+
+		syncpt_incrs = job->syncpt_end - syncpt_val;
+		dev_dbg(&dev->dev,
+			"%s: CPU incr (%d)\n", __func__, syncpt_incrs);
+
+		nvhost_job_dump(&dev->dev, job);
+
+		/* safe to use CPU to incr syncpts */
+		cdma_op().timeout_cpu_incr(cdma,
+				job->first_get,
+				syncpt_incrs,
+				job->syncpt_end,
+				job->num_slots);
+
+		syncpt_val += syncpt_incrs;
+	}
+
+	list_for_each_entry_from(job, &cdma->sync_queue, list)
+		if (job->clientid == cdma->timeout.clientid)
+			job->timeout = 500;
+
+	dev_dbg(&dev->dev,
+		"%s: finished sync_queue modification\n", __func__);
+
+	/* roll back DMAGET and start up channel again */
+	cdma_op().timeout_teardown_end(cdma, get_restart);
+}
+
+/**
+ * Create a cdma
+ */
+int nvhost_cdma_init(struct nvhost_cdma *cdma)
+{
+	int err;
+	struct push_buffer *pb = &cdma->push_buffer;
+	mutex_init(&cdma->lock);
+	sema_init(&cdma->sem, 0);
+
+	INIT_LIST_HEAD(&cdma->sync_queue);
+
+	cdma->event = CDMA_EVENT_NONE;
+	cdma->running = false;
+	cdma->torndown = false;
+
+	err = cdma_pb_op().init(pb);
+	if (err)
+		return err;
+	return 0;
+}
+
+/**
+ * Destroy a cdma
+ */
+void nvhost_cdma_deinit(struct nvhost_cdma *cdma)
+{
+	struct push_buffer *pb = &cdma->push_buffer;
+
+	if (cdma->running) {
+		pr_warn("%s: CDMA still running\n",
+				__func__);
+	} else {
+		cdma_pb_op().destroy(pb);
+		cdma_op().timeout_destroy(cdma);
+	}
+}
+
+/**
+ * Begin a cdma submit
+ */
+int nvhost_cdma_begin(struct nvhost_cdma *cdma, struct nvhost_job *job)
+{
+	mutex_lock(&cdma->lock);
+
+	if (job->timeout) {
+		/* init state on first submit with timeout value */
+		if (!cdma->timeout.initialized) {
+			int err;
+			err = cdma_op().timeout_init(cdma,
+				job->syncpt_id);
+			if (err) {
+				mutex_unlock(&cdma->lock);
+				return err;
+			}
+		}
+	}
+	if (!cdma->running)
+		cdma_op().start(cdma);
+
+	cdma->slots_free = 0;
+	cdma->slots_used = 0;
+	cdma->first_get = cdma_pb_op().putptr(&cdma->push_buffer);
+	return 0;
+}
+
+static void trace_write_gather(struct nvhost_cdma *cdma,
+		struct mem_handle *ref,
+		u32 offset, u32 words)
+{
+	void *mem = NULL;
+
+	if (nvhost_debug_trace_cmdbuf) {
+		mem = mem_op().mmap(ref);
+		if (IS_ERR_OR_NULL(mem))
+			mem = NULL;
+	};
+
+	if (mem) {
+		u32 i;
+		/*
+		 * Write in batches of 128 as there seems to be a limit
+		 * of how much you can output to ftrace at once.
+		 */
+		for (i = 0; i < words; i += TRACE_MAX_LENGTH) {
+			trace_nvhost_cdma_push_gather(
+				cdma_to_channel(cdma)->dev->name,
+				(u32)ref,
+				min(words - i, TRACE_MAX_LENGTH),
+				offset + i * sizeof(u32),
+				mem);
+		}
+		mem_op().munmap(ref, mem);
+	}
+}
+
+/**
+ * Push two words into a push buffer slot
+ * Blocks as necessary if the push buffer is full.
+ */
+void nvhost_cdma_push(struct nvhost_cdma *cdma, u32 op1, u32 op2)
+{
+	if (nvhost_debug_trace_cmdbuf)
+		trace_nvhost_cdma_push(cdma_to_channel(cdma)->dev->name,
+				op1, op2);
+
+	nvhost_cdma_push_gather(cdma, NULL, NULL, 0, op1, op2);
+}
+
+/**
+ * Push two words into a push buffer slot
+ * Blocks as necessary if the push buffer is full.
+ */
+void nvhost_cdma_push_gather(struct nvhost_cdma *cdma,
+		struct mem_mgr *client, struct mem_handle *handle,
+		u32 offset, u32 op1, u32 op2)
+{
+	u32 slots_free = cdma->slots_free;
+	struct push_buffer *pb = &cdma->push_buffer;
+
+	if (handle)
+		trace_write_gather(cdma, handle, offset, op1 & 0xffff);
+
+	if (slots_free == 0) {
+		cdma_op().kick(cdma);
+		slots_free = nvhost_cdma_wait_locked(cdma,
+				CDMA_EVENT_PUSH_BUFFER_SPACE);
+	}
+	cdma->slots_free = slots_free - 1;
+	cdma->slots_used++;
+	cdma_pb_op().push_to(pb, client, handle, op1, op2);
+}
+
+/**
+ * End a cdma submit
+ * Kick off DMA, add job to the sync queue, and a number of slots to be freed
+ * from the pushbuffer. The handles for a submit must all be pinned at the same
+ * time, but they can be unpinned in smaller chunks.
+ */
+void nvhost_cdma_end(struct nvhost_cdma *cdma,
+		struct nvhost_job *job)
+{
+	bool was_idle = list_empty(&cdma->sync_queue);
+
+	cdma_op().kick(cdma);
+
+	add_to_sync_queue(cdma,
+			job,
+			cdma->slots_used,
+			cdma->first_get);
+
+	/* start timer on idle -> active transitions */
+	if (job->timeout && was_idle)
+		cdma_start_timer_locked(cdma, job);
+
+	trace_nvhost_cdma_end(job->ch->dev->name);
+
+	mutex_unlock(&cdma->lock);
+}
+
+/**
+ * Update cdma state according to current sync point values
+ */
+void nvhost_cdma_update(struct nvhost_cdma *cdma)
+{
+	mutex_lock(&cdma->lock);
+	update_cdma_locked(cdma);
+	mutex_unlock(&cdma->lock);
+}
diff --git a/drivers/video/tegra/host/nvhost_cdma.h b/drivers/video/tegra/host/nvhost_cdma.h
new file mode 100644
index 0000000..2179f29
--- /dev/null
+++ b/drivers/video/tegra/host/nvhost_cdma.h
@@ -0,0 +1,116 @@ 
+/*
+ * drivers/video/tegra/host/nvhost_cdma.h
+ *
+ * Tegra Graphics Host Command DMA
+ *
+ * Copyright (c) 2010-2012, NVIDIA Corporation.
+ *
+ * This program is free software; you can redistribute it and/or modify it
+ * under the terms and conditions of the GNU General Public License,
+ * version 2, as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
+ * more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program.  If not, see <http://www.gnu.org/licenses/>.
+ */
+
+#ifndef __NVHOST_CDMA_H
+#define __NVHOST_CDMA_H
+
+#include <linux/sched.h>
+#include <linux/semaphore.h>
+
+#include <linux/nvhost.h>
+#include <linux/list.h>
+
+struct nvhost_syncpt;
+struct nvhost_userctx_timeout;
+struct nvhost_job;
+struct mem_mgr;
+struct mem_handle;
+
+/*
+ * cdma
+ *
+ * This is in charge of a host command DMA channel.
+ * Sends ops to a push buffer, and takes responsibility for unpinning
+ * (& possibly freeing) of memory after those ops have completed.
+ * Producer:
+ *	begin
+ *		push - send ops to the push buffer
+ *	end - start command DMA and enqueue handles to be unpinned
+ * Consumer:
+ *	update - call to update sync queue and push buffer, unpin memory
+ */
+
+struct mem_mgr_handle {
+	struct mem_mgr *client;
+	struct mem_handle *handle;
+};
+
+struct push_buffer {
+	u32 *mapped;			/* mapped pushbuffer memory */
+	dma_addr_t phys;		/* physical address of pushbuffer */
+	u32 fence;			/* index we've written */
+	u32 cur;			/* index to write to */
+	struct mem_mgr_handle *client_handle; /* handle for each opcode pair */
+};
+
+struct buffer_timeout {
+	struct delayed_work wq;		/* work queue */
+	bool initialized;		/* timer one-time setup flag */
+	u32 syncpt_id;			/* buffer completion syncpt id */
+	u32 syncpt_val;			/* syncpt value when completed */
+	ktime_t start_ktime;		/* starting time */
+	/* context timeout information */
+	int clientid;
+};
+
+enum cdma_event {
+	CDMA_EVENT_NONE,		/* not waiting for any event */
+	CDMA_EVENT_SYNC_QUEUE_EMPTY,	/* wait for empty sync queue */
+	CDMA_EVENT_PUSH_BUFFER_SPACE	/* wait for space in push buffer */
+};
+
+struct nvhost_cdma {
+	struct mutex lock;		/* controls access to shared state */
+	struct semaphore sem;		/* signalled when event occurs */
+	enum cdma_event event;		/* event that sem is waiting for */
+	unsigned int slots_used;	/* pb slots used in current submit */
+	unsigned int slots_free;	/* pb slots free in current submit */
+	unsigned int first_get;		/* DMAGET value, where submit begins */
+	unsigned int last_put;		/* last value written to DMAPUT */
+	struct push_buffer push_buffer;	/* channel's push buffer */
+	struct list_head sync_queue;	/* job queue */
+	struct buffer_timeout timeout;	/* channel's timeout state/wq */
+	bool running;
+	bool torndown;
+};
+
+#define cdma_to_channel(cdma) container_of(cdma, struct nvhost_channel, cdma)
+#define cdma_to_dev(cdma) nvhost_get_host(cdma_to_channel(cdma)->dev)
+#define cdma_to_memmgr(cdma) ((cdma_to_dev(cdma))->memmgr)
+#define pb_to_cdma(pb) container_of(pb, struct nvhost_cdma, push_buffer)
+
+int	nvhost_cdma_init(struct nvhost_cdma *cdma);
+void	nvhost_cdma_deinit(struct nvhost_cdma *cdma);
+void	nvhost_cdma_stop(struct nvhost_cdma *cdma);
+int	nvhost_cdma_begin(struct nvhost_cdma *cdma, struct nvhost_job *job);
+void	nvhost_cdma_push(struct nvhost_cdma *cdma, u32 op1, u32 op2);
+void	nvhost_cdma_push_gather(struct nvhost_cdma *cdma,
+		struct mem_mgr *client,
+		struct mem_handle *handle, u32 offset, u32 op1, u32 op2);
+void	nvhost_cdma_end(struct nvhost_cdma *cdma,
+		struct nvhost_job *job);
+void	nvhost_cdma_update(struct nvhost_cdma *cdma);
+void	nvhost_cdma_peek(struct nvhost_cdma *cdma,
+		u32 dmaget, int slot, u32 *out);
+unsigned int nvhost_cdma_wait_locked(struct nvhost_cdma *cdma,
+		enum cdma_event event);
+void nvhost_cdma_update_sync_queue(struct nvhost_cdma *cdma,
+		struct nvhost_syncpt *syncpt, struct platform_device *dev);
+#endif
diff --git a/drivers/video/tegra/host/nvhost_channel.c b/drivers/video/tegra/host/nvhost_channel.c
new file mode 100644
index 0000000..f95d8f2
--- /dev/null
+++ b/drivers/video/tegra/host/nvhost_channel.c
@@ -0,0 +1,129 @@ 
+/*
+ * drivers/video/tegra/host/nvhost_channel.c
+ *
+ * Tegra Graphics Host Channel
+ *
+ * Copyright (c) 2010-2012, NVIDIA Corporation.
+ *
+ * This program is free software; you can redistribute it and/or modify it
+ * under the terms and conditions of the GNU General Public License,
+ * version 2, as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
+ * more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program.  If not, see <http://www.gnu.org/licenses/>.
+ */
+
+#include "nvhost_channel.h"
+#include "dev.h"
+#include "nvhost_acm.h"
+#include "chip_support.h"
+
+#include <trace/events/nvhost.h>
+#include <linux/slab.h>
+#include <linux/module.h>
+
+#define NVHOST_CHANNEL_LOW_PRIO_MAX_WAIT 50
+
+int nvhost_channel_init(struct nvhost_channel *ch,
+		struct nvhost_master *dev, int index)
+{
+	int err;
+	struct nvhost_device_data *pdata = platform_get_drvdata(ch->dev);
+
+	/* Link platform_device to nvhost_channel */
+	err = channel_op().init(ch, dev, index);
+	if (err < 0) {
+		dev_err(&dev->dev->dev, "failed to init channel %d\n",
+				index);
+		return err;
+	}
+	pdata->channel = ch;
+
+	return 0;
+}
+
+int nvhost_channel_submit(struct nvhost_job *job)
+{
+	return channel_op().submit(job);
+}
+EXPORT_SYMBOL(nvhost_channel_submit);
+
+struct nvhost_channel *nvhost_getchannel(struct nvhost_channel *ch)
+{
+	int err = 0;
+	struct nvhost_device_data *pdata = platform_get_drvdata(ch->dev);
+
+	mutex_lock(&ch->reflock);
+	if (ch->refcount == 0) {
+		if (pdata->init)
+			pdata->init(ch->dev);
+		err = nvhost_cdma_init(&ch->cdma);
+	}
+	if (!err)
+		ch->refcount++;
+
+	mutex_unlock(&ch->reflock);
+
+	return err ? NULL : ch;
+}
+EXPORT_SYMBOL(nvhost_getchannel);
+
+void nvhost_putchannel(struct nvhost_channel *ch)
+{
+	mutex_lock(&ch->reflock);
+	if (ch->refcount == 1) {
+		channel_cdma_op().stop(&ch->cdma);
+		nvhost_cdma_deinit(&ch->cdma);
+		nvhost_module_suspend(ch->dev);
+	}
+	ch->refcount--;
+	mutex_unlock(&ch->reflock);
+}
+EXPORT_SYMBOL(nvhost_putchannel);
+
+int nvhost_channel_suspend(struct nvhost_channel *ch)
+{
+	int ret = 0;
+
+	mutex_lock(&ch->reflock);
+
+	if (ch->refcount) {
+		ret = nvhost_module_suspend(ch->dev);
+		if (!ret)
+			channel_cdma_op().stop(&ch->cdma);
+	}
+	mutex_unlock(&ch->reflock);
+
+	return ret;
+}
+
+struct nvhost_channel *nvhost_alloc_channel_internal(int chindex,
+	int max_channels, int *current_channel_count)
+{
+	struct nvhost_channel *ch = NULL;
+
+	if (chindex > max_channels ||
+	     (*current_channel_count + 1) > max_channels)
+		return NULL;
+	else {
+		ch = kzalloc(sizeof(*ch), GFP_KERNEL);
+		if (ch == NULL)
+			return NULL;
+		else {
+			(*current_channel_count)++;
+			return ch;
+		}
+	}
+}
+
+void nvhost_free_channel_internal(struct nvhost_channel *ch,
+	int *current_channel_count)
+{
+	kfree(ch);
+	(*current_channel_count)--;
+}
diff --git a/drivers/video/tegra/host/nvhost_channel.h b/drivers/video/tegra/host/nvhost_channel.h
new file mode 100644
index 0000000..65854b4
--- /dev/null
+++ b/drivers/video/tegra/host/nvhost_channel.h
@@ -0,0 +1,65 @@ 
+/*
+ * drivers/video/tegra/host/nvhost_channel.h
+ *
+ * Tegra Graphics Host Channel
+ *
+ * Copyright (c) 2010-2012, NVIDIA Corporation.
+ *
+ * This program is free software; you can redistribute it and/or modify it
+ * under the terms and conditions of the GNU General Public License,
+ * version 2, as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
+ * more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program.  If not, see <http://www.gnu.org/licenses/>.
+ */
+
+#ifndef __NVHOST_CHANNEL_H
+#define __NVHOST_CHANNEL_H
+
+#include <linux/cdev.h>
+#include <linux/io.h>
+#include "nvhost_cdma.h"
+
+#define NVHOST_MAX_WAIT_CHECKS		256
+#define NVHOST_MAX_GATHERS		512
+#define NVHOST_MAX_HANDLES		1280
+#define NVHOST_MAX_POWERGATE_IDS	2
+
+struct nvhost_master;
+struct platform_device;
+struct nvhost_channel;
+
+struct nvhost_channel {
+	int refcount;
+	int chid;
+	u32 syncpt_id;
+	struct mutex reflock;
+	struct mutex submitlock;
+	void __iomem *aperture;
+	struct device *node;
+	struct platform_device *dev;
+	struct cdev cdev;
+	struct nvhost_cdma cdma;
+};
+
+int nvhost_channel_init(struct nvhost_channel *ch,
+	struct nvhost_master *dev, int index);
+
+struct nvhost_channel *nvhost_getchannel(struct nvhost_channel *ch);
+void nvhost_putchannel(struct nvhost_channel *ch);
+int nvhost_channel_suspend(struct nvhost_channel *ch);
+
+struct nvhost_channel *nvhost_alloc_channel_internal(int chindex,
+	int max_channels, int *current_channel_count);
+
+void nvhost_free_channel_internal(struct nvhost_channel *ch,
+	int *current_channel_count);
+
+int nvhost_channel_save_context(struct nvhost_channel *ch);
+
+#endif
diff --git a/drivers/video/tegra/host/nvhost_intr.c b/drivers/video/tegra/host/nvhost_intr.c
new file mode 100644
index 0000000..9a0fa3e
--- /dev/null
+++ b/drivers/video/tegra/host/nvhost_intr.c
@@ -0,0 +1,391 @@ 
+/*
+ * drivers/video/tegra/host/nvhost_intr.c
+ *
+ * Tegra Graphics Host Interrupt Management
+ *
+ * Copyright (c) 2010-2012, NVIDIA Corporation.
+ *
+ * This program is free software; you can redistribute it and/or modify it
+ * under the terms and conditions of the GNU General Public License,
+ * version 2, as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
+ * more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program.  If not, see <http://www.gnu.org/licenses/>.
+ */
+
+#include "nvhost_intr.h"
+#include "dev.h"
+#include "nvhost_acm.h"
+#include <linux/interrupt.h>
+#include <linux/slab.h>
+#include <linux/irq.h>
+#include <trace/events/nvhost.h>
+#include "nvhost_channel.h"
+#include "chip_support.h"
+
+/*** Wait list management ***/
+
+struct nvhost_waitlist {
+	struct list_head list;
+	struct kref refcount;
+	u32 thresh;
+	enum nvhost_intr_action action;
+	atomic_t state;
+	void *data;
+	int count;
+};
+
+enum waitlist_state {
+	WLS_PENDING,
+	WLS_REMOVED,
+	WLS_CANCELLED,
+	WLS_HANDLED
+};
+
+static void waiter_release(struct kref *kref)
+{
+	kfree(container_of(kref, struct nvhost_waitlist, refcount));
+}
+
+/**
+ * add a waiter to a waiter queue, sorted by threshold
+ * returns true if it was added at the head of the queue
+ */
+static bool add_waiter_to_queue(struct nvhost_waitlist *waiter,
+				struct list_head *queue)
+{
+	struct nvhost_waitlist *pos;
+	u32 thresh = waiter->thresh;
+
+	list_for_each_entry_reverse(pos, queue, list)
+		if ((s32)(pos->thresh - thresh) <= 0) {
+			list_add(&waiter->list, &pos->list);
+			return false;
+		}
+
+	list_add(&waiter->list, queue);
+	return true;
+}
+
+/**
+ * run through a waiter queue for a single sync point ID
+ * and gather all completed waiters into lists by actions
+ */
+static void remove_completed_waiters(struct list_head *head, u32 sync,
+			struct list_head completed[NVHOST_INTR_ACTION_COUNT])
+{
+	struct list_head *dest;
+	struct nvhost_waitlist *waiter, *next, *prev;
+
+	list_for_each_entry_safe(waiter, next, head, list) {
+		if ((s32)(waiter->thresh - sync) > 0)
+			break;
+
+		dest = completed + waiter->action;
+
+		/* consolidate submit cleanups */
+		if (waiter->action == NVHOST_INTR_ACTION_SUBMIT_COMPLETE
+			&& !list_empty(dest)) {
+			prev = list_entry(dest->prev,
+					struct nvhost_waitlist, list);
+			if (prev->data == waiter->data) {
+				prev->count++;
+				dest = NULL;
+			}
+		}
+
+		/* PENDING->REMOVED or CANCELLED->HANDLED */
+		if (atomic_inc_return(&waiter->state) == WLS_HANDLED || !dest) {
+			list_del(&waiter->list);
+			kref_put(&waiter->refcount, waiter_release);
+		} else {
+			list_move_tail(&waiter->list, dest);
+		}
+	}
+}
+
+void reset_threshold_interrupt(struct nvhost_intr *intr,
+			       struct list_head *head,
+			       unsigned int id)
+{
+	u32 thresh = list_first_entry(head,
+				struct nvhost_waitlist, list)->thresh;
+
+	intr_op().set_syncpt_threshold(intr, id, thresh);
+	intr_op().enable_syncpt_intr(intr, id);
+}
+
+
+static void action_submit_complete(struct nvhost_waitlist *waiter)
+{
+	struct nvhost_channel *channel = waiter->data;
+	int nr_completed = waiter->count;
+
+	nvhost_cdma_update(&channel->cdma);
+	nvhost_module_idle_mult(channel->dev, nr_completed);
+
+	/*  Add nr_completed to trace */
+	trace_nvhost_channel_submit_complete(channel->dev->name,
+			nr_completed, waiter->thresh);
+
+}
+
+static void action_wakeup(struct nvhost_waitlist *waiter)
+{
+	wait_queue_head_t *wq = waiter->data;
+
+	wake_up(wq);
+}
+
+static void action_wakeup_interruptible(struct nvhost_waitlist *waiter)
+{
+	wait_queue_head_t *wq = waiter->data;
+
+	wake_up_interruptible(wq);
+}
+
+typedef void (*action_handler)(struct nvhost_waitlist *waiter);
+
+static action_handler action_handlers[NVHOST_INTR_ACTION_COUNT] = {
+	action_submit_complete,
+	action_wakeup,
+	action_wakeup_interruptible,
+};
+
+static void run_handlers(struct list_head completed[NVHOST_INTR_ACTION_COUNT])
+{
+	struct list_head *head = completed;
+	int i;
+
+	for (i = 0; i < NVHOST_INTR_ACTION_COUNT; ++i, ++head) {
+		action_handler handler = action_handlers[i];
+		struct nvhost_waitlist *waiter, *next;
+
+		list_for_each_entry_safe(waiter, next, head, list) {
+			list_del(&waiter->list);
+			handler(waiter);
+			WARN_ON(atomic_xchg(&waiter->state, WLS_HANDLED)
+					!= WLS_REMOVED);
+			kref_put(&waiter->refcount, waiter_release);
+		}
+	}
+}
+
+/**
+ * Remove & handle all waiters that have completed for the given syncpt
+ */
+static int process_wait_list(struct nvhost_intr *intr,
+			     struct nvhost_intr_syncpt *syncpt,
+			     u32 threshold)
+{
+	struct list_head completed[NVHOST_INTR_ACTION_COUNT];
+	unsigned int i;
+	int empty;
+
+	for (i = 0; i < NVHOST_INTR_ACTION_COUNT; ++i)
+		INIT_LIST_HEAD(completed + i);
+
+	spin_lock(&syncpt->lock);
+
+	remove_completed_waiters(&syncpt->wait_head, threshold, completed);
+
+	empty = list_empty(&syncpt->wait_head);
+	if (empty)
+		intr_op().disable_syncpt_intr(intr, syncpt->id);
+	else
+		reset_threshold_interrupt(intr, &syncpt->wait_head,
+					  syncpt->id);
+
+	spin_unlock(&syncpt->lock);
+
+	run_handlers(completed);
+
+	return empty;
+}
+
+/*** host syncpt interrupt service functions ***/
+/**
+ * Sync point threshold interrupt service thread function
+ * Handles sync point threshold triggers, in thread context
+ */
+irqreturn_t nvhost_syncpt_thresh_fn(int irq, void *dev_id)
+{
+	struct nvhost_intr_syncpt *syncpt = dev_id;
+	unsigned int id = syncpt->id;
+	struct nvhost_intr *intr = intr_syncpt_to_intr(syncpt);
+	struct nvhost_master *dev = intr_to_dev(intr);
+
+	(void)process_wait_list(intr, syncpt,
+				nvhost_syncpt_update_min(&dev->syncpt, id));
+
+	return IRQ_HANDLED;
+}
+
+/*** host general interrupt service functions ***/
+
+
+/*** Main API ***/
+
+int nvhost_intr_add_action(struct nvhost_intr *intr, u32 id, u32 thresh,
+			enum nvhost_intr_action action, void *data,
+			void *_waiter,
+			void **ref)
+{
+	struct nvhost_waitlist *waiter = _waiter;
+	struct nvhost_intr_syncpt *syncpt;
+	int queue_was_empty;
+
+	if (waiter == NULL) {
+		pr_warn("%s: NULL waiter\n", __func__);
+		return -EINVAL;
+	}
+
+	/* initialize a new waiter */
+	INIT_LIST_HEAD(&waiter->list);
+	kref_init(&waiter->refcount);
+	if (ref)
+		kref_get(&waiter->refcount);
+	waiter->thresh = thresh;
+	waiter->action = action;
+	atomic_set(&waiter->state, WLS_PENDING);
+	waiter->data = data;
+	waiter->count = 1;
+
+	syncpt = intr->syncpt + id;
+
+	spin_lock(&syncpt->lock);
+
+	queue_was_empty = list_empty(&syncpt->wait_head);
+
+	if (add_waiter_to_queue(waiter, &syncpt->wait_head)) {
+		/* added at head of list - new threshold value */
+		intr_op().set_syncpt_threshold(intr, id, thresh);
+
+		/* added as first waiter - enable interrupt */
+		if (queue_was_empty)
+			intr_op().enable_syncpt_intr(intr, id);
+	}
+
+	spin_unlock(&syncpt->lock);
+
+	if (ref)
+		*ref = waiter;
+	return 0;
+}
+
+void *nvhost_intr_alloc_waiter()
+{
+	return kzalloc(sizeof(struct nvhost_waitlist),
+			GFP_KERNEL|__GFP_REPEAT);
+}
+
+void nvhost_intr_put_ref(struct nvhost_intr *intr, u32 id, void *ref)
+{
+	struct nvhost_waitlist *waiter = ref;
+	struct nvhost_intr_syncpt *syncpt;
+	struct nvhost_master *host = intr_to_dev(intr);
+
+	while (atomic_cmpxchg(&waiter->state,
+				WLS_PENDING, WLS_CANCELLED) == WLS_REMOVED)
+		schedule();
+
+	syncpt = intr->syncpt + id;
+	(void)process_wait_list(intr, syncpt,
+				nvhost_syncpt_update_min(&host->syncpt, id));
+
+	kref_put(&waiter->refcount, waiter_release);
+}
+
+
+/*** Init & shutdown ***/
+
+int nvhost_intr_init(struct nvhost_intr *intr, u32 irq_gen, u32 irq_sync)
+{
+	unsigned int id;
+	struct nvhost_intr_syncpt *syncpt;
+	struct nvhost_master *host = intr_to_dev(intr);
+	u32 nb_pts = nvhost_syncpt_nb_pts(&host->syncpt);
+
+	mutex_init(&intr->mutex);
+	intr->host_syncpt_irq_base = irq_sync;
+	intr->wq = create_workqueue("host_syncpt");
+	intr_op().init_host_sync(intr);
+	intr->host_general_irq = irq_gen;
+	intr_op().request_host_general_irq(intr);
+
+	for (id = 0, syncpt = intr->syncpt;
+	     id < nb_pts;
+	     ++id, ++syncpt) {
+		syncpt->intr = &host->intr;
+		syncpt->id = id;
+		syncpt->irq = irq_sync + id;
+		spin_lock_init(&syncpt->lock);
+		INIT_LIST_HEAD(&syncpt->wait_head);
+		snprintf(syncpt->thresh_irq_name,
+			sizeof(syncpt->thresh_irq_name),
+			"host_sp_%02d", id);
+	}
+
+	return 0;
+}
+
+void nvhost_intr_deinit(struct nvhost_intr *intr)
+{
+	nvhost_intr_stop(intr);
+	destroy_workqueue(intr->wq);
+}
+
+void nvhost_intr_start(struct nvhost_intr *intr, u32 hz)
+{
+	mutex_lock(&intr->mutex);
+
+	intr_op().init_host_sync(intr);
+	intr_op().set_host_clocks_per_usec(intr,
+					       (hz + 1000000 - 1)/1000000);
+
+	intr_op().request_host_general_irq(intr);
+
+	mutex_unlock(&intr->mutex);
+}
+
+void nvhost_intr_stop(struct nvhost_intr *intr)
+{
+	unsigned int id;
+	struct nvhost_intr_syncpt *syncpt;
+	u32 nb_pts = nvhost_syncpt_nb_pts(&intr_to_dev(intr)->syncpt);
+
+	mutex_lock(&intr->mutex);
+
+	intr_op().disable_all_syncpt_intrs(intr);
+
+	for (id = 0, syncpt = intr->syncpt;
+	     id < nb_pts;
+	     ++id, ++syncpt) {
+		struct nvhost_waitlist *waiter, *next;
+		list_for_each_entry_safe(waiter, next,
+				&syncpt->wait_head, list) {
+			if (atomic_cmpxchg(&waiter->state,
+						WLS_CANCELLED, WLS_HANDLED)
+				== WLS_CANCELLED) {
+				list_del(&waiter->list);
+				kref_put(&waiter->refcount, waiter_release);
+			}
+		}
+
+		if (!list_empty(&syncpt->wait_head)) {  /* output diagnostics */
+			pr_warn("%s cannot stop syncpt intr id=%d\n",
+					__func__, id);
+			return;
+		}
+	}
+
+	intr_op().free_host_general_irq(intr);
+	intr_op().free_syncpt_irq(intr);
+
+	mutex_unlock(&intr->mutex);
+}
diff --git a/drivers/video/tegra/host/nvhost_intr.h b/drivers/video/tegra/host/nvhost_intr.h
new file mode 100644
index 0000000..7554864
--- /dev/null
+++ b/drivers/video/tegra/host/nvhost_intr.h
@@ -0,0 +1,110 @@ 
+/*
+ * drivers/video/tegra/host/nvhost_intr.h
+ *
+ * Tegra Graphics Host Interrupt Management
+ *
+ * Copyright (c) 2010-2012, NVIDIA Corporation.
+ *
+ * This program is free software; you can redistribute it and/or modify it
+ * under the terms and conditions of the GNU General Public License,
+ * version 2, as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
+ * more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program.  If not, see <http://www.gnu.org/licenses/>.
+ */
+
+#ifndef __NVHOST_INTR_H
+#define __NVHOST_INTR_H
+
+#include <linux/kthread.h>
+#include <linux/semaphore.h>
+#include <linux/interrupt.h>
+#include <linux/workqueue.h>
+
+struct nvhost_channel;
+
+enum nvhost_intr_action {
+	/**
+	 * Perform cleanup after a submit has completed.
+	 * 'data' points to a channel
+	 */
+	NVHOST_INTR_ACTION_SUBMIT_COMPLETE = 0,
+
+	/**
+	 * Wake up a  task.
+	 * 'data' points to a wait_queue_head_t
+	 */
+	NVHOST_INTR_ACTION_WAKEUP,
+
+	/**
+	 * Wake up a interruptible task.
+	 * 'data' points to a wait_queue_head_t
+	 */
+	NVHOST_INTR_ACTION_WAKEUP_INTERRUPTIBLE,
+
+	NVHOST_INTR_ACTION_COUNT
+};
+
+struct nvhost_intr;
+
+struct nvhost_intr_syncpt {
+	struct nvhost_intr *intr;
+	u8 id;
+	u16 irq;
+	spinlock_t lock;
+	struct list_head wait_head;
+	char thresh_irq_name[12];
+	struct work_struct work;
+};
+
+struct nvhost_intr {
+	struct nvhost_intr_syncpt *syncpt;
+	struct mutex mutex;
+	int host_general_irq;
+	int host_syncpt_irq_base;
+	struct workqueue_struct *wq;
+};
+#define intr_to_dev(x) container_of(x, struct nvhost_master, intr)
+#define intr_syncpt_to_intr(is) (is->intr)
+
+/**
+ * Schedule an action to be taken when a sync point reaches the given threshold.
+ *
+ * @id the sync point
+ * @thresh the threshold
+ * @action the action to take
+ * @data a pointer to extra data depending on action, see above
+ * @waiter waiter allocated with nvhost_intr_alloc_waiter - assumes ownership
+ * @ref must be passed if cancellation is possible, else NULL
+ *
+ * This is a non-blocking api.
+ */
+int nvhost_intr_add_action(struct nvhost_intr *intr, u32 id, u32 thresh,
+			enum nvhost_intr_action action, void *data,
+			void *waiter,
+			void **ref);
+
+/**
+ * Allocate a waiter.
+ */
+void *nvhost_intr_alloc_waiter(void);
+
+/**
+ * Unreference an action submitted to nvhost_intr_add_action().
+ * You must call this if you passed non-NULL as ref.
+ * @ref the ref returned from nvhost_intr_add_action()
+ */
+void nvhost_intr_put_ref(struct nvhost_intr *intr, u32 id, void *ref);
+
+int nvhost_intr_init(struct nvhost_intr *intr, u32 irq_gen, u32 irq_sync);
+void nvhost_intr_deinit(struct nvhost_intr *intr);
+void nvhost_intr_start(struct nvhost_intr *intr, u32 hz);
+void nvhost_intr_stop(struct nvhost_intr *intr);
+
+irqreturn_t nvhost_syncpt_thresh_fn(int irq, void *dev_id);
+#endif
diff --git a/drivers/video/tegra/host/nvhost_job.c b/drivers/video/tegra/host/nvhost_job.c
new file mode 100644
index 0000000..ebb071d
--- /dev/null
+++ b/drivers/video/tegra/host/nvhost_job.c
@@ -0,0 +1,398 @@ 
+/*
+ * drivers/video/tegra/host/nvhost_job.c
+ *
+ * Tegra Graphics Host Job
+ *
+ * Copyright (c) 2010-2012, NVIDIA Corporation.
+ *
+ * This program is free software; you can redistribute it and/or modify it
+ * under the terms and conditions of the GNU General Public License,
+ * version 2, as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
+ * more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program.  If not, see <http://www.gnu.org/licenses/>.
+ */
+
+#include <linux/module.h>
+#include <linux/slab.h>
+#include <linux/kref.h>
+#include <linux/err.h>
+#include <linux/vmalloc.h>
+#include <linux/scatterlist.h>
+#include <linux/nvhost.h>
+#include <trace/events/nvhost.h>
+#include "nvhost_channel.h"
+#include "nvhost_syncpt.h"
+#include "dev.h"
+#include "nvhost_memmgr.h"
+#include "chip_support.h"
+
+/* Magic to use to fill freed handle slots */
+#define BAD_MAGIC 0xdeadbeef
+
+static size_t job_size(u32 num_cmdbufs, u32 num_relocs, u32 num_waitchks)
+{
+	u32 num_unpins = num_cmdbufs + num_relocs;
+	s64 total;
+
+	if (num_relocs < 0 || num_waitchks < 0 || num_cmdbufs < 0)
+		return 0;
+
+	total = sizeof(struct nvhost_job)
+			+ num_relocs * sizeof(struct nvhost_reloc)
+			+ num_unpins * sizeof(struct nvhost_job_unpin_data)
+			+ num_waitchks * sizeof(struct nvhost_waitchk)
+			+ num_cmdbufs * sizeof(struct nvhost_job_gather);
+
+	if (total > ULONG_MAX)
+		return 0;
+	return (size_t)total;
+}
+
+
+static void init_fields(struct nvhost_job *job,
+		u32 num_cmdbufs, u32 num_relocs, u32 num_waitchks)
+{
+	u32 num_unpins = num_cmdbufs + num_relocs;
+	void *mem = job;
+
+	/* First init state to zero */
+
+	/*
+	 * Redistribute memory to the structs.
+	 * Overflows and negative conditions have
+	 * already been checked in job_alloc().
+	 */
+	mem += sizeof(struct nvhost_job);
+	job->relocarray = num_relocs ? mem : NULL;
+	mem += num_relocs * sizeof(struct nvhost_reloc);
+	job->unpins = num_unpins ? mem : NULL;
+	mem += num_unpins * sizeof(struct nvhost_job_unpin_data);
+	job->waitchk = num_waitchks ? mem : NULL;
+	mem += num_waitchks * sizeof(struct nvhost_waitchk);
+	job->gathers = num_cmdbufs ? mem : NULL;
+	mem += num_cmdbufs * sizeof(struct nvhost_job_gather);
+	job->addr_phys = (num_cmdbufs || num_relocs) ? mem : NULL;
+
+	job->reloc_addr_phys = job->addr_phys;
+	job->gather_addr_phys = &job->addr_phys[num_relocs];
+}
+
+struct nvhost_job *nvhost_job_alloc(struct nvhost_channel *ch,
+		int num_cmdbufs, int num_relocs, int num_waitchks,
+		struct mem_mgr *memmgr)
+{
+	struct nvhost_job *job = NULL;
+	size_t size = job_size(num_cmdbufs, num_relocs, num_waitchks);
+
+	if (!size)
+		return NULL;
+	job = vzalloc(size);
+	if (!job)
+		return NULL;
+
+	kref_init(&job->ref);
+	job->ch = ch;
+	job->memmgr = memmgr ? mem_op().get_mgr(memmgr) : NULL;
+
+	init_fields(job, num_cmdbufs, num_relocs, num_waitchks);
+
+	return job;
+}
+EXPORT_SYMBOL(nvhost_job_alloc);
+
+void nvhost_job_get(struct nvhost_job *job)
+{
+	kref_get(&job->ref);
+}
+
+static void job_free(struct kref *ref)
+{
+	struct nvhost_job *job = container_of(ref, struct nvhost_job, ref);
+
+	if (job->memmgr)
+		mem_op().put_mgr(job->memmgr);
+	vfree(job);
+}
+
+void nvhost_job_put(struct nvhost_job *job)
+{
+	kref_put(&job->ref, job_free);
+}
+EXPORT_SYMBOL(nvhost_job_put);
+
+void nvhost_job_add_gather(struct nvhost_job *job,
+		u32 mem_id, u32 words, u32 offset)
+{
+	struct nvhost_job_gather *cur_gather =
+			&job->gathers[job->num_gathers];
+
+	cur_gather->words = words;
+	cur_gather->mem_id = mem_id;
+	cur_gather->offset = offset;
+	job->num_gathers += 1;
+}
+EXPORT_SYMBOL(nvhost_job_add_gather);
+
+/*
+ * Check driver supplied waitchk structs for syncpt thresholds
+ * that have already been satisfied and NULL the comparison (to
+ * avoid a wrap condition in the HW).
+ */
+static int do_waitchks(struct nvhost_job *job, struct nvhost_syncpt *sp,
+		u32 patch_mem, struct mem_handle *h)
+{
+	int i;
+
+	/* compare syncpt vs wait threshold */
+	for (i = 0; i < job->num_waitchk; i++) {
+		struct nvhost_waitchk *wait = &job->waitchk[i];
+
+		/* validate syncpt id */
+		if (wait->syncpt_id > nvhost_syncpt_nb_pts(sp))
+			continue;
+
+		/* skip all other gathers */
+		if (patch_mem != wait->mem)
+			continue;
+
+		trace_nvhost_syncpt_wait_check(wait->mem, wait->offset,
+				wait->syncpt_id, wait->thresh,
+				nvhost_syncpt_read(sp, wait->syncpt_id));
+		if (nvhost_syncpt_is_expired(sp,
+					wait->syncpt_id, wait->thresh)) {
+			void *patch_addr = NULL;
+
+			/*
+			 * NULL an already satisfied WAIT_SYNCPT host method,
+			 * by patching its args in the command stream. The
+			 * method data is changed to reference a reserved
+			 * (never given out or incr) NVSYNCPT_GRAPHICS_HOST
+			 * syncpt with a matching threshold value of 0, so
+			 * is guaranteed to be popped by the host HW.
+			 */
+			dev_dbg(&syncpt_to_dev(sp)->dev->dev,
+			    "drop WAIT id %d (%s) thresh 0x%x, min 0x%x\n",
+			    wait->syncpt_id,
+			    syncpt_op().name(sp, wait->syncpt_id),
+			    wait->thresh,
+			    nvhost_syncpt_read_min(sp, wait->syncpt_id));
+
+			/* patch the wait */
+			patch_addr = mem_op().kmap(h,
+					wait->offset >> PAGE_SHIFT);
+			if (patch_addr) {
+				nvhost_syncpt_patch_wait(sp,
+					(patch_addr +
+					 (wait->offset & ~PAGE_MASK)));
+				mem_op().kunmap(h,
+						wait->offset >> PAGE_SHIFT,
+						patch_addr);
+			} else {
+				pr_err("Couldn't map cmdbuf for wait check\n");
+			}
+		}
+
+		wait->mem = 0;
+	}
+	return 0;
+}
+
+
+static int pin_job_mem(struct nvhost_job *job)
+{
+	int i;
+	int count = 0;
+	int result;
+	long unsigned *ids =
+			kmalloc(sizeof(u32 *) *
+				(job->num_relocs + job->num_gathers),
+				GFP_KERNEL);
+	if (!ids)
+		return -ENOMEM;
+
+	for (i = 0; i < job->num_relocs; i++) {
+		struct nvhost_reloc *reloc = &job->relocarray[i];
+		ids[count] = reloc->target;
+		count++;
+	}
+
+	for (i = 0; i < job->num_gathers; i++) {
+		struct nvhost_job_gather *g = &job->gathers[i];
+		ids[count] = g->mem_id;
+		count++;
+	}
+
+	/* validate array and pin unique ids, get refs for unpinning */
+	result = mem_op().pin_array_ids(job->memmgr, job->ch->dev,
+		ids, job->addr_phys,
+		count,
+		job->unpins);
+	kfree(ids);
+
+	if (result > 0)
+		job->num_unpins = result;
+
+	return result;
+}
+
+static int do_relocs(struct nvhost_job *job,
+		u32 cmdbuf_mem, struct mem_handle *h)
+{
+	int i = 0;
+	int last_page = -1;
+	void *cmdbuf_page_addr = NULL;
+
+	/* pin & patch the relocs for one gather */
+	while (i < job->num_relocs) {
+		struct nvhost_reloc *reloc = &job->relocarray[i];
+
+		/* skip all other gathers */
+		if (cmdbuf_mem != reloc->cmdbuf_mem) {
+			i++;
+			continue;
+		}
+
+		if (last_page != reloc->cmdbuf_offset >> PAGE_SHIFT) {
+			if (cmdbuf_page_addr)
+				mem_op().kunmap(h, last_page, cmdbuf_page_addr);
+
+			cmdbuf_page_addr = mem_op().kmap(h,
+					reloc->cmdbuf_offset >> PAGE_SHIFT);
+			last_page = reloc->cmdbuf_offset >> PAGE_SHIFT;
+
+			if (unlikely(!cmdbuf_page_addr)) {
+				pr_err("Couldn't map cmdbuf for relocation\n");
+				return -ENOMEM;
+			}
+		}
+
+		__raw_writel(
+			(job->reloc_addr_phys[i] +
+				reloc->target_offset) >> reloc->shift,
+			(cmdbuf_page_addr +
+				(reloc->cmdbuf_offset & ~PAGE_MASK)));
+
+		/* remove completed reloc from the job */
+		if (i != job->num_relocs - 1) {
+			struct nvhost_reloc *reloc_last =
+				&job->relocarray[job->num_relocs - 1];
+			reloc->cmdbuf_mem	= reloc_last->cmdbuf_mem;
+			reloc->cmdbuf_offset	= reloc_last->cmdbuf_offset;
+			reloc->target		= reloc_last->target;
+			reloc->target_offset	= reloc_last->target_offset;
+			reloc->shift		= reloc_last->shift;
+			job->reloc_addr_phys[i] =
+				job->reloc_addr_phys[job->num_relocs - 1];
+			job->num_relocs--;
+		} else {
+			break;
+		}
+	}
+
+	if (cmdbuf_page_addr)
+		mem_op().kunmap(h, last_page, cmdbuf_page_addr);
+
+	return 0;
+}
+
+
+int nvhost_job_pin(struct nvhost_job *job, struct platform_device *pdev)
+{
+	int err = 0, i = 0, j = 0;
+	struct nvhost_syncpt *sp = &nvhost_get_host(pdev)->syncpt;
+	unsigned long waitchk_mask[nvhost_syncpt_nb_pts(sp) / BITS_PER_LONG];
+
+	memset(&waitchk_mask[0], 0, sizeof(waitchk_mask));
+	for (i = 0; i < job->num_waitchk; i++) {
+		u32 syncpt_id = job->waitchk[i].syncpt_id;
+		if (syncpt_id < nvhost_syncpt_nb_pts(sp))
+			waitchk_mask[BIT_WORD(syncpt_id)] |=
+				BIT_MASK(syncpt_id);
+	}
+
+	/* get current syncpt values for waitchk */
+	for_each_set_bit(i, &waitchk_mask[0], sizeof(waitchk_mask))
+		nvhost_syncpt_update_min(sp, i);
+
+	/* pin memory */
+	err = pin_job_mem(job);
+	if (err <= 0)
+		goto fail;
+
+	/* patch gathers */
+	for (i = 0; i < job->num_gathers; i++) {
+		struct nvhost_job_gather *g = &job->gathers[i];
+
+		/* process each gather mem only once */
+		if (!g->ref) {
+			g->ref = mem_op().get(job->memmgr,
+				g->mem_id, job->ch->dev);
+			if (IS_ERR(g->ref)) {
+				err = PTR_ERR(g->ref);
+				g->ref = NULL;
+				break;
+			}
+
+			g->mem_base = job->gather_addr_phys[i];
+
+			for (j = 0; j < job->num_gathers; j++) {
+				struct nvhost_job_gather *tmp =
+					&job->gathers[j];
+				if (!tmp->ref && tmp->mem_id == g->mem_id) {
+					tmp->ref = g->ref;
+					tmp->mem_base = g->mem_base;
+				}
+			}
+			err = do_relocs(job, g->mem_id,  g->ref);
+			if (!err)
+				err = do_waitchks(job, sp,
+						g->mem_id, g->ref);
+			mem_op().put(job->memmgr, g->ref);
+			if (err)
+				break;
+		}
+	}
+fail:
+	wmb();
+
+	return err;
+}
+EXPORT_SYMBOL(nvhost_job_pin);
+
+void nvhost_job_unpin(struct nvhost_job *job)
+{
+	int i;
+
+	for (i = 0; i < job->num_unpins; i++) {
+		struct nvhost_job_unpin_data *unpin = &job->unpins[i];
+		mem_op().unpin(job->memmgr, unpin->h, unpin->mem);
+		mem_op().put(job->memmgr, unpin->h);
+	}
+	job->num_unpins = 0;
+}
+EXPORT_SYMBOL(nvhost_job_unpin);
+
+/**
+ * Debug routine used to dump job entries
+ */
+void nvhost_job_dump(struct device *dev, struct nvhost_job *job)
+{
+	dev_dbg(dev, "    SYNCPT_ID   %d\n",
+		job->syncpt_id);
+	dev_dbg(dev, "    SYNCPT_VAL  %d\n",
+		job->syncpt_end);
+	dev_dbg(dev, "    FIRST_GET   0x%x\n",
+		job->first_get);
+	dev_dbg(dev, "    TIMEOUT     %d\n",
+		job->timeout);
+	dev_dbg(dev, "    NUM_SLOTS   %d\n",
+		job->num_slots);
+	dev_dbg(dev, "    NUM_HANDLES %d\n",
+		job->num_unpins);
+}
diff --git a/drivers/video/tegra/host/nvhost_memmgr.c b/drivers/video/tegra/host/nvhost_memmgr.c
new file mode 100644
index 0000000..47e1944
--- /dev/null
+++ b/drivers/video/tegra/host/nvhost_memmgr.c
@@ -0,0 +1,252 @@ 
+/*
+ * drivers/video/tegra/host/nvhost_memmgr.c
+ *
+ * Tegra Graphics Host Memory Management Abstraction
+ *
+ * Copyright (c) 2012, NVIDIA Corporation.
+ *
+ * This program is free software; you can redistribute it and/or modify it
+ * under the terms and conditions of the GNU General Public License,
+ * version 2, as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
+ * more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program.  If not, see <http://www.gnu.org/licenses/>.
+ */
+
+#include <linux/kernel.h>
+#include <linux/err.h>
+
+#include "nvhost_memmgr.h"
+#ifdef CONFIG_TEGRA_GRHOST_USE_DMABUF
+#include "dmabuf.h"
+#endif
+#include "chip_support.h"
+
+struct mem_mgr *nvhost_memmgr_alloc_mgr(void)
+{
+	struct mem_mgr *mgr = NULL;
+#ifdef CONFIG_TEGRA_GRHOST_USE_DMABUF
+	mgr = (struct mem_mgr *)1;
+#endif
+
+	return mgr;
+}
+
+void nvhost_memmgr_put_mgr(struct mem_mgr *mgr)
+{
+#ifdef CONFIG_TEGRA_GRHOST_USE_DMABUF
+	mgr = (struct mem_mgr *)1;
+#endif
+}
+
+struct mem_mgr *nvhost_memmgr_get_mgr(struct mem_mgr *_mgr)
+{
+	struct mem_mgr *mgr = NULL;
+#ifdef CONFIG_TEGRA_GRHOST_USE_DMABUF
+	mgr = (struct mem_mgr *)1;
+#endif
+
+	return mgr;
+}
+
+struct mem_mgr *nvhost_memmgr_get_mgr_file(int fd)
+{
+	struct mem_mgr *mgr = NULL;
+#ifdef CONFIG_TEGRA_GRHOST_USE_DMABUF
+	mgr = (struct mem_mgr *)1;
+#endif
+
+	return mgr;
+}
+
+struct mem_handle *nvhost_memmgr_alloc(struct mem_mgr *mgr,
+		size_t size, size_t align, int flags)
+{
+	struct mem_handle *h = NULL;
+#ifdef CONFIG_TEGRA_GRHOST_USE_DMABUF
+	h = nvhost_dmabuf_alloc(mgr, size, align, flags);
+#endif
+
+	return h;
+}
+
+struct mem_handle *nvhost_memmgr_get(struct mem_mgr *mgr,
+		u32 id, struct platform_device *dev)
+{
+	struct mem_handle *h = NULL;
+
+	switch (nvhost_memmgr_type(id)) {
+#ifdef CONFIG_TEGRA_GRHOST_USE_DMABUF
+	case mem_mgr_type_dmabuf:
+		h = (struct mem_handle *) nvhost_dmabuf_get(id, dev);
+		break;
+#endif
+	default:
+		break;
+	}
+
+	return h;
+}
+
+void nvhost_memmgr_put(struct mem_mgr *mgr, struct mem_handle *handle)
+{
+	switch (nvhost_memmgr_type((u32)handle)) {
+#ifdef CONFIG_TEGRA_GRHOST_USE_DMABUF
+	case mem_mgr_type_dmabuf:
+		nvhost_dmabuf_put(handle);
+		break;
+#endif
+	default:
+		break;
+	}
+}
+
+struct sg_table *nvhost_memmgr_pin(struct mem_mgr *mgr,
+		struct mem_handle *handle)
+{
+	switch (nvhost_memmgr_type((u32)handle)) {
+#ifdef CONFIG_TEGRA_GRHOST_USE_DMABUF
+	case mem_mgr_type_dmabuf:
+		return nvhost_dmabuf_pin(handle);
+		break;
+#endif
+	default:
+		return 0;
+		break;
+	}
+}
+
+void nvhost_memmgr_unpin(struct mem_mgr *mgr,
+		struct mem_handle *handle, struct sg_table *sgt)
+{
+	switch (nvhost_memmgr_type((u32)handle)) {
+#ifdef CONFIG_TEGRA_GRHOST_USE_DMABUF
+	case mem_mgr_type_dmabuf:
+		nvhost_dmabuf_unpin(handle, sgt);
+		break;
+#endif
+	default:
+		break;
+	}
+}
+
+void *nvhost_memmgr_mmap(struct mem_handle *handle)
+{
+	switch (nvhost_memmgr_type((u32)handle)) {
+#ifdef CONFIG_TEGRA_GRHOST_USE_DMABUF
+	case mem_mgr_type_dmabuf:
+		return nvhost_dmabuf_mmap(handle);
+		break;
+#endif
+	default:
+		return 0;
+		break;
+	}
+}
+
+void nvhost_memmgr_munmap(struct mem_handle *handle, void *addr)
+{
+	switch (nvhost_memmgr_type((u32)handle)) {
+#ifdef CONFIG_TEGRA_GRHOST_USE_DMABUF
+	case mem_mgr_type_dmabuf:
+		nvhost_dmabuf_munmap(handle, addr);
+		break;
+#endif
+	default:
+		break;
+	}
+}
+
+void *nvhost_memmgr_kmap(struct mem_handle *handle, unsigned int pagenum)
+{
+	switch (nvhost_memmgr_type((u32)handle)) {
+#ifdef CONFIG_TEGRA_GRHOST_USE_DMABUF
+	case mem_mgr_type_dmabuf:
+		return nvhost_dmabuf_kmap(handle, pagenum);
+		break;
+#endif
+	default:
+		return 0;
+		break;
+	}
+}
+
+void nvhost_memmgr_kunmap(struct mem_handle *handle, unsigned int pagenum,
+		void *addr)
+{
+	switch (nvhost_memmgr_type((u32)handle)) {
+#ifdef CONFIG_TEGRA_GRHOST_USE_DMABUF
+	case mem_mgr_type_dmabuf:
+		nvhost_dmabuf_kunmap(handle, pagenum, addr);
+		break;
+#endif
+	default:
+		break;
+	}
+}
+
+int nvhost_memmgr_pin_array_ids(struct mem_mgr *mgr,
+		struct platform_device *dev,
+		long unsigned *ids,
+		dma_addr_t *phys_addr,
+		u32 count,
+		struct nvhost_job_unpin_data *unpin_data)
+{
+	int pin_count = 0;
+
+#ifdef CONFIG_TEGRA_GRHOST_USE_DMABUF
+	{
+		int dmabuf_count = 0;
+		dmabuf_count = nvhost_dmabuf_pin_array_ids(dev,
+			ids, MEMMGR_TYPE_MASK,
+			mem_mgr_type_dmabuf,
+			count, &unpin_data[pin_count],
+			phys_addr);
+
+		if (dmabuf_count < 0) {
+			/* clean up previous handles */
+			while (pin_count) {
+				pin_count--;
+				/* unpin, put */
+				nvhost_memmgr_unpin(mgr,
+						unpin_data[pin_count].h,
+						unpin_data[pin_count].mem);
+				nvhost_memmgr_put(mgr,
+						unpin_data[pin_count].h);
+			}
+			return dmabuf_count;
+		}
+		pin_count += dmabuf_count;
+	}
+#endif
+	return pin_count;
+}
+
+static const struct nvhost_mem_ops mem_ops = {
+	.alloc_mgr = nvhost_memmgr_alloc_mgr,
+	.put_mgr = nvhost_memmgr_put_mgr,
+	.get_mgr = nvhost_memmgr_get_mgr,
+	.get_mgr_file = nvhost_memmgr_get_mgr_file,
+	.alloc = nvhost_memmgr_alloc,
+	.get = nvhost_memmgr_get,
+	.put = nvhost_memmgr_put,
+	.pin = nvhost_memmgr_pin,
+	.unpin = nvhost_memmgr_unpin,
+	.mmap = nvhost_memmgr_mmap,
+	.munmap = nvhost_memmgr_munmap,
+	.kmap = nvhost_memmgr_kmap,
+	.kunmap = nvhost_memmgr_kunmap,
+	.pin_array_ids = nvhost_memmgr_pin_array_ids,
+};
+
+int nvhost_memmgr_init(struct nvhost_chip_support *chip)
+{
+	chip->mem = mem_ops;
+	return 0;
+}
+
diff --git a/drivers/video/tegra/host/nvhost_memmgr.h b/drivers/video/tegra/host/nvhost_memmgr.h
new file mode 100644
index 0000000..c983ee1
--- /dev/null
+++ b/drivers/video/tegra/host/nvhost_memmgr.h
@@ -0,0 +1,66 @@ 
+/*
+ * drivers/video/tegra/host/nvhost_memmgr.h
+ *
+ * Tegra Graphics Host Memory Management Abstraction header
+ *
+ * Copyright (c) 2012, NVIDIA Corporation.
+ *
+ * This program is free software; you can redistribute it and/or modify it
+ * under the terms and conditions of the GNU General Public License,
+ * version 2, as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
+ * more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program.  If not, see <http://www.gnu.org/licenses/>.
+ */
+
+#ifndef _NVHOST_MEM_MGR_H_
+#define _NVHOST_MEM_MGR_H_
+
+struct nvhost_chip_support;
+struct mem_mgr;
+struct mem_handle;
+struct platform_device;
+
+struct nvhost_job_unpin_data {
+	struct mem_handle *h;
+	struct sg_table *mem;
+};
+
+enum mem_mgr_flag {
+	mem_mgr_flag_uncacheable = 0,
+	mem_mgr_flag_write_combine = 1,
+};
+
+enum mem_mgr_type {
+	mem_mgr_type_dmabuf = 1,
+};
+
+#define MEMMGR_TYPE_MASK	0x3
+#define MEMMGR_ID_MASK		~0x3
+
+int nvhost_memmgr_init(struct nvhost_chip_support *chip);
+struct mem_mgr *nvhost_memmgr_alloc_mgr(void);
+void nvhost_memmgr_put_mgr(struct mem_mgr *);
+struct mem_mgr *nvhost_memmgr_get_mgr(struct mem_mgr *);
+struct mem_mgr *nvhost_memmgr_get_mgr_file(int fd);
+struct mem_handle *nvhost_memmgr_alloc(struct mem_mgr *,
+		size_t size, size_t align,
+		int flags);
+struct mem_handle *nvhost_memmgr_get(struct mem_mgr *,
+		u32 id, struct platform_device *dev);
+static inline int nvhost_memmgr_type(u32 id) { return id & MEMMGR_TYPE_MASK; }
+static inline int nvhost_memmgr_id(u32 id) { return id & MEMMGR_ID_MASK; }
+
+int nvhost_memmgr_pin_array_ids(struct mem_mgr *mgr,
+		struct platform_device *dev,
+		long unsigned *ids,
+		dma_addr_t *phys_addr,
+		u32 count,
+		struct nvhost_job_unpin_data *unpin_data);
+
+#endif
diff --git a/drivers/video/tegra/host/nvhost_syncpt.c b/drivers/video/tegra/host/nvhost_syncpt.c
new file mode 100644
index 0000000..4c72ea4
--- /dev/null
+++ b/drivers/video/tegra/host/nvhost_syncpt.c
@@ -0,0 +1,453 @@ 
+/*
+ * drivers/video/tegra/host/nvhost_syncpt.c
+ *
+ * Tegra Graphics Host Syncpoints
+ *
+ * Copyright (c) 2010-2012, NVIDIA Corporation.
+ *
+ * This program is free software; you can redistribute it and/or modify it
+ * under the terms and conditions of the GNU General Public License,
+ * version 2, as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
+ * more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program.  If not, see <http://www.gnu.org/licenses/>.
+ */
+
+#include <linux/platform_device.h>
+#include <linux/slab.h>
+#include <linux/stat.h>
+#include <trace/events/nvhost.h>
+#include "nvhost_syncpt.h"
+#include "nvhost_acm.h"
+#include "dev.h"
+#include "chip_support.h"
+
+#define MAX_SYNCPT_LENGTH	5
+
+/* Name of sysfs node for min and max value */
+static const char *min_name = "min";
+static const char *max_name = "max";
+
+/**
+ * Resets syncpoint and waitbase values to sw shadows
+ */
+void nvhost_syncpt_reset(struct nvhost_syncpt *sp)
+{
+	u32 i;
+
+	for (i = 0; i < nvhost_syncpt_nb_pts(sp); i++)
+		syncpt_op().reset(sp, i);
+	for (i = 0; i < nvhost_syncpt_nb_bases(sp); i++)
+		syncpt_op().reset_wait_base(sp, i);
+	wmb();
+}
+
+/**
+ * Updates sw shadow state for client managed registers
+ */
+void nvhost_syncpt_save(struct nvhost_syncpt *sp)
+{
+	u32 i;
+
+	for (i = 0; i < nvhost_syncpt_nb_pts(sp); i++) {
+		if (nvhost_syncpt_client_managed(sp, i))
+			syncpt_op().update_min(sp, i);
+		else
+			WARN_ON(!nvhost_syncpt_min_eq_max(sp, i));
+	}
+
+	for (i = 0; i < nvhost_syncpt_nb_bases(sp); i++)
+		syncpt_op().read_wait_base(sp, i);
+}
+
+/**
+ * Updates the last value read from hardware.
+ */
+u32 nvhost_syncpt_update_min(struct nvhost_syncpt *sp, u32 id)
+{
+	u32 val;
+
+	val = syncpt_op().update_min(sp, id);
+	trace_nvhost_syncpt_update_min(id, val);
+
+	return val;
+}
+
+/**
+ * Get the current syncpoint value
+ */
+u32 nvhost_syncpt_read(struct nvhost_syncpt *sp, u32 id)
+{
+	u32 val;
+	nvhost_module_busy(syncpt_to_dev(sp)->dev);
+	val = syncpt_op().update_min(sp, id);
+	nvhost_module_idle(syncpt_to_dev(sp)->dev);
+	return val;
+}
+
+/**
+ * Get the current syncpoint base
+ */
+u32 nvhost_syncpt_read_wait_base(struct nvhost_syncpt *sp, u32 id)
+{
+	u32 val;
+	nvhost_module_busy(syncpt_to_dev(sp)->dev);
+	syncpt_op().read_wait_base(sp, id);
+	val = sp->base_val[id];
+	nvhost_module_idle(syncpt_to_dev(sp)->dev);
+	return val;
+}
+
+/**
+ * Write a cpu syncpoint increment to the hardware, without touching
+ * the cache. Caller is responsible for host being powered.
+ */
+void nvhost_syncpt_cpu_incr(struct nvhost_syncpt *sp, u32 id)
+{
+	syncpt_op().cpu_incr(sp, id);
+}
+
+/**
+ * Increment syncpoint value from cpu, updating cache
+ */
+void nvhost_syncpt_incr(struct nvhost_syncpt *sp, u32 id)
+{
+	if (nvhost_syncpt_client_managed(sp, id))
+		nvhost_syncpt_incr_max(sp, id, 1);
+	nvhost_module_busy(syncpt_to_dev(sp)->dev);
+	nvhost_syncpt_cpu_incr(sp, id);
+	nvhost_module_idle(syncpt_to_dev(sp)->dev);
+}
+
+/**
+ * Updated sync point form hardware, and returns true if syncpoint is expired,
+ * false if we may need to wait
+ */
+static bool syncpt_update_min_is_expired(
+	struct nvhost_syncpt *sp,
+	u32 id,
+	u32 thresh)
+{
+	syncpt_op().update_min(sp, id);
+	return nvhost_syncpt_is_expired(sp, id, thresh);
+}
+
+/**
+ * Main entrypoint for syncpoint value waits.
+ */
+int nvhost_syncpt_wait_timeout(struct nvhost_syncpt *sp, u32 id,
+			u32 thresh, u32 timeout, u32 *value)
+{
+	DECLARE_WAIT_QUEUE_HEAD_ONSTACK(wq);
+	void *ref;
+	void *waiter;
+	int err = 0, check_count = 0, low_timeout = 0;
+	u32 val;
+
+	if (value)
+		*value = 0;
+
+	/* first check cache */
+	if (nvhost_syncpt_is_expired(sp, id, thresh)) {
+		if (value)
+			*value = nvhost_syncpt_read_min(sp, id);
+		return 0;
+	}
+
+	/* keep host alive */
+	nvhost_module_busy(syncpt_to_dev(sp)->dev);
+
+	/* try to read from register */
+	val = syncpt_op().update_min(sp, id);
+	if (nvhost_syncpt_is_expired(sp, id, thresh)) {
+		if (value)
+			*value = val;
+		goto done;
+	}
+
+	if (!timeout) {
+		err = -EAGAIN;
+		goto done;
+	}
+
+	/* schedule a wakeup when the syncpoint value is reached */
+	waiter = nvhost_intr_alloc_waiter();
+	if (!waiter) {
+		err = -ENOMEM;
+		goto done;
+	}
+
+	err = nvhost_intr_add_action(&(syncpt_to_dev(sp)->intr), id, thresh,
+				NVHOST_INTR_ACTION_WAKEUP_INTERRUPTIBLE, &wq,
+				waiter,
+				&ref);
+	if (err)
+		goto done;
+
+	err = -EAGAIN;
+	/* Caller-specified timeout may be impractically low */
+	if (timeout < SYNCPT_CHECK_PERIOD)
+		low_timeout = timeout;
+
+	/* wait for the syncpoint, or timeout, or signal */
+	while (timeout) {
+		u32 check = min_t(u32, SYNCPT_CHECK_PERIOD, timeout);
+		int remain = wait_event_interruptible_timeout(wq,
+				syncpt_update_min_is_expired(sp, id, thresh),
+				check);
+		if (remain > 0 || nvhost_syncpt_is_expired(sp, id, thresh)) {
+			if (value)
+				*value = nvhost_syncpt_read_min(sp, id);
+			err = 0;
+			break;
+		}
+		if (remain < 0) {
+			err = remain;
+			break;
+		}
+		if (timeout != NVHOST_NO_TIMEOUT)
+			timeout -= check;
+		if (timeout && check_count <= MAX_STUCK_CHECK_COUNT) {
+			dev_warn(&syncpt_to_dev(sp)->dev->dev,
+				"%s: syncpoint id %d (%s) stuck waiting %d, timeout=%d\n",
+				 current->comm, id, syncpt_op().name(sp, id),
+				 thresh, timeout);
+			syncpt_op().debug(sp);
+			if (check_count == MAX_STUCK_CHECK_COUNT) {
+				if (low_timeout) {
+					dev_warn(&syncpt_to_dev(sp)->dev->dev,
+						"is timeout %d too low?\n",
+						low_timeout);
+				}
+				nvhost_debug_dump(syncpt_to_dev(sp));
+			}
+			check_count++;
+		}
+	}
+	nvhost_intr_put_ref(&(syncpt_to_dev(sp)->intr), id, ref);
+
+done:
+	nvhost_module_idle(syncpt_to_dev(sp)->dev);
+	return err;
+}
+
+/**
+ * Returns true if syncpoint is expired, false if we may need to wait
+ */
+bool nvhost_syncpt_is_expired(
+	struct nvhost_syncpt *sp,
+	u32 id,
+	u32 thresh)
+{
+	u32 current_val;
+	u32 future_val;
+	smp_rmb();
+	current_val = (u32)atomic_read(&sp->min_val[id]);
+	future_val = (u32)atomic_read(&sp->max_val[id]);
+
+	/* Note the use of unsigned arithmetic here (mod 1<<32).
+	 *
+	 * c = current_val = min_val	= the current value of the syncpoint.
+	 * t = thresh			= the value we are checking
+	 * f = future_val  = max_val	= the value c will reach when all
+	 *				  outstanding increments have completed.
+	 *
+	 * Note that c always chases f until it reaches f.
+	 *
+	 * Dtf = (f - t)
+	 * Dtc = (c - t)
+	 *
+	 *  Consider all cases:
+	 *
+	 *	A) .....c..t..f.....	Dtf < Dtc	need to wait
+	 *	B) .....c.....f..t..	Dtf > Dtc	expired
+	 *	C) ..t..c.....f.....	Dtf > Dtc	expired	   (Dct very large)
+	 *
+	 *  Any case where f==c: always expired (for any t).	Dtf == Dcf
+	 *  Any case where t==c: always expired (for any f).	Dtf >= Dtc (because Dtc==0)
+	 *  Any case where t==f!=c: always wait.		Dtf <  Dtc (because Dtf==0,
+	 *							Dtc!=0)
+	 *
+	 *  Other cases:
+	 *
+	 *	A) .....t..f..c.....	Dtf < Dtc	need to wait
+	 *	A) .....f..c..t.....	Dtf < Dtc	need to wait
+	 *	A) .....f..t..c.....	Dtf > Dtc	expired
+	 *
+	 *   So:
+	 *	   Dtf >= Dtc implies EXPIRED	(return true)
+	 *	   Dtf <  Dtc implies WAIT	(return false)
+	 *
+	 * Note: If t is expired then we *cannot* wait on it. We would wait
+	 * forever (hang the system).
+	 *
+	 * Note: do NOT get clever and remove the -thresh from both sides. It
+	 * is NOT the same.
+	 *
+	 * If future valueis zero, we have a client managed sync point. In that
+	 * case we do a direct comparison.
+	 */
+	if (!nvhost_syncpt_client_managed(sp, id))
+		return future_val - thresh >= current_val - thresh;
+	else
+		return (s32)(current_val - thresh) >= 0;
+}
+
+void nvhost_syncpt_debug(struct nvhost_syncpt *sp)
+{
+	syncpt_op().debug(sp);
+}
+/* remove a wait pointed to by patch_addr */
+int nvhost_syncpt_patch_wait(struct nvhost_syncpt *sp, void *patch_addr)
+{
+	return syncpt_op().patch_wait(sp, patch_addr);
+}
+
+/* Displays the current value of the sync point via sysfs */
+static ssize_t syncpt_min_show(struct kobject *kobj,
+		struct kobj_attribute *attr, char *buf)
+{
+	struct nvhost_syncpt_attr *syncpt_attr =
+		container_of(attr, struct nvhost_syncpt_attr, attr);
+
+	return snprintf(buf, PAGE_SIZE, "%u",
+			nvhost_syncpt_read(&syncpt_attr->host->syncpt,
+				syncpt_attr->id));
+}
+
+static ssize_t syncpt_max_show(struct kobject *kobj,
+		struct kobj_attribute *attr, char *buf)
+{
+	struct nvhost_syncpt_attr *syncpt_attr =
+		container_of(attr, struct nvhost_syncpt_attr, attr);
+
+	return snprintf(buf, PAGE_SIZE, "%u",
+			nvhost_syncpt_read_max(&syncpt_attr->host->syncpt,
+				syncpt_attr->id));
+}
+
+int nvhost_syncpt_init(struct platform_device *dev,
+		struct nvhost_syncpt *sp)
+{
+	int i;
+	struct nvhost_master *host = syncpt_to_dev(sp);
+	int err = 0;
+
+	/* Allocate structs for min, max and base values */
+	sp->min_val = kzalloc(sizeof(atomic_t) * nvhost_syncpt_nb_pts(sp),
+			GFP_KERNEL);
+	sp->max_val = kzalloc(sizeof(atomic_t) * nvhost_syncpt_nb_pts(sp),
+			GFP_KERNEL);
+	sp->base_val = kzalloc(sizeof(u32) * nvhost_syncpt_nb_bases(sp),
+			GFP_KERNEL);
+	sp->lock_counts =
+		kzalloc(sizeof(atomic_t) * nvhost_syncpt_nb_mlocks(sp),
+			GFP_KERNEL);
+
+	if (!(sp->min_val && sp->max_val && sp->base_val && sp->lock_counts)) {
+		/* frees happen in the deinit */
+		err = -ENOMEM;
+		goto fail;
+	}
+
+	sp->kobj = kobject_create_and_add("syncpt", &dev->dev.kobj);
+	if (!sp->kobj) {
+		err = -EIO;
+		goto fail;
+	}
+
+	/* Allocate two attributes for each sync point: min and max */
+	sp->syncpt_attrs = kzalloc(sizeof(*sp->syncpt_attrs)
+			* nvhost_syncpt_nb_pts(sp) * 2, GFP_KERNEL);
+	if (!sp->syncpt_attrs) {
+		err = -ENOMEM;
+		goto fail;
+	}
+
+	/* Fill in the attributes */
+	for (i = 0; i < nvhost_syncpt_nb_pts(sp); i++) {
+		char name[MAX_SYNCPT_LENGTH];
+		struct kobject *kobj;
+		struct nvhost_syncpt_attr *min = &sp->syncpt_attrs[i*2];
+		struct nvhost_syncpt_attr *max = &sp->syncpt_attrs[i*2+1];
+
+		/* Create one directory per sync point */
+		snprintf(name, sizeof(name), "%d", i);
+		kobj = kobject_create_and_add(name, sp->kobj);
+		if (!kobj) {
+			err = -EIO;
+			goto fail;
+		}
+
+		min->id = i;
+		min->host = host;
+		min->attr.attr.name = min_name;
+		min->attr.attr.mode = S_IRUGO;
+		min->attr.show = syncpt_min_show;
+		if (sysfs_create_file(kobj, &min->attr.attr)) {
+			err = -EIO;
+			goto fail;
+		}
+
+		max->id = i;
+		max->host = host;
+		max->attr.attr.name = max_name;
+		max->attr.attr.mode = S_IRUGO;
+		max->attr.show = syncpt_max_show;
+		if (sysfs_create_file(kobj, &max->attr.attr)) {
+			err = -EIO;
+			goto fail;
+		}
+	}
+
+	return err;
+
+fail:
+	nvhost_syncpt_deinit(sp);
+	return err;
+}
+
+void nvhost_syncpt_deinit(struct nvhost_syncpt *sp)
+{
+	kobject_put(sp->kobj);
+
+	kfree(sp->min_val);
+	sp->min_val = NULL;
+
+	kfree(sp->max_val);
+	sp->max_val = NULL;
+
+	kfree(sp->base_val);
+	sp->base_val = NULL;
+
+	kfree(sp->lock_counts);
+	sp->lock_counts = 0;
+
+	kfree(sp->syncpt_attrs);
+	sp->syncpt_attrs = NULL;
+}
+
+int nvhost_syncpt_client_managed(struct nvhost_syncpt *sp, u32 id)
+{
+	return BIT(id) & syncpt_to_dev(sp)->info.client_managed;
+}
+
+int nvhost_syncpt_nb_pts(struct nvhost_syncpt *sp)
+{
+	return syncpt_to_dev(sp)->info.nb_pts;
+}
+
+int nvhost_syncpt_nb_bases(struct nvhost_syncpt *sp)
+{
+	return syncpt_to_dev(sp)->info.nb_bases;
+}
+
+int nvhost_syncpt_nb_mlocks(struct nvhost_syncpt *sp)
+{
+	return syncpt_to_dev(sp)->info.nb_mlocks;
+}
diff --git a/drivers/video/tegra/host/nvhost_syncpt.h b/drivers/video/tegra/host/nvhost_syncpt.h
new file mode 100644
index 0000000..7cdf749
--- /dev/null
+++ b/drivers/video/tegra/host/nvhost_syncpt.h
@@ -0,0 +1,148 @@ 
+/*
+ * drivers/video/tegra/host/nvhost_syncpt.h
+ *
+ * Tegra Graphics Host Syncpoints
+ *
+ * Copyright (c) 2010-2012, NVIDIA Corporation.
+ *
+ * This program is free software; you can redistribute it and/or modify it
+ * under the terms and conditions of the GNU General Public License,
+ * version 2, as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
+ * more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program.  If not, see <http://www.gnu.org/licenses/>.
+ */
+
+#ifndef __NVHOST_SYNCPT_H
+#define __NVHOST_SYNCPT_H
+
+#include <linux/kernel.h>
+#include <linux/sched.h>
+#include <linux/nvhost.h>
+#include <linux/atomic.h>
+
+/* host managed and invalid syncpt id */
+#define NVSYNCPT_GRAPHICS_HOST		     (0)
+
+/* Attribute struct for sysfs min and max attributes */
+struct nvhost_syncpt_attr {
+	struct kobj_attribute attr;
+	struct nvhost_master *host;
+	int id;
+};
+
+struct nvhost_syncpt {
+	struct kobject *kobj;
+	atomic_t *min_val;
+	atomic_t *max_val;
+	u32 *base_val;
+	atomic_t *lock_counts;
+	const char **syncpt_names;
+	struct nvhost_syncpt_attr *syncpt_attrs;
+};
+
+int nvhost_syncpt_init(struct platform_device *, struct nvhost_syncpt *);
+void nvhost_syncpt_deinit(struct nvhost_syncpt *);
+
+#define syncpt_to_dev(sp) container_of(sp, struct nvhost_master, syncpt)
+#define SYNCPT_CHECK_PERIOD (2 * HZ)
+#define MAX_STUCK_CHECK_COUNT 15
+
+/**
+ * Updates the value sent to hardware.
+ */
+static inline u32 nvhost_syncpt_incr_max(struct nvhost_syncpt *sp,
+					u32 id, u32 incrs)
+{
+	return (u32)atomic_add_return(incrs, &sp->max_val[id]);
+}
+
+/**
+ * Updated the value sent to hardware.
+ */
+static inline u32 nvhost_syncpt_set_max(struct nvhost_syncpt *sp,
+					u32 id, u32 val)
+{
+	atomic_set(&sp->max_val[id], val);
+	smp_wmb();
+	return val;
+}
+
+static inline u32 nvhost_syncpt_read_max(struct nvhost_syncpt *sp, u32 id)
+{
+	smp_rmb();
+	return (u32)atomic_read(&sp->max_val[id]);
+}
+
+static inline u32 nvhost_syncpt_read_min(struct nvhost_syncpt *sp, u32 id)
+{
+	smp_rmb();
+	return (u32)atomic_read(&sp->min_val[id]);
+}
+
+int nvhost_syncpt_client_managed(struct nvhost_syncpt *sp, u32 id);
+int nvhost_syncpt_nb_pts(struct nvhost_syncpt *sp);
+int nvhost_syncpt_nb_bases(struct nvhost_syncpt *sp);
+int nvhost_syncpt_nb_mlocks(struct nvhost_syncpt *sp);
+
+static inline bool nvhost_syncpt_check_max(struct nvhost_syncpt *sp,
+		u32 id, u32 real)
+{
+	u32 max;
+	if (nvhost_syncpt_client_managed(sp, id))
+		return true;
+	max = nvhost_syncpt_read_max(sp, id);
+	return (s32)(max - real) >= 0;
+}
+
+/**
+ * Returns true if syncpoint min == max
+ */
+static inline bool nvhost_syncpt_min_eq_max(struct nvhost_syncpt *sp, u32 id)
+{
+	int min, max;
+	smp_rmb();
+	min = atomic_read(&sp->min_val[id]);
+	max = atomic_read(&sp->max_val[id]);
+	return (min == max);
+}
+
+void nvhost_syncpt_cpu_incr(struct nvhost_syncpt *sp, u32 id);
+
+u32 nvhost_syncpt_update_min(struct nvhost_syncpt *sp, u32 id);
+bool nvhost_syncpt_is_expired(struct nvhost_syncpt *sp, u32 id, u32 thresh);
+
+void nvhost_syncpt_save(struct nvhost_syncpt *sp);
+
+void nvhost_syncpt_reset(struct nvhost_syncpt *sp);
+
+u32 nvhost_syncpt_read(struct nvhost_syncpt *sp, u32 id);
+u32 nvhost_syncpt_read_wait_base(struct nvhost_syncpt *sp, u32 id);
+
+void nvhost_syncpt_incr(struct nvhost_syncpt *sp, u32 id);
+
+int nvhost_syncpt_wait_timeout(struct nvhost_syncpt *sp, u32 id, u32 thresh,
+			u32 timeout, u32 *value);
+
+static inline int nvhost_syncpt_wait(struct nvhost_syncpt *sp,
+		u32 id, u32 thresh)
+{
+	return nvhost_syncpt_wait_timeout(sp, id, thresh,
+					  MAX_SCHEDULE_TIMEOUT, NULL);
+}
+
+int nvhost_syncpt_patch_wait(struct nvhost_syncpt *sp, void *patch_addr);
+
+void nvhost_syncpt_debug(struct nvhost_syncpt *sp);
+
+static inline int nvhost_syncpt_is_valid(struct nvhost_syncpt *sp, u32 id)
+{
+	return id != NVSYNCPT_INVALID && id < nvhost_syncpt_nb_pts(sp);
+}
+
+#endif
diff --git a/drivers/video/tegra/host/t20/Makefile b/drivers/video/tegra/host/t20/Makefile
new file mode 100644
index 0000000..9524668
--- /dev/null
+++ b/drivers/video/tegra/host/t20/Makefile
@@ -0,0 +1,6 @@ 
+ccflags-y = -Idrivers/video/tegra/host
+
+nvhost-t20-objs  = \
+	t20.o
+
+obj-$(CONFIG_TEGRA_GRHOST) += nvhost-t20.o
diff --git a/drivers/video/tegra/host/t20/t20.c b/drivers/video/tegra/host/t20/t20.c
new file mode 100644
index 0000000..0a47f79
--- /dev/null
+++ b/drivers/video/tegra/host/t20/t20.c
@@ -0,0 +1,78 @@ 
+/*
+ * drivers/video/tegra/host/t20/t20.c
+ *
+ * Tegra Graphics Init for T20 Architecture Chips
+ *
+ * Copyright (c) 2011-2012, NVIDIA Corporation.
+ *
+ * This program is free software; you can redistribute it and/or modify it
+ * under the terms and conditions of the GNU General Public License,
+ * version 2, as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
+ * more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program.  If not, see <http://www.gnu.org/licenses/>.
+ */
+
+#include <linux/init.h>
+#include <linux/clk.h>
+#include <linux/of.h>
+#include <linux/of_platform.h>
+#include <linux/nvhost.h>
+
+#include <mach/powergate.h>
+
+#include "t20.h"
+#include "host1x/host1x.h"
+#include "nvhost_channel.h"
+#include "nvhost_memmgr.h"
+#include "host1x/host1x01_hardware.h"
+#include "chip_support.h"
+
+static int t20_num_alloc_channels;
+
+static void t20_free_nvhost_channel(struct nvhost_channel *ch)
+{
+	nvhost_free_channel_internal(ch, &t20_num_alloc_channels);
+}
+
+static
+struct nvhost_channel *t20_alloc_nvhost_channel(struct platform_device *dev)
+{
+	struct nvhost_device_data *pdata = platform_get_drvdata(dev);
+	return nvhost_alloc_channel_internal(pdata->index,
+		nvhost_get_host(dev)->info.nb_channels,
+		&t20_num_alloc_channels);
+}
+
+#include "host1x/host1x_channel.c"
+#include "host1x/host1x_cdma.c"
+#include "host1x/host1x_debug.c"
+#include "host1x/host1x_syncpt.c"
+#include "host1x/host1x_intr.c"
+
+int nvhost_init_t20_support(struct nvhost_master *host,
+	struct nvhost_chip_support *op)
+{
+	int err;
+
+	op->channel = host1x_channel_ops;
+	op->cdma = host1x_cdma_ops;
+	op->push_buffer = host1x_pushbuffer_ops;
+	op->debug = host1x_debug_ops;
+	host->sync_aperture = host->aperture + HOST1X_CHANNEL_SYNC_REG_BASE;
+	op->syncpt = host1x_syncpt_ops;
+	op->intr = host1x_intr_ops;
+	err = nvhost_memmgr_init(op);
+	if (err)
+		return err;
+
+	op->nvhost_dev.alloc_nvhost_channel = t20_alloc_nvhost_channel;
+	op->nvhost_dev.free_nvhost_channel = t20_free_nvhost_channel;
+
+	return 0;
+}
diff --git a/drivers/video/tegra/host/t20/t20.h b/drivers/video/tegra/host/t20/t20.h
new file mode 100644
index 0000000..729f9d8
--- /dev/null
+++ b/drivers/video/tegra/host/t20/t20.h
@@ -0,0 +1,29 @@ 
+/*
+ * drivers/video/tegra/host/t20/t20.h
+ *
+ * Tegra Graphics Chip support for T20
+ *
+ * Copyright (c) 2011-2012, NVIDIA Corporation.
+ *
+ * This program is free software; you can redistribute it and/or modify it
+ * under the terms and conditions of the GNU General Public License,
+ * version 2, as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
+ * more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program.  If not, see <http://www.gnu.org/licenses/>.
+ */
+#ifndef _NVHOST_T20_H_
+#define _NVHOST_T20_H_
+
+struct nvhost_master;
+struct nvhost_chip_support;
+
+int nvhost_init_t20_support(struct nvhost_master *,
+	struct nvhost_chip_support *);
+
+#endif /* _NVHOST_T20_H_ */
diff --git a/drivers/video/tegra/host/t30/Makefile b/drivers/video/tegra/host/t30/Makefile
new file mode 100644
index 0000000..08dd3f5
--- /dev/null
+++ b/drivers/video/tegra/host/t30/Makefile
@@ -0,0 +1,6 @@ 
+ccflags-y += -Idrivers/video/tegra/host
+
+nvhost-t30-objs  = \
+	t30.o
+
+obj-$(CONFIG_TEGRA_GRHOST) += nvhost-t30.o
diff --git a/drivers/video/tegra/host/t30/t30.c b/drivers/video/tegra/host/t30/t30.c
new file mode 100644
index 0000000..5d5c43e
--- /dev/null
+++ b/drivers/video/tegra/host/t30/t30.c
@@ -0,0 +1,80 @@ 
+/*
+ * drivers/video/tegra/host/t30/t30.c
+ *
+ * Tegra Graphics Init for T30 Architecture Chips
+ *
+ * Copyright (c) 2011-2012, NVIDIA Corporation.
+ *
+ * This program is free software; you can redistribute it and/or modify it
+ * under the terms and conditions of the GNU General Public License,
+ * version 2, as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
+ * more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program.  If not, see <http://www.gnu.org/licenses/>.
+ */
+
+#include <linux/mutex.h>
+#include <linux/kernel.h>
+#include <linux/clk.h>
+#include <linux/of.h>
+#include <linux/of_platform.h>
+#include <linux/nvhost.h>
+
+#include <mach/powergate.h>
+
+#include "t20/t20.h"
+#include "t30.h"
+#include "host1x/host1x.h"
+#include "host1x/host1x01_hardware.h"
+#include "chip_support.h"
+#include "nvhost_channel.h"
+#include "nvhost_memmgr.h"
+
+static int t30_num_alloc_channels;
+
+static void t30_free_nvhost_channel(struct nvhost_channel *ch)
+{
+	nvhost_free_channel_internal(ch, &t30_num_alloc_channels);
+}
+
+static
+struct nvhost_channel *t30_alloc_nvhost_channel(struct platform_device *dev)
+{
+	struct nvhost_device_data *pdata = platform_get_drvdata(dev);
+	return nvhost_alloc_channel_internal(pdata->index,
+		nvhost_get_host(dev)->info.nb_channels,
+		&t30_num_alloc_channels);
+}
+
+#include "host1x/host1x_channel.c"
+#include "host1x/host1x_cdma.c"
+#include "host1x/host1x_debug.c"
+#include "host1x/host1x_syncpt.c"
+#include "host1x/host1x_intr.c"
+
+int nvhost_init_t30_support(struct nvhost_master *host,
+	struct nvhost_chip_support *op)
+{
+	int err;
+
+	op->channel = host1x_channel_ops;
+	op->cdma = host1x_cdma_ops;
+	op->push_buffer = host1x_pushbuffer_ops;
+	op->debug = host1x_debug_ops;
+	host->sync_aperture = host->aperture + HOST1X_CHANNEL_SYNC_REG_BASE;
+	op->syncpt = host1x_syncpt_ops;
+	op->intr = host1x_intr_ops;
+	err = nvhost_memmgr_init(op);
+	if (err)
+		return err;
+
+	op->nvhost_dev.alloc_nvhost_channel = t30_alloc_nvhost_channel;
+	op->nvhost_dev.free_nvhost_channel = t30_free_nvhost_channel;
+
+	return 0;
+}
diff --git a/drivers/video/tegra/host/t30/t30.h b/drivers/video/tegra/host/t30/t30.h
new file mode 100644
index 0000000..80838a5
--- /dev/null
+++ b/drivers/video/tegra/host/t30/t30.h
@@ -0,0 +1,29 @@ 
+/*
+ * drivers/video/tegra/host/t30/t30.h
+ *
+ * Tegra Graphics Chip support for Tegra3
+ *
+ * Copyright (c) 2011-2012, NVIDIA Corporation.
+ *
+ * This program is free software; you can redistribute it and/or modify it
+ * under the terms and conditions of the GNU General Public License,
+ * version 2, as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
+ * more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program.  If not, see <http://www.gnu.org/licenses/>.
+ */
+#ifndef _NVHOST_T30_H_
+#define _NVHOST_T30_H_
+
+struct nvhost_master;
+struct nvhost_chip_support;
+
+int nvhost_init_t30_support(struct nvhost_master *host,
+	struct nvhost_chip_support *);
+
+#endif /* _NVHOST_T30_H_ */
diff --git a/include/linux/nvhost.h b/include/linux/nvhost.h
new file mode 100644
index 0000000..f74e5fc
--- /dev/null
+++ b/include/linux/nvhost.h
@@ -0,0 +1,302 @@ 
+/*
+ * include/linux/nvhost.h
+ *
+ * Tegra graphics host driver
+ *
+ * Copyright (c) 2009-2012, NVIDIA Corporation. All rights reserved.
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation; either version 2 of the License, or
+ * (at your option) any later version.
+ *
+ * This program is distributed in the hope that it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
+ * more details.
+ *
+ * You should have received a copy of the GNU General Public License along
+ * with this program; if not, write to the Free Software Foundation, Inc.,
+ * 51 Franklin Street, Fifth Floor, Boston, MA  02110-1301, USA.
+ */
+
+#ifndef __LINUX_NVHOST_H
+#define __LINUX_NVHOST_H
+
+#include <linux/device.h>
+#include <linux/types.h>
+#include <linux/platform_device.h>
+
+struct nvhost_job;
+struct nvhost_device_power_attr;
+struct mem_mgr;
+struct nvhost_job_unpin_data;
+
+#define NVHOST_MODULE_MAX_CLOCKS		3
+#define NVHOST_MODULE_MAX_POWERGATE_IDS		2
+#define NVHOST_MODULE_NO_POWERGATE_IDS		.powergate_ids = {-1, -1}
+#define NVHOST_DEFAULT_CLOCKGATE_DELAY		.clockgate_delay = 25
+#define NVHOST_NAME_SIZE			24
+#define NVSYNCPT_INVALID			(-1)
+#define NVHOST_NO_TIMEOUT (-1)
+#define NVHOST_PRIORITY_LOW 50
+#define NVHOST_PRIORITY_MEDIUM 100
+#define NVHOST_PRIORITY_HIGH 150
+
+#define NVSYNCPT_2D_0		(18)
+#define NVSYNCPT_2D_1		(19)
+#define NVSYNCPT_VBLANK0	(26)
+#define NVSYNCPT_VBLANK1	(27)
+
+/* sync points that are wholly managed by the client */
+#define NVSYNCPTS_CLIENT_MANAGED (\
+	BIT(NVSYNCPT_VBLANK0) | \
+	BIT(NVSYNCPT_VBLANK1) | \
+	BIT(NVSYNCPT_2D_1))
+
+#define NVWAITBASE_2D_0 (1)
+#define NVWAITBASE_2D_1 (2)
+enum nvhost_power_sysfs_attributes {
+	NVHOST_POWER_SYSFS_ATTRIB_CLOCKGATE_DELAY = 0,
+	NVHOST_POWER_SYSFS_ATTRIB_POWERGATE_DELAY,
+	NVHOST_POWER_SYSFS_ATTRIB_REFCOUNT,
+	NVHOST_POWER_SYSFS_ATTRIB_MAX
+};
+
+struct nvhost_clock {
+	char *name;
+	unsigned long default_rate;
+	int reset;
+};
+
+enum nvhost_device_powerstate_t {
+	NVHOST_POWER_STATE_DEINIT,
+	NVHOST_POWER_STATE_RUNNING,
+	NVHOST_POWER_STATE_CLOCKGATED,
+	NVHOST_POWER_STATE_POWERGATED
+};
+
+struct host1x_device_info {
+	int		nb_channels;	/* host1x: num channels supported */
+	int		nb_pts;		/* host1x: num syncpoints supported */
+	int		nb_bases;	/* host1x: num syncpoints supported */
+	u32		client_managed; /* host1x: client managed syncpts */
+	int		nb_mlocks;	/* host1x: number of mlocks */
+	const char	**syncpt_names;	/* names of sync points */
+};
+
+struct nvhost_device_data {
+	int		version;	/* ip version number of device */
+	int		id;		/* Separates clients of same hw */
+	int		index;		/* Hardware channel number */
+	void __iomem	*aperture;	/* Iomem mapped to kernel */
+
+	u32		syncpts;	/* Bitfield of sync points used */
+	u32		modulemutexes;	/* Bit field of module mutexes */
+
+	u32		class;		/* Device class */
+	bool		serialize;	/* Serialize submits in the channel */
+
+	int		powergate_ids[NVHOST_MODULE_MAX_POWERGATE_IDS];
+	bool		can_powergate;	/* True if module can be power gated */
+	int		clockgate_delay;/* Delay before clock gated */
+	int		powergate_delay;/* Delay before power gated */
+	struct nvhost_clock clocks[NVHOST_MODULE_MAX_CLOCKS];/* Clock names */
+
+	struct delayed_work powerstate_down;/* Power state management */
+	int		num_clks;	/* Number of clocks opened for dev */
+	struct clk	*clk[NVHOST_MODULE_MAX_CLOCKS];
+	struct mutex	lock;		/* Power management lock */
+	int		powerstate;	/* Current power state */
+	int		refcount;	/* Number of tasks active */
+	wait_queue_head_t idle_wq;	/* Work queue for idle */
+
+	struct nvhost_channel *channel;	/* Channel assigned for the module */
+	struct kobject *power_kobj;	/* kobj to hold power sysfs entries */
+	struct nvhost_device_power_attr *power_attrib;	/* sysfs attributes */
+	struct dentry *debugfs;		/* debugfs directory */
+
+	void *private_data;		/* private platform data */
+	struct platform_device *pdev;	/* owner platform_device */
+
+	/* Finalize power on. Can be used for context restore. */
+	void (*finalize_poweron)(struct platform_device *dev);
+
+	/* Device is busy. */
+	void (*busy)(struct platform_device *);
+
+	/* Device is idle. */
+	void (*idle)(struct platform_device *);
+
+	/* Device is going to be suspended */
+	void (*suspend_ndev)(struct platform_device *);
+
+	/* Device is initialized */
+	void (*init)(struct platform_device *dev);
+
+	/* Device is de-initialized. */
+	void (*deinit)(struct platform_device *dev);
+
+	/* Preparing for power off. Used for context save. */
+	int (*prepare_poweroff)(struct platform_device *dev);
+
+	/* Clock gating callbacks */
+	int (*prepare_clockoff)(struct platform_device *dev);
+	void (*finalize_clockon)(struct platform_device *dev);
+};
+
+struct nvhost_device_power_attr {
+	struct platform_device *ndev;
+	struct kobj_attribute power_attr[NVHOST_POWER_SYSFS_ATTRIB_MAX];
+};
+
+/* public host1x power management APIs */
+bool nvhost_module_powered_ext(struct platform_device *dev);
+void nvhost_module_busy_ext(struct platform_device *dev);
+void nvhost_module_idle_ext(struct platform_device *dev);
+
+/* public host1x sync-point management APIs */
+u32 host1x_syncpt_incr_max(u32 id, u32 incrs);
+void host1x_syncpt_incr(u32 id);
+u32 host1x_syncpt_read(u32 id);
+int host1x_syncpt_wait(u32 id, u32 thresh, u32 timeout, u32 *value);
+
+/* Register device */
+int nvhost_client_device_init(struct platform_device *dev);
+int nvhost_client_device_suspend(struct platform_device *dev);
+struct nvhost_channel *nvhost_getchannel(struct nvhost_channel *ch);
+void nvhost_putchannel(struct nvhost_channel *ch);
+int nvhost_channel_submit(struct nvhost_job *job);
+
+enum host1x_class {
+	NV_HOST1X_CLASS_ID		= 0x1,
+	NV_GRAPHICS_2D_CLASS_ID		= 0x51,
+};
+
+struct nvhost_job_gather {
+	u32 words;
+	struct sg_table *mem_sgt;
+	dma_addr_t mem_base;
+	u32 mem_id;
+	int offset;
+	struct mem_handle *ref;
+};
+
+struct nvhost_cmdbuf {
+	__u32 mem;
+	__u32 offset;
+	__u32 words;
+};
+
+struct nvhost_reloc {
+	__u32 cmdbuf_mem;
+	__u32 cmdbuf_offset;
+	__u32 target;
+	__u32 target_offset;
+	__u32 shift;
+};
+
+struct nvhost_waitchk {
+	__u32 mem;
+	__u32 offset;
+	__u32 syncpt_id;
+	__u32 thresh;
+};
+
+/*
+ * Each submit is tracked as a nvhost_job.
+ */
+struct nvhost_job {
+	/* When refcount goes to zero, job can be freed */
+	struct kref ref;
+
+	/* List entry */
+	struct list_head list;
+
+	/* Channel where job is submitted to */
+	struct nvhost_channel *ch;
+
+	int clientid;
+
+	/* Nvmap to be used for pinning & unpinning memory */
+	struct mem_mgr *memmgr;
+
+	/* Gathers and their memory */
+	struct nvhost_job_gather *gathers;
+	int num_gathers;
+
+	/* Wait checks to be processed at submit time */
+	struct nvhost_waitchk *waitchk;
+	int num_waitchk;
+	u32 waitchk_mask;
+
+	/* Array of handles to be pinned & unpinned */
+	struct nvhost_reloc *relocarray;
+	int num_relocs;
+	struct nvhost_job_unpin_data *unpins;
+	int num_unpins;
+
+	dma_addr_t *addr_phys;
+	dma_addr_t *gather_addr_phys;
+	dma_addr_t *reloc_addr_phys;
+
+	/* Sync point id, number of increments and end related to the submit */
+	u32 syncpt_id;
+	u32 syncpt_incrs;
+	u32 syncpt_end;
+
+	/* Maximum time to wait for this job */
+	int timeout;
+
+	/* Null kickoff prevents submit from being sent to hardware */
+	bool null_kickoff;
+
+	/* Index and number of slots used in the push buffer */
+	int first_get;
+	int num_slots;
+};
+/*
+ * Allocate memory for a job. Just enough memory will be allocated to
+ * accomodate the submit.
+ */
+struct nvhost_job *nvhost_job_alloc(struct nvhost_channel *ch,
+		int num_cmdbufs, int num_relocs, int num_waitchks,
+		struct mem_mgr *memmgr);
+
+/*
+ * Add a gather to a job.
+ */
+void nvhost_job_add_gather(struct nvhost_job *job,
+		u32 mem_id, u32 words, u32 offset);
+
+/*
+ * Increment reference going to nvhost_job.
+ */
+void nvhost_job_get(struct nvhost_job *job);
+
+/*
+ * Decrement reference job, free if goes to zero.
+ */
+void nvhost_job_put(struct nvhost_job *job);
+
+/*
+ * Pin memory related to job. This handles relocation of addresses to the
+ * host1x address space. Handles both the gather memory and any other memory
+ * referred to from the gather buffers.
+ *
+ * Handles also patching out host waits that would wait for an expired sync
+ * point value.
+ */
+int nvhost_job_pin(struct nvhost_job *job, struct platform_device *pdev);
+
+/*
+ * Unpin memory related to job.
+ */
+void nvhost_job_unpin(struct nvhost_job *job);
+
+/*
+ * Dump contents of job to debug output.
+ */
+void nvhost_job_dump(struct device *dev, struct nvhost_job *job);
+
+#endif
diff --git a/include/trace/events/nvhost.h b/include/trace/events/nvhost.h
new file mode 100644
index 0000000..8cde821
--- /dev/null
+++ b/include/trace/events/nvhost.h
@@ -0,0 +1,249 @@ 
+/*
+ * include/trace/events/nvhost.h
+ *
+ * Nvhost event logging to ftrace.
+ *
+ * Copyright (c) 2010-2012, NVIDIA Corporation.
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation; either version 2 of the License, or
+ * (at your option) any later version.
+ *
+ * This program is distributed in the hope that it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
+ * more details.
+ *
+ * You should have received a copy of the GNU General Public License along
+ * with this program; if not, write to the Free Software Foundation, Inc.,
+ * 51 Franklin Street, Fifth Floor, Boston, MA  02110-1301, USA.
+ */
+
+#undef TRACE_SYSTEM
+#define TRACE_SYSTEM nvhost
+
+#if !defined(_TRACE_NVHOST_H) || defined(TRACE_HEADER_MULTI_READ)
+#define _TRACE_NVHOST_H
+
+#include <linux/ktime.h>
+#include <linux/tracepoint.h>
+
+DECLARE_EVENT_CLASS(nvhost,
+	TP_PROTO(const char *name),
+	TP_ARGS(name),
+	TP_STRUCT__entry(__field(const char *, name)),
+	TP_fast_assign(__entry->name = name;),
+	TP_printk("name=%s", __entry->name)
+);
+
+DEFINE_EVENT(nvhost, nvhost_channel_open,
+	TP_PROTO(const char *name),
+	TP_ARGS(name)
+);
+
+DEFINE_EVENT(nvhost, nvhost_channel_release,
+	TP_PROTO(const char *name),
+	TP_ARGS(name)
+);
+
+TRACE_EVENT(nvhost_cdma_end,
+	TP_PROTO(const char *name),
+
+	TP_ARGS(name),
+
+	TP_STRUCT__entry(
+		__field(const char *, name)
+	),
+
+	TP_fast_assign(
+		__entry->name = name;
+	),
+
+	TP_printk("name=%s",
+		__entry->name)
+);
+
+TRACE_EVENT(nvhost_cdma_flush,
+	TP_PROTO(const char *name, int timeout),
+
+	TP_ARGS(name, timeout),
+
+	TP_STRUCT__entry(
+		__field(const char *, name)
+		__field(int, timeout)
+	),
+
+	TP_fast_assign(
+		__entry->name = name;
+		__entry->timeout = timeout;
+	),
+
+	TP_printk("name=%s, timeout=%d",
+		__entry->name, __entry->timeout)
+);
+
+TRACE_EVENT(nvhost_cdma_push,
+	TP_PROTO(const char *name, u32 op1, u32 op2),
+
+	TP_ARGS(name, op1, op2),
+
+	TP_STRUCT__entry(
+		__field(const char *, name)
+		__field(u32, op1)
+		__field(u32, op2)
+	),
+
+	TP_fast_assign(
+		__entry->name = name;
+		__entry->op1 = op1;
+		__entry->op2 = op2;
+	),
+
+	TP_printk("name=%s, op1=%08x, op2=%08x",
+		__entry->name, __entry->op1, __entry->op2)
+);
+
+TRACE_EVENT(nvhost_cdma_push_gather,
+	TP_PROTO(const char *name, u32 mem_id,
+			u32 words, u32 offset, void *cmdbuf),
+
+	TP_ARGS(name, mem_id, words, offset, cmdbuf),
+
+	TP_STRUCT__entry(
+		__field(const char *, name)
+		__field(u32, mem_id)
+		__field(u32, words)
+		__field(u32, offset)
+		__field(bool, cmdbuf)
+		__dynamic_array(u32, cmdbuf, words)
+	),
+
+	TP_fast_assign(
+		if (cmdbuf) {
+			memcpy(__get_dynamic_array(cmdbuf), cmdbuf+offset,
+					words * sizeof(u32));
+		}
+		__entry->cmdbuf = cmdbuf;
+		__entry->name = name;
+		__entry->mem_id = mem_id;
+		__entry->words = words;
+		__entry->offset = offset;
+	),
+
+	TP_printk("name=%s, mem_id=%08x, words=%u, offset=%d, contents=[%s]",
+	  __entry->name, __entry->mem_id,
+	  __entry->words, __entry->offset,
+	  __print_hex(__get_dynamic_array(cmdbuf),
+		  __entry->cmdbuf ? __entry->words * 4 : 0))
+);
+
+TRACE_EVENT(nvhost_channel_submitted,
+	TP_PROTO(const char *name, u32 syncpt_base, u32 syncpt_max),
+
+	TP_ARGS(name, syncpt_base, syncpt_max),
+
+	TP_STRUCT__entry(
+		__field(const char *, name)
+		__field(u32, syncpt_base)
+		__field(u32, syncpt_max)
+	),
+
+	TP_fast_assign(
+		__entry->name = name;
+		__entry->syncpt_base = syncpt_base;
+		__entry->syncpt_max = syncpt_max;
+	),
+
+	TP_printk("name=%s, syncpt_base=%d, syncpt_max=%d",
+		__entry->name, __entry->syncpt_base, __entry->syncpt_max)
+);
+
+TRACE_EVENT(nvhost_channel_submit_complete,
+	TP_PROTO(const char *name, int count, u32 thresh),
+
+	TP_ARGS(name, count, thresh),
+
+	TP_STRUCT__entry(
+		__field(const char *, name)
+		__field(int, count)
+		__field(u32, thresh)
+	),
+
+	TP_fast_assign(
+		__entry->name = name;
+		__entry->count = count;
+		__entry->thresh = thresh;
+	),
+
+	TP_printk("name=%s, count=%d, thresh=%d",
+		__entry->name, __entry->count, __entry->thresh)
+);
+
+TRACE_EVENT(nvhost_wait_cdma,
+	TP_PROTO(const char *name, u32 eventid),
+
+	TP_ARGS(name, eventid),
+
+	TP_STRUCT__entry(
+		__field(const char *, name)
+		__field(u32, eventid)
+	),
+
+	TP_fast_assign(
+		__entry->name = name;
+		__entry->eventid = eventid;
+	),
+
+	TP_printk("name=%s, event=%d", __entry->name, __entry->eventid)
+);
+
+TRACE_EVENT(nvhost_syncpt_update_min,
+	TP_PROTO(u32 id, u32 val),
+
+	TP_ARGS(id, val),
+
+	TP_STRUCT__entry(
+		__field(u32, id)
+		__field(u32, val)
+	),
+
+	TP_fast_assign(
+		__entry->id = id;
+		__entry->val = val;
+	),
+
+	TP_printk("id=%d, val=%d", __entry->id, __entry->val)
+);
+
+TRACE_EVENT(nvhost_syncpt_wait_check,
+	TP_PROTO(u32 mem_id, u32 offset, u32 syncpt_id, u32 thresh, u32 min),
+
+	TP_ARGS(mem_id, offset, syncpt_id, thresh, min),
+
+	TP_STRUCT__entry(
+		__field(u32, mem_id)
+		__field(u32, offset)
+		__field(u32, syncpt_id)
+		__field(u32, thresh)
+		__field(u32, min)
+	),
+
+	TP_fast_assign(
+		__entry->mem_id = mem_id;
+		__entry->offset = offset;
+		__entry->syncpt_id = syncpt_id;
+		__entry->thresh = thresh;
+		__entry->min = min;
+	),
+
+	TP_printk("mem_id=%08x, offset=%05x, id=%d, thresh=%d, current=%d",
+		__entry->mem_id, __entry->offset,
+		__entry->syncpt_id, __entry->thresh,
+		__entry->min)
+);
+
+#endif /*  _TRACE_NVHOST_H */
+
+/* This part must be outside protection */
+#include <trace/define_trace.h>