Hook up the default ->reset() and ->atomic_duplicate_state() helpers.
This ensures that state objects are properly created and framebuffer
reference counts correctly maintained.
Signed-off-by: Thierry Reding <treding@nvidia.com>

Implement initial atomic state handling. Hook up the CRTCs, planes' and
connectors' ->atomic_destroy_state() callback to ensure that the atomic
state objects don't leak.

Furthermore the CRTC now implements the ->mode_set_nofb() callback that
is used by new helpers to implement ->mode_set() and ->mode_set_base().
These new helpers also make use of the new plane helper functions which
the driver now provides.
Signed-off-by: Thierry Reding <treding@nvidia.com>

Commit 9c012700 ("drm/tegra: dc: Add powergate support") changed the
driver's ->probe() implementation to deassert the module reset, and with
there being nobody else to assert it until ->remove() there is no need
to deassert again later on.
Signed-off-by: Thierry Reding <treding@nvidia.com>

The tegra_output_exit() and tegra_output_remove() functions cannot fail,
so make them return void.
Signed-off-by: Thierry Reding <treding@nvidia.com>

The tegra_output midlayer is now completely gone and output drivers use
it purely as a helper library.
Signed-off-by: Thierry Reding <treding@nvidia.com>

The debugfs cleanup code never fails, so no error is returned. Therefore
the functions can all return void instead.
Signed-off-by: Thierry Reding <treding@nvidia.com>

Implement encoder and connector within the eDP driver itself using the
Tegra output helpers rather than using the Tegra output as midlayer. By
doing so one level of indirection is removed and output drivers become
more flexible while keeping the majority of the advantages provided by
the common output helpers.
Signed-off-by: Thierry Reding <treding@nvidia.com>

Implement encoder and connector within the DSI driver itself using the
Tegra output helpers rather than using the Tegra output as midlayer. By
doing so one level of indirection is removed and output drivers become
more flexible while keeping the majority of the advantages provided by
the common output helpers.
Signed-off-by: Thierry Reding <treding@nvidia.com>

Implement encoder and connector within the HDMI driver itself using the
Tegra output helpers rather than using the Tegra output as midlayer. By
doing so one level of indirection is removed and output drivers become
more flexible while keeping the majority of the advantages provided by
the common output helpers.
Signed-off-by: Thierry Reding <treding@nvidia.com>

Implement encoder and connector within the RGB driver itself using the
Tegra output helpers rather than using the Tegra output as midlayer. By
doing so one level of indirection is removed and output drivers become
more flexible while keeping the majority of the advantages provided by
the common output helpers.
Signed-off-by: Thierry Reding <treding@nvidia.com>

This is a small helper that performs the basic steps required by all
output drivers to prepare the display controller for use with a given
encoder.
Signed-off-by: Thierry Reding <treding@nvidia.com>

In order to transition output drivers to using the struct tegra_output
as a helper rather than midlayer, make this callback optional. Instead
drivers should implement the equivalent as part of ->mode_fixup(). For
the conversion to atomic modesetting a new callback ->atomic_check()
should be implemented that updates the display controller's state with
the corresponding parent clock, rate and shift clock divider.
Signed-off-by: Thierry Reding <treding@nvidia.com>

The output layer was initially designed to help reduce the amount of
code duplicated in output drivers. An unfortunate side-effect of that
was that it turned into a midlayer and it became difficult to make the
output drivers work without bending over backwards to fit into the
midlayer.

This commit starts to convert the midlayer into a helper library by
exporting most of the common functions so that they can be used by the
output drivers directly. Doing so will allow output drivers to reuse
common code paths but more easily override them where necessary.
Signed-off-by: Thierry Reding <treding@nvidia.com>

The DRM core should take care of disabling all unneeded planes, so there
is no need to do this explicitly.
Signed-off-by: Thierry Reding <treding@nvidia.com>

This structure will be extended using non-framebuffer related callbacks
in subsequent patches, so it should move to a more central location.
Signed-off-by: Thierry Reding <treding@nvidia.com>

When disabling the display controller, stop it and wait for it to become
idle. Doing so ensures that no further accesses to the framebuffer occur
and the buffers can be safely unmapped or freed.
Signed-off-by: Thierry Reding <treding@nvidia.com>

Previously output drivers would all stop the display controller in their
disable path. However with the transition to atomic modesetting the
display controller needs to be kept running until all planes have been
disabled so that software can properly determine (using VBLANK counts)
when it is safe to remove the framebuffers associated with the planes.

Moving this code into the display controller's disable path also gets
rid of the duplication of this into all output drivers.
Signed-off-by: Thierry Reding <treding@nvidia.com>

All output drivers have open-coded variants of this function, so export
it to remove some code duplication.
Signed-off-by: Thierry Reding <treding@nvidia.com>

This callback can be used instead of the legacy ->mode_fixup() and is
passed the CRTC and connector states. It can thus use these states to
validate the modeset and cache values in the state to be used during
the actual modeset.
Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Thierry Reding <treding@nvidia.com>

In order to prevent drivers from having to perform the same checks over
and over again, add an optional ->atomic_disable callback which the core
calls under the right circumstances.

v2: pass old state and detect edges to avoid calling ->atomic_disable on
already disabled planes, remove redundant comment (Daniel Vetter)

v3: rename helper to drm_atomic_plane_disabling() to clarify that it is
checking for transitions, move helper to drm_atomic_helper.h, clarify
check for !old_state and its relation to transitional helpers

Here's an extract from some discussion rationalizing the behaviour (for
a full version, see the reference below):

    > > Hm, thinking about this some more this will result in a slight difference
    > > in behaviour, at least when drivers just use the helper ->reset functions
    > > but don't disable everything:
    > > - With transitional helpers we assume we know nothing and call
    > >   ->atomic_disable.
    > > - With atomic old_state->crtc == NULL in the same situation right after
    > >   boot-up, but we asssume the plane is really off and _dont_ call
    > >   ->atomic_disable.
    > >
    > > Should we instead check for (old_state && old_state->crtc) and state that
    > > drivers need to make sure they don't have stuff hanging around?
    >
    > I don't think we can check for old_state because otherwise this will
    > always return false, whereas we really want it to force-disable planes
    > that could be on (lacking any more accurate information). For
    > transitional helpers anyway.
    >
    > For the atomic helpers, old_state will never be NULL, but I'd assume
    > that the driver would reconstruct the current state in ->reset().

    By the way, the reason for why old_state can be NULL with transitional
    helpers is the ordering of the steps in the atomic transition. Currently
    the Tegra patches do this (based on your blog post and the Exynos proto-
    type):

        1) atomic conversion, phase 1:
           - implement ->atomic_{check,update,disable}()
           - use drm_plane_helper_{update,disable}()

        2) atomic conversion, phase 2:
           - call drm_mode_config_reset() from ->load()
           - implement ->reset()

    That's only a partial list of what's done in these steps, but that's the
    only relevant pieces for why old_state is NULL.

    What happens is that without ->reset() implemented there won't be any
    initial state, hence plane->state (the old_state here) will be NULL the
    first time atomic state is applied.

    We could of course reorder the sequence such that drivers are required
    to hook up ->reset() before they can (or at the same as they) hook up
    the transitional helpers. We could add an appropriate WARN_ON to this
    helper to make that more obvious.

    However, that will not solve the problem because it only gets rid of the
    special case. We still don't know whether old_state->crtc == NULL is the
    current state or just the initial default.

    So no matter which way we do this, I don't see a way to get away without
    requiring specific semantics from drivers. They would be that:

        - drivers recreate the correct state in ->reset() so that
          old_state->crtc != NULL if the plane is really enabled

    or

        - drivers have to ensure that the real state in fact mirrors the
          initial default as encoded in the state (plane disabled)

References: http://lists.freedesktop.org/archives/dri-devel/2015-January/075578.htmlReviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Reviewed-by: Gustavo Padovan <gustavo.padovan@collabora.co.uk>
Signed-off-by: Thierry Reding <treding@nvidia.com>

There is no use-case where it would be useful for drivers not to
implement this function and the transitional plane helpers already
require drivers to provide an implementation.

v2: add new requirement to kerneldoc
Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Thierry Reding <treding@nvidia.com>

kfree(ERR_PTR(-ENOMEM)) will not work very well.
Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: Thierry Reding <treding@nvidia.com>

We can't save two values to the IRQ flags at the same time so the IRQs
are not enabled at the end.  This kind of bug is easy to miss in testing
if the function is normally called with IRQs disabled so we wouldn't
enable IRQs anyway.
Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: Thierry Reding <treding@nvidia.com>

Compliance testing shows that HS Trail is off by -12%. Increase the HS
Trail time to make this test pass.
Signed-off-by: David Ung <davidu@nvidia.com>
[treding@nvidia.com: update specification references, add comment]
Signed-off-by: Thierry Reding <treding@nvidia.com>

This allows a DRM driver unload/reload cycle to completely reset the DSI
controller and may help in situations where it's broken.
Signed-off-by: Thierry Reding <treding@nvidia.com>

This reset is necessary to properly clean up the internal state of the
controller.
Signed-off-by: Thierry Reding <treding@nvidia.com>

Use a sized unsigned 32-bit data type (u32) to store register contents.
The DSI registers are 32 bits wide irrespective of the architecture's
data width.
Signed-off-by: Thierry Reding <treding@nvidia.com>

Use a sized unsigned 32-bit data type (u32) to store register contents.
The HDMI registers are 32 bits wide irrespective of the architecture's
data width.
Signed-off-by: Thierry Reding <treding@nvidia.com>

This prevents the compiler from warning about using a variable that is
possibly uninitialized.
Signed-off-by: Thierry Reding <treding@nvidia.com>

Most of the display controller's registers are double-buffered, a few of
them are triple-buffered. The ASSEMBLY shadow copy is latched intto the
ACTIVE copy for double-buffered registers. For triple-buffered registers
the ASSEMBLY copy is first latched into the ARM copy.

Latching into the ACTIVE copy happens immediately if the controller is
inactive. Otherwise the latching happens on the next frame boundary. The
latching of the ASSEMBLY into the ARM copy happens immediately. Latching
is controlled by a set of *_ACT_REQ and *_UPDATE bits in the
DC_CMD_STATE_CONTROL register.
Signed-off-by: Thierry Reding <treding@nvidia.com>

Tegra114 and earlier support specifying the color of the border (i.e.
the active area of the screen that is not covered by any of the overlay
windows). By default this is set to a light blue, so set it to black to
comply with the requirements set by atomic modesetting.
Signed-off-by: Thierry Reding <treding@nvidia.com>

iommu_domain_alloc() returns NULL on error, it never returns error
pointers.
Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: Thierry Reding <treding@nvidia.com>

The possible_crtcs mask needs to be a mask of CRTC indices. There is no
guarantee that the DRM indices match the hardware pipe number, so the
mask must be computed from the CRTC index.
Signed-off-by: Thierry Reding <treding@nvidia.com>

The DRM core now zeroes out the memory associated with CRTC, encoder and
connector objects upon cleanup, so there's no need to explicitly do that
in drivers anymore.
Signed-off-by: Thierry Reding <treding@nvidia.com>

The iommu_map_sg() function is now available in the IOMMU API, so drop
the open-coded variant.
Signed-off-by: Thierry Reding <treding@nvidia.com>

Previously the struct bus_type exported by the host1x infrastructure was
only a very basic skeleton. Turn that implementation into a more full-
fledged bus to support proper probe ordering and power management.

Note that the bus infrastructure needs to be available before any of the
drivers can be registered. This is automatically ensured if all drivers
are built as loadable modules (via symbol dependencies). If all drivers
are built-in there are no such guarantees and the link order determines
the initcall ordering. Adjust drivers/gpu/Makefile to make sure that the
host1x bus infrastructure is initialized prior to any of its users (only
drm/tegra currently).

v2: Fix building host1x and tegra-drm as modules
Reported-by: Dave Airlie <airlied@gmail.com>
Reviewed-by: Sean Paul <seanpaul@chromium.org>
Reviewed-by: Mark Zhang <markz@nvidia.com>
Signed-off-by: Thierry Reding <treding@nvidia.com>

This function is needed in several places, so factor it out.
Signed-off-by: Thierry Reding <treding@nvidia.com>

Instead of locking within host1x_device_add(), call it under the lock to
make the locking more consistent.
Signed-off-by: Thierry Reding <treding@nvidia.com>

When a driver's ->probe() function fails, the host1x bus must not call
its ->remove() function because the driver will already have cleaned up
in the error handling path in ->probe().
Signed-off-by: Thierry Reding <treding@nvidia.com>

Backmerge Linus tree after rc5 + drm-fixes went in.

There were a few amdkfd conflicts I wanted to avoid,
and Ben requested this for nouveau also.

Conflicts:
	drivers/gpu/drm/amd/amdkfd/Makefile
	drivers/gpu/drm/amd/amdkfd/kfd_chardev.c
	drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager.c
	drivers/gpu/drm/amd/amdkfd/kfd_priv.h
	drivers/gpu/drm/amd/include/kgd_kfd_interface.h
	drivers/gpu/drm/i915/intel_runtime_pm.c
	drivers/gpu/drm/radeon/radeon_kfd.c