Commit 917de180 ("drm: rcar-du: Implement universal plane support")
made the number of planes per group dynamic, but didn't update all loops
over the planes array, resulting in out-of-bound accesses on DU
instances that have an odd number of CRTCs (such as the R8A7790). Fix
it.

Fixes: 917de180 ("drm: rcar-du: Implement universal plane support")
Cc: stable@vger.kernel.org
Signed-off-by: NLaurent Pinchart <laurent.pinchart+renesas@ideasonboard.com>

The DSnPR plane configuration registers are updated on vblank, and no
vblank will occur once the CRTC is stopped. We thus can't only disable
planes right before starting the CRTC as it would start scanning out
immediately from old frame buffers until the next vblank.

Fix the problem by disabling all planes when stopping the CRTC and wait
for the change to take effect. This increases the CRTC stop delay,
especially when multiple CRTCs are stopped in one operation as we now
wait for one vblank per CRTC. Whether this can be improved needs to be
researched.
Signed-off-by: NLaurent Pinchart <laurent.pinchart+renesas@ideasonboard.com>

This helps debugging probe failures.
Signed-off-by: NLaurent Pinchart <laurent.pinchart+renesas@ideasonboard.com>

If we have more than one CRTCs in a group pre-associate planes 0-3 with
CRTC 0 and planes 4-7 with CRTC 1 to minimize flicker occurring when the
association is changed. The pre-association could be controlled by a
module parameter if needed.
Signed-off-by: NLaurent Pinchart <laurent.pinchart+renesas@ideasonboard.com>

The number of CRTCs in a group is only used to implement plane
initialization for now, but is also needed to implement pre-association
of planes to CRTCs. Store it in the group structure instead of computing
it on demand.
Signed-off-by: NLaurent Pinchart <laurent.pinchart+renesas@ideasonboard.com>

Hardware planes are driven by the timing generator of the CRTC they are
associated to. Changing the association requires restarting the CRTC
group that the plane belongs to, resulting in flicker on the other CRTC.

To avoid flicker as much as possible, try to allocate planes first from
the free planes already associated with the target CRTC. If allocation
fails then fall back to allocation from all free planes.
Signed-off-by: NLaurent Pinchart <laurent.pinchart+renesas@ideasonboard.com>

Changing the plane to CRTC associations requires restarting the CRTC
group, creating visible flicker. Mitigate the issue by changing plane
association only when a plane becomes enabled, not when it get disabled.
Signed-off-by: NLaurent Pinchart <laurent.pinchart+renesas@ideasonboard.com>

Plane allocation is a complex process, add debugging statements to help
finding out what could might wrong.
Signed-off-by: NLaurent Pinchart <laurent.pinchart+renesas@ideasonboard.com>

All other cast functions are named without using "du", make the plane
state cast consistent with them.
Signed-off-by: NLaurent Pinchart <laurent.pinchart+renesas@ideasonboard.com>

The rcar_du_planes structure contains a single field and is only
instantiated in the rcar_du_group structure. Embed it directly and
remove the rcar_du_planes structure.
Signed-off-by: NLaurent Pinchart <laurent.pinchart+renesas@ideasonboard.com>

The plane property objects are instantiated once per CRTC group, while
they should be instantiated once globally for the device. Fix this and
move them to the rcar_du_device structure.
Signed-off-by: NLaurent Pinchart <laurent.pinchart+renesas@ideasonboard.com>

Document the structure fields using kerneldoc.
Signed-off-by: NLaurent Pinchart <laurent.pinchart+renesas@ideasonboard.com>

Document the structure fields using kerneldoc.
Signed-off-by: NLaurent Pinchart <laurent.pinchart+renesas@ideasonboard.com>

Using the for_each_... macro should make the code a bit shorter and
easier to read.
Signed-off-by: NPhilipp Zabel <p.zabel@pengutronix.de>
Acked-by: NLaurent Pinchart <laurent.pinchart@ideasonboard.com>

Set the DRIVER_ATOMIC flag to enable usage of the atomic updates API
with the driver.
Signed-off-by: NLaurent Pinchart <laurent.pinchart+renesas@ideasonboard.com>

The field is set by drm_send_vblank_event(), there's no need to
preinitialize it.
Signed-off-by: NLaurent Pinchart <laurent.pinchart+renesas@ideasonboard.com>

Plane state duplication takes a reference to the framebuffer stored in
the state, but state destroy doesn't release it. This causes a reference
leak. Fix it.
Signed-off-by: NLaurent Pinchart <laurent.pinchart+renesas@ideasonboard.com>

The plane allocator has been inherently racy since the beginning of the
transition to atomic updates, as the allocator lock is released between
free plane check (at .atomic_check() time) and the reservation (at
.atomic_update() time).

To fix it, create a new allocator solely based on the atomic plane
states without keeping any external state and perform allocation in the
.atomic_check() handler. The core idea is to replace the free planes
bitmask with a collective knowledge based on the allocated hardware
plane(s) for each KMS plane. The allocator then loops over all plane
states to compute the free planes bitmask, allocates hardware planes
based on that bitmask, and stores the result back in the plane states.

For this to work we need to access the current state of planes not
touched by the atomic update. To ensure that it won't be modified, we
need to lock all planes using drm_atomic_get_plane_state(). This
effectively serializes atomic updates from .atomic_check() up to
completion, either when swapping the states if the check step has
succeeded, or when freeing the states if the check step has failed.
Suggested-by: NDaniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: NLaurent Pinchart <laurent.pinchart+renesas@ideasonboard.com>

Only the planes to CRTCs association control register DPTSR needs to be
protected by custom locking, don't hold the mutex around the whole code.
Signed-off-by: NLaurent Pinchart <laurent.pinchart+renesas@ideasonboard.com>

As the DRM core will commit plane states when performing atomic updates,
those don't need to be committed manually when the CRTC is started except
in the system resume code path.

However, the atomic plane commit step is currently performed between
mode set disable and mode set enable to mimick the legacy mode setting
operations order. This causes the device clocks to be disabled after
applying plane settings and reenabled when enabling the CRTC,
potentially losing hardware in between.

Reorder the operations to enable the CRTC first and only then apply
plane settings, removing the need to manage clocks in the atomic begin
and flush handlers. We can then move the plane state commit code out of
the CRTC start handler to the system resume handler.
Signed-off-by: NLaurent Pinchart <laurent.pinchart+renesas@ideasonboard.com>

The format stored in the rcar_du_plane structure is part of the plane
state. Move it to the rcar_du_plane_state structure and precompute it in
the .atomic_check() handler.
Signed-off-by: NLaurent Pinchart <laurent.pinchart+renesas@ideasonboard.com>

The rcar_du_crtc plane field is only used to check for an error that
can't occur. Remove it.
Signed-off-by: NLaurent Pinchart <laurent.pinchart+renesas@ideasonboard.com>

The crtc and enabled fields duplicates information stored in the plane
state. Use the plane state instead and remove the fields.
Signed-off-by: NLaurent Pinchart <laurent.pinchart+renesas@ideasonboard.com>

Now that the plane setup code isn't called outside of the plane
implementation, it can be simplified by merging the
rcar_du_plane_compute_base() and rcar_du_plane_update_base() functions.
Signed-off-by: NLaurent Pinchart <laurent.pinchart+renesas@ideasonboard.com>

Allow setting up plane properties atomically using the plane
set_property atomic helper. The properties are now stored in the plane
state (requiring subclassing it) and applied when updating the planes.
Signed-off-by: NLaurent Pinchart <laurent.pinchart+renesas@ideasonboard.com>

The atomic page flip helper implements the page flip operation using
asynchronous commits.

As the legacy page flip was the last CRTC operation that needed direct
access to plane setup, the plane setup functions can now become private
to the plane implementation.
Signed-off-by: NLaurent Pinchart <laurent.pinchart+renesas@ideasonboard.com>

Implement a custom .atomic_commit() handler that supports asynchronous
commits using a work queue. This can be used for userspace-driven
asynchronous commits, as well as for an atomic page flip implementation.
Signed-off-by: NLaurent Pinchart <laurent.pinchart+renesas@ideasonboard.com>

The encoder .mode_fixup() operation is legacy, atomic updates uses the
new .atomic_check() operation. Convert the encoders drivers.
Signed-off-by: NLaurent Pinchart <laurent.pinchart+renesas@ideasonboard.com>

The atomic connector DPMS helper implements the connector DPMS operation
using atomic commit, removing the need for DPMS helper operations on
CRTCs and encoders.
Signed-off-by: NLaurent Pinchart <laurent.pinchart+renesas@ideasonboard.com>

This removes the legacy mode config code. The CRTC and encoder prepare
and commit operations are not used anymore, remove them.
Signed-off-by: NLaurent Pinchart <laurent.pinchart+renesas@ideasonboard.com>

This removes the legacy plane update code. Wire up the default atomic
check and atomic commit mode config helpers as needed by the plane
update atomic helpers.
Signed-off-by: NLaurent Pinchart <laurent.pinchart+renesas@ideasonboard.com>

When using atomic updates the CRTC .enable() and .disable() helper
operations are preferred over the (then legacy) .prepare() and .commit()
operations. Implement .enable() and rework .disable() to not depend on
DPMS, easing DPMS removal later on.
Signed-off-by: NLaurent Pinchart <laurent.pinchart+renesas@ideasonboard.com>

When using atomic updates the encoder .enable() and .disable() helper
operations are preferred over the (then legacy) .prepare() and .commit()
operations. Implement .enable() and .disable() and rework .prepare(),
.commit() and .dpms() as wrappers around .enable() and .disable(),
easing their future removal.
Signed-off-by: NLaurent Pinchart <laurent.pinchart+renesas@ideasonboard.com>

When using atomic updates the encoder .enable() and .disable() helper
operations are preferred over the (then legacy) .prepare() and .commit()
operations. Implement .enable() and .disable() and rework .prepare(),
.commit() and .dpms() as wrappers around .enable() and .disable(),
easing their future removal.
Signed-off-by: NLaurent Pinchart <laurent.pinchart+renesas@ideasonboard.com>

The LVDS encoder doesn't support DPMS states, replace the DPMS operation
by enable/disable to avoid propagating DPMS states down to the encoder
code.
Signed-off-by: NLaurent Pinchart <laurent.pinchart+renesas@ideasonboard.com>

The plane source and destination size and positions are stored in the
plane state, and a private copy is kept in the rcar_du_plane objects.
Remove the private copy as it just duplicates the state.
Signed-off-by: NLaurent Pinchart <laurent.pinchart+renesas@ideasonboard.com>

Hook up the default .reset(), .atomic_duplicate_state() and
.atomic_free_state() helpers to ensure that state objects are properly
created and destroyed, and call drm_mode_config_reset() at init time to
create the initial state objects.

Framebuffer reference count also gets maintained automatically by the
transitional helpers except for the legacy page flip operation. Maintain
it explicitly there.
Signed-off-by: NLaurent Pinchart <laurent.pinchart+renesas@ideasonboard.com>

Use the new CRTC atomic transitional helpers drm_helper_crtc_mode_set()
and drm_helper_crtc_mode_set_base() to implement the CRTC .mode_set and
.mode_set_base operations. This delegates primary plane configuration to
the plane .atomic_update and .atomic_disable operations, removing
duplicate code from the CRTC implementation.

There is now no code path available to the driver in which to drop the
reference to the CRTC acquired in the .prepare() operation if an error
then occurs. The driver thus now leaks a reference if an error occurs
during mode set. So be it, this will be fixed in a further step of the
atomic update transition.
Signed-off-by: NLaurent Pinchart <laurent.pinchart+renesas@ideasonboard.com>

Implement the CRTC .atomic_begin() and .atomic_flush() operations, the
plane .atomic_check(), .atomic_update() and operations, and use the
transitional atomic helpers to implement the plane update and disable
operations on top of the new atomic operations.

The plane setup code can't be moved out of the CRTC start function
completely yet, as the atomic code paths are not taken every time the
CRTC needs to be started. This results in some code duplication that
will be fixed after switching to atomic updates completely.
Signed-off-by: NLaurent Pinchart <laurent.pinchart+renesas@ideasonboard.com>

The hardware plane allocator loops over all planes to find free
candidates. However, instead of looping over the number of hardware
planes, it loops over the number of software planes, which happens to be
larger by one unit. This has no effect in practise as the extra plane is
always cleared in the mask of free planes, but it should still be fixed
for correctness.
Signed-off-by: NLaurent Pinchart <laurent.pinchart+renesas@ideasonboard.com>