Document the R8A7795-specific DT bindings and support them in the
driver. The HDMI and LVDS outputs are currently not supported.
Signed-off-by: NKoji Matsuoka <koji.matsuoka.xm@renesas.com>
Signed-off-by: NLaurent Pinchart <laurent.pinchart+renesas@ideasonboard.com>

The ODDF signal, output by default on the ODDF pin, isn't used on any
board supported in the kernel. As the Gen3 Salvator-X board uses the
ODDF pin as a DISP signal, hardcode that configuration in the driver.

Use of the ODDF signal will be implemented later through proper DT-based
configuration of the DU pins.
Signed-off-by: NLaurent Pinchart <laurent.pinchart+renesas@ideasonboard.com>

The DE signal is currently configured to be identical to the DISP
signal and is used for the same purpose. To make it clearer that the
DISP pin outputs the DISP signal, select it explicitly.
Signed-off-by: NLaurent Pinchart <laurent.pinchart+renesas@ideasonboard.com>

The Gen3 R8A7795 DU has 4 CRTCs, support them all.
Signed-off-by: NKoji Matsuoka <koji.matsuoka.xm@renesas.com>
Signed-off-by: NLaurent Pinchart <laurent.pinchart+renesas@ideasonboard.com>

LVDS support depends on DRM_RCAR_DU which already depends on OF. Drop
the explicit dependency.
Signed-off-by: NLaurent Pinchart <laurent.pinchart+renesas@ideasonboard.com>

The R8A7795 SoC is ARM64-based and include a DU. Enable driver
compilation on ARM64.
Signed-off-by: NKoji Matsuoka <koji.matsuoka.xm@renesas.com>
Signed-off-by: NLaurent Pinchart <laurent.pinchart+renesas@ideasonboard.com>

Use %tu instead of %u to print difference between pointers.
Signed-off-by: NKoji Matsuoka <koji.matsuoka.xm@renesas.com>
Signed-off-by: NLaurent Pinchart <laurent.pinchart+renesas@ideasonboard.com>

On R-Car Gen3 SoCs the DU lost its ability to access memory directly and
needs to work in conjunction with the VSP to do so. This commit handles
the VSP internally to hide it from the user.
Signed-off-by: NLaurent Pinchart <laurent.pinchart+renesas@ideasonboard.com>

The plane allocator is specific to DU planes and won't be used for
VSP-based planes, move it with the rest of the DU planes code where it
belongs.
Signed-off-by: NLaurent Pinchart <laurent.pinchart+renesas@ideasonboard.com>

Plane sources are configured by the VSPS bit in the PnDDCR4 register.
Although the datasheet states that the bit is updated during vertical
blanking, it seems that updates only occur when the DU group is held in
reset through the DSYSR.DRES bit. Restart the group if the source
changes.
Signed-off-by: NLaurent Pinchart <laurent.pinchart+renesas@ideasonboard.com>

Configure the plane source at plane setup time to source frames from
memory or from the VSP1.
Signed-off-by: NLaurent Pinchart <laurent.pinchart+renesas@ideasonboard.com>

The R8A7790 DU can source frames directly from the VSP1 devices VSPD0
and VSPD1. VSPD0 feeds DU0/1 plane 0, and VSPD1 feeds either DU2 plane 0
or DU0/1 plane 1.

Allocate the correct fixed plane when sourcing frames from VSPD0 or
VSPD1, and allocate planes in reverse index order otherwise to ensure
maximum availability of planes 0 and 1.
Signed-off-by: NLaurent Pinchart <laurent.pinchart+renesas@ideasonboard.com>

Rename plane setup functions using more explicit names, and use local
variables to store intermediate values to increase readability. This
also prepares for VSP compositor support.
Signed-off-by: NLaurent Pinchart <laurent.pinchart+renesas@ideasonboard.com>

There's no need for a read-modify-write pattern, all register bits can
be set explicitly.
Signed-off-by: NLaurent Pinchart <laurent.pinchart+renesas@ideasonboard.com>

A disabled CRTC can't display planes, the driver shouldn't try to
configure it when updating planes. The DRM core will store the plane
state for us, and the plane will be configured appropriately the next
time the CRTC is enabled.
Signed-off-by: NLaurent Pinchart <laurent.pinchart+renesas@ideasonboard.com>

The atomic framework guarantees that the CRTC enable and disable
functions will only be called when needed, there's no need to duplicate
the CRTC state check. By replacing the local CRTC enabled state check at
resume time with the CRTC active state we can remove the local CRTC
enabled state altogether.
Signed-off-by: NLaurent Pinchart <laurent.pinchart+renesas@ideasonboard.com>

The drm driver .load() operation is prone to race conditions as it
initializes the driver after registering the device nodes. Its usage is
deprecated, inline it in the probe function and call drm_dev_alloc() and
drm_dev_register() explicitly.

For consistency inline the .unload() handler in the remove function as
well.
Signed-off-by: NLaurent Pinchart <laurent.pinchart+renesas@ideasonboard.com>
Acked-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

The Innosilicon HDMI is a low power HDMI 1.4 transmitter
IP, and it have been integrated on some rockchip CPUs
(like RK3036, RK312x).
Signed-off-by: NYakir Yang <ykk@rock-chips.com>
Tested-by: NHeiko Stuebner <heiko@sntech.de>

rbtree_postorder_for_each_entry_safe can skip over some entries if
the tree is rebalanced in interval_tree_remove. interval_tree_remove
is also redundant when the tree is just about to be freed.
Signed-off-by: NFelix Kuehling <Felix.Kuehling@amd.com>
Reviewed-by: NChristian König <christian.koenig@amd.com>

Exchange locking order of adev->mn_lock and mm_sem, so that
rmn->mm->mmap_sem can be taken safely, protected by adev->mn_lock,
when amdgpu_mn_destroy runs concurrently.
Signed-off-by: NFelix Kuehling <Felix.Kuehling@amd.com>
Reviewed-by: NChristian König <christian.koenig@amd.com>

Macro module_usb_driver is used for drivers whose init and exit paths
only register and unregister to usb API. So remove boilerplate code to
make code simpler by using module_usb_driver.

This change was made with the help of the following Coccinelle
semantic patch:

//<smpl>
@a@
identifier f, x;
@@
-static f(...) { return usb_register(&x); }

@b depends on a@
identifier e, a.x;
@@
-static e(...) { usb_deregister(&x); }

@c depends on a && b@
identifier a.f;
declarer name module_init;
@@
-module_init(f);

@d depends on a && b && c@
identifier b.e, a.x;
declarer name module_exit;
declarer name module_usb_driver;
@@
-module_exit(e);
+module_usb_driver(x);
//</smpl>
Signed-off-by: NAmitoj Kaur Chawla <amitoj1606@gmail.com>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>
Link: http://patchwork.freedesktop.org/patch/msgid/20160217121327.GA29682@amitoj-Inspiron-3542

Remove the double housekeeping and use something sane to
forcefuly delete BOs on unload.
Signed-off-by: NChristian König <christian.koenig@amd.com>
Reviewed-by: NAlex Deucher <alexander.deucher@amd.com>

Print BOs grouped per client.
Signed-off-by: NChristian König <christian.koenig@amd.com>
Reviewed-by: NAlex Deucher <alexander.deucher@amd.com>

To easily find which memory is used.
Signed-off-by: NChristian König <christian.koenig@amd.com>
Reviewed-by: NAlex Deucher <alexander.deucher@amd.com>

Usefull when debugging page flipping.
Signed-off-by: NChristian König <christian.koenig@amd.com>
Reviewed-by: NAlex Deucher <alexander.deucher@amd.com>

Splitting it into KB/MB is just confusing.
Signed-off-by: NChristian König <christian.koenig@amd.com>
Reviewed-by: NAlex Deucher <alexander.deucher@amd.com>

Not sure why somebody thought that this is a long.
Signed-off-by: NChristian König <christian.koenig@amd.com>
Reviewed-by: NAlex Deucher <alexander.deucher@amd.com>

Just use the system queue now that we don't block any more.

v2: handle DAL as well.
v3: agd: split DAL changes out
Signed-off-by: NChristian König <christian.koenig@amd.com>
Reviewed-by: NMykola Lysenko <mykola.lysenko@amd.com>
Acked-by: Alex Deucher <alexander.deucher@amd.com> (v1)

Just register an callback and reschedule the work item if necessary.
Signed-off-by: NChristian König <christian.koenig@amd.com>
Acked-by: NAlex Deucher <alexander.deucher@amd.com>

We don't return -EDEADLK any more.
Signed-off-by: NChristian König <christian.koenig@amd.com>
Acked-by: NAlex Deucher <alexander.deucher@amd.com>

wait_event() never returns before the fence was signaled.
Signed-off-by: NChristian König <christian.koenig@amd.com>
Acked-by: NAlex Deucher <alexander.deucher@amd.com>

The release notifier can get called a second time from
mmu_notifier_unregister depending on a race between
__mmu_notifier_release and amdgpu_mn_destroy. Use
mmu_notifier_unregister_no_release to avoid this.
Signed-off-by: NFelix Kuehling <Felix.Kuehling@amd.com>
Reviewed-by: NChristian König <christian.koenig@amd.com>

This supports 420 and 422 subsampling with 2 or 3 planes, tested with
modetest.  It doesn't set up chroma subsampling position (which it
appears KMS doesn't deal with yet).

The LBM memory is overallocated in many cases, but apparently the docs
aren't quite correct and I'll probably need to look at the hardware
source to really figure it out.
Signed-off-by: NEric Anholt <eric@anholt.net>

These were all touch-tested with modetest.
Signed-off-by: NEric Anholt <eric@anholt.net>

This implements a simple policy for choosing scaling modes
(trapezoidal for decimation, PPF for magnification), and a single PPF
filter (Mitchell/Netravali's recommendation).
Signed-off-by: NEric Anholt <eric@anholt.net>

This doesn't matter yet since we only allow 1:1 scaling, but the
comment clearly says we should be using the source size.
Signed-off-by: NEric Anholt <eric@anholt.net>

Previously we only did the primary and cursor plane, but overlay
planes are useful and just require this setup to add, since all planes
go into the HVS display list in the same way.
Signed-off-by: NEric Anholt <eric@anholt.net>

So far, we've only ever lit up one CRTC, so this has been fine.  To
extend to more displays or more planes, we need to make sure we don't
run our display lists into each other.
Signed-off-by: NEric Anholt <eric@anholt.net>

Previously, on every modeset we would allocate new display list
memory, recompute changed planes, write all of them to the new memory,
and pointed scanout at the new list (which will latch approximately at
the next line of scanout).  We let
drm_atomic_helper_wait_for_vblanks() decide whether we needed to wait
for a vblank after a modeset before cleaning up the old state and
letting the next modeset proceed, and on legacy cursor updates we
wouldn't wait.  If you moved the cursor fast enough, we could
potentially wrap around the display list memory area and overwrite the
existing display list while it was still being scanned out, resulting
in the HVS scanning out garbage or just halting.

Instead of making cursor updates wait for scanout to move to the new
display list area (which introduces significant cursor lag in X), we
just rewrite our current display list.
Signed-off-by: NEric Anholt <eric@anholt.net>

As we add actual scaling, this is going to get way more complicated.
Signed-off-by: NEric Anholt <eric@anholt.net>