Commit eab3bbef ("drm/atomic: Add drm_crtc_state->active") added the
field to track the DPMS state. However, the Tegra driver was in modified
in parallel and subclasses the CRTC atomic state, so needed to duplicate
the code in the atomic helpers. After the addition of the active_changed
field it became out of sync and doesn't reset it when duplicating state.

This causes a full modeset on things like page-flips, which will in turn
cause warnings due to the VBLANK machinery being disabled when it really
should remain on.
Tested-by: NTomeu Vizoso <tomeu.vizoso@collabora.com>
Signed-off-by: NThierry Reding <treding@nvidia.com>

Recent changes in the clock framework have caused a behavioural change
in that clocks that have not had their rate set explicitly will now be
reset to their initial rate (or 0) when the clock is released. This is
triggered in the deferred probing path, resulting in the clock running
at a wrong frequency after the successful probe.

This can be easily fixed by setting the rate explicitly rather than by
relying on the implicit rate inherited by the parent.
Tested-by: NTomeu Vizoso <tomeu.vizoso@collabora.com>
Signed-off-by: NThierry Reding <treding@nvidia.com>

Missing parameter when fetching the real voltage values
from atom.  Fixes problems with dynamic clocking on
certain boards.

bug:
https://bugs.freedesktop.org/show_bug.cgi?id=87457Signed-off-by: NAlex Deucher <alexander.deucher@amd.com>
Cc: stable@vger.kernel.org

Don't restrict it to just eDP panels.  Some LVDS bridge chips require
this.  Fixes blank panels on resume on certain laptops.  Noticed
by mrnuke on IRC.

bug:
https://bugs.freedesktop.org/show_bug.cgi?id=42960Signed-off-by: NAlex Deucher <alexander.deucher@amd.com>
Cc: stable@vger.kernel.org

Enable at init and disable on fini. Workaround for hardware problems.

v2 (chk): extend commit message
v3: add new function
Signed-off-by: NAlex Deucher <alexander.deucher@amd.com>
Signed-off-by: Christian König <christian.koenig@amd.com> (v2)
Cc: stable@vger.kernel.org

Emit the EOP twice to avoid cache flushing problems.
Signed-off-by: NChristian König <christian.koenig@amd.com>
Cc: stable@vger.kernel.org
Signed-off-by: NAlex Deucher <alexander.deucher@amd.com>

Doing so can cause things to become slow.

Print a warning at compile time and an informative message at runtime in
that case.

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=88758
Cc: stable@vger.kernel.org
Reviewed-by: NChristian König <christian.koenig@amd.com>
Signed-off-by: NMichel Dänzer <michel.daenzer@amd.com>
Signed-off-by: NAlex Deucher <alexander.deucher@amd.com>

0-255 seems to be the preferred range for the pwm interface.
Reviewed-by: NChristian König <christian.koenig@amd.com>
Signed-off-by: NAlex Deucher <alexander.deucher@amd.com>

The efficient frequency (RPe) should stay in the range
RPn <= RPe <= RP0.  The pcode clamps the returned value
internally on Broadwell but not on Haswell.

Fix for missing range check in
commit 93ee2920
Author: Tom O'Rourke <Tom.O'Rourke@intel.com>
Date:   Wed Nov 19 14:21:52 2014 -0800

    drm/i915: Use efficient frequency for HSW/BDW

Reference: http://lists.freedesktop.org/archives/intel-gfx/2015-February/059802.htmlReported-by: NMichael Auchter <a@phire.org>
Suggested-by: NChris Wilson <chris@chris-wilson.co.uk>
Cc: stable@vger.kernel.org # v3.19
Signed-off-by: NTom O'Rourke <Tom.O'Rourke@intel.com>
Signed-off-by: NJani Nikula <jani.nikula@intel.com>

Return IRQ_HANDLED from intel_dp_hpd_pulse() to properly
ignore the long HPD pulse on eDP to avoid the never ending
VDD off->HPD->VDD on->VDD off->HPD... cycle.

This fixes a regression intoduced by
 commit b2c5c181
 Author: Daniel Vetter <daniel.vetter@ffwll.ch>
 Date:   Fri Jan 23 06:00:31 2015 +0100

    drm/i915: Use symbolic irqreturn for ->hpd_pulse

Cc: Daniel Vetter <daniel.vetter@ffwll.ch>
Cc: Jani Nikula <jani.nikula@linux.intel.com>
Reviewed-by: NDaniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: NVille Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: NJani Nikula <jani.nikula@intel.com>

This patch is based on exynos-drm-next branch of Inki Dae's tree at:
git://git.kernel.org/pub/scm/linux/kernel/git/daeinki/drm-exynos.git

DECON(Display and Enhancement Controller) is the new IP
in exynos7 SOC for generating video signals using pixel data.

DECON driver can be used to drive 2 different interfaces on Exynos7:
DECON-INT(video controller) and DECON-EXT(Mixer for HDMI)

The existing FIMD driver code was used as a template to create
DECON driver. Only DECON-INT is supported as of now, and
DECON-EXT support will be added later.

The current version of the driver supports video mode displays.

Changelog v2:
- Change config name, DRM_EXYNOS_DECON to DRM_EXYNOS7_DECON.
Signed-off-by: NAkshu Agrawal <akshua@gmail.com>
Signed-off-by: NAjay Kumar <ajaykumar.rs@samsung.com>
Signed-off-by: NInki Dae <inki.dae@samsung.com>

LP_OUTPUT_HOLD is only in MIPI_PORT_CTRL(PORT_A) even for PORT_C in case
of dual link. In the dual link implementation, the bit is correctly set
or unset for hardcoded PORT_A, but for bit update the register base value
is read by using MIPI_PORT_CTRL(port) in a loop. The second iteration will
read base value from PORT_C and program for PORT_A. Mostly in case of dual
link all other bit values should be same, but logically we should read from
PORT_A. So hardcode to read initial value from PORT_A as well.
Signed-off-by: NShobhit Kumar <shobhit.kumar@intel.com>
Signed-off-by: NJani Nikula <jani.nikula@intel.com>

This looked like an odd regression from

commit ec5cc0f9
Author: Chris Wilson <chris@chris-wilson.co.uk>
Date:   Thu Jun 12 10:28:55 2014 +0100

    drm/i915: Restrict GPU boost to the RCS engine

but in reality it undercovered a much older coherency bug. The issue that
boosting the GPU frequency on the BCS ring was masking was that we could
wake the CPU up after completion of a BCS batch and inspect memory prior
to the write cache being fully evicted. In order to serialise the
breadcrumb interrupt (and so ensure that the CPU's view of memory is
coherent) we need to perform a post-sync operation in the MI_FLUSH_DW.

v2: Fix all the MI_FLUSH_DW (bsd plus the duplication in execlists).

Also fix the invalidate_domains mask in gen8_emit_flush() for ring !=
VCS.

Testcase: gpuX-rcs-gpu-read-after-write
Signed-off-by: NChris Wilson <chris@chris-wilson.co.uk>
Cc: stable@vger.kernel.org
Acked-by: NDaniel Vetter <daniel@ffwll.ch>
Signed-off-by: NJani Nikula <jani.nikula@intel.com>

Nothing in Bspec seems to indicate that we actually needs this, and it
looks like can't work since by this point the pipe is off and so
vblanks won't really happen any more.

Note that Bspec mentions that it takes a vblank for this bit to
change, but _only_ when enabling.

Dropping this code quenches an annoying backtrace introduced by the
more anal checking since

commit 51e31d49
Author: Daniel Vetter <daniel.vetter@ffwll.ch>
Date:   Mon Sep 15 12:36:02 2014 +0200

    drm/i915: Use generic vblank wait

Note: This fixes the fallout from the above commit, but does not address
the shortcomings of the IBX transcoder select workaround implementation
discussed during review [1].

[1] http://mid.gmane.org/87y4o7usxf.fsf@intel.com

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=86095Signed-off-by: NDaniel Vetter <daniel.vetter@intel.com>
Reviewed-by: NPaulo Zanoni <paulo.r.zanoni@intel.com>
Cc: stable@vger.kernel.org # 3.19
Signed-off-by: NJani Nikula <jani.nikula@intel.com>

There is a case called disable_plane callback function even if
plane->crtc is NULL from exynos_drm_encoder_disable and it will cause
NULL pointer reference error.
Signed-off-by: NJoonyoung Shim <jy0922.shim@samsung.com>
Signed-off-by: NInki Dae <inki.dae@samsung.com>

The exynos_plane_dpms function handles enabled flag of exynos plane and
calls internal hw driver callback function for hw overlay on/off. But
it causes state disharmory problem currently and is will be obstacle to
apply atomic operation later to keep non-standard per-plane dpms state
like enabled flag.

Let's remove enabled flag, it just stop to recall internal callback
function but hw drivers can handle it properly. And call internal
callback function directly then we can remove unnecessary
exynos_plane_dpms function
Signed-off-by: NJoonyoung Shim <jy0922.shim@samsung.com>
Signed-off-by: NInki Dae <inki.dae@samsung.com>

This was added by commit 3b8d1cf8 ("drm/exynos: add property for
crtc mode"). Currently we can control a plane used for crtc using
primary plane by universal plane feature. Stop to use non-standard
property to control primary plane.
Signed-off-by: NJoonyoung Shim <jy0922.shim@samsung.com>
Signed-off-by: NInki Dae <inki.dae@samsung.com>

exynos_plane_dpms(DRM_MODE_DPMS_ON) calls the win_enable()'s callback
from the underlying layer. However neither one of these layers implement
win_enable() - FIMD, Mixer and VIDI. Thus the call to exynos_plane_dpms()
is pointless.
Signed-off-by: NGustavo Padovan <gustavo.padovan@collabora.co.uk>
Acked-by: NDaniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: NInki Dae <inki.dae@samsung.com>

We added this WARN_ON to guard against using uninitialized
forcewake domains. But forgot blissfully that not all
gens have forcewake domains in the first place.

v2: Move WARN_ON to fw_domains_init (Chris)

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=88911
Tested-by: Ding Heng <hengx.ding@intel.com> (v1)
Signed-off-by: NMika Kuoppala <mika.kuoppala@intel.com>
Reviewed-by: NChris Wilson <chris@chris-wilson.co.uk>
[Jani: add comment above WARN_ON as suggested by Chris]
Signed-off-by: NJani Nikula <jani.nikula@intel.com>

We read the coherent current seqno and actual head from ring.
For hardware access we need to take runtime_pm reference.

Get hardware specific values with runtime reference held
and print them first to emphasize hw state vs bookkeepping.

v2: Reorder output according to hw access (Chris)
    remove superfluous locking (Daniel)

Testcase: igt/pm_rpm/debugfs-read
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=88910
Tested-by: Ding Heng <hengx.ding@intel.com> (v1)
Signed-off-by: NMika Kuoppala <mika.kuoppala@intel.com>
Reviewed-by: NChris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: NJani Nikula <jani.nikula@intel.com>

As per the specififcation, the SB_DevFn is the PCI_DEVFN of the target
device and not the source. So PCI_DEVFN(2,0) is not correct. Further the
port ID should be enough to identify devices unless they are MFD. The
SB_DevFn was intended to remove ambiguity in case of these MFD devices.

For non MFD devices the recommendation for the target device IP was to
ignore these fields, but not all of them followed the recommendation.
Some like CCK ignore these fields and hence PCI_DEVFN(2, 0) works and so
does PCI_DEVFN(0, 0) as it works for DPIO. The issue came to light because
of GPIONC which was not getting programmed correctly with PCI_DEVFN(2, 0).
It turned out that this did not follow the recommendation and expected 0
in this field.

In general the recommendation is to use SB_DevFn as PCI_DEVFN(0, 0) for
all devices except target PCI devices.
Signed-off-by: NShobhit Kumar <shobhit.kumar@intel.com>
Reviewed-by: NVille Syrjälä <ville.syrjala@linux.intel.com>
Cc: stable@vger.kernel.org
Signed-off-by: NJani Nikula <jani.nikula@intel.com>

The Exynos DRM driver doesn't follow the correct API when dealing with
dma_{alloc, mmap, free}_attrs functions and the
DMA_ATTR_NO_KERNEL_MAPPING attribute.

When a IOMMU is not available and the DMA_ATTR_NO_KERNEL_MAPPING is
used, the driver should use the pointer returned by dma_alloc_attr() as
a cookie.

The Exynos DRM driver directly uses the non-requested virtual kernel
address returned by the DMA mapping subsystem. This just works now
because the non-IOMMU codepath doesn't obey DMA_ATTR_NO_KERNEL_MAPPING
but we need to fix it before fixing the DMA layer.
Signed-off-by: NCarlo Caione <carlo@caione.org>
Acked-by: NMarek Szyprowski <m.szyprowski@samsung.com>
Acked-by: NJoonyoung Shim <jy0922.shim@samsung.com>
Signed-off-by: NInki Dae <inki.dae@samsung.com>

For default graphic window, mixer_win_commit() sets display size
register as fb size. Calling setplane with smaller fb size than
mode size to default window causes distorted display result. So
this patch replaces fb size with mode size for display size from
the mixer_win_commit().
Signed-off-by: NSeung-Woo Kim <sw0312.kim@samsung.com>
Acked-by: NJoonyoung Shim <jy0922.shim@samsung.com>
Signed-off-by: NInki Dae <inki.dae@samsung.com>

The hdmi outputs black screen only even though under the hood Xorg and
framebuffer console  are fine : devices found and initialized, but
not a pixel out.

Commit 93bca243 ("drm/exynos: remove struct exynos_drm_manager")
changed the call order of mixer_initialize with regards to
 exynos_drm_crtc_create.

This changes breaks hdmi out on Odroid U2 (linux-next with added
 Marek Szyprowski v4 hdmi patchset from linux-samsung-soc ML).

Restore the previous call ordering get hdmi to ouput proper pixels:
ie call mixer_initialize first then exynos_drm_crtc_create.

Fixes: 93bca243 ("drm/exynos: remove struct exynos_drm_manager")
Signed-off-by: NAlban Browaeys <prahal@yahoo.com>
Signed-off-by: NInki Dae <inki.dae@samsung.com>

Use driver internal struct as argument instead of struct exynos_drm_crtc
except functions of exynos_drm_crtc_ops and instead of struct
exynos_drm_display except functions of exynos_drm_display_ops.

It can reduce unnecessary variable declaration.
Signed-off-by: NJoonyoung Shim <jy0922.shim@samsung.com>
Signed-off-by: NInki Dae <inki.dae@samsung.com>

We get wrong pipe value for crtc since commit 93bca243 ("drm/exynos:
remove struct exynos_drm_manager"). We should should increase pipe value
before call exynos_drm_crtc_create.
Signed-off-by: NJoonyoung Shim <jy0922.shim@samsung.com>
Signed-off-by: NInki Dae <inki.dae@samsung.com>

The exynos_drm_dmabuf.c file doesn't include any module feature and it
isn't built to module.
Signed-off-by: NJoonyoung Shim <jy0922.shim@samsung.com>
Reviewed-by: NGustavo Padovan <gustavo.padovan@collabora.co.uk>
Signed-off-by: NInki Dae <inki.dae@samsung.com>

The exynos drm driver has DRIVER_PRIME capability, then it's reasonable
to support dmabuf as default. Remove DRM_EXYNOS_DMABUF config, it will
prevent that user selects the option unnecessarily.
Signed-off-by: NJoonyoung Shim <jy0922.shim@samsung.com>
Reviewed-by: NGustavo Padovan <gustavo.padovan@collabora.co.uk>
Signed-off-by: NInki Dae <inki.dae@samsung.com>

If system provides IOMMU feature, Exynos DRM should use it by default,
because the Exynos DRM subdrivers don't work correctly when Exynos IOMMU
driver has been enabled and no IOMMU support has been compiled into Exynos
DRM driver.
Signed-off-by: NMarek Szyprowski <m.szyprowski@samsung.com>
Signed-off-by: NInki Dae <inki.dae@samsung.com>

Mixed need to have hdmi clock enabled to properly perform power on/off
sequences, so add handling of this clock directly to the mixer driver.
Dependency between hdmi clock and mixer module has been observed on
Exynos4 based boards.
Suggested-by: NAndrzej Hajda <a.hajda@samsung.com>
Reviewed-by: NJavier Martinez Canillas <javier.martinez@collabora.co.uk>
Tested-by: NJavier Martinez Canillas <javier.martinez@collabora.co.uk>
Signed-off-by: NMarek Szyprowski <m.szyprowski@samsung.com>
Signed-off-by: NInki Dae <inki.dae@samsung.com>

Add a default audio infoframe for HDMI compliance
Signed-off-by: NArnaud Pouliquen <arnaud.pouliquen@st.com>

It's possible for invalidate_range_start mmu notifier callback to race
against userptr object release. If the gem object was released prior to
obtaining the spinlock in invalidate_range_start we're hitting null
pointer dereference.

Testcase: igt/gem_userptr_blits/stress-mm-invalidate-close
Testcase: igt/gem_userptr_blits/stress-mm-invalidate-close-overlap
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: NMichał Winiarski <michal.winiarski@intel.com>
Reviewed-by: NChris Wilson <chris@chris-wilson.co.uk>
Cc: stable@vger.kernel.org
[Jani: added code comment suggested by Chris]
Signed-off-by: NJani Nikula <jani.nikula@intel.com>

Use GDP capabilities to support DRM_FORMAT_XBGR8888 (XB24)
Signed-off-by: NFabien Dessenne <fabien.dessenne@st.com>

Use GDP capabilities to support DRM_FORMAT_ABGR8888 (AB24)
Signed-off-by: NBenjamin Gaignard <benjamin.gaignard@linaro.org>

The shift and the mask done on arg value is useless
since arg is null.
Signed-off-by: NVincent Abriou <vincent.abriou@st.com>

copy-paste wasn't followed by editing, do it.
Signed-off-by: NJassi Brar <jaswinder.singh@linaro.org>

Instead of manual calls of device_create_file() and
device_remove_file(), assign the static attribute groups to the device
with device_create_with_groups().  The conditionally built sysfs
entries are handled via is_visible callback.

This simplifies the code and also avoids the possible races.
Signed-off-by: NTakashi Iwai <tiwai@suse.de>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

So this has been merged originally in

commit 83052d4d
Author: Seung-Woo Kim <sw0312.kim@samsung.com>
Date:   Thu Dec 15 15:40:55 2011 +0900

    drm: Add multi buffer plane pixel formats

which hasn't seen a lot of review really. The problem is that it's not
a real pixel format, but just a different way to lay out NV12 pixels
in macroblocks, i.e. a tiling format.

The new way of doing this is with the soon-to-be-merged fb modifiers.

This was brough up in some long irc discussion around the entire
topic, as an example of where things have gone wrong. Luckily we can
correct the mistake:
- The kms side support for NV12MT is all dead code because
  format_check in drm_crtc.c never accepted NV12MT.
- The gem side for the gsc support doesn't look better: The code
  forgets to set the pixel format and makes a big mess with the tiling
  mode bits, inadvertedly setting them all.

Conclusion: This never really worked (at least not in upstream) and
hence we can safely correct our mistake here.

Cc: Seung-Woo Kim <sw0312.kim@samsung.com>
Cc: Inki Dae <inki.dae@samsung.com>
Cc: Kyungmin Park <kyungmin.park@samsung.com>
Cc: Rob Clark <robclark@freedesktop.org>
Cc: Daniel Stone <daniel@fooishbar.org>
Cc: Damien Lespiau <damien.lespiau@intel.com>
Reviewed-by: NRob Clark <robclark@freedesktop.org>
Reviewed-by: NGustavo Padovan <gustavo.padovan@collabora.co.uk>
Acked-by: NJoonyoung Shim <jy0922.shim@samsung.com>
Acked-by: NSeung-Woo Kim <sw0312.kim@samsung.com>
Signed-off-by: NDaniel Vetter <daniel.vetter@intel.com>

You can _never_ assert that a lock is not held, except in some very
restricted corner cases where it's guranteed that your code is running
single-threade (e.g. driver load before you've published any pointers
leading to that lock).

In addition the early return breaks a bunch of testcases since with
highly concurrent hangcheck stress tests the reset fails to work and
the test doesn't recover and time out.

This regression has been introduced in

commit b8d24a06
Author: Mika Kuoppala <mika.kuoppala@linux.intel.com>
Date:   Wed Jan 28 17:03:14 2015 +0200

    drm/i915: Remove nested work in gpu error handling

Aside: It is possible to check whether a given task doesn't hold a
lock, but only when lockdep is enabled, using the lockdep_assert_held
stuff.

Cc: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Mika Kuoppala <mika.kuoppala@intel.com>
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=88908Reviewed-by: NMika Kuoppala <mika.kuoppala@intel.com>
Signed-off-by: NDaniel Vetter <daniel.vetter@intel.com>

Currently when a mode is rejected the reason is printed as a raw number.
Having to manually decode that to a enum drm_mode_status value is
tiresome. Have the code do the decoding instead and print the result
in a human readable format.

Just having an array of strings indexed with the mode status doesn't
work since the enum includes negative values. So we offset the status
by +3 which makes all the indexes non-negative. Also add a bit of
paranoia into the code to catch out of bounds accesses in case
someone adds more enum values but forgets to update the code.
Signed-off-by: NVille Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: NAlex Deucher <alexander.deucher@amd.com>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>