drm/tegra: Fix crash caused by reference count imbalance

Commit d2307dea14a4 ("drm/atomic: use connector references (v3)") added
reference counting for DRM connectors and this caused a crash when
exercising system suspend on Tegra114 Dalmore.

The Tegra DSI driver implements a Tegra specific function,
tegra_dsi_connector_duplicate_state(), to duplicate the connector state
and destroys the state using the generic helper function,
drm_atomic_helper_connector_destroy_state(). Following commit
d2307dea14a4 ("drm/atomic: use connector references (v3)") there is
now an imbalance in the connector reference count because the Tegra
function to duplicate state does not take a reference when duplicating
the state information. However, the generic helper function to destroy
the state information assumes a reference has been taken and during
system suspend, when the connector state is destroyed, this leads to a
crash because we attempt to put the reference for an object that has
already been freed.

Fix this by calling __drm_atomic_helper_connector_duplicate_state() from
tegra_dsi_connector_duplicate_state() to ensure that we take a reference
on a connector if crtc is set. Note that this will also copy the
connector state a 2nd time, but this should be harmless.

By fixing tegra_dsi_connector_duplicate_state() to take a reference,
although a crash was no longer seen, it was then observed that after
each system suspend-resume cycle, the reference would be one greater
than before the suspend-resume cycle. Following commit d2307dea14a4
("drm/atomic: use connector references (v3)"), it was found that we
also need to put the reference when calling the function
tegra_dsi_connector_reset() before freeing the state. Fix this by
updating tegra_dsi_connector_reset() to call the function
__drm_atomic_helper_connector_destroy_state() in order to put the
reference for the connector.

Fixes: d2307dea14a4 ("drm/atomic: use connector references (v3)")

Signed-off-by: Jon Hunter <jonathanh@nvidia.com>
Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Acked-by: Thierry Reding <treding@nvidia.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Link: http://patchwork.freedesktop.org/patch/msgid/1463585856-16606-1-git-send-email-jonathanh@nvidia.com

drm: Fix error handling in drm_connector_register

When debugfs or sysfs registration failed, we failed to clean up the
idr registration. Reorder to fix this.

Cc: Dave Airlie <airlied@gmail.com>
Signed-off-by: Daniel Vetter <daniel.vetter@intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Link: http://patchwork.freedesktop.org/patch/msgid/1462539302-27764-1-git-send-email-daniel.vetter@ffwll.ch

drm: Avoid connector reference imbalance on error path

Whilst looking at the fallout from using connector references for
atomic, I noticed that there is an early return buried in
drm_atomic_set_crtc_for_connector() that if hit could cause us to leak a
reference on the connector.

Fixes: d2307dea14 (drm/atomic: use connector references (v3))
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Daniel Stone <daniels@collabora.com>
Cc: Daniel Vetter <daniel.vetter@ffwll.ch>
Cc: Dave Airlie <airlied@redhat.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Link: http://patchwork.freedesktop.org/patch/msgid/1462535265-13058-1-git-send-email-chris@chris-wilson.co.uk

drm/fb_helper: Fix references to dev->mode_config.num_connector

During boot, MST hotplugs are generally expected (even if no physical
hotplugging occurs) and result in DRM's connector topology changing.
This means that using num_connector from the current mode configuration
can lead to the number of connectors changing under us. This can lead to
some nasty scenarios in fbcon:

- We allocate an array to the size of dev->mode_config.num_connectors.
- MST hotplug occurs, dev->mode_config.num_connectors gets incremented.
- We try to loop through each element in the array using the new value
  of dev->mode_config.num_connectors, and end up going out of bounds
  since dev->mode_config.num_connectors is now larger then the array we
  allocated.

fb_helper->connector_count however, will always remain consistent while
we do a modeset in fb_helper.

Note: This is just polish for 4.7, Dave Airlie's drm_connector
refcounting fixed these bugs for real. But it's good enough duct-tape
for stable kernel backporting, since backporting the refcounting
changes is way too invasive.

Cc: stable@vger.kernel.org
Signed-off-by: Lyude <cpaul@redhat.com>
[danvet: Clarify why we need this. Also remove the now unused "dev"
local variable to appease gcc.]
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Link: http://patchwork.freedesktop.org/patch/msgid/1463065021-18280-3-git-send-email-cpaul@redhat.com

drm/i915/fbdev: Fix num_connector references in intel_fb_initial_config()

During boot time, MST devices usually send a ton of hotplug events
irregardless of whether or not any physical hotplugs actually occurred.
Hotplugs mean connectors being created/destroyed, and the number of DRM
connectors changing under us. This isn't a problem if we use
fb_helper->connector_count since we only set it once in the code,
however if we use num_connector from struct drm_mode_config we risk it's
value changing under us. On top of that, there's even a chance that
dev->mode_config.num_connector != fb_helper->connector_count. If the
number of connectors happens to increase under us, we'll end up using
the wrong array size for memcpy and start writing beyond the actual
length of the array, occasionally resulting in kernel panics.

Note: This is just polish for 4.7, Dave Airlie's drm_connector
refcounting fixed these bugs for real. But it's good enough duct-tape
for stable kernel backporting, since backporting the refcounting
changes is way too invasive.

Cc: stable@vger.kernel.org
Signed-off-by: Lyude <cpaul@redhat.com>
[danvet: Clarify why we need this.]
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Link: http://patchwork.freedesktop.org/patch/msgid/1463065021-18280-2-git-send-email-cpaul@redhat.com

qxl: catch qxlfb_create_pinned_object failures

Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Link: http://patchwork.freedesktop.org/patch/msgid/1463072816-11788-1-git-send-email-kraxel@redhat.com

drm/exynos/hdmi: add a missing tab

Smatch warns that the if statement isn't indented.

Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Link: http://patchwork.freedesktop.org/patch/msgid/20160512195457.GA19095@mwanda

drm/fb-cma-helper: Add function drm_fb_cma_create_with_funcs()

Add drm_fb_cma_create_with_funcs() for drivers that need to set the
dirty() callback.

Signed-off-by: Noralf Trønnes <noralf@tronnes.org>
Acked-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Link: http://patchwork.freedesktop.org/patch/msgid/1463077523-23959-3-git-send-email-noralf@tronnes.org

drm/fb-cma-helper: Use const for drm_framebuffer_funcs argument

drm_framebuffer_init() uses const for the drm_framebuffer_funcs
argument so use that on drm_fb_cma_alloc() and
drm_fbdev_cma_create_with_funcs() as well.

Cc: laurent.pinchart@ideasonboard.com
Signed-off-by: Noralf Trønnes <noralf@tronnes.org>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Link: http://patchwork.freedesktop.org/patch/msgid/1463077523-23959-2-git-send-email-noralf@tronnes.org

drm: Drop connector argument from __drm_atomic_helper_connector_destroy_state

It's unused, and really this helper should only look at the state
structure and nothing else. Note that this conflicts with a patch from
Dave that adds refcounting to drm_connectors. It's not yet clear
whether the check Dave adds for connector != NULL is really needed or
the right check.

v2: Fix commmit message (Laurent).

Cc: Dave Airlie <airlied@gmail.com>
Acked-by: Thierry Reding <thierry.reding@gmail.com>
Acked-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Acked-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: Daniel Vetter <daniel.vetter@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Link: http://patchwork.freedesktop.org/patch/msgid/1462804451-15318-3-git-send-email-daniel.vetter@ffwll.ch

drm: Drop plane argument from __drm_atomic_helper_plane_destroy_state

It's unused, and really this helper should only look at the state
structure and nothing else.

v2: Fix commit message (Laurent).

v3: Rebase onto mtk driver merge.

Cc: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Cc: Thierry Reding <thierry.reding@gmail.com>
Cc: Eric Anholt <eric@anholt.net>
Cc: Laurent Pinchart <laurent.pinchart+renesas@ideasonboard.com>
Cc: Inki Dae <inki.dae@samsung.com>
Cc: Tomi Valkeinen <tomi.valkeinen@ti.com>
Cc: Mark Yao <mark.yao@rock-chips.com>
Acked-by: Thierry Reding <thierry.reding@gmail.com>
Acked-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Acked-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Acked-by: Eric Anholt <eric@anholt.net>
Signed-off-by: Daniel Vetter <daniel.vetter@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Link: http://patchwork.freedesktop.org/patch/msgid/1462804451-15318-2-git-send-email-daniel.vetter@ffwll.ch

drm: Drop crtc argument from __drm_atomic_helper_crtc_destroy_state

It's unused, and really this helper should only look at the state
structure and nothing else.

v2: Rebase on top of rockchip changes

v3: Drop unrelated hunk, spotted by Laurent.

v4: Rebase onto mtk driver merge.

Cc: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Cc: Thierry Reding <thierry.reding@gmail.com>
Cc: Eric Anholt <eric@anholt.net>
Cc: Mark Yao <mark.yao@rock-chips.com>
Acked-by: Thierry Reding <thierry.reding@gmail.com>
Acked-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Acked-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Acked-by: Eric Anholt <eric@anholt.net>
Signed-off-by: Daniel Vetter <daniel.vetter@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Link: http://patchwork.freedesktop.org/patch/msgid/1462804451-15318-1-git-send-email-daniel.vetter@ffwll.ch

drm: Remove unused drm_device from drm_gem_object_lookup()

drm_gem_object_lookup() has never required the drm_device for its file
local translation of the user handle to the GEM object. Let's remove the
unused parameter and save some space.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: dri-devel@lists.freedesktop.org
Cc: Dave Airlie <airlied@redhat.com>
Cc: Daniel Vetter <daniel.vetter@ffwll.ch>
[danvet: Fixup kerneldoc too.]
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

Merge tag 'topic/drm-misc-2016-05-13' of git://anongit.freedesktop.org/drm-intel into drm-next

I kinda hoped that I could still sneak in Noralf's
drm_simple_display_pipe, since there's intereset by others now (for tilcdc
at least). But it wasn't ready by a hair. Oh well.

Otherwise random stuff plus prep patches from Noralf.

* tag 'topic/drm-misc-2016-05-13' of git://anongit.freedesktop.org/drm-intel:
  drm/atomic: Add drm_atomic_helper_best_encoder()
  drm/atomic: Don't skip drm_bridge_*() calls if !drm_encoder_helper_funcs
  drm/fb-cma-helper: Hook up to DocBook and fix some docs
  drm/fb-helper: Remove mention of CONFIG_FB_DEFERRED_IO in docs
  drm/sti: include linux/seq_file.h where needed
  drm/tegra: Use lockless gem BO free callback
  drm/exynos: Use lockless gem BO free callback
  drm: Make drm_encoder_helper_funcs optional

Merge branch 'topic-arcpgu-updates' of https://github.com/foss-for-synopsys-dwc-arc-processors/linux into drm-next

Please pull this mini-series that allows ARC PGU to use
dedicated memory location as framebuffer backing storage.

* 'topic-arcpgu-updates' of https://github.com/foss-for-synopsys-dwc-arc-processors/linux:
  ARC: [axs10x] Specify reserved memory for frame buffer
  drm/arcpgu: use dedicated memory area for frame buffer

Merge tag 'drm/panel/for-4.7-rc1' of git://anongit.freedesktop.org/tegra/linux into drm-next

drm/panel: Changes for v4.7-rc1

This contains support for a bunch of new panels in the simple panel
driver along with some cleanup and support for a new Analogix HDMI to DP
bridge.

* tag 'drm/panel/for-4.7-rc1' of git://anongit.freedesktop.org/tegra/linux:
  drm/panel: simple: Add support for TPK U.S.A. LLC Fusion 7" and 10.1" panels
  drm/bridge: Add Analogix anx78xx support
  devicetree: Add ANX7814 SlimPort transmitter binding
  of: Add vendor prefix for Analogix Semiconductor
  drm/dp: Add define to set 0.5% down-spread in MAX_DOWNSPREAD register
  drm/panel: simple: Add support for Innolux AT070TN92
  drm/panel: simple: Remove useless drm_mode_set_name()
  drm/panel: simple: Set appropriate mode type
  drm/panel: simple: Add timings for the Olimex LCD-OLinuXino-4.3TS
  drm/panel: simple: Add the 7" DPI panel from Adafruit
  of: Add vendor prefix for On Tat Industrial Company.

drm/panel: simple: Add support for TPK U.S.A. LLC Fusion 7" and 10.1" panels

Add support for TPK U.S.A. LLC Fusion 7", 10.1" panels to the DRM simple
panel driver.

Signed-off-by: Bhuvanchandra DV <bhuvanchandra.dv@toradex.com>
Acked-by: Rob Herring <robh@kernel.org>
Signed-off-by: Thierry Reding <treding@nvidia.com>

drm/bridge: Add Analogix anx78xx support

Although there are other chips from the same family that can reuse this
driver, at the moment we only tested ANX7814 chip.

The ANX7814 is an ultra-low power Full-HD (1080p60) SlimPort transmitter
designed for portable devices. This driver adds initial support for HDMI
to DP pass-through mode.

Signed-off-by: Enric Balletbo i Serra <enric.balletbo@collabora.com>
Tested-by: Nicolas Boichat <drinkcat@chromium.org>
Reviewed-by: Nicolas Boichat <drinkcat@chromium.org>
Cc: Emil Velikov <emil.l.velikov@gmail.com>
Cc: Rob Herring <robh@kernel.org>
Cc: Dan Carpenter <dan.carpenter@oracle.com>
Cc: Daniel Kurtz <djkurtz@chromium.org>
Cc: Nicolas Boichat <drinkcat@chromium.org>
Cc: Thierry Reding <treding@nvidia.com>
[treding@nvidia.com: coding style, propagate regulator_get() errors]
Signed-off-by: Thierry Reding <treding@nvidia.com>

devicetree: Add ANX7814 SlimPort transmitter binding

The ANX7814 is an ultra-low power Full-HD (1080p60) SlimPort transmitter
designed for portable devices.

Signed-off-by: Enric Balletbo i Serra <enric.balletbo@collabora.com>
Acked-by: Rob Herring <robh@kernel.org>
Signed-off-by: Thierry Reding <treding@nvidia.com>

of: Add vendor prefix for Analogix Semiconductor

Analogix Semiconductor Inc. develops analog and mixed-signal devices for
digital media and communications interconnect applications.

Signed-off-by: Enric Balletbo i Serra <enric.balletbo@collabora.com>
Acked-by: Rob Herring <robh@kernel.org>
Signed-off-by: Thierry Reding <treding@nvidia.com>

drm/dp: Add define to set 0.5% down-spread in MAX_DOWNSPREAD register

Support of 0.5% down-spread is required for DisplayPort Specification
Version 1.1 Sink.

Signed-off-by: Enric Balletbo i Serra <enric.balletbo@collabora.com>
[treding@nvidia.com: rename to DP_MAX_DOWNSPREAD_0_5]
Signed-off-by: Thierry Reding <treding@nvidia.com>

drm/panel: simple: Add support for Innolux AT070TN92

Add support for the Innolux AT070TN92 panel.

Signed-off-by: Riccardo Bortolato <bortolato@navaltechitalia.it>
Signed-off-by: Boris Brezillon <boris.brezillon@free-electrons.com>
Signed-off-by: Thierry Reding <treding@nvidia.com>

drm/panel: simple: Remove useless drm_mode_set_name()

drm_display_mode_from_videomode() already calls drm_mode_set_name() on
the provided mode.

Signed-off-by: Boris Brezillon <boris.brezillon@free-electrons.com>
[treding@nvidia.com: slightly reword commit message]
Signed-off-by: Thierry Reding <treding@nvidia.com>

drm/panel: simple: Set appropriate mode type

All modes exposed by simple panels should be tagged as driver defined
modes. Moreover, if a panel supports only one mode, this mode is
obviously the preferred one.

Doing this also fix a problem occurring when a 'video=' parameter is
passed on the kernel command line. In some cases the user provided mode
will be preferred over the simple panel ones, which might result in
unpredictable behavior.

Signed-off-by: Boris Brezillon <boris.brezillon@free-electrons.com>
Reviewed-by: Nicolas Ferre <nicolas.ferre@atmel.com>
Tested-by: Nicolas Ferre <nicolas.ferre@atmel.com>
[treding@nvidia.com: reshuffle some code for consistency]
Signed-off-by: Thierry Reding <treding@nvidia.com>

drm/panel: simple: Add timings for the Olimex LCD-OLinuXino-4.3TS

Add support for the Olimex LCD-OLinuXino-4.3TS panel to the DRM simple
panel driver.

It is a 480x272 panel connected through a 24-bits RGB interface.

Signed-off-by: Maxime Ripard <maxime.ripard@free-electrons.com>
Acked-by: Rob Herring <robh@kernel.org>
Signed-off-by: Thierry Reding <treding@nvidia.com>

drm/panel: simple: Add the 7" DPI panel from Adafruit

This is a basic TFT panel with a 40-pin FPC connector on it.  The
specification doesn't define timings, but the Adafruit instructions
were setting up 800x480 CVT.

v2: Add .bus_format and vsync/hsync flags.

Signed-off-by: Eric Anholt <eric@anholt.net>
Acked-by: Rob Herring <robh@kernel.org>
[treding@nvidia.com: keep entries properly sorted]
Signed-off-by: Thierry Reding <treding@nvidia.com>

of: Add vendor prefix for On Tat Industrial Company.

This is the vendor for a 7" DPI panel sold by Adafruit which I'd like
to describe in DT.

Signed-off-by: Eric Anholt <eric@anholt.net>
Acked-by: Rob Herring <robh@kernel.org>
Signed-off-by: Thierry Reding <treding@nvidia.com>

Merge branch 'drm-next-4.7' of git://people.freedesktop.org/~agd5f/linux into drm-next

More amdgpu fixes for 4.7.  Highlights:
- enable async pageflips
- UVD fixes for polaris
- lots of GPUVM fixes
- whitespace and code cleanups
- misc bug fixes

* 'drm-next-4.7' of git://people.freedesktop.org/~agd5f/linux: (32 commits)
  drm/amd/powerplay: rewrite pp_sw_init to make code readable
  drm/amdgpu/dce11: fix audio offset for asics with >7 audio pins
  drm/amdgpu: fix and cleanup user fence handling v2
  drm/amdgpu: move VM fields into job
  drm/amdgpu: move the context from the IBs into the job
  drm/amdgpu: move context switch handling into common code v2
  drm/amdgpu: move preamble IB handling into common code
  drm/amdgpu/gfx7: fix pipeline sync
  amdgpu/uvd: separate context buffer from DPB
  drm/amdgpu: use fence_context to judge ctx switch v2
  drm/amd/amdgpu:  Added more named DRM info messages for debugging
  drm/amd/amdgpu: Add name field to amd_ip_funcs (v2)
  drm/amdgpu: Support DRM_MODE_PAGE_FLIP_ASYNC (v2)
  drm/amdgpu/dce11: don't share PLLs on Polaris
  drm/amdgpu: Drop unused parameter for *get_sleep_divider_id_from_clock
  drm/amdgpu: Simplify calculation in *get_sleep_divider_id_from_clock
  drm/amdgpu: Use max macro in *get_sleep_divider_id_from_clock
  drm/amd/powerplay: Use defined constants for minium engine clock
  drm/amdgpu: add missing licenses on a couple of files
  drm/amdgpu: fetch cu_info once at init
  ...

Merge tag 'drm/tegra/for-4.7-rc1' of git://anongit.freedesktop.org/tegra/linux into drm-next

drm/tegra: Changes for v4.7-rc1

Two small changes, one getting rid of the bogus gamma table size and
another removing Terje from the MAINTAINERS file since he no longer does
any work on host1x or display.

* tag 'drm/tegra/for-4.7-rc1' of git://anongit.freedesktop.org/tegra/linux:
  MAINTAINERS: Remove Terje Bergström as Tegra DRM maintainer
  drm/tegra: Don't set a gamma table size

Merge branch 'drm-hisilicon-next' of github.com:xin3liang/linux into drm-next

misc hisilicon cleanups.

* 'drm-hisilicon-next' of github.com:xin3liang/linux:
  drm/hisilicon: Fix DRM_INFO printed issue
  drm/hisilicon: Make kirin_drm_unbind sufficient
  drm/hisilicon: Use drm_connector_register_all

Merge branch 'exynos-drm-next' of git://git./linux/kernel/git/daeinki/drm-exynos into drm-next

 Summary:
   - expose HDMI-PHY clock to other drivers.
     . this patch was included in below patch series but I missed.
             http://www.spinics.net/lists/dri-devel/msg103097.html
   - some fixups about DECON5433 driver
     . this patch corrects vblank handling and fixes up trigger
       configuration.
   - use generic functions - gem_prime_mmap and dma_buf_mmap.
   - use DMA-Mapping API instead of specific one.
   - some code cleanups and fixeups.

* 'exynos-drm-next' of git://git.kernel.org/pub/scm/linux/kernel/git/daeinki/drm-exynos:
  drm/exynos/decon5433: fix trigger configuration
  drm/exynos/dsi: use of_graph_get_endpoint_by_regs helper
  drm/exynos/dpi: use of_graph_get_endpoint_by_regs helper
  drm/exynos: Nuke dummy fb->dirty callback
  drm/exynos: use directly DMA mapping APIs on g2d
  drm/exynos/hdmi: Don't print error on deferral due to regulators
  drm/exynos: fix imported dma-buf to be mapped
  drm/exynos: support gem_prime_mmap
  drm/exynos: fimd: harden fimd_calc_clkdiv()
  drm/exynos: fix cancel page flip code
  drm/exynos/decon5433: do not use unnecessary software trigger
  drm/exynos/decon5433: handle vblank in vblank interrupt
  drm/exynos/hdmi: expose HDMI-PHY clock as pipeline clock

drm/amd/powerplay: rewrite pp_sw_init to make code readable

Actually, pp_sw_init executes pptable_init and backend_init orderly if
they are initialized successfully. So rewrite it to make code more
readable.

Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Huang Rui <ray.huang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu/dce11: fix audio offset for asics with >7 audio pins

Missing offset in the audio offset array.

Reported-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: fix and cleanup user fence handling v2

We leaked the BO in the error pass, additional to that we only have
one user fence for all IBs in a job.

v2: remove white space changes

Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: move VM fields into job

They are the same for all IBs.

Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: move the context from the IBs into the job

We only have one context for all IBs.

Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: move context switch handling into common code v2

It was a source of bugs to repeat that in each IP version.

v2: rename parameter

Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: move preamble IB handling into common code

This fixes the handling which was completely broken when you
ad more than one preamble IB.

Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu/gfx7: fix pipeline sync

Need to wait on the fence as well.

Signed-off-by: Chunming Zhou <David1.Zhou@amd.com>
Reviewed-by: Monk Liu <monk.liu@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

amdgpu/uvd: separate context buffer from DPB

Updated to handle latest UVD ucode.

Signed-off-by: Sonny Jiang <sonny.jiang@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/atomic: Add drm_atomic_helper_best_encoder()

Add (struct drm_connector_helper_funcs *)->best_encoder callback helper
for connectors that support exactly 1 encoder, statically determined at
driver init time.

Signed-off-by: Noralf Trønnes <noralf@tronnes.org>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Link: http://patchwork.freedesktop.org/patch/msgid/1462982962-10530-6-git-send-email-noralf@tronnes.org

drm/atomic: Don't skip drm_bridge_*() calls if !drm_encoder_helper_funcs

Don't skip drm_bridge_*() calls if encoder->helper_private is NULL.

Signed-off-by: Noralf Trønnes <noralf@tronnes.org>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Link: http://patchwork.freedesktop.org/patch/msgid/1462982962-10530-5-git-send-email-noralf@tronnes.org

drm/fb-cma-helper: Hook up to DocBook and fix some docs

Hook up fb_cma_helper to DocBook. Remove mention of
CONFIG_FB_DEFERRED_IO in the docs, which was forgotten in the latest
version of the deferred_io patch.
Use & when referencing drm_mode_config_funcs in docs.

Signed-off-by: Noralf Trønnes <noralf@tronnes.org>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Link: http://patchwork.freedesktop.org/patch/msgid/1462982962-10530-3-git-send-email-noralf@tronnes.org

drm/fb-helper: Remove mention of CONFIG_FB_DEFERRED_IO in docs

This was forgotten to fixup in the latest version of the deferred_io
patch which made FB_DEFERRED_IO mandatory.

Signed-off-by: Noralf Trønnes <noralf@tronnes.org>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Link: http://patchwork.freedesktop.org/patch/msgid/1462982962-10530-2-git-send-email-noralf@tronnes.org

drm/amdgpu: use fence_context to judge ctx switch v2

Use of the ctx pointer is not safe, because they are likely already
be assigned to another ctx when doing comparing.

v2: recreate from scratch, avoid all unnecessary changes.

Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Monk.Liu <monk.liu@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/amdgpu:  Added more named DRM info messages for debugging

Signed-off-by: Tom St Denis <tom.stdenis@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/amdgpu: Add name field to amd_ip_funcs (v2)

Add name that we can print out in kernel messages
to aid in debugging.

v2: drop DAL changes for upstream

Signed-off-by: Tom St Denis <tom.stdenis@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: Support DRM_MODE_PAGE_FLIP_ASYNC (v2)

When this flag is set, we program the hardware to execute the flip
during horizontal blank (i.e. for the next scanline) instead of during
vertical blank (i.e. for the next frame).

Ported from radeon commit:
drm/radeon: Support DRM_MODE_PAGE_FLIP_ASYNC

v2: drop DAL change for upstream

Reviewed-by: Michel Dänzer <michel.daenzer@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu/dce11: don't share PLLs on Polaris

They are part of the phy so you can't share them.

Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: Drop unused parameter for *get_sleep_divider_id_from_clock

Signed-off-by: Nils Wallménius <nils.wallmenius@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: Simplify calculation in *get_sleep_divider_id_from_clock

a / (1 << b) is equivalent to a >> b for unsigned values

Signed-off-by: Nils Wallménius <nils.wallmenius@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: Use max macro in *get_sleep_divider_id_from_clock

Signed-off-by: Nils Wallménius <nils.wallmenius@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/powerplay: Use defined constants for minium engine clock

Replacing magic numbers in calculation of sleep divider id for fiji
and polaris.

Signed-off-by: Nils Wallménius <nils.wallmenius@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: add missing licenses on a couple of files

Noticed by n1s on IRC.

Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: fetch cu_info once at init

Fetch this info once at init and just store the results
for future requests.

Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd: cleanup remaining spaces and tabs v2

This is the result of running the following commands:
find drivers/gpu/drm/amd/ -name "*.h" -exec sed -i 's/[ \t]\+$//' {} \;
find drivers/gpu/drm/amd/ -name "*.c" -exec sed -i 's/[ \t]\+$//' {} \;
find drivers/gpu/drm/amd/ -name "*.h" -exec sed -i 's/ \+\t/\t/' {} \;
find drivers/gpu/drm/amd/ -name "*.c" -exec sed -i 's/ \+\t/\t/' {} \;

v2: drop changes to DAL and internal headers

Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: remove define for reserved client ID

Just set it to zero instead.

Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Chunming Zhou <david1.zhou@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: remove owner cleanup v2

The client ID is now unique, so no need to resert the owner fields any more.

v2: remove unused variables as well

Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Chunming Zhou <david1.zhou@amd.com> (v1)
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: make the VMID owner always 64bit

Otherwise we could (in theory) run into problems on 32bit systems.

Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Chunming Zhou <david1.zhou@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: two minor 80 char fixes

Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Tom St Denis <tom.stdenis@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: hdp flush&inval should always do

This fixes Tonga vm-fault issue when running disaster
(a multiple context GL heavy tests),
We should always flush & invalidate hdp no matter vm
used or not.

Signed-off-by: Monk Liu <Monk.Liu@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Chunming Zhou <David1.Zhou@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Cc: stable@vger.kernel.org

drm/amd/amdgpu:  Enable CG for UVD6 on Carrizo

Tested via vdpau/mpv.

Signed-off-by: Tom St Denis <tom.stdenis@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: add pipeline sync for compute job

hardware ring is async processed, the job is executed in parallel.
In some case, this will result vm fault, like jobs with different vmids.

This works around a CPC hw issue which will eventually be fixed in fw.

Signed-off-by: Chunming Zhou <David1.Zhou@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Monk Liu <monk.liu@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: keep vm in job instead of ib (v2)

ib.vm is a legacy way to get vm, after scheduler
implemented vm should be get from job, and all ibs
from one job share the same vm, no need to keep ib.vm
just move vm field to job.

this patch as well add job as paramter to ib_schedule
so it can get vm from job->vm.

v2: agd: sqaush in:
drm/amdgpu: check if ring emit_vm_flush exists in vm flush

No vm flush on engines that don't support VM.

bug:
https://bugs.freedesktop.org/show_bug.cgi?id=95195

Signed-off-by: Monk Liu <Monk.Liu@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: make vmid owner be client_id

Using the pointer is not adequate.

Signed-off-by: Chunming Zhou <David1.Zhou@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Reviewed-by: Monk Liu <monk.liu@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: add client id for every vm

This adds a unique id for each vm client so we can
properly track them.

Signed-off-by: Chunming Zhou <David1.Zhou@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Reviewed-by: Monk Liu <monk.liu@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: fix wrong release of vmid owner

The release of the vmid owner was not handled
correctly.  We need to take the lock and walk
the lru list.

Signed-off-by: Chunming Zhou <David1.Zhou@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Reviewed-by: Monk Liu <monk.liu@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/hisilicon: Fix DRM_INFO printed issue

This patch fixed the bellow no DRM_INFO is printed issue:

if (!delay_count)
DRM_INFO("phylock and phystopstateclklane is not ready.\n");

There will some printed issues with above info, under certain
circumstances:

If ((BIT(0) | BIT(2)) & val) is never true, break will not happen and
delay_count will be max u32 value (?), and no DRM_INFO is printed.

Also if ((BIT(0) | BIT(2)) & val) is true at the last possible
loop round, break happens, but now delay_count is already zero
( because of earlier delay_count-- ) and DRM_INFO is erroneously
printed.

Thanks to Juha Leppänen, he reports to me this issue.

Signed-off-by: Xinliang Liu <xinliang.liu@linaro.org>
Reported-by: Juha Leppänen <juha_efku@dnainternet.net>

drm/hisilicon: Make kirin_drm_unbind sufficient

Remove deprecated drm_put_dev.
Clean up everything needed in unbind.
Thanks to Daniel Vetter, this issue is reported by him.

Acked-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Reported-by: Daniel Vetter <daniel.vetter@intel.com>
Signed-off-by: Xinliang Liu <xinliang.liu@linaro.org>

drm/hisilicon: Use drm_connector_register_all

Use drm_connector_register_all helper to register connectors.

Cc: Xinliang Liu <xinliang.liu@linaro.org>
Cc: Xinwei Kong <kong.kongxinwei@hisilicon.com>
Signed-off-by: Daniel Vetter <daniel.vetter@intel.com>

drm/sti: include linux/seq_file.h where needed

The sti drm driver has a lot of debugfs interface that cause
build errors in some configurations when seq_file.h is not
included implicitly:

drm/sti/sti_mixer.c: In function 'mixer_dbg_ctl':
drm/sti/sti_mixer.c:88:2: error: implicit declaration of function 'seq_puts' [-Werror=implicit-function-declaration]
drm/sti/sti_mixer.c:91:4: error: implicit declaration of function 'seq_printf' [-Werror=implicit-function-declaration]
drm/sti/sti_gdp.c: In function 'gdp_dbg_ctl':
drm/sti/sti_gdp.c:146:2: error: implicit declaration of function 'seq_puts' [-Werror=implicit-function-declaration]
drm/sti/sti_gdp.c:149:4: error: implicit declaration of function 'seq_printf' [-Werror=implicit-function-declaration]
drm/sti/sti_gdp.c: In function 'gdp_dbg_show':
drm/sti/sti_gdp.c:208:32: error: dereferencing pointer to incomplete type 'struct seq_file'

This adds an explicit #include statement in all of the affected files.

Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Acked-by: Benjamin Gaignard <benjamin.gaignard@linaro.org>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Link: http://patchwork.freedesktop.org/patch/msgid/1462830733-1710590-2-git-send-email-arnd@arndb.de

drm/tegra: Use lockless gem BO free callback

No dev->struct_mutex anywhere to be seen.

Cc: Thierry Reding <thierry.reding@gmail.com>
Cc: Terje Bergström <tbergstrom@nvidia.com>
Cc: linux-tegra@vger.kernel.org
Signed-off-by: Daniel Vetter <daniel.vetter@intel.com>
Acked-by: Thierry Reding <treding@nvidia.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Link: http://patchwork.freedesktop.org/patch/msgid/1461691808-12414-28-git-send-email-daniel.vetter@ffwll.ch

drm/exynos/decon5433: fix trigger configuration

It seems trigger cannot be configured too early, otherwise it does not work in
case of panel. The patch fixes also trigger flag logic, previously HW-TRIGGER
flag was cleared in case of panel - as a result panel used always software
trigger.

Signed-off-by: Andrzej Hajda <a.hajda@samsung.com>
Signed-off-by: Inki Dae <inki.dae@samsung.com>

drm/exynos/dsi: use of_graph_get_endpoint_by_regs helper

This allows to remove the local of_graph_get_port_by_reg(),
of_graph_get_endpoint_by_reg(), and of_get_child_by_name_reg()
functions.

Signed-off-by: Philipp Zabel <p.zabel@pengutronix.de>
Reviewed-by: Andrzej Hajda <a.hajda@samsung.com>
Reviewed-by: Inki Dae <inki.dae@samsung.com>

drm/exynos/dpi: use of_graph_get_endpoint_by_regs helper

This allows to remove the local of_graph_get_port_by_reg(),
of_graph_get_endpoint_by_reg(), of_get_child_by_name_reg(),
and of_graph_get_remote_port_parent() functions.

Signed-off-by: Philipp Zabel <p.zabel@pengutronix.de>
Reviewed-by: Andrzej Hajda <a.hajda@samsung.com>
Signed-off-by: Inki Dae <inki.dae@samsung.com>

drm/exynos: Nuke dummy fb->dirty callback

It's an optional hook. Might be needed for frontbuffer rendering on
manual upload displays, but a simple TODO doesn't explain at all what
needs to be done or why.

Cc: Inki Dae <inki.dae@samsung.com>
Signed-off-by: Daniel Vetter <daniel.vetter@intel.com>
Signed-off-by: Inki Dae <inki.dae@samsung.com>

drm/exynos: use directly DMA mapping APIs on g2d

There is no reason to be wapper functions to use DMA mapping APIs. Use
directly DMA mapping APIs and remove the wapper functions.

Signed-off-by: Joonyoung Shim <jy0922.shim@samsung.com>
Signed-off-by: Inki Dae <inki.dae@samsung.com>

drm/exynos/hdmi: Don't print error on deferral due to regulators

The regulators may not be available just because their driver's probe
function was just not executed and so the regulators not registered.

So, in this case the Exynos HDMI driver should not print logs since
a -EPROBE_DEFER is not really an error and that will just pollute
the kernel log and confuse users.

This patch prevents the following misleading messages to be printed:

[    1.443638] [drm:hdmi_probe] *ERROR* failed to get regulators
[    1.449326] [drm:hdmi_probe] *ERROR* hdmi_resources_init failed

Reported-by: Krzysztof Kozlowski <k.kozlowski@samsung.com>
Signed-off-by: Javier Martinez Canillas <javier@osg.samsung.com>
Signed-off-by: Inki Dae <inki.dae@samsung.com>

drm/exynos: fix imported dma-buf to be mapped

The imported dma-buf should be mapped by sub-system exporting it.

Signed-off-by: Joonyoung Shim <jy0922.shim@samsung.com>
Signed-off-by: Inki Dae <inki.dae@samsung.com>

drm/exynos: support gem_prime_mmap

This allows exported dma-bufs to be mapped using gem_prime_mmap.

Signed-off-by: Joonyoung Shim <jy0922.shim@samsung.com>
Signed-off-by: Inki Dae <inki.dae@samsung.com>

drm/exynos: fimd: harden fimd_calc_clkdiv()

Don't use the vrefresh field of the DRM mode since this
one is supposed to only be used for debug purpose.
Instead use the clock field which should also provide
much more precise information.

Also sanitize the case in which the clock value
should be zero. We then just default to the maximum
clock divisor.

Signed-off-by: Tobias Jakobi <tjakobi@math.uni-bielefeld.de>
Reviewed-by: Andrzej Hajda <a.hajda@samsung.com>
Signed-off-by: Inki Dae <inki.dae@samsung.com>

drm/exynos: fix cancel page flip code

Driver code did not remove event from the list of pending events before destroy.
As a result drm core later tried to inspect invalid memory location.
The patch replaces removal code with call to core helper.

The bug was detected using KASAN:

[   10.107249] ==================================================================
[   10.107518] BUG: KASAN: use-after-free in drm_release+0xe9c/0x1000 at addr ffffffc089154a18
[   10.107784] Read of size 8 by task modetest/103
[   10.107931] =============================================================================
[   10.113191] BUG kmalloc-128 (Not tainted): kasan: bad access detected
[   10.119608] -----------------------------------------------------------------------------
[   10.119608]
[   10.129243] Disabling lock debugging due to kernel taint
[   10.134551] INFO: Allocated in drm_mode_page_flip_ioctl+0x500/0xa98 age=4 cpu=0 pid=103
[   10.142532]  alloc_debug_processing+0x18c/0x198
[   10.147043]  ___slab_alloc.constprop.28+0x360/0x380
[   10.151906]  __slab_alloc.isra.25.constprop.27+0x54/0xa0
[   10.157197]  kmem_cache_alloc_trace+0x370/0x3b0
[   10.161709]  drm_mode_page_flip_ioctl+0x500/0xa98
[   10.166400]  drm_ioctl+0x4c4/0xb68
[   10.169787]  do_vfs_ioctl+0x16c/0xeb8
[   10.173429]  SyS_ioctl+0x8c/0xa0
[   10.176642]  el0_svc_naked+0x24/0x28
[   10.180204] INFO: Freed in exynos_drm_crtc_cancel_page_flip+0xe0/0x160 age=0 cpu=0 pid=103
[   10.188447]  free_debug_processing+0x174/0x388
[   10.192871]  __slab_free+0x2e8/0x438
[   10.196431]  kfree+0x350/0x360
[   10.199469]  exynos_drm_crtc_cancel_page_flip+0xe0/0x160
[   10.204762]  exynos_drm_preclose+0x58/0xa0
[   10.208844]  drm_release+0x1f0/0x1000
[   10.212491]  __fput+0x1c4/0x5b8
[   10.215613]  ____fput+0xc/0x18
[   10.218654]  task_work_run+0x130/0x198
[   10.222385]  do_exit+0x700/0x2278
[   10.225681]  do_group_exit+0xe4/0x2c8
[   10.229327]  SyS_exit_group+0x1c/0x20
[   10.232973]  el0_svc_naked+0x24/0x28
[   10.236532] INFO: Slab 0xffffffbdc2a45500 objects=32 used=10 fp=0xffffffc089154a00 flags=0x4080
[   10.245210] INFO: Object 0xffffffc089154a00 @offset=2560 fp=0xffffffc089157600
[   10.245210]
...
[   10.384532] CPU: 0 PID: 103 Comm: modetest Tainted: G    B           4.5.0-rc3-00748-gd5e2881 #271
[   10.398325] Call trace:
[   10.400764] [<ffffffc000091428>] dump_backtrace+0x0/0x328
[   10.406141] [<ffffffc000091764>] show_stack+0x14/0x20
[   10.411176] [<ffffffc00089c550>] dump_stack+0xb0/0xe8
[   10.416210] [<ffffffc000395778>] print_trailer+0xf8/0x160
[   10.421592] [<ffffffc00039b5cc>] object_err+0x3c/0x50
[   10.426626] [<ffffffc00039d630>] kasan_report_error+0x248/0x550
[   10.432527] [<ffffffc00039da50>] __asan_report_load8_noabort+0x40/0x48
[   10.439039] [<ffffffc000b5b724>] drm_release+0xe9c/0x1000
[   10.444419] [<ffffffc0003d340c>] __fput+0x1c4/0x5b8
[   10.449280] [<ffffffc0003d3884>] ____fput+0xc/0x18
[   10.454055] [<ffffffc000101aa8>] task_work_run+0x130/0x198
[   10.459522] [<ffffffc0000bc058>] do_exit+0x700/0x2278
[   10.464557] [<ffffffc0000bdcfc>] do_group_exit+0xe4/0x2c8
[   10.469939] [<ffffffc0000bdefc>] SyS_exit_group+0x1c/0x20
[   10.475320] [<ffffffc000087530>] el0_svc_naked+0x24/0x28

Signed-off-by: Andrzej Hajda <a.hajda@samsung.com>
Signed-off-by: Inki Dae <inki.dae@samsung.com>

drm/exynos/decon5433: do not use unnecessary software trigger

Software trigger should not be used if hardware trigger is configured.

Signed-off-by: Andrzej Hajda <a.hajda@samsung.com>
Signed-off-by: Inki Dae <inki.dae@samsung.com>

drm/exynos/decon5433: handle vblank in vblank interrupt

vblank should be signaled to userspace after reading framebuffers not before,
signaling it in TE interrupt looks wrong. TE triggers reading framebuffers
so it is the worst moment. Tearing is not observable because hardware prevents
it, but there are frequently skipped vblank events.

Signed-off-by: Andrzej Hajda <a.hajda@samsung.com>
Signed-off-by: Inki Dae <inki.dae@samsung.com>

drm/exynos/hdmi: expose HDMI-PHY clock as pipeline clock

HDMI-PHY clock should be accessible from other components in the pipeline.

Signed-off-by: Andrzej Hajda <a.hajda@samsung.com>
Signed-off-by: Inki Dae <inki.dae@samsung.com>

drm/exynos: Use lockless gem BO free callback

No dev->struct_mutex anywhere to be seen.

Cc: Inki Dae <inki.dae@samsung.com>
Signed-off-by: Daniel Vetter <daniel.vetter@intel.com>
Acked-by: Inki Dae <inki.dae@samsung.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Link: http://patchwork.freedesktop.org/patch/msgid/1461691808-12414-18-git-send-email-daniel.vetter@ffwll.ch

drm: Make drm_encoder_helper_funcs optional

Make drm_encoder_helper_funcs and it's functions optional to avoid
having dummy functions.

Signed-off-by: Noralf Trønnes <noralf@tronnes.org>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Link: http://patchwork.freedesktop.org/patch/msgid/1462454674-2246-3-git-send-email-noralf@tronnes.org

Merge branch 'drm-etnaviv-next' of git://git.pengutronix.de:/git/lst/linux into drm-next

not much new stuff this time. A (micro-)optimization to allow the
hangcheck timer to be coalesced with other wakeups in the system and a
fix to handle mmaping of prime imported and userptr buffers correctly. I
don't think we have seen any actual issues going back to this yet, so I
figured it's safer to get this in via drm-next rather than smashing it
into fixes.

* 'drm-etnaviv-next' of git://git.pengutronix.de:/git/lst/linux:
  drm/etnaviv: fix mmap operations for userptr and dma-buf objects
  drm/etnaviv: take etnaviv_gem_obj in etnaviv_gem_mmap_obj
  drm/etnaviv: use deferrable timer for hangcheck handler
  timer: add setup_deferrable_timer macro

Merge tag 'mediatek-drm-2016-05-09' of git://git.pengutronix.de/git/pza/linux into drm-next

MT8173 DRM support

- device tree binding documentation for all MT8173 display
  subsystem components
- basic mediatek-drm driver for MT8173 with two optional,
  currently fixed output paths:
- DSI encoder support for DSI and (via bridge) eDP panels
- DPI encoder support for output to HDMI bridge
- necessary clock tree changes for the DPI->HDMI path
- export mtk-smi functions used by mediatek-drm

* tag 'mediatek-drm-2016-05-09' of git://git.pengutronix.de/git/pza/linux:
  clk: mediatek: remove hdmitx_dig_cts from TOP clocks
  clk: mediatek: Add hdmi_ref HDMI PHY PLL reference clock output
  clk: mediatek: make dpi0_sel propagate rate changes
  drm/mediatek: Add DPI sub driver
  drm/mediatek: Add DSI sub driver
  drm/mediatek: Add DRM Driver for Mediatek SoC MT8173.
  dt-bindings: drm/mediatek: Add Mediatek display subsystem dts binding
  memory: mtk-smi: export mtk_smi_larb_get/put

Merge tag 'v4.6-rc7' into drm-next

Merge this back as we've built up a fair few conflicts, and I have
some newer trees to pull in.

drm/amdgpu: fix build on aarch64.

Same fix went in previously for some other files.

Signed-off-by: Dave Airlie <airlied@redhat.com>

Merge branch 'msm-next' of git://people.freedesktop.org/~robclark/linux into drm-next

Mostly cleanups, fixes, and 'struct fence' conversion this time
around, with one reservation patch which is a-b Sumit (which the fence
conversion patches depend on).

* 'msm-next' of git://people.freedesktop.org/~robclark/linux: (25 commits)
  drm/msm: Drop load/unload drm_driver ops
  drm/msm: Centralize connector registration/unregistration
  drm/msm/hdmi: Prevent gpio_free related kernel warnings
  drm/msm: print offender task name on hangcheck recovery
  drm/msm: fix leak in failed submit path
  drm/msm: de-indent submit_create()
  drm/msm: drop return from gpu->submit()
  drm/msm/mdp4: Don't manage DSI PLL regulators in MDP driver
  drm/msm/edp: Drop regulator_set_voltage call
  drm/msm/dsi: Fix regulator API abuse
  drm/msm: Move call to PTR_ERR_OR_ZERO after reassignment
  drm/msm/mdp: Add support for more RGBX formats
  drm: msm: remove unused variable
  drm/msm: fix ->last_fence() after recover
  drm/msm: 'struct fence' conversion
  drm/msm: remove fence_cbs
  drm/msm: introduce msm_fence_context
  drm/msm: split locking and pinning BO's
  drm/msm/gpu: simplify tracking in-flight bo's
  drm/msm: split out timeout_to_jiffies helper
  ...

Merge tag 'topic/drm-misc-2016-05-08' of git://anongit.freedesktop.org/drm-intel into drm-next

Refcounting is hard, so here's a quick pull request with the one-liner to
fix up i915. Otherwise just a few other small things I picked up. Plus the
regression fix from Marten for rmfb behaviour that lingered around forever
since no testers. Feel free to cherry-pick that over to drm-fixes, but
given that there's not many who seemed to have cared, meh.

* tag 'topic/drm-misc-2016-05-08' of git://anongit.freedesktop.org/drm-intel:
  drm/i915: Correctly refcount connectors in hw state readou
  drm/panel: Flesh out kerneldoc
  drm: Add gpu.tmpl docbook to MAINTAINERS entry
  drm/core: Do not preserve framebuffer on rmfb, v4.
  drm: Fix up markup fumble
  drm/fb_helper: Fix a few typos

Linux 4.6-rc7

drm/msm: Drop load/unload drm_driver ops

The load/unload drm_driver ops are deprecated. They should be removed as
they result in creation of devices visible to userspace even before
the drm_device is registered.

Drop these ops and use drm_dev_alloc/register and drm_dev_unregister/unref
to explicitly create and destroy the drm device in the msm platform
driver's bind and unbind ops. With this in use, the drm connectors are
only registered once the drm_device is registered.

It also fixes the issue of stray debugfs files after the msm module is
removed. With this, all the debugfs files are removed, and allows
successive module insertions/removals.

Signed-off-by: Archit Taneja <architt@codeaurora.org>

drm/msm: Centralize connector registration/unregistration

Move the drm_connector registration from the encoder(HDMI/DSI etc) drivers
to the msm platform driver. This will simplify the task of ensuring that
the connectors are registered only after the drm_device itself is
registered.

The connectors' destroy ops are made to use kzalloc instead of
devm_kzalloc to ensure that that the connectors can be successfully
unregistered when the msm driver module is removed. The memory for the
connectors is unallocated when drm_mode_config_cleanup() is called
during either during an error or during driver remove.

Signed-off-by: Archit Taneja <architt@codeaurora.org>
Signed-off-by: Rob Clark <robdclark@gmail.com>

drm/msm/hdmi: Prevent gpio_free related kernel warnings

Calling the legacy gpio_free on an invalid GPIO (a GPIO numbered -1)
results in kernel warnings. This causes a lot of backtraces when
we try to unload the drm/msm module.

Call gpio_free only on valid GPIOs.

Signed-off-by: Archit Taneja <architt@codeaurora.org>
Signed-off-by: Rob Clark <robdclark@gmail.com>

drm/msm: print offender task name on hangcheck recovery

Track the pid per submit, so we can print the name of the task which
submitted the batch that caused the gpu to hang.

Signed-off-by: Rob Clark <robdclark@gmail.com>

drm/msm: fix leak in failed submit path

Signed-off-by: Rob Clark <robdclark@gmail.com>

drm/msm: de-indent submit_create()

Signed-off-by: Rob Clark <robdclark@gmail.com>