drm/i915/gen9: fix plane_blocks_per_line on watermarks calculations

The confusing thing is that plane_blocks_per_line is listed as part of
the method 2 calculation but is also used for other things. We
calculated it in two different places and different ways: one inside
skl_wm_method2() and the other inside skl_compute_plane_wm(). The
skl_wm_method2() implementation is the one that matches the
specification.

With this patch we fix the skl_compute_plane_wm() calculation and just
pass it as a parameter to skl_wm_method2(). We also take care to not
modify the value of plane_bytes_per_line since we're going to rely on
it having a correct value in later patches.

This should affect the watermarks for Linear and Y-tiled.

From my analysis, it looks like the two plane_blocks_per_line
variables got out of sync on 0fda65680e92, but we can't really say
that commit was a regression, it looks like just an incomplete fix.
There's always the possibility that 0fda65680e92 matched our
specification at that time, and then later the specification changed.

v2: Try to add a "Fixes" tag (Maarten).

Fixes: 0fda65680e92 ("drm/i915/skl: Update watermarks for Y tiling")
Cc: stable@vger.kernel.org
Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Reviewed-by: Lyude <cpaul@redhat.com>
Reviewed-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Signed-off-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1474578035-424-7-git-send-email-paulo.r.zanoni@intel.com
(cherry picked from commit 7a1a8aed67e0a60772defe3f6499eb340da48634)
Signed-off-by: Jani Nikula <jani.nikula@intel.com>

drm/i915/gen9: minimum scanlines for Y tile is not always 4

During watermarks calculations, this value is used in 3 different
places. Only one of them was not using a hardcoded 4. Move the code up
so everybody can benefit from the actual value.

This should only help on situations with Y tiling + 90/270 rotation +
1 or 2 bpp or NV12.

Cc: stable@vger.kernel.org
Reviewed-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Signed-off-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1474578035-424-6-git-send-email-paulo.r.zanoni@intel.com
(cherry picked from commit 1186fa85eb9b3cc0589990fbc39617e50e38759a)
Signed-off-by: Jani Nikula <jani.nikula@intel.com>

drm/i915/gen9: fix the WaWmMemoryReadLatency implementation

Bspec says:
  "The mailbox response data may not account for memory read latency.
   If the mailbox response data for level 0 is 0us, add 2 microseconds
   to the result for each valid level."

This means we should only do the +2 in case wm[0] == 0, not always.

So split the sanitizing implementation from the WA implementation and
fix the WA implementation.

v2: Add Fixes tag (Maarten).

Fixes: 367294be7c25 ("drm/i915/gen9: Add 2us read latency to WM level")
Cc: stable@vger.kernel.org
Cc: Vandana Kannan <vandana.kannan@intel.com>
Reviewed-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Signed-off-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1474578035-424-5-git-send-email-paulo.r.zanoni@intel.com
(cherry picked from commit 0727e40a48a1d08cf54ce2c01e120864b92e59bf)
Signed-off-by: Jani Nikula <jani.nikula@intel.com>

drm/i915/kbl: KBL also needs to run the SAGV code

According to BSpec, it's the "core CPUs" that need the code, which
means SKL and KBL, but not BXT.

I don't have a KBL to test this patch on it.

v2: Only SKL should have I915_SAGV_NOT_CONTROLLED.

Cc: stable@vger.kernel.org
Reviewed-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Signed-off-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1474578035-424-4-git-send-email-paulo.r.zanoni@intel.com
(cherry picked from commit 6e3100ec21e7c774a0fc01e36a1e0739530c2f71)
Signed-off-by: Jani Nikula <jani.nikula@intel.com>

drm/i915: introduce intel_has_sagv()

And use it to move knowledge about the SAGV-supporting platforms from
the callers to the SAGV code.

We'll add more platforms to intel_has_sagv(), so IMHO it makes more
sense to move all this to a single function instead of patching all
the callers every time we add SAGV support to a new platform.

v2: Move I915_SAGV_NOT_CONTROLLED to the new function (Lyude).

Cc: stable@vger.kernel.org
Reviewed-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Signed-off-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1474578035-424-3-git-send-email-paulo.r.zanoni@intel.com
(cherry picked from commit 56feca91973459d0b62cbb2610b62d341025ed89)
Signed-off-by: Jani Nikula <jani.nikula@intel.com>

drm/i915: SAGV is not SKL-only, so rename a few things

The plan is to introduce intel_has_sagv() and then use it to discover
which platforms actually support it.

I thought about keeping the functions with their current skl names,
but found two problems: (i) skl_has_sagv() would become a very
confusing name, and (ii) intel_atomic_commit_tail() doesn't seem to be
calling any functions whose name start with a platform name, so the
"intel_" naming scheme seems make more sense than the "firstplatorm_"
naming scheme here.

Cc: stable@vger.kernel.org
Reviewed-by: Lyude <cpaul@redhat.com>
Reviewed-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Signed-off-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1474578035-424-2-git-send-email-paulo.r.zanoni@intel.com
(cherry picked from commit 16dcdc4edbcf5cb130004737f2548401776170f1)
Signed-off-by: Jani Nikula <jani.nikula@intel.com>

drm/i915: don't forget to set intel_crtc->dspaddr_offset on SKL+

We never remembered to set it (so it was zero), but this was not a
problem in the past due to the way handled the hardware registers.
Unfortunately we changed how we set the hardware and forgot to set
intel_crtc->dspaddr_offset.

This started to reflect on a few kms_frontbuffer_tracking subtests
that relied on page flips with CRTCs that don't point to the x:0,y:0
coordinates of the frontbuffer. After the page flip the CRTC was
showing the x:0,y:0 coordinate of the frontbuffer instead of
x:500,y:500. This problem is present even if we don't enable FBC or
PSR.

While trying to bisect it I realized that the first bad commit
actually just gives me a black screen for the mentioned tests instead
of showing the wrong x:0,y:0 offsets. A few commits later the black
screen problem goes away and we get to the point where the code is
today, but I'll consider the black screen as the first bad commit
since it's the point where the IGT subtests start to fail.

Fixes: 6687c9062c46 ("drm/i915: Rewrite fb rotation GTT handling")
Testcase: kms_frontbuffer_tracking/fbc-1p-primscrn-shrfb-pgflip-blt
Testcase: kms_frontbuffer_tracking/fbc-1p-primscrn-shrfb-evflip-blt
Testcase: kms_frontbuffer_tracking/fbc-1p-shrfb-fliptrack
Cc: Ville Syrjälä <ville.syrjala@linux.intel.com>
Cc: Sivakumar Thulasimani <sivakumar.thulasimani@intel.com>
Cc: drm-intel-fixes@lists.freedesktop.org
Signed-off-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1471644203-23463-1-git-send-email-paulo.r.zanoni@intel.com
(cherry picked from commit 4c0b8a8bc49c477be9467f614b6b4ec479736019)
Signed-off-by: Jani Nikula <jani.nikula@intel.com>

drm/i915: Only shrink the unbound objects during freeze

At the point of creating the hibernation image, the runtime power manage
core is disabled - and using the rpm functions triggers a warn.
i915_gem_shrink_all() tries to unbind objects, which requires device
access and so tries to how an rpm reference triggering a warning:

[   44.235420] ------------[ cut here ]------------
[   44.235424] WARNING: CPU: 2 PID: 2199 at drivers/gpu/drm/i915/intel_runtime_pm.c:2688 intel_runtime_pm_get_if_in_use+0xe6/0xf0
[   44.235426] WARN_ON_ONCE(ret < 0)
[   44.235445] Modules linked in: ctr ccm arc4 rt2800usb rt2x00usb rt2800lib rt2x00lib crc_ccitt mac80211 cmac cfg80211 btusb rfcomm bnep btrtl btbcm btintel bluetooth dcdbas x86_pkg_temp_thermal intel_powerclamp coretemp snd_hda_codec_realtek crct10dif_pclmul crc32_pclmul ghash_clmulni_intel snd_hda_codec_generic aesni_intel snd_hda_codec_hdmi aes_x86_64 lrw gf128mul snd_hda_intel glue_helper ablk_helper cryptd snd_hda_codec hid_multitouch joydev snd_hda_core binfmt_misc i2c_hid serio_raw snd_pcm acpi_pad snd_timer snd i2c_designware_platform 8250_dw nls_iso8859_1 i2c_designware_core lpc_ich mfd_core soundcore usbhid hid psmouse ahci libahci
[   44.235447] CPU: 2 PID: 2199 Comm: kworker/u8:8 Not tainted 4.8.0-rc5+ #130
[   44.235447] Hardware name: Dell Inc. XPS 13 9343/0310JH, BIOS A07 11/11/2015
[   44.235450] Workqueue: events_unbound async_run_entry_fn
[   44.235453]  0000000000000000 ffff8801b2f7fb98 ffffffff81306c2f ffff8801b2f7fbe8
[   44.235454]  0000000000000000 ffff8801b2f7fbd8 ffffffff81056c01 00000a801f50ecc0
[   44.235456]  ffff88020ce50000 ffff88020ce59b60 ffffffff81a60b5c ffffffff81414840
[   44.235456] Call Trace:
[   44.235459]  [<ffffffff81306c2f>] dump_stack+0x4d/0x6e
[   44.235461]  [<ffffffff81056c01>] __warn+0xd1/0xf0
[   44.235464]  [<ffffffff81414840>] ? i915_pm_suspend_late+0x30/0x30
[   44.235465]  [<ffffffff81056c6f>] warn_slowpath_fmt+0x4f/0x60
[   44.235468]  [<ffffffff814e73ce>] ? pm_runtime_get_if_in_use+0x6e/0xa0
[   44.235469]  [<ffffffff81433526>] intel_runtime_pm_get_if_in_use+0xe6/0xf0
[   44.235471]  [<ffffffff81458a26>] i915_gem_shrink+0x306/0x360
[   44.235473]  [<ffffffff81343fd4>] ? pci_platform_power_transition+0x24/0x90
[   44.235475]  [<ffffffff81414840>] ? i915_pm_suspend_late+0x30/0x30
[   44.235476]  [<ffffffff81458dfb>] i915_gem_shrink_all+0x1b/0x30
[   44.235478]  [<ffffffff814560b3>] i915_gem_freeze_late+0x33/0x90
[   44.235479]  [<ffffffff81414877>] i915_pm_freeze_late+0x37/0x40
[   44.235481]  [<ffffffff814e9b8e>] dpm_run_callback+0x4e/0x130
[   44.235483]  [<ffffffff814ea5db>] __device_suspend_late+0xdb/0x1f0
[   44.235484]  [<ffffffff814ea70f>] async_suspend_late+0x1f/0xa0
[   44.235486]  [<ffffffff81077557>] async_run_entry_fn+0x37/0x150
[   44.235488]  [<ffffffff8106f518>] process_one_work+0x148/0x3f0
[   44.235490]  [<ffffffff8106f8eb>] worker_thread+0x12b/0x490
[   44.235491]  [<ffffffff8106f7c0>] ? process_one_work+0x3f0/0x3f0
[   44.235492]  [<ffffffff81074d09>] kthread+0xc9/0xe0
[   44.235495]  [<ffffffff816e257f>] ret_from_fork+0x1f/0x40
[   44.235496]  [<ffffffff81074c40>] ? kthread_park+0x60/0x60
[   44.235497] ---[ end trace e438706b97c7f132 ]---

Alternatively, to actually shrink everything we have to do so slightly
earlier in the hibernation process.

To keep lockdep silent, we need to take struct_mutex for the shrinker
even though we know that we are the only user during the freeze.

Fixes: 7aab2d534e35 ("drm/i915: Shrink objects prior to hibernation")
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20160921135108.29574-2-chris@chris-wilson.co.uk
(cherry picked from commit 6a800eabba34945c2986d70114b41d564bad52a8)
Signed-off-by: Jani Nikula <jani.nikula@intel.com>

drm/i915: Restore current RPS state after reset

Following commit 821ed7df6e2a ("drm/i915: Update reset path to fix
incomplete requests") we no longer mark the context as lost on reset as
we keep the requests (and contexts) alive. However, RPS remains reset
and we need to restore the current state to match the in-flight
requests.

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=97824
Fixes: 821ed7df6e2a ("drm/i915: Update reset path to fix incomplete requests")
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Mika Kuoppala <mika.kuoppala@intel.com>
Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Cc: Arun Siluvery <arun.siluvery@linux.intel.com>
Reviewed-by: Mika Kuoppala <mika.kuoppala@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20160921135108.29574-1-chris@chris-wilson.co.uk
(cherry picked from commit f2a91d1a6f5960c08f1ca60bd076f4dc020c50c6)
Signed-off-by: Jani Nikula <jani.nikula@intel.com>

drm/i915: Unlock PPS registers after GPU reset

Reapply the PPS register unlock workaround after GPU reset on platforms
where the reset clobbers the display HW state. This at least gets rid of
the related WARN during LVDS encoder enabling on PNV.

Fixes: ed6143b8f75 ("drm/i915/lvds: Restore initial HW state during encoder enabling")
Reported-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Imre Deak <imre.deak@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1473847453-4771-1-git-send-email-imre.deak@intel.com
Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
(cherry picked from commit 51f592050a523fc5882f9b8b4e9259422e41e848)
Signed-off-by: Jani Nikula <jani.nikula@intel.com>

drm/i915/backlight: setup backlight pwm alternate increment on backlight enable

Backlight enable is supposed to do a full setup of the backlight. We
were missing the PWM alternate increment bit in the south chicken
registers on lpt+ pch. This potentially caused a PWM frequency change
when the chicken register value was lost e.g. on suspend.

v2 by Jani, rebase on the patch caching alt increment

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=97486
References: https://bugs.freedesktop.org/show_bug.cgi?id=67454
Cc: Cooper Chiou <cooper.chiou@intel.com>
Cc: Wei Shun Chen <wei.shun.chang@intel.com>
Cc: Gary C Wang <gary.c.wang@intel.com>
Cc: stable@vger.kernel.org # v4.4+ 16e1203db8ab drm/i915/backlight: setup and cache...
Cc: stable@vger.kernel.org # v4.4+
Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Shawn Lee <shawn.c.lee@intel.com>
Signed-off-by: Jani Nikula <jani.nikula@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/8265f5935bd31c039ddfc82819d26c2ca1ae9cba.1474281249.git.jani.nikula@intel.com
(cherry picked from commit e29aff05f239f8dd24e9ee7816fd96726e20105a)
Signed-off-by: Jani Nikula <jani.nikula@intel.com>

drm/i915/backlight: setup and cache pwm alternate increment value

This will also be needed later on when setting up the alternate
increment in backlight enable.

Cc: Shawn Lee <shawn.c.lee@intel.com>
Cc: <stable@vger.kernel.org>
Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Jani Nikula <jani.nikula@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/9984b20bc59aee90b83caf59ce91f3fb122c9627.1474281249.git.jani.nikula@intel.com
(cherry picked from commit 32b421e79e6b546da1d469f1229403ac9142d695)
Signed-off-by: Jani Nikula <jani.nikula@intel.com>

Merge branch 'drm-next-4.9' of git://people.freedesktop.org/~agd5f/linux into drm-next

Just some misc bug fixes for 4.9.

* 'drm-next-4.9' of git://people.freedesktop.org/~agd5f/linux:
  drm/amdgpu: revert "use more than 64KB fragment size if possible"
  drm/amdgpu: warn if dp aux is still attached on free
  drm/amdgpu/dce11: add missing drm_mode_config_cleanup call
  drm/amdgpu: also track late init state
  drm/amdgpu/virtual_dce: adjust config ifdef
  drm/amdgpu/vce: add support for hw config packet (v2)
  drm/amdgpu: clean up to set fw_offset as 0 twice
  drm/amdgpu: remove DRM_AMD_POWERPLAY
  drm/radeon: Prevent races on pre DCE4 between flip submission and completion.
  drm/radeon: Slightly more robust flip completion handling for < DCE-4

Merge tag 'topic/drm-misc-2016-10-05' of git://anongit.freedesktop.org/drm-intel into drm-next

Another attempt, this time rebased and without the pipe crc patches:
- display_info cleanups from Ville
- make prime/gem lookups faster with rbtrees (Chris)
- misc stuff all over

* tag 'topic/drm-misc-2016-10-05' of git://anongit.freedesktop.org/drm-intel:
  drm/rockchip: analogix_dp: Refuse to enable PSR if panel doesn't support it
  drm/bridge: analogix_dp: Add analogix_dp_psr_supported
  drm/fb-helper: add DRM_FB_HELPER_DEFAULT_OPS for fb_ops
  drm: Document caveats around atomic event handling
  uapi: add missing install of sync_file.h
  drm: Simplify drm_printk to reduce object size quite a bit
  drm/i915: Account for sink max TMDS clock when checking the port clock
  drm/i915: Replace a bunch of connector->base.display_info with a local variable
  drm/edid: Move dvi_dual/max_tmds_clock parsing out from drm_edid_to_eld()
  drm/edid: Clear the old cea_rev when there's no CEA extension in the new EDID
  drm/edid: Reduce the number of times we parse the CEA extension block
  drm/edid: Don't pass around drm_display_info needlessly
  drm/edid: Move dvi_dual/max_tmds_clock to drm_display_info
  drm/edid: Make max_tmds_clock kHz instead of MHz
  drm/edid: Clear old dvi_dual/max_tmds_clock before parsing the new EDID
  drm/edid: Clear old audio latency values before parsing the new EDID
  drm: Convert prime dma-buf <-> handle to rbtree
  drm/mediatek: mark symbols static where possible
  drm/rockchip: mark symbols static where possible
  drm/rockchip: add missing header dependencies

Merge tag 'drm-vc4-next-2016-10-06' of https://github.com/anholt/linux into drm-next

This pull request brings in several fixes for drm-next, mostly for
HDMI.

* tag 'drm-vc4-next-2016-10-06' of https://github.com/anholt/linux:
  drm/vc4: Add support for double-clocked modes.
  drm/vc4: Set up the AVI and SPD infoframes.
  drm/vc4: Fix support for interlaced modes on HDMI.
  drm/vc4: Increase timeout for HDMI_SCHEDULER_CONTROL changes.
  drm/vc4: Fall back to using an EDID probe in the absence of a GPIO.
  drm/vc4: Enable limited range RGB output on HDMI with CEA modes.
  drm/vc4: Fix races when the CS reads from render targets.
  drm/vc4: cleanup with list_first_entry_or_null()

drm/vc4: Add support for double-clocked modes.

Now that we have infoframes to report the pixel repeat flag, we can
start using it.  Fixes locking the 720x480i and 720x576i modes on my
Dell 2408WFP.  Like the 1920x1080i case, they don't fit properly on
the screen, though.

Signed-off-by: Eric Anholt <eric@anholt.net>

drm/vc4: Set up the AVI and SPD infoframes.

Fixes a purple bar on the left side of the screen with my Dell
2408WFP.  It will also be required for supporting the double-clocked
video modes.

Signed-off-by: Eric Anholt <eric@anholt.net>

drm/vc4: Fix support for interlaced modes on HDMI.

We really do need to be using the halved V fields.  I had been
confused by the code I was using as a reference because it stored
halved vsync fields but not halved vdisplay, so it looked like I only
needed to divide vdisplay by 2.

This reverts part of Mario's timestamping fixes that prevented
CRTC_HALVE_V from applying, and instead adjusts the timestamping code
to not use the crtc field in that case.

Fixes locking of 1920x1080x60i on my Dell 2408WFP.  There are black
bars on the top and bottom, but I suspect that might be an
under/overscan flags problem as opposed to video timings.

Signed-off-by: Eric Anholt <eric@anholt.net>

drm/vc4: Increase timeout for HDMI_SCHEDULER_CONTROL changes.

Fixes occasional debug spew at boot when connected directly through
HDMI, and probably confusing the HDMI state machine when we go trying
to poke registers for the enable sequence too soon.

Signed-off-by: Eric Anholt <eric@anholt.net>

drm/vc4: Fall back to using an EDID probe in the absence of a GPIO.

On Pi0/1/2, we use an external GPIO line for hotplug detection, since
the HDMI_HOTPLUG register isn't connected to anything.  However, with
the Pi3 the HPD GPIO line has moved off to a GPIO expander that will
be tricky to get to (the firmware is constantly polling the expander
using i2c0, so we'll need to coordinate with it).

As a stop-gap, if we don't have a GPIO line, use an EDID probe to
detect connection.  Fixes HDMI display on the pi3.

Signed-off-by: Eric Anholt <eric@anholt.net>

drm/vc4: Enable limited range RGB output on HDMI with CEA modes.

Fixes broken grayscale ramps on many HDMI monitors, where large areas
at the ends of the ramp would all appear as black or white.

Signed-off-by: Eric Anholt <eric@anholt.net>

drm/vc4: Fix races when the CS reads from render targets.

With the introduction of bin/render pipelining, the previous job may
not be completed when we start binning the next one.  If the previous
job wrote our VBO, IB, or CS textures, then the binning stage might
get stale or uninitialized results.

Fixes the major rendering failure in glmark2 -b terrain.

Signed-off-by: Eric Anholt <eric@anholt.net>
Fixes: ca26d28bbaa3 ("drm/vc4: improve throughput by pipelining binning and rendering jobs")
Cc: stable@vger.kernel.org

drm/vc4: cleanup with list_first_entry_or_null()

The combo of list_empty() check and return list_first_entry()
can be replaced with list_first_entry_or_null().

Signed-off-by: Masahiro Yamada <yamada.masahiro@socionext.com>
Signed-off-by: Eric Anholt <eric@anholt.net>

drm/amdgpu: revert "use more than 64KB fragment size if possible"

This reverts commit 1dcd32fb9c54334ec948a0f18174a748d6b14364.

The block size is indeed an equal match, so this can cause performance regressions.

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

drm/amdgpu: warn if dp aux is still attached on free

If this happens (and it recently did), we free a structure while part of
it is still in use, which results in non-obvious crashes. The way it's
detached is not trivial (DRM core has to call the connector .destroy
callback and things must be torn down in the right order), so better
detect it and warn early.

Signed-off-by: Grazvydas Ignotas <notasas@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu/dce11: add missing drm_mode_config_cleanup call

All other amdgpu/dce_v* files have this call, it's only mysteriously
missing from dce_v11_0.c since the file was added and causes leaks.

Fixes: aaa36a976bbb ("drm/amdgpu: Add initial VI support")
Signed-off-by: Grazvydas Ignotas <notasas@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Cc: stable@vger.kernel.org

drm/amdgpu: also track late init state

Successful sw_init() and hw_init() states are tracked, but not
late_init(). Various error paths may result in amdgpu_fini() being
called before .late init is done, so late_init needs to be tracked
to avoid unexpected or multiple .late_fini() calls.

Signed-off-by: Grazvydas Ignotas <notasas@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu/virtual_dce: adjust config ifdef

Include the CIK asics in the ifdef.

Reviewed-By: Emily Deng <Emily.Deng@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu/vce: add support for hw config packet (v2)

This is needed for proper VCE DPM on some APUs.

v2: fix the asic list

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

drm/amdgpu: clean up to set fw_offset as 0 twice

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

drm/amdgpu: remove DRM_AMD_POWERPLAY

Powerplay is no longer optional after the recently cleanups

Reviewed-by: Edward O'Callaghan <funfunctor@folklore1984.net>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/radeon: Prevent races on pre DCE4 between flip submission and completion.

Pre DCE4 hw doesn't have reliable pageflip completion
interrupts, so instead polling for flip completion is
used from within the vblank irq handler to complete
page flips.

This causes a race if pageflip ioctl is called close to
vblank:

1. pageflip ioctl queues execution of radeon_flip_work_func.

2. vblank irq fires, radeon_crtc_handle_vblank checks for
   flip_status == FLIP_SUBMITTED finds none, no-ops.

3. radeon_flip_work_func runs inside vblank, decides to
   set flip_status == FLIP_SUBMITTED and programs the
   flip into hw.

4. hw executes flip immediately (because in vblank), but
   as 2 already happened, the flip completion routine only
   emits the flip completion event one refresh later ->
   wrong vblank count/timestamp for completion and no
   performance gain, as instead of delaying the flip until
   next vblank, we now delay the next flip by 1 refresh
   while waiting for the delayed flip completion event.

Given we often don't gain anything due to this race, but
lose precision, prevent the programmed flip from executing
in vblank on pre DCE4 asics to avoid this race.

On pre-AVIVO hw we can't program the hw for edge-triggered
flips, they always execute anywhere in vblank. Therefore delay
the actual flip programming until after vblank on pre-AVIVO.

Reviewed-by: Michel Dänzer <michel.daenzer@amd.com>
Signed-off-by: Mario Kleiner <mario.kleiner.de@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/radeon: Slightly more robust flip completion handling for < DCE-4

Pre DCE4 hardware doesn't have (reliable) pageflip completion
irqs, therefore we have to use the old polling method for flip
completion handling in vblank irq.

As vblank irqs fire a bit before start of vblank (when the
linebuffer fifo read position reaches end of scanout), we
have some fudge for flip completion handling in the last
lines of active scanout. Old code assumed the threshold to
be 99% of active scanout height, a ballpark estimate which
worked ok. Since we know since a while how to calculate the
actual threshold from linebuffer size, lets make use of it
to get a more accurate threshold.

This completion path is still prone to some races in corner
cases, especially on pre-AVIVO hardware, so document them
a bit better in the code comments.

Acked-by: Michel Dänzer <michel.daenzer@amd.com>
Signed-off-by: Mario Kleiner <mario.kleiner.de@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/rockchip: analogix_dp: Refuse to enable PSR if panel doesn't support it

There's no point in enabling PSR when the panel doesn't support it.

This also avoids a problem when PSR gets enabled when a CRTC is being
disabled, because sometimes in that situation the DSP_HOLD_VALID_INTR
interrupt on which we wait will never arrive. This was observed on
RK3288 with a panel without PSR (veyron-jaq Chromebook).

It's very easy to reproduce by running the kms_rmfb test in IGT a few
times.

Cc: Yakir Yang <ykk@rock-chips.com>
Reviewed-by: Sean Paul <seanpaul@chromium.org>
Signed-off-by: Tomeu Vizoso <tomeu.vizoso@collabora.com>
Signed-off-by: Archit Taneja <architt@codeaurora.org>
Link: http://patchwork.freedesktop.org/patch/msgid/1474639600-30090-2-git-send-email-tomeu.vizoso@collabora.com

drm/bridge: analogix_dp: Add analogix_dp_psr_supported

So users know whether PSR should be enabled or not.

Cc: Yakir Yang <ykk@rock-chips.com>
Reviewed-by: Sean Paul <seanpaul@chromium.org>
Signed-off-by: Tomeu Vizoso <tomeu.vizoso@collabora.com>
Signed-off-by: Archit Taneja <architt@codeaurora.org>
Link: http://patchwork.freedesktop.org/patch/msgid/1474639600-30090-1-git-send-email-tomeu.vizoso@collabora.com

drm/fb-helper: add DRM_FB_HELPER_DEFAULT_OPS for fb_ops

The define DRM_FB_HELPER_DEFAULT_OPS provides the drm_fb_helper default
implementations for functions in struct fb_ops. A drm driver can use it
like:

    static struct fb_ops drm_fbdev_cma_ops = {
        .owner          = THIS_MODULE,
        DRM_FB_HELPER_DEFAULT_OPS,
        /* driver specific implementations */
    };

Suggested-by: Daniel Vetter <daniel@ffwll.ch>
Signed-off-by: Stefan Christ <contact@stefanchrist.eu>
Reviewed-by: Gustavo Padovan <gustavo.padovan@collabora.co.uk>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Link: http://patchwork.freedesktop.org/patch/msgid/1475182136-15191-2-git-send-email-contact@stefanchrist.eu

drm: Document caveats around atomic event handling

It's not that obvious how a driver can all race the atomic commit with
handling the completion event. And there's unfortunately a pile of
drivers with rather bad event handling which misdirect people into the
wrong direction.

Try to remedy this by documenting everything better.

v2: Type fixes Alex spotted.

v3: More typos Alex spotted.

Cc: Andrzej Hajda <a.hajda@samsung.com>
Cc: Alex Deucher <alexdeucher@gmail.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Daniel Vetter <daniel.vetter@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1475229896-6047-1-git-send-email-daniel.vetter@ffwll.ch

uapi: add missing install of sync_file.h

As part of the sync framework destaging, the sync_file.h header
was moved, but an entry was not added on Kbuild to install it.
This patch resolves this omission so that "make headers_install"
installs this header.

Fixes: 460bfc41fd52 ("dma-buf/sync_file: de-stage sync_file headers")
Reported-by: Michael Ellerman <mpe@ellerman.id.au>
Reviewed-by: Gustavo Padovan <gustavo.padovan@collabora.co.uk>
Signed-off-by: Emilio López <emilio.lopez@collabora.co.uk>
Signed-off-by: Sean Paul <seanpaul@chromium.org>
Link: http://patchwork.freedesktop.org/patch/msgid/20160927143142.8975-1-emilio.lopez@collabora.co.uk

drm: Simplify drm_printk to reduce object size quite a bit

Remove function name and special " *ERROR*" from argument list

$ size drivers/gpu/drm/built-in.o* (x86-32 defconfig, most drm selected)
   text    data     bss     dec     hex filename
5635366  182579   14328 5832273  58fe51 drivers/gpu/drm/built-in.o.new
5779552  182579   14328 5976459  5b318b drivers/gpu/drm/built-in.o.old

Using "%ps", __builtin_return_address(0) is the same as "%s", __func__
except for static inlines, but it's more or less the same output.

Miscellanea:

o Convert args... to ##__VA_ARGS__
o The equivalent DRM_DEV_<FOO> macros are rarely used and not
  worth conversion

Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Joe Perches <joe@perches.com>
Signed-off-by: Sean Paul <seanpaul@chromium.org>
Link: http://patchwork.freedesktop.org/patch/msgid/01f976d5ab93c985756fc1b2e83656fb0a2a28c8.1474856262.git.joe@perches.com

drm/i915: Account for sink max TMDS clock when checking the port clock

It's perfectly legal for the sink to support 12bpc only for
some lower resolution modes, while the higher resolution modes
can only be used with 8bpc. So let's take the sink's max TMDS clock
into account before we go and decide that a particular mode can
be used with 12bpc.

Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Ander Conselvan de Oliveira <conselvan2@gmail.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Link: http://patchwork.freedesktop.org/patch/msgid/1475070703-6435-11-git-send-email-ville.syrjala@linux.intel.com

drm/i915: Replace a bunch of connector->base.display_info with a local variable

Reduce the eyesore with a local variable.

Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Ander Conselvan de Oliveira <conselvan2@gmail.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Link: http://patchwork.freedesktop.org/patch/msgid/1475070703-6435-10-git-send-email-ville.syrjala@linux.intel.com

drm/edid: Move dvi_dual/max_tmds_clock parsing out from drm_edid_to_eld()

drm_edid_to_eld() is just mean to cook up the ELD for the audio driver,
so having it parse non-audio related stuff seems just wrong, and
potentially could lead to that information not being even filled out
if the function doesn't even get called. Let's move that stuff to the
place where we parse the color formats and whatnot from the CEA ext
block.

Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Acked-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Link: http://patchwork.freedesktop.org/patch/msgid/1475070703-6435-9-git-send-email-ville.syrjala@linux.intel.com

drm/edid: Clear the old cea_rev when there's no CEA extension in the new EDID

It's not a good idea to leave stale cea_rev in the drm_display_info. The
current EDID might not even have a CEA ext block in which case we'd end
up leaving the stale value in place.

Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Acked-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Link: http://patchwork.freedesktop.org/patch/msgid/1475070703-6435-8-git-send-email-ville.syrjala@linux.intel.com

drm/edid: Reduce the number of times we parse the CEA extension block

Instead of parsing parts of the CEA extension block in two places
to determine supported color formats and whatnot, let's just
consolidate it to one function. This also makes it possible to neatly
flatten drm_assign_hdmi_deep_color_info().

Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Acked-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Link: http://patchwork.freedesktop.org/patch/msgid/1475070703-6435-7-git-send-email-ville.syrjala@linux.intel.com

drm/edid: Don't pass around drm_display_info needlessly

We already pass the connector to drm_add_display_info() and
drm_assign_hdmi_deep_color_info(), so passing the
connector->display_info also is pointless.

Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Acked-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Link: http://patchwork.freedesktop.org/patch/msgid/1475070703-6435-6-git-send-email-ville.syrjala@linux.intel.com

drm/edid: Move dvi_dual/max_tmds_clock to drm_display_info

We have the drm_display_info for storing information about the sink, so
let's move dvi_dual and max_tmds_clock in there.

v2: Deal with superfluous code shuffling
    Document dvi_dual and max_tmds_clock too

Cc: Alex Deucher <alexander.deucher@amd.com>
Cc: "Christian König" <christian.koenig@amd.com>
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Christian König <christian.koenig@amd.com> (v1)
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Link: http://patchwork.freedesktop.org/patch/msgid/1475070703-6435-5-git-send-email-ville.syrjala@linux.intel.com

drm/edid: Make max_tmds_clock kHz instead of MHz

We generally store clocks in kHz, so let's do that for the
HDMI max TMDS clock value as well. Less surpising.

v2: Deal with superfluous code shuffling

Cc: Alex Deucher <alexander.deucher@amd.com>
Cc: "Christian König" <christian.koenig@amd.com>
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Christian König <christian.koenig@amd.com> (v1)
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Link: http://patchwork.freedesktop.org/patch/msgid/1475070703-6435-4-git-send-email-ville.syrjala@linux.intel.com

drm/edid: Clear old dvi_dual/max_tmds_clock before parsing the new EDID

Clear out old max_tmds_clock and dvi_dual information (possibly from a
previous EDID) before parsing the current EDID. Tne current EDID might
not even have these in its HDMI VSDB, which would mean that we'd leave
the old stale values in place.

Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Acked-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Link: http://patchwork.freedesktop.org/patch/msgid/1475070703-6435-3-git-send-email-ville.syrjala@linux.intel.com

drm/edid: Clear old audio latency values before parsing the new EDID

Clear out stale audio latency information (potentially from a previous
EDID) before constructing the ELD from the EDID.

Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Acked-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Link: http://patchwork.freedesktop.org/patch/msgid/1475070703-6435-2-git-send-email-ville.syrjala@linux.intel.com

drm: Convert prime dma-buf <-> handle to rbtree

Currently we use a linear walk to lookup a handle and return a dma-buf,
and vice versa. A long overdue TODO task is to convert that to a
hashtable. Since the initial implementation of dma-buf/prime, we now
have resizeable hashtables we can use (and now a future task is to RCU
enable the lookup!). However, this patch opts to use an rbtree instead
to provide O(lgN) lookups (and insertion, deletion). rbtrees were chosen
over using the RCU backed resizable hashtable to firstly avoid the
reallocations (rbtrees can be embedded entirely within the parent
struct) and to favour simpler code with predictable worst case
behaviour. In simple testing, the difference between using the constant
lookup and insertion of the rhashtable and the rbtree was less than 10%
of the wall time (igt/benchmarks/prime_lookup) - both are dramatic
improvements over the existing linear lists.

v2: Favour rbtree over rhashtable

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=94631
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Sean Paul <seanpaul@chromium.org>
Cc: David Herrmann <dh.herrmann@gmail.com>
Reviewed-by: David Herrmann <dh.herrmann@gmail.com>
Reviewed-by: Sean Paul <seanpaul@chromium.org>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Link: http://patchwork.freedesktop.org/patch/msgid/20160926204414.23222-1-chris@chris-wilson.co.uk

drm/mediatek: mark symbols static where possible

We get 4 warnings when building kernel with W=1:
drivers/gpu/drm/mediatek/mtk_hdmi.c:1089:6: warning: no previous prototype for 'mtk_hdmi_audio_enable' [-Wmissing-prototypes]
drivers/gpu/drm/mediatek/mtk_hdmi.c:1095:6: warning: no previous prototype for 'mtk_hdmi_audio_disable' [-Wmissing-prototypes]
drivers/gpu/drm/mediatek/mtk_hdmi.c:1101:5: warning: no previous prototype for 'mtk_hdmi_audio_set_param' [-Wmissing-prototypes]
drivers/gpu/drm/mediatek/mtk_hdmi.c:1627:5: warning: no previous prototype for 'mtk_hdmi_audio_digital_mute' [-Wmissing-prototypes]

In fact, both functions are only used in the file in which they are
declared and don't need a declaration, but can be made static.
So this patch marks both functions with 'static'.

Signed-off-by: Baoyou Xie <baoyou.xie@linaro.org>
[seanpaul fixed checkpatch warning for argument alignment]
Signed-off-by: Sean Paul <seanpaul@chromium.org>
Link: http://patchwork.freedesktop.org/patch/msgid/1474789109-22010-2-git-send-email-baoyou.xie@linaro.org

drm/rockchip: mark symbols static where possible

We get 2 warnings when building kernel with W=1:
drivers/gpu/drm/rockchip/rockchip_drm_drv.c:309:6: warning: no previous prototype for 'rockchip_drm_fb_suspend' [-Wmissing-prototypes]
drivers/gpu/drm/rockchip/rockchip_drm_drv.c:318:6: warning: no previous prototype for 'rockchip_drm_fb_resume' [-Wmissing-prototypes]

In fact, these functions are only used in the file in which they are
declared and don't need a declaration, but can be made static.
So this patch marks these functions with 'static'.

Signed-off-by: Baoyou Xie <baoyou.xie@linaro.org>
Signed-off-by: Sean Paul <seanpaul@chromium.org>
Link: http://patchwork.freedesktop.org/patch/msgid/1474789388-3284-1-git-send-email-baoyou.xie@linaro.org

drm/rockchip: add missing header dependencies

We get 2 warnings when building kernel with W=1:
drivers/gpu/drm/rockchip/rockchip_drm_fbdev.c:130:5: warning: no previous prototype for 'rockchip_drm_fbdev_init' [-Wmissing-prototypes]
drivers/gpu/drm/rockchip/rockchip_drm_fbdev.c:173:6: warning: no previous prototype for 'rockchip_drm_fbdev_fini' [-Wmissing-prototypes]

In fact, these functions are declared
in drivers/gpu/drm/rockchip/rockchip_drm_fbdev.h,
so this patch adds missing header dependencies.

Signed-off-by: Baoyou Xie <baoyou.xie@linaro.org>
Signed-off-by: Sean Paul <seanpaul@chromium.org>
Link: http://patchwork.freedesktop.org/patch/msgid/1474789109-22010-1-git-send-email-baoyou.xie@linaro.org

drm: virtio: reinstate drm_virtio_set_busid()

Before commit a325725633c2 ("drm: Lobotomize set_busid nonsense for !pci
drivers"), several DRM drivers for platform devices used to expose an
explicit "drm_driver.set_busid" callback, invariably backed by
drm_platform_set_busid().

Commit a325725633c2 removed drm_platform_set_busid(), along with the
referring .set_busid field initializations. This was justified because
interchangeable functionality had been implemented in drm_dev_alloc() /
drm_dev_init(), which DRM_IOCTL_SET_VERSION would rely on going forward.

However, commit a325725633c2 also removed drm_virtio_set_busid(), for
which the same consolidation was not appropriate: this .set_busid callback
had been implemented with drm_pci_set_busid(), and not
drm_platform_set_busid(). The error regressed Xorg/xserver on QEMU's
"virtio-vga" card; the drmGetBusid() function from libdrm would no longer
return stable PCI identifiers like "pci:0000:00:02.0", but rather unstable
platform ones like "virtio0".

Reinstate drm_virtio_set_busid() with judicious use of

  git checkout -p a325725633c2^ -- drivers/gpu/drm/virtio

Cc: Daniel Vetter <daniel.vetter@intel.com>
Cc: Emil Velikov <emil.l.velikov@gmail.com>
Cc: Gerd Hoffmann <kraxel@redhat.com>
Cc: Gustavo Padovan <gustavo.padovan@collabora.co.uk>
Cc: Hans de Goede <hdegoede@redhat.com>
Cc: Joachim Frieben <jfrieben@hotmail.com>
Cc: stable@vger.kernel.org # v4.8
Reported-by: Joachim Frieben <jfrieben@hotmail.com>
Fixes: a325725633c26aa66ab940f762a6b0778edf76c0
Ref: https://bugzilla.redhat.com/show_bug.cgi?id=1366842
Signed-off-by: Laszlo Ersek <lersek@redhat.com>
Reviewed-by: Emil Velikov <emil.l.velikov@gmail.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>

drm: Undo damage to page_flip_ioctl

I screwed up rebasing of my patch in

commit 43968d7b806d7a7e021261294c583a216fddf0e5
Author: Daniel Vetter <daniel.vetter@ffwll.ch>
Date:   Wed Sep 21 10:59:24 2016 +0200

    drm: Extract drm_plane.[hc]

which meant on error paths drm_crtc_vblank_put could be called without
a get, leading to an underrun of the refcount.

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=98020
Reported-and-tested-by: Andy Furniss <adf.lists@gmail.com>
Cc: Sean Paul <seanpaul@chromium.org>
Cc: Michel Dänzer <michel@daenzer.net>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Link: http://patchwork.freedesktop.org/patch/msgid/20161003082827.11586-1-daniel.vetter@ffwll.ch
Signed-off-by: Dave Airlie <airlied@redhat.com>

drm: Restore lost drm_framebuffer_unreference in drm_mode_page_flip_ioctl

Commit 43968d7b806d ("drm: Extract drm_plane.[hc]") was not the simple
cut'n'paste we presumed, somehow it introduced a leak of the page flip
target's framebuffer.

Fixes: 43968d7b806d ("drm: Extract drm_plane.[hc]")
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Daniel Vetter <daniel.vetter@intel.com>
Cc: Sean Paul <seanpaul@chromium.org>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Link: http://patchwork.freedesktop.org/patch/msgid/20160928222500.11827-1-chris@chris-wilson.co.uk
Signed-off-by: Dave Airlie <airlied@redhat.com>

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

This pull request includes,
    - Code refactoring on HDMI DDC and PHY.
    - Regression fixup on deadlock issue with G2D pm integration.
    - Fixup on page fault issue with wait_for_vblank mechianism specific to Exynos drm.
    - And some cleanups.

* 'exynos-drm-next' of git://git.kernel.org/pub/scm/linux/kernel/git/daeinki/drm-exynos:
  drm/exynos: g2d: simplify g2d_free_runqueue_node()
  drm/exynos: g2d: use autosuspend mode for PM runtime
  drm/exynos: g2d: wait for engine to finish
  drm/exynos: g2d: remove runqueue nodes in g2d_{close,remove}()
  drm/exynos: g2d: move PM management to runqueue worker
  Revert "drm/exynos: g2d: fix system and runtime pm integration"
  drm/exynos: use drm core to handle page-flip event
  drm/exynos: mark exynos_dp_crtc_clock_enable() static
  drm/exynos/fimd: add clock rate checking
  drm/exynos: fix pending update handling
  drm/exynos/vidi: use timer for vblanks instead of sleeping worker
  drm/exynos: g2d: beautify probing message
  drm/exynos: mixer: simplify loop in vp_win_reset()
  drm/exynos: mixer: convert booleans to flags in mixer context
  gpu: drm: exynos_hdmi: Remove duplicate initialization of regulator bulk consumer
  gpu: drm: exynos_hdmi: Move PHY logic into single function
  gpu: drm: exynos_hdmi: Move DDC logic into single function

drm/exynos: g2d: simplify g2d_free_runqueue_node()

The function is never called with zero 'runqueue_node'.

Signed-off-by: Tobias Jakobi <tjakobi@math.uni-bielefeld.de>
Acked-by: Marek Szyprowski <m.szyprowski@samsung.com>
Signed-off-by: Inki Dae <inki.dae@samsung.com>

drm/exynos: g2d: use autosuspend mode for PM runtime

The runqueue worker currently issues a get() when a new
node is processed, and a put() once a node is completed.

The corresponding suspend and resume calls currently only
do clock gating, but with the upcoming introduction of
IOMMU runpm also the corresponding IOMMU domain gets
enabled (for get()) and disabled (for put()). This
introduces performance regressions with we mitigate here.

Switch PM runtime to autosuspend, such that clock gating
and IOMMU control only happens when the engine is idle for
a 'long' time.

Signed-off-by: Tobias Jakobi <tjakobi@math.uni-bielefeld.de>
Acked-by: Marek Szyprowski <m.szyprowski@samsung.com>
Signed-off-by: Inki Dae <inki.dae@samsung.com>

drm/exynos: g2d: wait for engine to finish

While the engine works on a runqueue node it does memory access to
the buffers associated with that node.
Make sure that the engine is idle when g2d_close() and/or
g2d_remove() are called, i.e. buffer associated with the process (for
g2d_close()), or all buffers (for g2d_remove()) can be safely be
unmapped.

We have to take into account that the engine might be in an undefined
state, i.e. it hangs and doesn't become idle. In this case, we issue
a hardware reset to return the hardware and the driver context into a
proper state.

Signed-off-by: Tobias Jakobi <tjakobi@math.uni-bielefeld.de>
Acked-by: Marek Szyprowski <m.szyprowski@samsung.com>
Signed-off-by: Inki Dae <inki.dae@samsung.com>

drm/exynos: g2d: remove runqueue nodes in g2d_{close,remove}()

The driver might be closed (and/or removed) while there are still
nodes queued for processing.
Make sure to remove these nodes, which means all of them in
the case of g2d_remove() and only those belonging to the
corresponding process in g2d_close().

Signed-off-by: Tobias Jakobi <tjakobi@math.uni-bielefeld.de>
Acked-by: Marek Szyprowski <m.szyprowski@samsung.com>
Signed-off-by: Inki Dae <inki.dae@samsung.com>

drm/exynos: g2d: move PM management to runqueue worker

Do all pm_runtime_{get,put}() calls in the runqueue worker.
Also keep track of the engine's idle/busy state.

Signed-off-by: Tobias Jakobi <tjakobi@math.uni-bielefeld.de>
Acked-by: Marek Szyprowski <m.szyprowski@samsung.com>
Signed-off-by: Inki Dae <inki.dae@samsung.com>

Revert "drm/exynos: g2d: fix system and runtime pm integration"

This reverts commit b05984e21a7e000bf5074ace00d7a574944b2c16.

The change, i.e. merging the sleep and runpm operations, produces
a deadlock situation:
(1) exynos_g2d_exec_ioctl() prepares a runqueue node and
    calls g2d_exec_runqueue()
(2) g2d_exec_runqueue() calls g2d_dma_start() which gets
    runtime PM sync
(3) runtime PM core calls g2d_runtime_resume()
(4) g2d_runtime_resume() calls g2d_exec_runqueue(), which
    loops back to (2)

Due to mutexes that are in place, a deadlock situation is created.

Signed-off-by: Tobias Jakobi <tjakobi@math.uni-bielefeld.de>
Acked-by: Marek Szyprowski <m.szyprowski@samsung.com>
Signed-off-by: Inki Dae <inki.dae@samsung.com>

drm/exynos: use drm core to handle page-flip event

Exynos DRM framework handled page-flip event with custom code.
The patch replaces it with drm-core vblank queue.

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

drm/exynos: mark exynos_dp_crtc_clock_enable() static

We get 1 warning when building kernel with W=1:
drivers/gpu/drm/exynos/exynos_dp.c:46:5: warning: no previous prototype for 'exynos_dp_crtc_clock_enable' [-Wmissing-prototypes]

In fact, this function is only used in the file in which it is
declared and don't need a declaration, but can be made static.
So this patch marks it 'static'.

Signed-off-by: Baoyou Xie <baoyou.xie@linaro.org>
Signed-off-by: Inki Dae <inki.dae@samsung.com>

drm/exynos/fimd: add clock rate checking

In case of some platforms fimd clocks can be configured to
very low values, as a result refresh rate can be very low and
driver/drm-core will timeout waiting for vblanks, it will result
in premature removal of framebuffers and will cause oopses.
The patch adds atomic_check callback to fimd to prevent setting
such modes.

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

drm/exynos: fix pending update handling

Exynos DRM devices update their registers at vblank time. Exynos-DRM uses
custom mechanism to wait for vblank. This mechanism is error prone -
variables are not updated atomically. As a result in certain circumstances
user space can try to free buffers which are still in use by hardware,
in such cases IOMMU can throw OOPS.
The patch instead of fixing the mechanism replaces it with drm core helper.

Signed-off-by: Andrzej Hajda <a.hajda@samsung.com>
Reviewed-by: Gustavo Padovan <gustavo.padovan@collabora.co.uk>
Signed-off-by: Inki Dae <inki.dae@samsung.com>

drm/exynos/vidi: use timer for vblanks instead of sleeping worker

VIDI driver uses fake vblank handler to generate vblank events.
It was implemented using worker which slept for vblank time, additionally
it did not work if there were no page flips. The patch replaces it with
timer, uses drm_crtc_vblank_(on|off) helpers to manage it and fixes
behavior for non-page-flip cases.
This change allows further improvements of vblank in exynos-drm framework.

Signed-off-by: Andrzej Hajda <a.hajda@samsung.com>
Reviewed-by: Gustavo Padovan <gustavo.padovan@collabora.co.uk>
Signed-off-by: Inki Dae <inki.dae@samsung.com>

drm/exynos: g2d: beautify probing message

Apply some 'make-up' in g2d_probe().

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

drm/exynos: mixer: simplify loop in vp_win_reset()

A simple while loop should do the same here.

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

drm/exynos: mixer: convert booleans to flags in mixer context

The mixer context struct already has a 'flags' field, so
we can use it to store the 'interlace', 'vp_enabled' and
'has_sclk' booleans.
We use the non-atomic helper functions to access these bits.

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>

gpu: drm: exynos_hdmi: Remove duplicate initialization of regulator bulk consumer

The helper, devm_regulator_bulk_get() initializes the consumer as NULL,
so this code can be ignored.

Signed-off-by: Milo Kim <woogyom.kim@gmail.com>
Reviewed-by: Andrzej Hajda <a.hajda@samsung.com>
Signed-off-by: Inki Dae <inki.dae@samsung.com>

gpu: drm: exynos_hdmi: Move PHY logic into single function

Paring DT properties and getting PHY IO (memory mapped or I2C) in one
function.

Signed-off-by: Milo Kim <woogyom.kim@gmail.com>
Reviewed-by: Andrzej Hajda <a.hajda@samsung.com>
Signed-off-by: Inki Dae <inki.dae@samsung.com>

gpu: drm: exynos_hdmi: Move DDC logic into single function

Paring DT properties and getting the I2C adapter in one function.

Signed-off-by: Milo Kim <woogyom.kim@gmail.com>
Reviewed-by: Andrzej Hajda <a.hajda@samsung.com>
Signed-off-by: Inki Dae <inki.dae@samsung.com>

Merge branch 'drm-next-4.9' of git://people.freedesktop.org/~agd5f/linux into drm-next

Some additional fixes for 4.9:
- The rest of Christian's GTT rework which fixes a long standing bug
  in the GPUVM code among other things
- Changes to the pci shutdown callbacks for certain hypervisors
- Fix hpd interrupt storms on eDP panels which have the hpd interrupt
  enabled by the bios
- misc cleanups and bug fixes

* 'drm-next-4.9' of git://people.freedesktop.org/~agd5f/linux: (33 commits)
  drm/radeon: always apply pci shutdown callbacks
  drm/amdgpu: always apply pci shutdown callbacks (v2)
  drm/amdgpu: improve VM PTE trace points
  drm/amdgpu: fix GART_DEBUGFS define
  drm/amdgpu: free userptrs even if GTT isn't bound
  drm/amd/amdgpu: Various cleanups for DCEv6
  drm/amdgpu: fix BO move offsets
  drm/amdgpu: fix amdgpu_move_blit on 32bit systems
  drm/amdgpu: fix gtt_mgr bo's offset
  drm/amdgpu: fix initializing the VM BO shadow
  drm/amdgpu: fix initializing the VM last eviction counter
  drm/amdgpu: cleanup VM shadow BO unreferencing
  drm/amdgpu: allocate GTT space for shadow VM page tables
  drm/amdgpu: rename all rbo variable to abo v2
  drm/amdgpu: remove unused member from struct amdgpu_bo
  drm/amdgpu: add a custom GTT memory manager v2
  drm/amdgpu/dce6: disable hpd on local panels
  drm/amdgpu/dce8: disable hpd on local panels
  drm/amdgpu/dce11: disable hpd on local panels
  drm/amdgpu/dce10: disable hpd on local panels
  ...

Merge tag 'drm/tegra/for-4.9-rc1' of git://anongit.freedesktop.org/tegra/linux into drm-next

drm/tegra: Changes for v4.9-rc1

One bugfix that avoids overwriting the Y plane base address when
displaying buffers with one of the YUV/YVU formats.

* tag 'drm/tegra/for-4.9-rc1' of git://anongit.freedesktop.org/tegra/linux:
  drm/tegra: Fix window[0] base address corruption

Merge tag 'drm/panel/for-4.9-rc1' of git://anongit.freedesktop.org/tegra/linux into drm-next

drm/panel: Changes for v4.9-rc1

Adds support for one more panel to the simple-panel driver, fixes up a
couple of delays and flags for existing panels and finally adds a new
driver for the DSI panel found on Nexus 7 devices.

* tag 'drm/panel/for-4.9-rc1' of git://anongit.freedesktop.org/tegra/linux:
  drm/panel: Add JDI LT070ME05000 WUXGA DSI Panel
  dt-bindings: Add JDI LT070ME05000 panel bindings
  drm/panel: simple: Fix bus_format for the Olimex LCD-OLinuXino-4.3TS
  drm/panel: simple-panel: Add delay timings for Starry KR122EA0SRA
  drm/panel: simple: Fix bus flags for Ortustech com43h4m85ulc
  drm/panel: simple: Add Innolux G101ICE-L01 panel
  drm/panel: simple: Add delay timing for Sharp LQ123P1JX31
  drm/dsi: Implement DCS set/get display brightness
  drm/dsi: Order DCS helpers by command code

drm/radeon: always apply pci shutdown callbacks

We can't properly detect all hypervisors and we
need this to properly tear down the hardware.

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

drm/amdgpu: always apply pci shutdown callbacks (v2)

We can't properly detect all hypervisors and we
need this to properly tear down the hardware.

v2: trivial warning fix

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

drm/amdgpu: improve VM PTE trace points

Use a separate one for the copy operation and
log all the interesting parameters.

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: fix GART_DEBUGFS define

Obviously missed during the rename.

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: free userptrs even if GTT isn't bound

This fixes a memory leak since binding GTT only on demand.

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/amd/amdgpu: Various cleanups for DCEv6

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: fix BO move offsets

It's pretty pointless to get the offset first and then initialize it.

Should fix issues with the new GTT manager.

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

drm/amdgpu: fix amdgpu_move_blit on 32bit systems

This bug seems to be present for a very long time.

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>
Cc: stable@vger.kernel.org

drm/amdgpu: fix gtt_mgr bo's offset

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

drm/amdgpu: fix initializing the VM BO shadow

We need to clear the shadows as well.

Signed-off-by: Christian König <christian.koenig@amd.com>
Acked-by: Edward O'Callaghan <funfunctor@folklore1984.net>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: fix initializing the VM last eviction counter

Close a very small window where something can go wrong.

Signed-off-by: Christian König <christian.koenig@amd.com>
Acked-by: Edward O'Callaghan <funfunctor@folklore1984.net>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: cleanup VM shadow BO unreferencing

Unreference the shadow BOs in the error path as well and drop the NULL checks.

Signed-off-by: Christian König <christian.koenig@amd.com>
Acked-by: Edward O'Callaghan <funfunctor@folklore1984.net>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: allocate GTT space for shadow VM page tables

We need to access those with the system domain.

Fixes fallout from only allocating GTT space on demand.

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: rename all rbo variable to abo v2

Just to cleanup some radeon leftovers.

sed -i "s/rbo/abo/g" drivers/gpu/drm/amd/amdgpu/*.c
sed -i "s/rbo/abo/g" drivers/gpu/drm/amd/amdgpu/*.h

v2: rebased

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 unused member from struct amdgpu_bo

Not used in a while.

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: add a custom GTT memory manager v2

Only allocate address space when we really need it.

v2: fix a typo, add correct function description,
    stop leaking the node in the error case.

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

drm/amdgpu/dce6: disable hpd on local panels

Otherwise we can get a hotplug interrupt storm when
we turn the panel off if hpd interrupts were enabled
by the bios.

bug:
https://bugs.freedesktop.org/show_bug.cgi?id=97471

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

drm/amdgpu/dce8: disable hpd on local panels

Otherwise we can get a hotplug interrupt storm when
we turn the panel off if hpd interrupts were enabled
by the bios.

bug:
https://bugs.freedesktop.org/show_bug.cgi?id=97471

Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Cc: stable@vger.kernel.org

drm/amdgpu/dce11: disable hpd on local panels

Otherwise we can get a hotplug interrupt storm when
we turn the panel off if hpd interrupts were enabled
by the bios.

bug:
https://bugs.freedesktop.org/show_bug.cgi?id=97471

Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Cc: stable@vger.kernel.org

drm/amdgpu/dce10: disable hpd on local panels

Otherwise we can get a hotplug interrupt storm when
we turn the panel off if hpd interrupts were enabled
by the bios.

bug:
https://bugs.freedesktop.org/show_bug.cgi?id=97471

Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Cc: stable@vger.kernel.org

drm/amdgpu/si/dpm: fix phase shedding setup

Used the wrong index to setup the phase shedding mask.

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

drm/radeon/si/dpm: fix phase shedding setup

Used the wrong index to setup the phase shedding mask.

Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Cc: stable@vger.kernel.org

drm/amdgpu/si/dpm: sync up quirks from radeon

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