drm/i915: Convert wait_for(I915_READ(reg)) to intel_wait_for_register()

By using the out-of-line intel_wait_for_register() not only do we can
efficiency from using the hybrid wait_for() contained within, but we
avoid code bloat from the numerous inlined loops, in total (all patches):

   text    data     bss     dec     hex filename
1078551    4557     416 1083524  108884 drivers/gpu/drm/i915/i915.ko
1070775    4557     416 1075748  106a24 drivers/gpu/drm/i915/i915.ko

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1467297225-21379-18-git-send-email-chris@chris-wilson.co.uk

drm/i915: Convert wait_for(I915_READ(reg)) to intel_wait_for_register()

By using the out-of-line intel_wait_for_register() not only do we can
efficiency from using the hybrid wait_for() contained within, but we
avoid code bloat from the numerous inlined loops, in total (all patches):

   text    data     bss     dec     hex filename
1078551    4557     416 1083524  108884 drivers/gpu/drm/i915/i915.ko
1070775    4557     416 1075748  106a24 drivers/gpu/drm/i915/i915.ko

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1467297225-21379-17-git-send-email-chris@chris-wilson.co.uk

drm/i915: Convert wait_for(I915_READ(reg)) to intel_wait_for_register()

By using the out-of-line intel_wait_for_register() not only do we can
efficiency from using the hybrid wait_for() contained within, but we
avoid code bloat from the numerous inlined loops, in total (all patches):

   text    data     bss     dec     hex filename
1078551    4557     416 1083524  108884 drivers/gpu/drm/i915/i915.ko
1070775    4557     416 1075748  106a24 drivers/gpu/drm/i915/i915.ko

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1467297225-21379-16-git-send-email-chris@chris-wilson.co.uk

drm/i915: Convert wait_for(I915_READ(reg)) to intel_wait_for_register()

By using the out-of-line intel_wait_for_register() not only do we can
efficiency from using the hybrid wait_for() contained within, but we
avoid code bloat from the numerous inlined loops, in total (all patches):

   text    data     bss     dec     hex filename
1078551    4557     416 1083524  108884 drivers/gpu/drm/i915/i915.ko
1070775    4557     416 1075748  106a24 drivers/gpu/drm/i915/i915.ko

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1467297225-21379-15-git-send-email-chris@chris-wilson.co.uk

drm/i915: Convert wait_for(I915_READ(reg)) to intel_wait_for_register()

By using the out-of-line intel_wait_for_register() not only do we can
efficiency from using the hybrid wait_for() contained within, but we
avoid code bloat from the numerous inlined loops, in total (all patches):

   text    data     bss     dec     hex filename
1078551    4557     416 1083524  108884 drivers/gpu/drm/i915/i915.ko
1070775    4557     416 1075748  106a24 drivers/gpu/drm/i915/i915.ko

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1467297225-21379-14-git-send-email-chris@chris-wilson.co.uk

drm/i915: Convert wait_for(I915_READ(reg)) to intel_wait_for_register()

By using the out-of-line intel_wait_for_register() not only do we can
efficiency from using the hybrid wait_for() contained within, but we
avoid code bloat from the numerous inlined loops, in total (all patches):

   text    data     bss     dec     hex filename
1078551    4557     416 1083524  108884 drivers/gpu/drm/i915/i915.ko
1070775    4557     416 1075748  106a24 drivers/gpu/drm/i915/i915.ko

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1467297225-21379-13-git-send-email-chris@chris-wilson.co.uk

drm/i915: Convert wait_for(I915_READ(reg)) to intel_wait_for_register()

By using the out-of-line intel_wait_for_register() not only do we can
efficiency from using the hybrid wait_for() contained within, but we
avoid code bloat from the numerous inlined loops, in total (all patches):

   text    data     bss     dec     hex filename
1078551    4557     416 1083524  108884 drivers/gpu/drm/i915/i915.ko
1070775    4557     416 1075748  106a24 drivers/gpu/drm/i915/i915.ko

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1467297225-21379-12-git-send-email-chris@chris-wilson.co.uk

drm/i915: Convert wait_for(I915_READ(reg)) to intel_wait_for_register()

By using the out-of-line intel_wait_for_register() not only do we can
efficiency from using the hybrid wait_for() contained within, but we
avoid code bloat from the numerous inlined loops, in total (all patches):

   text    data     bss     dec     hex filename
1078551    4557     416 1083524  108884 drivers/gpu/drm/i915/i915.ko
1070775    4557     416 1075748  106a24 drivers/gpu/drm/i915/i915.ko

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1467297225-21379-11-git-send-email-chris@chris-wilson.co.uk

drm/i915: Convert wait_for(I915_READ(reg)) to intel_wait_for_register()

By using the out-of-line intel_wait_for_register() not only do we can
efficiency from using the hybrid wait_for() contained within, but we
avoid code bloat from the numerous inlined loops, in total (all patches):

   text    data     bss     dec     hex filename
1078551    4557     416 1083524  108884 drivers/gpu/drm/i915/i915.ko
1070775    4557     416 1075748  106a24 drivers/gpu/drm/i915/i915.ko

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1467297225-21379-10-git-send-email-chris@chris-wilson.co.uk

drm/i915: Convert wait_for(I915_READ(reg)) to intel_wait_for_register()

By using the out-of-line intel_wait_for_register() not only do we can
efficiency from using the hybrid wait_for() contained within, but we
avoid code bloat from the numerous inlined loops, in total (all patches):

   text    data     bss     dec     hex filename
1078551    4557     416 1083524  108884 drivers/gpu/drm/i915/i915.ko
1070775    4557     416 1075748  106a24 drivers/gpu/drm/i915/i915.ko

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1467297225-21379-9-git-send-email-chris@chris-wilson.co.uk

drm/i915: Convert wait_for(I915_READ(reg)) to intel_wait_for_register()

By using the out-of-line intel_wait_for_register() not only do we can
efficiency from using the hybrid wait_for() contained within, but we
avoid code bloat from the numerous inlined loops, in total (all patches):

   text    data     bss     dec     hex filename
1078551    4557     416 1083524  108884 drivers/gpu/drm/i915/i915.ko
1070775    4557     416 1075748  106a24 drivers/gpu/drm/i915/i915.ko

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1467297225-21379-8-git-send-email-chris@chris-wilson.co.uk

drm/i915: Convert wait_for(I915_READ(reg)) to intel_wait_for_register()

By using the out-of-line intel_wait_for_register() not only do we can
efficiency from using the hybrid wait_for() contained within, but we
avoid code bloat from the numerous inlined loops, in total (all patches):

   text    data     bss     dec     hex filename
1078551    4557     416 1083524  108884 drivers/gpu/drm/i915/i915.ko
1070775    4557     416 1075748  106a24 drivers/gpu/drm/i915/i915.ko

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1467297225-21379-7-git-send-email-chris@chris-wilson.co.uk

drm/i915: Convert wait_for(I915_READ(reg)) to intel_wait_for_register()

By using the out-of-line intel_wait_for_register() not only do we can
efficiency from using the hybrid wait_for() contained within, but we
avoid code bloat from the numerous inlined loops, in total (all patches):

   text    data     bss     dec     hex filename
1078551    4557     416 1083524  108884 drivers/gpu/drm/i915/i915.ko
1070775    4557     416 1075748  106a24 drivers/gpu/drm/i915/i915.ko

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1467297225-21379-6-git-send-email-chris@chris-wilson.co.uk

drm/i915: Convert wait_for(I915_READ(reg)) to intel_wait_for_register()

By using the out-of-line intel_wait_for_register() not only do we can
efficiency from using the hybrid wait_for() contained within, but we
avoid code bloat from the numerous inlined loops, in total (all patches):

   text    data     bss     dec     hex filename
1078551    4557     416 1083524  108884 drivers/gpu/drm/i915/i915.ko
1070775    4557     416 1075748  106a24 drivers/gpu/drm/i915/i915.ko

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1467297225-21379-5-git-send-email-chris@chris-wilson.co.uk

drm/i915: Convert wait_for(I915_READ(reg)) to intel_wait_for_register()

By using the out-of-line intel_wait_for_register() not only do we can
efficiency from using the hybrid wait_for() contained within, but we
avoid code bloat from the numerous inlined loops, in total (all patches):

   text    data     bss     dec     hex filename
1078551    4557     416 1083524  108884 drivers/gpu/drm/i915/i915.ko
1070775    4557     416 1075748  106a24 drivers/gpu/drm/i915/i915.ko

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1467297225-21379-4-git-send-email-chris@chris-wilson.co.uk

drm/i915: Convert wait_for(I915_READ(reg)) to intel_wait_for_register()

By using the out-of-line intel_wait_for_register() not only do we can
efficiency from using the hybrid wait_for() contained within, but we
avoid code bloat from the numerous inlined loops, in total (all patches):

   text    data     bss     dec     hex filename
1078551    4557     416 1083524  108884 drivers/gpu/drm/i915/i915.ko
1070775    4557     416 1075748  106a24 drivers/gpu/drm/i915/i915.ko

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1467297225-21379-3-git-send-email-chris@chris-wilson.co.uk

drm/i915: Convert sandybridge_pcode_*() to use intel_wait_for_register()

We want to replace the inline wait_for() with an out-of-line hybrid
busy/sleep wait_for() in the hopes of speeding up the communication wit
the PCode unit.

Indeed, on my i5-2500s, __gen6_update_ring_freq improves from
6,080,661ns to 8172ns.

v2: Missed using _fw variants for sandybridge_pcode_read()

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Mika Kuoppala <mika.kuoppala@intel.com>
Cc: Ville Syrjälä <ville.syrjala@linux.intel.com>
Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1467297225-21379-2-git-send-email-chris@chris-wilson.co.uk

drm/i915: Use a hybrid scheme for fast register waits

Ville Syrjälä reported that in the majority of wait_for(I915_READ()) he
inspect, most completed within the first couple of reads and that the
delay between those wait_for() reads was the ratelimiting step for many
code paths. For example, __gen6_update_ring_freq() was blamed for
slowing down boot by many milliseconds, but under Ville's scrutiny the
issue was just excessive delay waiting for sandybridge_pcode_write().

We can eliminate the wait by initially using a busyspin upon the register
read and only fallback to the sleeping loop in cases where the hardware
is indeed too slow. A threshold of 2 microseconds is used as the initial
ballpark.

To avoid excessive code bloating from converting every wait_for() into a
hybrid busy/sleep loop, we extend wait_for_register_fw() and export it
for use by other callers.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Mika Kuoppala <mika.kuoppala@intel.com>
Cc: Ville Syrjälä <ville.syrjala@linux.intel.com>
Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1467297225-21379-1-git-send-email-chris@chris-wilson.co.uk

Revert "drm/i915: Use atomic commits for legacy page_flips"

This reverts commit ee042aa40b66d18d465206845b0752c6a617ba3f.

Something appears to be off in the timing, but as far as I can tell it
is not along the event delivery path. The net effect appears to be
rendering flicker (the current render buffer appears on the scanout,
with what appears to be active rendering for a fraction of a frame) and
is causing me a headache.

The cursor is also being stalled by page flips, causing a "heavy mouse"
and jitter.

Daniel Stone did find what appears to the cause of the tearing, in
https://lists.freedesktop.org/archives/intel-gfx/2016-June/099466.html
That is the parameter passed to intel_atomic_commit_tail is the
old_state but we need the new_state to wait upon.

That leaves the question of how the CRC based tests didn't spot the
error (how can we improve our tests?), the issue of legacy cursor
stalling flips, and the issue of flips stalling the cursor. For the
moment, step back until the condundrum of new/old state is reviewed
along with more tests!

Reported-and-tested-by: Steven Newbury <steve@snewbury.org.uk>
Reported-by: Rafael Ristovski <rafael.ristovski@gmail.com>
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=96593
Testcase: igt/kms_cursor_legacy/basic-cursor-vs-flip
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Daniel Vetter <daniel.vetter@ffwll.ch>
Cc: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1466772243-21879-1-git-send-email-chris@chris-wilson.co.uk
Acked-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>

drm/i915: Remove request->reset_counter

Since commit 2ed53a94d8cb ("drm/i915: On GPU reset, set the HWS
breadcrumb to the last seqno") once a hang is completed, the seqno is
advanced past all current requests. With this we know that if we wake up
from waiting for a request, if a hang has occurred and reset completed,
our request will be considered complete (i.e.
i915_gem_request_completed() returns true). Therefore we only need to
worry about the situation where a hang has occurred, but not yet reset,
where we may need to release our struct_mutex. Since we don't need to
detect the completed reset using the global gpu_error->reset_counter
anymore, we do not need to track the reset_counter epoch inside the
request.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Arun Siluvery <arun.siluvery@linux.intel.com>
Cc: Mika Kuoppala <mika.kuoppala@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1467211874-11552-1-git-send-email-chris@chris-wilson.co.uk
Reviewed-by: Arun Siluvery <arun.siluvery@linux.intel.com>

drm/i915: Use atomic waits for short non-atomic ones

usleep_range is not recommended for waits shorten than 10us.

Make the wait_for_us use the atomic variant for such waits.

To do so we need to reimplement the _wait_for_atomic macro to
be safe with regards to preemption and interrupts.

v2: Reimplement _wait_for_atomic to be irq and preemption safe.
    (Chris Wilson and Imre Deak)

v3: Fixed in_atomic check due rebase error.
v4: Build bug on non-constant timeouts.
v5: Compile away cpu migration code in atomic paths.

Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Imre Deak <imre.deak@intel.com>
Cc: Mika Kuoppala <mika.kuoppala@intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: http://patchwork.freedesktop.org/patch/msgid/1467114710-29989-1-git-send-email-tvrtko.ursulin@linux.intel.com

drm/i915/debug: Select PREEMPT_COUNT when enabling debugging

Required to enable correct wait_for_atomic checks.

Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>

drm/i915/opregion: handle missing connector types for acpi display types

Most notably eDP, DSI, and TV. Add MISSING_CASE handling so we won't
miss this in the future.

Reviewed-and-tested-by: Peter Wu <peter@lekensteyn.nl>
Signed-off-by: Jani Nikula <jani.nikula@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/5773794027ea0d699052a343491b52343ba30504.1465810007.git.jani.nikula@intel.com

drm/i915/opregion: abstract acpi display type getter for a connector

No functional changes.

Reviewed-and-tested-by: Peter Wu <peter@lekensteyn.nl>
Signed-off-by: Jani Nikula <jani.nikula@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/ef19172e5a00d8abd8190a5389283ef6b5f7eaa9.1465810007.git.jani.nikula@intel.com

drm/i915/opregion: add acpi defines from the spec

It's easier to read the code when the macro names match the spec. Also
add a bunch of missing ones. No functional changes.

Reviewed-and-tested-by: Peter Wu <peter@lekensteyn.nl>
Signed-off-by: Jani Nikula <jani.nikula@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/7cfdd2733b8cc7abae43cfa8c81aed902b69ecd5.1465810007.git.jani.nikula@intel.com

drm/i915: Removing PCI IDs that are no longer listed as Kabylake.

This is unusual. Usually IDs listed on early stages of platform
definition are kept there as reserved for later use.

However these IDs here are not listed anymore in any of steppings
and devices IDs tables for Kabylake on configurations overview
section of BSpec.

So it is better removing them before they become used in any
other future platform.

Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Reviewed-by: Dhinakaran Pandiyan <dhinakaran.pandiyan@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1466718636-19675-2-git-send-email-rodrigo.vivi@intel.com

drm/i915: Add more Kabylake PCI IDs.

The spec has been updated adding new PCI IDs.

Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Reviewed-by: Dhinakaran Pandiyan <dhinakaran.pandiyan@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1466718636-19675-1-git-send-email-rodrigo.vivi@intel.com

drm/i915: Avoid early timeout during AUX transfers

Since wait_for_atomic doesn't re-check the wait-for condition after
expiry of the timeout it can fail when called from non-atomic context
even if the condition is set correctly before the expiry. Fix this by
using the non-atomic wait_for instead.

Due to the relatively long 10ms timeout, probably this didn't cause any
real problems, but fix it in any case for consistency.

Fixes: 0351b93992aa ("drm/i915: Do not lie about atomic timeout granularity")
CC: Chris Wilson <chris@chris-wilson.co.uk>
CC: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Signed-off-by: Imre Deak <imre.deak@intel.com>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
CC: drm-intel-fixes@lists.freedesktop.org
Link: http://patchwork.freedesktop.org/patch/msgid/1467110253-16046-5-git-send-email-imre.deak@intel.com

drm/i915/hsw: Avoid early timeout during LCPLL disable/restore

Since wait_for_atomic doesn't re-check the wait-for condition after
expiry of the timeout it can fail when called from non-atomic context
even if the condition is set correctly before the expiry. Fix this by
using the non-atomic wait_for instead.

Fixes: 0351b93992aa ("drm/i915: Do not lie about atomic timeout granularity")
CC: Chris Wilson <chris@chris-wilson.co.uk>
CC: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Signed-off-by: Imre Deak <imre.deak@intel.com>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
CC: drm-intel-fixes@lists.freedesktop.org
Link: http://patchwork.freedesktop.org/patch/msgid/1467110253-16046-4-git-send-email-imre.deak@intel.com

drm/i915/lpt: Avoid early timeout during FDI PHY reset

Since wait_for_atomic doesn't re-check the wait-for condition after
expiry of the timeout it can fail when called from non-atomic context
even if the condition is set correctly before the expiry. Fix this by
using the non-atomic wait_for instead.

Fixes: 0351b93992aa ("drm/i915: Do not lie about atomic timeout granularity")
CC: Chris Wilson <chris@chris-wilson.co.uk>
CC: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Signed-off-by: Imre Deak <imre.deak@intel.com>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
CC: drm-intel-fixes@lists.freedesktop.org
Link: http://patchwork.freedesktop.org/patch/msgid/1467110253-16046-3-git-send-email-imre.deak@intel.com

drm/i915/bxt: Avoid early timeout during PLL enable

Since wait_for_atomic doesn't re-check the wait-for condition after
expiry of the timeout it can fail when called from non-atomic context
even if the condition is set correctly before the expiry. Fix this by
using the non-atomic wait_for instead.

I noticed this via the PLL locking timing out incorrectly, with this fix
I couldn't reproduce the problem.

Fixes: 0351b93992aa ("drm/i915: Do not lie about atomic timeout granularity")
CC: Chris Wilson <chris@chris-wilson.co.uk>
CC: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Signed-off-by: Imre Deak <imre.deak@intel.com>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
CC: drm-intel-fixes@lists.freedesktop.org
Link: http://patchwork.freedesktop.org/patch/msgid/1467110253-16046-2-git-send-email-imre.deak@intel.com

drm/i915/gen9: Re-allocate DDB only for changed pipes

When a display update triggers a DDB re-allocation, we should start by
assuming that only the updated pipes need to be re-allocated (we have
logic later that may add additional pipes if, e.g., a modeset triggers a
change to the global allocation).

We were erroneously using the _active_ pipes as our starting point
rather than the changed pipes.  This causes us to grab CRTC locks that
we didn't actually need, reducing parallelism.  Given the recent
non-blocking atomic changes, it also causes legacy pageflips against one
CRTC to return -EBUSY if there's an outstanding pageflip against a
different CRTC (a situation easily triggered via compositors like
Weston).

Fixes: 98d39494d3 ("drm/i915/gen9: Compute DDB allocation at atomic check time (v4)")
Cc: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Signed-off-by: Matt Roper <matthew.d.roper@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1467070964-14864-1-git-send-email-matthew.d.roper@intel.com
Reviewed-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>

drm/i915: fix build errors when ACPI is not enabled

Fix build errors when ACPI is not enabled by adding function stubs:

../drivers/gpu/drm/i915/i915_drv.c: In function 'i915_drm_suspend':
../drivers/gpu/drm/i915/i915_drv.c:635:2: error: implicit declaration of function 'intel_opregion_unregister' [-Werror=implicit-function-declaration]
  intel_opregion_unregister(dev_priv);
../drivers/gpu/drm/i915/i915_drv.c: In function 'i915_drm_resume':
../drivers/gpu/drm/i915/i915_drv.c:798:2: error: implicit declaration of function 'intel_opregion_register' [-Werror=implicit-function-declaration]
  intel_opregion_register(dev_priv);

Signed-off-by: Randy Dunlap <rdunlap@infradead.org>
Cc: Daniel Vetter <daniel.vetter@intel.com>
Cc: Jani Nikula <jani.nikula@linux.intel.com>
Cc: intel-gfx@lists.freedesktop.org
Cc: dri-devel@lists.freedesktop.org
Fixes: 03d92e4779b8 ("drm/i915/opregion: Rename init/fini functions to register/unregister")
Cc: drm-intel-fixes@lists.freedesktop.org
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
[Jani: dropped the stale init/fini declarations]
Signed-off-by: Jani Nikula <jani.nikula@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1467028399-9965-1-git-send-email-jani.nikula@intel.com

drm/i915/dmc: Step away from symbolic links

Load specific firmware versions for the DMC instead of using symbolic
links. The currently recommended versions are: SKL 1.26, KBL 1.01 and
BXT 1.07.

Certain DMC versions need workarounds in the driver which forces us to
have a tight dependency between firmware and driver. In order to be able
to provide a tested and known working configuration we must lock down on
a specific DMC firmware version.

Cc: Rodrigo Vivi <rodrigo.vivi@intel.com>
Cc: Imre Deak <imre.deak@intel.com>
Cc: Mika Kuoppala <mika.kuoppala@linux.intel.com>
Signed-off-by: Patrik Jakobsson <patrik.jakobsson@linux.intel.com>
Acked-by: Jani Nikula <jani.nikula@intel.com>
Reviewed-by: Mika Kuoppala <mika.kuoppala@intel.com>
Acked-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1463391057-32350-1-git-send-email-patrik.jakobsson@linux.intel.com

drm/i915/guc: don't ever forward VBlank to the GuC

If a context waiting for VBlank were switched out, switching
in the next context and generating a CSB event in the process,
then the GuC would have to put the context back in the queue,
and then observe the subsequent VBlank interrupt so that it
could resubmit the suspended context.

However, we always set the CTX_CTRL_INHIBIT_SYN_CTX_SWITCH bit
in the RING_CONTEXT_CONTROL register, so this case cannot occur.
Furthermore we don't use the GuC's internal scheduler or allow
it to auto-resubmit workloads.  Consequently, the GuC doesn't
need to see VBlanks, and by sending them to it we may be waking
it up unnecessarily, which might reduce RC6 residency and
increase power consumption.

So this patch removes the setting of the GFC_FORWARD_VBLANK
field from the code that diverts interrupts towards the GuC.
(The code to direct interrupts to the host, OTOH, continues to
explicitly set the field to "never send VBlanks to the GuC".)

v3:
    Remove the line of code completely (original set the field
    to ALWAYS forward, v1 changed it to CONDITIONAL forwarding,
    v2 explicitly set it to NEVER, v3 just doesn't touch it at
    all, as we know it's already set to NEVER).

Signed-off-by: Dave Gordon <david.s.gordon@intel.com>
Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch> (previous version)
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1466780277-23435-1-git-send-email-david.s.gordon@intel.com

drm/i915: tweak gen6_for_{each_pde, all_pdes} macros

Gen8 versions of these macros were updated a few months ago
(e8ebd8e drm/i915: eliminate 'temp' in gen8_for_each macros)
originally because at least one iterator could generate an
out of bounds access, but also because eliminating the 'temp'
parameter generated smaller and faster code.

Matthew Auld recently noticed the same problem with the gen6
versions and provided a patch
https://lists.freedesktop.org/archives/intel-gfx/2016-June/099334.html
but while we're changing these, we might as well make them as
much like the gen8 versions as possible, including the style
of using "&& (..., true)" rather than ": (..., 1) : 0", and
of course eliminating the redundant 'temp'.

Furthermore, the "all_pdes" version is only used in one place,
so we can improve code efficiency by changing both the macro
parameters and the calling code to reduce extra dereferences.

Signed-off-by: Dave Gordon <david.s.gordon@intel.com>
Cc: Matthew Auld <matthew.auld@intel.com>
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Matthew Auld <matthew.auld@intel.com>
Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1466793466-23500-1-git-send-email-david.s.gordon@intel.com

drm/i915: Only switch to default context when evicting from GGTT

The contexts only pin space within the global GTT. Therefore forcing the
switch to the perma-pinned kernel context only has an effect when trying
to evict from and find room within the global GTT. We can then restrict
the switch to only when operating on the default context. This is mostly
a no-op as full-ppgtt only exists with execlists at present which skips
the context switch anyway.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1466776558-21516-7-git-send-email-chris@chris-wilson.co.uk

drm/i915: Split idling from forcing context switch

We only need to force a switch to the kernel context placeholder during
eviction. All other uses of i915_gpu_idle() just want to wait until
existing work on the GPU is idle. Rename i915_gpu_idle() to
i915_gem_wait_for_idle() to avoid any implications about "parking" the
context first.

v2: Tweak an error message if the wait fails for the ilk vtd w/a

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Reviewed-by: Mika Kuoppala <mika.kuoppala@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1466776558-21516-6-git-send-email-chris@chris-wilson.co.uk

drm/i915: No need to wait for idle on L3 remap

As the L3 remapping is applied before the next execution, there is no
need to wait until all previous uses are idle, the application will not
occur any sooner.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Reviewed-by: Mika Kuoppala <mika.kuoppala@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1466776558-21516-5-git-send-email-chris@chris-wilson.co.uk

drm/i915: Mark all default contexts as uninitialised after context loss

When the GPU is reset or state lost through suspend, every default
legacy context needs to reload their state - both the golden render
state and the L3 mapping. Only context images explicitly saved to memory
(i.e. all execlists and non-default legacy contexts) will retain their
state across the reset.

v2: Rebase

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1466776558-21516-4-git-send-email-chris@chris-wilson.co.uk

drm/i915: Treat kernel context as initialised

The kernel context exists simply as a placeholder and should never be
executed with a render context. It does not need the golden render
state, as that will always be applied to a user context. By skipping the
initialisation we can avoid issues in attempting to program the golden
render context when trying to make the hardware idle.

v2: Rebase

Testcase: igt/drm_module_reload_basic #byt
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=95634
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1466776558-21516-3-git-send-email-chris@chris-wilson.co.uk

drm/i915: Move legacy kernel context pinning to intel_ringbuffer.c

This is so that we have symmetry with intel_lrc.c and avoid a source of
if (i915.enable_execlists) layering violation within i915_gem_context.c -
that is we move the specific handling of the dev_priv->kernel_context
for legacy submission into the legacy submission code.

This depends upon the init/fini ordering between contexts and engines
already defined by intel_lrc.c, and also exporting the context alignment
required for pinning the legacy context.

v2: Separate out pin/unpin context funcs for greater symmetry with
intel_lrc. One more step towards unifying behaviour between the two
classes of engines and towards fixing another bug in i915_switch_context
vs requests.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1466776558-21516-2-git-send-email-chris@chris-wilson.co.uk

drm/i915: Skip idling an idle engine

During suspend (or module unload), if we have never accessed the engine
(i.e. userspace never submitted a batch to it), the engine is idle. Then
we attempt to idle the engine by forcing it to the default context,
which actually means we submit a render batch to setup the golden
context state and then wait for it to complete. We can skip this
entirely as we know the engine is idle.

v2: Drop incorrect comment.

References: https://bugs.freedesktop.org/show_bug.cgi?id=95634
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1466776558-21516-1-git-send-email-chris@chris-wilson.co.uk

drm/i915: Move module init/exit to i915_pci.c

The module init/exit routines are a wrapper around the PCI device
init/exit, so move them across.

Note that in order to avoid exporting the driver struct, instead of
manipulating driver.features inside i915_init we instead opt to simply
exit if i915.modeset is disabled.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Daniel Vetter <daniel.vetter@ffwll.ch>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1466773227-7994-15-git-send-email-chris@chris-wilson.co.uk

drm/i915: Split out the PCI driver interface to i915_pci.c

To reclaim a bit of space from i915_drv.c, we can move the routines that
just hook us into the PCI device tree into i915_pci.c

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Daniel Vetter <daniel.vetter@ffwll.ch>
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1466773227-7994-14-git-send-email-chris@chris-wilson.co.uk

drm/i915: Fix misleading driver debug message

Stop claiming that UMS support is disabled when it's not actually
supported anymore.

Signed-off-by: Frank Binns <frank.binns@imgtec.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: http://patchwork.freedesktop.org/patch/msgid/1466763836-27772-1-git-send-email-frank.binns@imgtec.com
Link: http://patchwork.freedesktop.org/patch/msgid/1466773227-7994-13-git-send-email-chris@chris-wilson.co.uk

drm/i915: Remove user controllable DRM_ERROR for intel_get_pipe_from_crtc_id()

Don't emit a driver DRM_ERROR for a user passing in an invalid CRTC id,
simply report it is missing back to the user.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1466773227-7994-12-git-send-email-chris@chris-wilson.co.uk

drm/i915: Remove user controllable DRM_ERROR for i915_getparam()

The GETPARAM ioctl writes to a user supplied address. If that address is
invalid, it is the user's error and not the driver's, so quietly report
EFAULT and don't blame ourselves with a DRM_ERROR.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1466773227-7994-11-git-send-email-chris@chris-wilson.co.uk

drm/i915: Merge i915_dma.c into i915_drv.c

i915_dma.c used to contain the DRI1/UMS horror show, but now all that
remains are the out-of-place driver level interfaces (such as
allocating, initialising and registering the driver). These should be in
i915_drv.c alongside similar routines for suspend/resume.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Daniel Vetter <daniel.vetter@ffwll.ch>
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1466773227-7994-10-git-send-email-chris@chris-wilson.co.uk

drm/i915: Start exploiting drm_device subclassing

Baby step, update to_i915() conversion from drm_device to
drm_i915_private:

   text    data     bss     dec     hex filename
1108812   23207     416 1132435  114793 i915.ko (before)
1104999   23207     416 1128622  1138ae i915.ko (after)

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Cc: Daniel Vetter <daniel.vetter@ffwll.ch>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1466773227-7994-9-git-send-email-chris@chris-wilson.co.uk

drm/i915: Remove redundant drm_connector_register_all()

drm_connector_register_all() is now automatically called by
drm_dev_register(), and so we no longer have to do so ourselves (via
intel_modeset_register() after calling drm_dev_register()). Similarly
for unregistering.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Link: http://patchwork.freedesktop.org/patch/msgid/1466773227-7994-8-git-send-email-chris@chris-wilson.co.uk

drm/i915: Demidlayer driver unloading

To complete the transition to manual control of load/unload, we need to
take over unloading from i915_pci_remove(). This allows us to correctly
order our unregister vs shutdown phases, which currently are inverted
due to the midlayer.

However, the unload sequence is still invalid as we shutdown the driver
with the last reference. Ideally, all we want to do is remove the
userspace access on device removal, deferring the cleanup to the
drm_dev_release() - breaking the reference cycles is then left as an
exercise for the reader.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Daniel Vetter <daniel.vetter@ffwll.ch>
Cc: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Link: http://patchwork.freedesktop.org/patch/msgid/1466773227-7994-7-git-send-email-chris@chris-wilson.co.uk

drm/i915: Demidlayer driver loading

Take control over allocating, loading and registering the driver from the
DRM midlayer by performing it manually from i915_pci_probe. This allows
us to carefully control the order of when we setup the hardware vs when
it becomes visible to third parties (including userspace). The current
ordering makes the driver visible to userspace first (in order to
coordinate with removed DRI1 userspace), but that ordering incurs risk.
The risk increases as we strive for more asynchronous loading.

One side effect of controlling the allocation is that we can allocate
both the drm_device + drm_i915_private in one block, the next step
towards subclassing.

Unload is still left as before, a mix of midlayer and driver.

v2: After drm_dev_init(), we should call drm_dev_unref() so that we call
drm_dev_release() and free everything from drm_dev_init().
v3: Fixup missed error code for failing to allocate dev_priv

Suggested-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Daniel Vetter <daniel.vetter@ffwll.ch>
Cc: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Link: http://patchwork.freedesktop.org/patch/msgid/1466773227-7994-6-git-send-email-chris@chris-wilson.co.uk

drm/i915: Register debugfs interface last

Currently debugfs files are created before the driver is even loads.
This gives the opportunity for userspace to open that interface and poke
around before the backing data structures are initialised - with the
possibility of oopsing or worse.

Move the creation of the debugfs files to our registration phase, where
we announce our presence to the world when we are ready, i.e the
sequence changes from

drm_dev_register()
 -> drm_minor_register()
  -> drm_debugfs_init()
   -> i915_debugfs_init()
 -> i915_driver_load()

to

drm_dev_register()
 -> drm_minor_register()
  -> drm_debugfs_init()
 -> i915_driver_load()
  -> i915_debugfs_register()

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Link: http://patchwork.freedesktop.org/patch/msgid/1466773227-7994-5-git-send-email-chris@chris-wilson.co.uk

drm/i915: Move connector registration to driver registration

Defer connector registration from during construction to the driver
registration phase. This is important for ordering the action correctly,
e.g. not using debugfs before it is ready.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Link: http://patchwork.freedesktop.org/patch/msgid/1466773227-7994-4-git-send-email-chris@chris-wilson.co.uk

drm/i915: Move backlight registration to connector registration

Currently the backlight is being registered in the load phase (before
the display and its objects are registered). Move the backlight
registration into the analogous phase by performing it from the
connector registration, just after its creation.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Jani Nikula <jani.nikula@linux.intel.com>
Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Link: http://patchwork.freedesktop.org/patch/msgid/1466773227-7994-3-git-send-email-chris@chris-wilson.co.uk

drm/i915: Move registration actions to connector->late_register

With the introduction of a connector->func for callback from
drm_connector_register() we can move all the tasks that we want to do
upon registration into that callback. Later, this will allow us to
reorder the registration and defer it until after the device is setup
and ready for userspace.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Link: http://patchwork.freedesktop.org/patch/msgid/1466773227-7994-2-git-send-email-chris@chris-wilson.co.uk

drm/i915: Move panel's backlight setup next to panel init

Currently setting up the backlight for a panel is sometimes done
together with initialising the panel, and sometimes after the connector
is registered. The backlight setup does not depend upon connector
registration (i.e. access to sysfs/debugfs and the kobject hierachy) so
perform it consistently just after panel initialisation.

Note the discrepancy here as destroying the panel is done during
connector unregistration...

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Jani Nikula <jani.nikula@linux.intel.com>
Cc: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Link: http://patchwork.freedesktop.org/patch/msgid/1466773227-7994-1-git-send-email-chris@chris-wilson.co.uk

drm/i915: Small compaction of the engine init code

Effectively removes one layer of indirection between the mask of
possible engines and the engine constructors. Instead of spelling
out in code the mapping of HAS_<engine> to constructors, makes
more use of the recently added data driven approach by putting
engine constructor vfuncs into the table as well.

Effect is fewer lines of source and smaller binary.

At the same time simplify the error handling since engine
destructors can run on unitialized engines anyway.

Similar approach could be done for legacy submission is wanted.

v2: Removed ugly BUILD_BUG_ONs in favour of newly introduced
    ENGINE_MASK and HAS_ENGINE macros.
    Also removed the forward declarations by shuffling functions
    around.

v3: Warn when logical_rings table does not contain enough data
    and disable the engines which could not be initialized.
    (Chris Wilson)

v4: Chris Wilson suggested a nicer engine init loop.

Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: http://patchwork.freedesktop.org/patch/msgid/1466689961-23232-1-git-send-email-tvrtko.ursulin@linux.intel.com

Merge remote-tracking branch 'airlied/drm-next' into drm-intel-next-queued

Backmerge drm-next for the reworked device register/unregistering.
Chris Wilson needs that to be able to land his i915 load/unload
demidlayering.

Signed-off-by: Daniel Vetter <daniel.vetter@intel.com>

Merge tag 'mediatek-drm-2016-06-20' of git://git.pengutronix.de/git/pza/linux into drm-next

MT8173 HDMI support

- device tree binding documentation for MT8173 HDMI encoder, CEC, DDC,
  and PHY
- drivers for MT8173 HDMI encoder, CEC (HPD only for now), DDC, and PHY
- enable HDMI output via a custom SMCCC call
- add ddc-i2c-bus property to HDMI connector device tree binding

* tag 'mediatek-drm-2016-06-20' of git://git.pengutronix.de/git/pza/linux:
  dt-bindings: hdmi-connector: add DDC I2C bus phandle documentation
  drm/mediatek: enable hdmi output control bit
  drm/mediatek: Add HDMI support
  dt-bindings: drm/mediatek: Add Mediatek HDMI dts binding

Merge branch 'drm/next/du' of git://linuxtv.org/pinchartl/media into drm-next

some rcar-du fixes.

* 'drm/next/du' of git://linuxtv.org/pinchartl/media:
  drm: rcar-du: error message is not needed for EPROBE_DEFER
  drm: rcar-du: error message is not needed for drm_vblank_init()
  rcar-du: add/rename DEFR6 TCON bits

Merge tag 'drm-intel-next-2016-06-20' of git://anongit.freedesktop.org/drm-intel into drm-next

- Infrastructure for GVT-g (paravirtualized gpu on gen8+), from Zhi Wang
- another attemp at nonblocking atomic plane updates
- bugfixes and refactoring for GuC doorbell code (Dave Gordon)
- GuC command submission enabled by default, if fw available (Dave Gordon)
- more bxt w/a (Arun Siluvery)
- bxt phy improvements (Imre Deak)
- prep work for stolen objects support (Ankitprasa Sharma & Chris Wilson)
- skl/bkl w/a update from Mika Kuoppala
- bunch of small improvements and fixes all over, as usual

* tag 'drm-intel-next-2016-06-20' of git://anongit.freedesktop.org/drm-intel: (81 commits)
  drm/i915: Update DRIVER_DATE to 20160620
  drm/i915: Introduce GVT context creation API
  drm/i915: Support LRC context single submission
  drm/i915: Introduce execlist context status change notification
  drm/i915: Make addressing mode bits in context descriptor configurable
  drm/i915: Make ring buffer size of a LRC context configurable
  drm/i915: gvt: Introduce the basic architecture of GVT-g
  drm/i915: Fold vGPU active check into inner functions
  drm/i915: Use offsetof() to calculate the offset of members in PVINFO page
  drm/i915: Factor out i915_pvinfo.h
  drm/i915: Serialise presentation with imported dmabufs
  drm/i915: Use atomic commits for legacy page_flips
  drm/i915: Move fb_bits updating later in atomic_commit
  drm/i915: nonblocking commit
  Reapply "drm/i915: Pass atomic states to fbc update, functions."
  drm/i915: Roll out the helper nonblock tracking
  drm/i915: Signal drm events for atomic
  drm/i915/ilk: Don't disable SSC source if it's in use
  drm/i915/guc: (re)initialise doorbell h/w when enabling GuC submission
  drm/i915/guc: replace assign_doorbell() with select_doorbell_register()
  ...

Merge tag 'topic/drm-misc-2016-06-22-updated' of git://anongit.freedesktop.org/drm-intel into drm-next

Again a pile of things all over
- Conversion to rst from docbook from Jani. Looks real pretty, and the
  source is now actually readable (compared to horrible, horrible docbook
  xml)! https://01.org/linuxgraphics/gfx-docs/drm/
- device register/unregister rework from Chris, with follow-up work from
  Benjamin. Allows more drivers to demidlayer load/unload and others to
  remove a bit of boilerplate.
- master/auth related cleanup, with docs
- some dma-buf polish, merged by Sumit
- small stuff all over (like build fixes from Arnd)

Group maintainership seems to slowly take off, with both Thierry and Sumit
pushing a few things. No hiccups thus far.

* tag 'topic/drm-misc-2016-06-22-updated' of git://anongit.freedesktop.org/drm-intel: (68 commits)
  drm/vc4: Remove unused connector
  drm/fb-helper: Reduce READ_ONCE(master) to lockless_dereference
  drm/sun4i: Remove open-coded drm_connector_register_all()
  drm/vc4: Remove open-coded drm_connector_register_all()
  drm/atmel-hlcdc: Remove redundant call to drm_connector_unregister_all()
  drm: document drm_auth.c
  drm: Clear up master tracking booleans
  drm: Extract drm_is_current_master
  drm: Refactor drop/set master code a bit
  drm: Lobotomize set_busid nonsense for !pci drivers
  drm: Nuke SET_UNIQUE ioctl
  drm: Don't call drm_dev_set_unique from platform drivers
  drm/vgem: Stop calling drm_drv_set_unique
  drm: Use dev->name as fallback for dev->unique
  drm: Clean up drm_crtc.h
  drm: Move master pointer from drm_minor to drm_device
  drm: sti: rework init sequence
  drm: sti: use late_register and early_unregister callbacks
  drm/amdkfd: Clean up inline handling
  drm: Add callbacks for late registering
  ...

Merge tag 'drm-sii902x' of github.com:bbrezillon/linux-at91 into drm-next

Add basic support for the sii902x RGB -> HDMI bridge.

* tag 'drm-sii902x' of github.com:bbrezillon/linux-at91:
  drm/bridge: Add sii902x DT bindings doc
  drm/bridge: Add sii902x driver

drm/i915: Refresh cached DP port register value on resume

During hibernation the cached DP port register value will be left with
whatever value we have there when we create the hibernation image.
Currently that means the port (and eDP PLL) will be off in the cached
value. However when we resume there is no guarantee that the value
in the actual register will match the cached value. If i915 isn't
loaded in the kernel that loads the hibernation image, the port may
well be on (eg. left on by the BIOS). The encoder state readout
does the right thing in this case and updates our encoder state
to reflect the actual hardware state. However the post-resume modeset
will then use the stale cached port register value in
intel_dp_link_down() and potentially confuse the hardware.

This was caught by the following assert
 WARNING: CPU: 3 PID: 5288 at ../drivers/gpu/drm/i915/intel_dp.c:2184 assert_edp_pll+0x99/0xa0 [i915]
 eDP PLL state assertion failure (expected on, current off)
on account of the eDP PLL getting prematurely turned off when
shutting down the port, since the DP_PLL_ENABLE bit wasn't set
in the cached register value.

Presumably I introduced this problem in
commit 6fec76628333 ("drm/i915: Use intel_dp->DP in eDP PLL setup")
as before that we didn't update the cached value after shuttting the
port down. That's assuming the port got enabled at least once prior
to hibernating. If that didn't happen then the cached value would
still have been totally out of sync with reality (eg. first boot w/o
eDP on, then hibernate, and then resume with eDP on).

So, let's fix this properly and refresh the cached register value from
the hardware register during resume.

DDI platforms shouldn't use the cached value during port disable at
least, so shouldn't have this particular issue. They might still have
issues if we skip the initial modeset and then try to retrain the link
or something. But untangling this DP vs. DDI mess is a bigger topic,
so let's jut punt on DDI for now.

Cc: Jani Nikula <jani.nikula@intel.com>
Cc: stable@vger.kernel.org
Fixes: 6fec76628333 ("drm/i915: Use intel_dp->DP in eDP PLL setup")
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1463162036-27931-1-git-send-email-ville.syrjala@linux.intel.com
Reviewed-by: Imre Deak <imre.deak@intel.com>

drm/i915: Sanity check PPS HW state

The wait for panel status helper will only function correctly if the
HW panel timings are programmed correctly. Returning prematurely from
this helper may lead to obscure bugs later, so sanity check the HW
timing registers.

v2:
- Check the T8, T9 fields too, we do program them (Ville)

CC: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Imre Deak <imre.deak@intel.com>
Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1466096506-11937-1-git-send-email-imre.deak@intel.com

drm/i915: Factor out helper to read out PPS HW state

This will be needed by the next patch too so factor it out.

No functional change.

Signed-off-by: Imre Deak <imre.deak@intel.com>
Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1466084243-5388-4-git-send-email-imre.deak@intel.com

drm/i915: Deduplicate PPS register retrieval

No functional change.

Signed-off-by: Imre Deak <imre.deak@intel.com>
Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1466084243-5388-3-git-send-email-imre.deak@intel.com

drm/i915/bxt: Fix PPS lost state after suspend breaking eDP link training

The PPS registers are backed by power well #0 and as such may be reset
after system or runtime suspend (both implying a possible DC9
transition). Fix this by reusing the VLV/CHV PPS pipe-reassignment
logic. The difference on BXT is that the PPS instances are not pipe but
port (or more accurately pin) specific, so we only need to care about
the lost HW state. As opposed to VLV/CHV the SW state is fixed and
initialized during connector init.

This also paves the way towards using the actual port->PPS instance
mapping based on VBT.

This fixes eDP link training errors on BXT after suspend, where we
started the link training too early due to an incorrect T3 (panel power
on) register value.

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=96436
Signed-off-by: Imre Deak <imre.deak@intel.com>
Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1466084243-5388-2-git-send-email-imre.deak@intel.com

drm/i915: Group all the PPS init steps to one place

Move the early PPS initialization calls next to the rest of PPS
initialization steps. This allows us to forgo a duplicated call to
intel_dp_init_panel_power_sequencer_registers() on VLV/CHV.

This will swap the order of DP AUX registration wrt. PPS initialization.
There is an existing race here in case of a user space access via the
DPAUX device node after DP AUX registration and before calling
intel_dp_init_panel_power_sequencer_registers(), but this change won't
make this worse. The fix for this is to separate DP AUX initialization
and registration, that's a separate work already underway.

The order of MST wrt. PPS init as well as the order of
intel_dp_init_panel_power_sequencer_registers() wrt.
intel_edp_panel_vdd_sanitize() also swap, which is ok, there are no
dependencies between these steps.

Suggested by Ville.

CC: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Imre Deak <imre.deak@intel.com>
Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1466499109-20240-4-git-send-email-imre.deak@intel.com

drm/i915: Initialize the PPS HW before its first use

The initial DPCD read for eDP detection involves using the PPS, but so
far we only initialized the PPS registers after the DPCD read. The
reason this was done so far is to preserve a possible LVDS PPS HW setup
if LVDS is detected but eDP is not. This is not an issue any more after
the previous patch, so we can move the init earlier now.

This was caught by CI with the PPS sanity checks in place and the
initial eDP DPCD readout waiting for the panel power cycle timeout
without the PPS registers being initialized.

CC: Ville Syrjälä <ville.syrjala@linux.intel.com>
CC: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Imre Deak <imre.deak@intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1466499109-20240-3-git-send-email-imre.deak@intel.com

drm/i915/ibx, cpt: Don't attempt to register eDP if LVDS was detected

Atm on IBX/CPT we attempt to detect if eDP is present even if LVDS was
already detected and an encoder for it was registered. This involves
trying to read out the eDP DPCD, which in turn needs the same power
sequencer that LVDS uses. Poking at the VDD line at an unexpected time
may or may not interfere with the LVDS panel, but it's probably safer to
prevent this. Registering both an LVDS and an eDP connector would also
present a similar problem accessing the shared PPS at any point later in
an unexpected way.

We also need this to be able fix PPS initialization before its first use
in the next patch. For that we want to be sure that PPS is not in use
by LVDS.

v2:
- Split out the PPS init fix to a separate patch. (Chris)
- Add comment about eDP init depending on LVDS init. (Chris)
- Make the use of the intel_encoder ptr less error prone.
v3:
- Use IBX/CPT reference instead of the incorrect ILK, add a WARN about
  this. (Ville)
v4:
- Use a helper to get the lvds encoder instead of opencoding the same.
  (Ville)

CC: Ville Syrjälä <ville.syrjala@linux.intel.com>
CC: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Imre Deak <imre.deak@intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk> (v2)
Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com> (v3)
Link: http://patchwork.freedesktop.org/patch/msgid/1466499109-20240-2-git-send-email-imre.deak@intel.com

drm/vc4: Remove unused connector

Somehow I didn't spot this when pushing :(

Fixes: 398e97994f6d ("drm/vc4: Remove open-coded drm_connector_register_all()")
Acked-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Daniel Vetter <daniel.vetter@intel.com>

drm/fb-helper: Reduce READ_ONCE(master) to lockless_dereference

We are only documenting that the read is outside of the lock, and do not
require strict ordering on the operation. In this case the more relaxed
lockless_dereference() will suffice.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Daniel Vetter <daniel.vetter@intel.com>
Cc: Julia Lawall <julia.lawall@lip6.fr>
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Emil Velikov <emil.l.velikov@gmail.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Link: http://patchwork.freedesktop.org/patch/msgid/1466581572-16608-1-git-send-email-chris@chris-wilson.co.uk

drm/sun4i: Remove open-coded drm_connector_register_all()

drm_dev_register() will now register all known connectors, so we no
longer have to do so manually.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Maxime Ripard <maxime.ripard@free-electrons.com>
Cc: David Airlie <airlied@linux.ie>
Cc: Chen-Yu Tsai <wens@csie.org>
Cc: Daniel Vetter <daniel.vetter@ffwll.ch>
Cc: dri-devel@lists.freedesktop.org
Cc: linux-arm-kernel@lists.infradead.org
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Link: http://patchwork.freedesktop.org/patch/msgid/1466501283-19976-4-git-send-email-chris@chris-wilson.co.uk

drm/vc4: Remove open-coded drm_connector_register_all()

drm_dev_register() will now register all known connectors, so we no
longer have to do so manually.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Eric Anholt <eric@anholt.net>
Cc: David Airlie <airlied@linux.ie>
Cc: Daniel Vetter <daniel.vetter@ffwll.ch>
Cc: dri-devel@lists.freedesktop.org
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Link: http://patchwork.freedesktop.org/patch/msgid/1466501283-19976-3-git-send-email-chris@chris-wilson.co.uk

drm/atmel-hlcdc: Remove redundant call to drm_connector_unregister_all()

drm_connector_unregister_all() is not automatically called by
drm_dev_unregister() so we can drop the local call.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Daniel Vetter <daniel.vetter@ffwll.ch>
Cc: Boris Brezillon <boris.brezillon@free-electrons.com>
Cc: David Airlie <airlied@linux.ie>
Cc: dri-devel@lists.freedesktop.org
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Link: http://patchwork.freedesktop.org/patch/msgid/1466501283-19976-2-git-send-email-chris@chris-wilson.co.uk

drm: document drm_auth.c

Also extract drm_auth.h for nicer grouping.

v2: Nuke the other comments since they don't really explain a lot, and
within the drm core we generally only document functions exported to
drivers: The main audience for these docs are driver writers.

v3: Limit the exposure of drm_master internals by only including
drm_auth.h where it is neede (Chris).

v4: Spelling polish (Emil).

Cc: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Emil Velikov <emil.l.velikov@gmail.com>
Signed-off-by: Daniel Vetter <daniel.vetter@intel.com>

drm: Clear up master tracking booleans

- is_master can be removed, we can compute this by checking allowed_master
  (which really just tracks whether a master struct has been allocated
  for this fpriv in either open or set_master), and whether the fpriv is
  the current master on the device.

- that frees up is_master as a good replacement name for allowed_master.
  With that it's clear that it tracks whether the fpriv is a master (with
  possibly clients attached to it and authenticated against it), and that
  one of those fprivs with is_master set is the current master.

v2: Fix kerneldoc for is_master (Emil).

Cc: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Thomas Hellstrom <thellstrom@vmware.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Emil Velikov <emil.l.velikov@gmail.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Link: http://patchwork.freedesktop.org/patch/msgid/1466499262-18717-10-git-send-email-daniel.vetter@ffwll.ch

drm: Extract drm_is_current_master

Just rolling out a bit of abstraction to be able to clean
up the master logic in the next step.

Cc: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Thomas Hellstrom <thellstrom@vmware.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Emil Velikov <emil.l.velikov@gmail.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm: Refactor drop/set master code a bit

File open/set_maseter ioctl and file close/drop_master ioctl share the
same master handling code. Extract it.

Note that vmwgfx's master_set callback needs to know whether the
master is a new one or has been used already, so thread this through.
On the close/drop side a similar parameter existed, but wasnt used.
Drop it to simplify the flow.

v2: Try to make it not leak so much (Emil).

v3: Send out the right version ...

Cc: Emil Velikov <emil.l.velikov@gmail.com>
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Thomas Hellstrom <thellstrom@vmware.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Emil Velikov <emil.l.velikov@gmail.com>
Signed-off-by: Daniel Vetter <daniel.vetter@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1466511638-9885-1-git-send-email-daniel.vetter@ffwll.ch

drm: Lobotomize set_busid nonsense for !pci drivers

We already have a fallback in place to fill out the unique from
dev->unique, which is set to something reasonable in drm_dev_alloc.

Which means we only need to have a special set_busid for pci devices,
to be able to care the backwards compat code for drm 1.1 around, which
libdrm still needs.

While developing and testing this patch things blew up in really
interesting ways, and the code is rather confusing in naming things
between the kernel code, ioctl #defines and libdrm. For the next brave
dragon slayer, document all this madness properly in the userspace
interface section of gpu.tmpl.

v2: Make drm_dev_set_unique static and update kerneldoc.

v3: Entire rewrite, plus document what's going on for posterity in the
gpu docbook uapi section.

v4: Drop accidental amdgpu hunk (Emil).

v5: Drop accidental omapdrm vblank counter change (Emil).

v6: Rebase on top of the sphinx conversion.

Cc: Gustavo Padovan <gustavo.padovan@collabora.co.uk>
Cc: Emil Velikov <emil.l.velikov@gmail.com>
Tested-by: Gustavo Padovan <gustavo.padovan@collabora.co.uk> (virt_gpu)
Reviewed-by: Emil Velikov <emil.l.velikov@gmail.com>
Signed-off-by: Daniel Vetter <daniel.vetter@intel.com>

drm: Nuke SET_UNIQUE ioctl

Ever since

commit 2e1868b560315a8b20d688e646c489a5ad93eeae
Author: Eric Anholt <anholt@freebsd.org>
Date:   Wed Jun 16 09:25:21 2004 +0000

    DRI trunk-20040613 import

the X server supports drm 1.1, thus doesn't call call libdrm's
drmSetBusid - the sole user of this ioctl. When reviewing this note
that for hilarity both the kernel-internal functions (set_busid) and
the libdrm wrapper (drmSetBusid) have names not matching this ioctl
(SET_UNIQUE).

v2: Polish commit message (Emil).

Cc: Emil Velikov <emil.l.velikov@gmail.com>
Reviewed-by: Emil Velikov <emil.l.velikov@gmail.com>
Signed-off-by: Daniel Vetter <daniel.vetter@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1466499262-18717-6-git-send-email-daniel.vetter@ffwll.ch

drm: Don't call drm_dev_set_unique from platform drivers

Since

commit e112e593b215c394c0303dbf0534db0928e87967
Author: Nicolas Iooss <nicolas.iooss_linux@m4x.org>
Date:   Fri Dec 11 11:20:28 2015 +0100

    drm: use dev_name as default unique name in drm_dev_alloc()

we're using a reasonable default which should work for everyone. Only
mtk, rcar-du and sun4i are affected, and as kms-only drivers without
any rendering support no one should ever care about the unique name

v2: Rebase on top of mediatek.

Cc: Philipp Zabel <p.zabel@pengutronix.de>
Cc: Maxime Ripard <maxime.ripard@free-electrons.com>
Cc: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Cc: Emil Velikov <emil.l.velikov@gmail.com>
Reviewed-by: Emil Velikov <emil.l.velikov@gmail.com>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Acked-by: Maxime Ripard <maxime.ripard@free-electrons.com>
Acked-by: Philipp Zabel <p.zabel@pengutronix.de>
Signed-off-by: Daniel Vetter <daniel.vetter@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1466499262-18717-5-git-send-email-daniel.vetter@ffwll.ch

drm/vgem: Stop calling drm_drv_set_unique

With the previous patch this is now redudant, the core always
sets a reasonable dev->unique string.

Cc: Sean Paul <seanpaul@chromium.org>
Reviewed-by: Emil Velikov <emil.l.velikov@gmail.com>
Signed-off-by: Daniel Vetter <daniel.vetter@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1466499262-18717-4-git-send-email-daniel.vetter@ffwll.ch

drm: Use dev->name as fallback for dev->unique

Lots of arm drivers get this wrong and for most arm boards this is the
right thing actually. And anyway with most loaders you want to chase
sysfs links anyway to figure out which dri device you want.

This will fix dmesg noise for rockchip and sti.

Also add a fallback to driver->name for entirely virtual drivers like
vgem.

v2: Rebase on top of

commit e112e593b215c394c0303dbf0534db0928e87967
Author: Nicolas Iooss <nicolas.iooss_linux@m4x.org>
Date:   Fri Dec 11 11:20:28 2015 +0100

    drm: use dev_name as default unique name in drm_dev_alloc()

and simplify a bit. Plus add a comment.

v3: WARN_ON(!dev->unique) as discussed with Emil.

Cc: Ilia Mirkin <imirkin@alum.mit.edu>
Reported-by: Ilia Mirkin <imirkin@alum.mit.edu>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk> (v2)
Reviewed-by: Emil Velikov <emil.l.velikov@gmail.com>
Signed-off-by: Daniel Vetter <daniel.vetter@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1466499262-18717-3-git-send-email-daniel.vetter@ffwll.ch

drm: Clean up drm_crtc.h

- Group declarations for separate files (drm_bridge.c, drm_edid.c)
- Move declarations only used within drm.ko to drm_crtc_internal.h
- drm_property_type_valid to drm_crtc.c, its only callsite

Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Emil Velikov <emil.l.velikov@gmail.com>
Signed-off-by: Daniel Vetter <daniel.vetter@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1466499262-18717-2-git-send-email-daniel.vetter@ffwll.ch

drm: Move master pointer from drm_minor to drm_device

There can only be one current master, and it's for the overall device.
Render/control minors don't support master-based auth at all.

This simplifies the master logic a lot, at least in my eyes: All these
additional pointer chases are just confusing.

While doing the conversion I spotted some locking fail:
- drm_lock/drm_auth check dev->master without holding the
  master_mutex. This is fallout from

  commit c996fd0b956450563454e7ccc97a82ca31f9d043
  Author: Thomas Hellstrom <thellstrom@vmware.com>
  Date:   Tue Feb 25 19:57:44 2014 +0100

      drm: Protect the master management with a drm_device::master_mutex v3

  but I honestly don't care one bit about those old legacy drivers
  using this.

- debugfs name info should just grab master_mutex.

- And the fbdev helper looked at it to figure out whether someone is
  using KMS. We just need a consistent value, so READ_ONCE. Aside: We
  should probably check if anyone has opened a control node too, but I
  guess current userspace doesn't really do that yet.

v2: Balance locking, reported by Julia.

v3: Rebase on top of Chris' oops fixes.

Cc: Julia Lawall <julia.lawall@lip6.fr>
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk> (v2)
Reviewed-by: Emil Velikov <emil.l.velikov@gmail.com> (v2)
Signed-off-by: Daniel Vetter <daniel.vetter@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1466499262-18717-1-git-send-email-daniel.vetter@ffwll.ch

drm: sti: rework init sequence

Use drm_dev_alloc() and drm_dev_register() instead of .load()
To simplify init sequence only create fbdev when requested
in output_poll_changed().

version 2:
remove call to drm_connector_unregister_all() and
drm_dev_set_unique()

Signed-off-by: Benjamin Gaignard <benjamin.gaignard@linaro.org>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Link: http://patchwork.freedesktop.org/patch/msgid/1466514580-15194-4-git-send-email-benjamin.gaignard@linaro.org

drm: sti: use late_register and early_unregister callbacks

Make sti driver use register callback to move debugfs
initialization out of sub-components creation.
This will allow to convert driver .load() to
drm_dev_alloc() and drm_dev_register().

sti_compositor bring up 2 crtc but only one debugfs init is
needed so use drm_crtc_index to do it on the first one.
This can't be done in sti_drv because only sti_compositor have
access to the devices.
It is almost the same for sti_encoder which handle multiple
encoder while one only debugfs entry is needed so add a boolean
to avoid multiple debugfs initialization

Signed-off-by: Benjamin Gaignard <benjamin.gaignard@linaro.org>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Link: http://patchwork.freedesktop.org/patch/msgid/1466514580-15194-3-git-send-email-benjamin.gaignard@linaro.org

drm/amdkfd: Clean up inline handling

- inline functions need to be static inline, otherwise gcc can opt to
  not inline and the linker gets unhappy.
- no forward decls for inline functions, just include the right headers.

Cc: Oded Gabbay <oded.gabbay@gmail.com>
Cc: Ben Goz <ben.goz@amd.com>
Reviewed-by: Oded Gabbay <oded.gabbay@gmail.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Link: http://patchwork.freedesktop.org/patch/msgid/1466500235-21282-2-git-send-email-daniel.vetter@ffwll.ch

drm: Add callbacks for late registering

Like what has been done for connectors add callbacks on encoder,
crtc and plane to let driver do actions after drm device registration.

Correspondingly, add callbacks called before unregister drm device.

version 2:
add drm_modeset_register_all() and drm_modeset_unregister_all()
to centralize all calls

version 3:
in error case unwind registers in drm_modeset_register_all
fix uninitialed return value
inverse order of unregistration in drm_modeset_unregister_all

version 4:
move function definitions in drm_crtc_internal.h
remove not needed documentation

Signed-off-by: Benjamin Gaignard <benjamin.gaignard@linaro.org>
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/1466519829-4000-1-git-send-email-benjamin.gaignard@linaro.org

drm/i915/guc: Remove one unnecessary variable

No need for local struct drm_device * since dev_priv is the
correct thing to pass in to NEEDS_WaRsDisableCoarsePowerGating
anyway. Changed the macro definition for the latter to reflect
that as well.

v2: Alignment bikeshed.

Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Cc: Dave Gordon <david.s.gordon@intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Link: http://patchwork.freedesktop.org/patch/msgid/1466518034-24838-1-git-send-email-tvrtko.ursulin@linux.intel.com

drm/fsl-dcu: use drm_mode_config_cleanup on initialization errors

Commit 7566e247672d ("drm/fsl-dcu: handle initialization errors properly")
introduced error handling during initialization, but with a wrong cleanup
order.

Replace the error handling with the generic cleanup function
drm_mode_config_cleanup.

Signed-off-by: Stefan Agner <stefan@agner.ch>
Signed-off-by: Thierry Reding <treding@nvidia.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20160619021543.23587-1-stefan@agner.ch

drm/mediatek: Remove IOMMU_DMA select

We get a harmless build warning when trying to use the mediatek DRM
driver with IOMMU support disabled:

warning: (DRM_MEDIATEK) selects IOMMU_DMA which has unmet direct dependencies (IOMMU_SUPPORT)

However, the IOMMU_DMA symbol is not meant to be used by drivers at all,
and this driver doesn't seem to have a strict dependency on it other
than using the mediatek IOMMU driver that does.

Since we also want to be able to do compile tests with the driver on
other platforms, the IOMMU_DMA symbol should not be selected here.

Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Acked-by: Philipp Zabel <p.zabel@pengutronix.de>
Signed-off-by: Thierry Reding <treding@nvidia.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1462997501-982363-1-git-send-email-arnd@arndb.de

Documentation/DocBook: remove gpu.tmpl

The gpu documentation has now been converted to reStructuredText files
under Documentation/gpu. Remove the obsolete DocBook template. Also
remove it from MAINTAINERS.

Good riddance.

Signed-off-by: Jani Nikula <jani.nikula@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Link: http://patchwork.freedesktop.org/patch/msgid/8d673f75fe686371ed9838682c368a4e3b96bf54.1466506505.git.jani.nikula@intel.com

Documentation/gpu: split up mm, kms and kms-helpers from internals

Make the documents more manageable.

Signed-off-by: Jani Nikula <jani.nikula@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Link: http://patchwork.freedesktop.org/patch/msgid/be992e56eb8442d6e03b52444df5a42525085718.1466506505.git.jani.nikula@intel.com

Documentation/gpu: convert the KMS properties table to CSV

Pandoc really did a bad job of converting the big KMS properties table
to RST. Instead, put the properties into a separate plain text CSV file,
and include it in the RST file. The generated output isn't very pretty,
but at least the information is there, and it's stored in a format
that's easier to process and improve upon at a later time.

The CSV file was generated by copy-pasting the table from the HTML
generated by the DocBook toolchain into LibreOffice Calc, and then saved
as CSV, unmodified.

Signed-off-by: Jani Nikula <jani.nikula@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Link: http://patchwork.freedesktop.org/patch/msgid/17053ff08caf5cfac4f478437ef796f83a31d772.1466506505.git.jani.nikula@intel.com

Documentation/gpu: use recommended order of heading markers

While splitting the document up, the headings "shifted" from what pandoc
generated. Use the following order for headings for consistency:

==============
Document title
==============

First
=====

Second
------

Third
~~~~~

Leave the lower level headings as they are; I think those are less
important. Although RST doesn't mandate a specific order ("Rather than
imposing a fixed number and order of section title adornment styles, the
order enforced will be the order as encountered."), having the higher
levels the same overall makes it easier to follow the documents.

[I'm sort of kind of writing the recommendation for docs-next in the
mean time, but this order seems sensible, and is what I'm proposing.]

Signed-off-by: Jani Nikula <jani.nikula@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Link: http://patchwork.freedesktop.org/patch/msgid/126f42734defac6cbb8496a481d58db7b38461dd.1466506505.git.jani.nikula@intel.com