Noticed while playing with coccinelle.
Reviewed-by: NDamien Lespiau <damien.lespiau@intel.com>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

During initial probing of the modes to assign to the fbdev console, we
use the CRTC and connector ids. These are much harder for us to
understand than if we used their actual names (or pipe in the CRTC
case). Similarly, we want to manually print the mode size rather than
rely on mode->name being set.
Signed-off-by: NChris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

Patch done using the following semantic patch (thanks Daniel for the
help!)

  @@
  iterator name list_for_each_entry;
  iterator name for_each_crtc;
  struct drm_crtc * crtc;
  struct drm_device * dev;
  @@
  -list_for_each_entry(crtc,&dev->mode_config.crtc_list, head) {
  +for_each_crtc(dev,crtc) {
  	...
  }

Followed by a couple of fixups by hand (that spatch doesn't match the
cases where list_for_each_entry() is not followed by a set of '{', '}',
but I couldn't figure out a way to leave the '{' out of the iterator
match).
Signed-off-by: NDamien Lespiau <damien.lespiau@intel.com>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

Signed-off-by: NDamien Lespiau <damien.lespiau@intel.com>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

Generated using the semantic patch:

  @@
  iterator name list_for_each_entry;
  iterator name for_each_intel_crtc;
  struct intel_crtc * crtc;
  struct drm_device * dev;
  @@
  -list_for_each_entry(crtc,&dev->mode_config.crtc_list,...) {
  +for_each_intel_crtc(dev,crtc) {
	...
  }

Followed by a couple of fixups by hand (that spatch doesn't match the
cases where list_for_each_entry() is not followed by a set of '{', '}',
but I couldn't figure out a way to leave the '{' out of the iterator
match).
Signed-off-by: NDamien Lespiau <damien.lespiau@intel.com>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

Fed up with having that long list_for_each_entry() invocation?

Use for_each_intel_crtc()!
Signed-off-by: NDamien Lespiau <damien.lespiau@intel.com>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

Signed-off-by: NDamien Lespiau <damien.lespiau@intel.com>
[danvet: Squash in patch that exported ilk_wm_max_level.]
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

That's not necessary and makes the code not as neat as it could be.
Signed-off-by: NDamien Lespiau <damien.lespiau@intel.com>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

Somehow UXA submits a completely bogus DR4 value since essentially
forever. It was originally introduced in

commit bade7d7d2505a10a8a7d24b084aff9742e2d6d64
Author: Eric Anholt <eric@anholt.net>
Date:   Fri Jun 6 14:03:25 2008 -0700

    Use the DRM for submitting batchbuffers when available.

and dutifully copied around ever since. Since we want to keep the
general dirt catching around just special case the UXA value.

This regression was introduced in

commit 9cb34664
Author: Daniel Vetter <daniel.vetter@ffwll.ch>
Date:   Thu Apr 24 08:09:11 2014 +0200

    drm/i915: Catch dirt in unused execbuffer fields

Comment from Chris' review:

"To be fair, it is a sensible value if one supposes a Region style API to
cliprects. Under that API, DR[14] define the extents of the clip region,
and ((0,0), (0,0)) [DR1==DR4==0] would mean all clipped, do not draw
anything."

v2: Pimp commit message a bit and remove the double space.

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=78494
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Jörg Otte <jrg.otte@gmail.com>
Acked-by: NChris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

The fence pin count should always be <= the bo pin count. If that's
not the case then we have a funny problem and are leaking references
somewhere.

Which means we can catch fence pin leaks by checking for the same
upper limit as we do for the bo pin count. Inspired by a discussion
with Ville about a fence leak igt testcase.

v2: Also check for fence->pin_count <= ggtt_vma->pin_count, since that
might catch a leak even quicker. Also de-inline them, they're getting
too big.

v3: Don't separately check for MAX_PIN_COUNT since the > vma->pin_count
check will catch that already (Chris).

Cc: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

The pipe might not start to actually run until the port has been enabled
(depends on the platform and port type). So don't try to wait for vblank
after we enabled the pipe but haven't yet enabled the port.
Signed-off-by: NVille Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: NChris Wilson <chris@chris-wilson.co.uk>
References: https://bugs.freedesktop.org/show_bug.cgi?id=77297Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

We already moved the plane disable/enable to happen as the first/last
thing on every other platforms. Follow suit with gmch platforms.
Signed-off-by: NVille Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: NChris Wilson <chris@chris-wilson.co.uk>
[danvet: Frob drm_vblank_on conflict, as usual.]
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

This is missing in:

commit 78325f2d
Author: Ben Widawsky <benjamin.widawsky@intel.com>
Date:   Tue Apr 29 14:52:29 2014 -0700

    drm/i915: Virtualize the ringbuffer signal func

Looks to me like a rebase side-effect...
Signed-off-by: NOscar Mateo <oscar.mateo@intel.com>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

gen8_stolen_size() is missing __init, so add it.

Also all the intel_stolen_funcs structures can be marked
__initconst.

intel_stolen_ids[] can also be made const if we replace the
__initdata with __initconst.

Cc: Ingo Molnar <mingo@kernel.org>
Cc: H. Peter Anvin <hpa@zytor.com>
Signed-off-by: NVille Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

CHV uses the same bits as SNB/VLV to code the Graphics Mode Select field
(GFX stolen memory size) with the addition of finer granularity modes:
4MB increments from 0x11 (8MB) to 0x1d.

Values strictly above 0x1d are either reserved or not supported.

v2: 4MB increments, not 8MB. 32MB has been omitted from the list of new
    values (Ville Syrjälä)

v3: Also correctly interpret GGMS (GTT Graphics Memory Size) (Ville
    Syrjälä)

v4: Don't assign a value that needs 20bits or more to a u16 (Rafael
    Barbalho)

[vsyrjala: v5: Split from i915 changes and add chv_stolen_funcs]

Cc: Ingo Molnar <mingo@kernel.org>
Cc: H. Peter Anvin <hpa@zytor.com>
Reviewed-by: NJani Nikula <jani.nikula@intel.com>
Reviewed-by: NVille Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: NRafael Barbalho <rafael.barbalho@intel.com>
Tested-by: NRafael Barbalho <rafael.barbalho@intel.com>
Signed-off-by: NDamien Lespiau <damien.lespiau@intel.com>
Signed-off-by: NVille Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

CHV uses the same bits as SNB/VLV to code the Graphics Mode Select field
(GFX stolen memory size) with the addition of finer granularity modes:
4MB increments from 0x11 (8MB) to 0x1d.

Values strictly above 0x1d are either reserved or not supported.

v2: 4MB increments, not 8MB. 32MB has been omitted from the list of new
    values (Ville Syrjälä)

v3: Also correctly interpret GGMS (GTT Graphics Memory Size) (Ville
    Syrjälä)

v4: Don't assign a value that needs 20bits or more to a u16 (Rafael
    Barbalho)

[vsyrjala: v5: Split the early quirks to another patch]
Reviewed-by: NJani Nikula <jani.nikula@intel.com>
Reviewed-by: NVille Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: NRafael Barbalho <rafael.barbalho@intel.com>
Tested-by: NRafael Barbalho <rafael.barbalho@intel.com>
Signed-off-by: NDamien Lespiau <damien.lespiau@intel.com>
Signed-off-by: NVille Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

No CRT output on CHV, so don't call intel_crt_init().

v2: Don't disable CRT on HAS.

Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com> (v1)
Reviewed-by: NJani Nikula <jani.nikula@intel.com>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

Add chv_crtc_clock_get() to read out the DPLL settings.
Signed-off-by: NVille Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: NImre Deak <imre.deak@intel.com>
[danvet: Fix compile due to bikeshedded headers in an earlier patch.]
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

With additional of pipe C, current 1 bit registers for pipe select
for HDMI and DP are no longer able to gather for 3 pipes. As a result,
new bits location in the same registers are added.

For HDMI, VLV uses bit 30, CHV uses bit 24-25.

For DP, VLV uses bit 30, CHV uses bit 16-17.
Reviewed-by: NVille Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: NChon Ming Lee <chon.ming.lee@intel.com>
Reviewed-by: NImre Deak <imre.deak@intel.com>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

Added programming phy layer for CHV based on "Application note for 1273
CHV Display phy".

v2: Rebase the code and do some cleanup.
v3: Rework based on Ville review.
    -Fix the macro where the ch info need to swap, and add parens to ?
	 operator.
	-Fix wrong bit define for DPIO_PCS_SWING_CALC_0 and
	 DPIO_PCS_SWING_CALC_1 and rename for meaningful.
    -Add some comments for CHV specific DPIO registers.
    -Change the dp margin registery value to decimal to align with the
	 doc.
	-Fix the not clearing some value in vlv_dpio_read before write again.
    -Create new hdmi/dp encoder function for chv instead of share with
	valleyview.
v4: Rebase the code after rename the DPIO registers define and upstream
	change.
    Based on Ville review.
    -For unique transition scale selection, after Ville point out, look
	 like the doc might wrong for the bit 26.  Use bit 27 for ch0 and
	 ch1.
	-Break up some dpio write value into two/three steps for readability.
	-Remove unrelated change.
    -Add some shift define for some registers instead just give the hex
	value.
    -Fix a bug where write to wrong VLV_TX_DW3.
v5: Based on Ville review.
	- Move tx lane latency optimal setting from chv_dp_pre_pll_enable to
	  chv_pre_enable_dp, and chv_hdmi_pre_pll_enable to
	  chv_hdmi_pre_enable respectively.
 	- Fix typo in one margin_reg_value for DP_TRAIN_VOLTAGE_SWING_400.
	- Clear DPIO_TX_UNIQ_TRANS_SCALE_EN for DP and HDMI.
	- Mask the old deemph and swing bits for hdmi.
v6: Remove stub for pre_pll_enable for dp and hdmi.
Signed-off-by: NChon Ming Lee <chon.ming.lee@intel.com>
Reviewed-by: NVille Syrjälä <ville.syrjala@linux.intel.com>
[vsyrjala: Don't touch panel power sequencing on DP]
Signed-off-by: NVille Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: NImre Deak <imre.deak@intel.com>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

Added programming PLL for CHV based on "Application note for 1273 CHV
Display phy".

v2:  -Break the common lane reset into another patch.
     -Break the clock calculation into another patch.

    -The changes are based on Ville review.
    -Rework based on DPIO register define naming convention change.
    -Break the dpio write into few lines to improve readability.
    -Correct the udelay during chv_enable_pll.
    -clean up some magic numbers with some new define.
    -program the afc recal bit which was missed.

v3: Based on Ville review
	-  minor correction of the bit defination
    - add deassert/propagate data lane reset

v4: Corrected the udelay between dclkp enable and pll enable.
	Minor comment and better way to clear the TX lane reset.

v5: Squash in fixup from Rafael Barbalho.

[vsyrjala: v6: Polish the defines (Imre)]
Reviewed-by: NVille Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: NImre Deak <imre.deak@intel.com>
Signed-off-by: NChon Ming Lee <chon.ming.lee@intel.com>
Signed-off-by: NVille Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

Based on the chv clock limit, find the best divisor.

The divisor data has been verified with this spreadsheet.
P1273_DPLL_Programming Spreadsheet.

v2: Rebase the code and change the chv_find_best_dpll based on new
standard way to use intel_PLL_is_valid.  Besides, clean up some extra
variables.

v3: Ville suggest better fixed point for m2 calculation.

v4: -Add comment for the limit is compute using fast clock. (Ville)
	-Don't pass the request clock to chv_clock, as the same function will
	 be use clock readout, which doens't have request clock. (Ville)
	-Add and use DIV_ROUND_CLOSEST_ULL to consistent with other clock
	calculation. (Ville)
	-Fix the dp m2 after m2 has stored fixed point. (Ville)
Signed-off-by: NChon Ming Lee <chon.ming.lee@intel.com>
[vsyrjala: Avoid div-by-zero in chv_clock()]
Signed-off-by: NVille Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: NImre Deak <imre.deak@intel.com>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

During cold boot, the display controller needs to deassert the common
lane reset.  Only do it once during intel_init_dpio for both PHYx2 and
PHYx1.

Besides, assert the common lane reset when disable pll.  This still
to be determined whether need to do it by driver.
Signed-off-by: NChon Ming Lee <chon.ming.lee@intel.com>
[vsyrjala: Don't disable DPIO PLL when using DSI]
[vsyrjala: Don't call vlv_disable_pll() by accident on CHV]
Signed-off-by: NVille Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: NImre Deak <imre.deak@intel.com>
[danvet: Move part of a moved comment back as suggested by Imre since
it's valid for both byt and chv.]
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

Cherryview has 3 pipes.  Some of the pll dpio offset calculation is
based on pipe number.  Need to use vlv_pipe_to_channel to calculate the
correct phy channel to use for the pipe.
Signed-off-by: NChon Ming Lee <chon.ming.lee@intel.com>
Reviewed-by: NImre Deak <imre.deak@intel.com>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

The additional DPLL registers added to support Port D.  Besides, add
some new PHY control and status registers based on B-spec.

v2: Based on Ville review
	- Corrected DPIO_PHY_STATUS offset and name.
    - Rebase based on upstream change after introduce enum dpio_phy and
      enum dpio_channel.

v3: Rebased on top of Antti's 3-pipe prep patch. Note that the new offsets for
the DPLL registers aren't in place yet, so this introduces a slight regression.
But since 3 pipe support isn't fully enabled yet anyaway in -internal this
shouldn't matter too much.
Signed-off-by: NChon Ming Lee <chon.ming.lee@intel.com>
Reviewed-by: NImre Deak <imre.deak@intel.com>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

CHV has 2 display phys.  First phy (IOSF offset 0x1A) has two channels,
and second phy (IOSF offset 0x12) has single channel.  The first phy is
used for port B and port C, while second phy is only for port D.

v2: Move the pipe to determine which phy to select for
vlv_dpio_read/vlv_dpio_write to another patch. (Daniel)
v3: Rebase the code based on rework on how to calculate DPIO offset.
Signed-off-by: NChon Ming Lee <chon.ming.lee@intel.com>
Reviewed-by: NImre Deak <imre.deak@intel.com>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

Fill in the sprite bits for DDL1/DDL2 registers, and add DDL3.

Still need to write the code to use these...
Signed-off-by: NVille Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: NJani Nikula <jani.nikula@intel.com>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

v2: Update to also fill in the new num_pipes field.

v3: Rebase on top of the pciid extraction.

v4: Switch from info->has*ring to info->ring mask. Also add VEBOX support whiel
at it.

v5: s/CHV_PCI_IDS/CHV_IDS/, and drop the trailing '\'
Signed-off-by: NVille Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: NJani Nikula <jani.nikula@intel.com>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

CHV clock gating isn't identical to VLV, so add a new function
for it. This is only a start, and further changes are needed as
the details become available.
Signed-off-by: NVille Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: NJani Nikula <jani.nikula@intel.com>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

Make i915_gem_interrupt debugfs file functional on CHV.

FIXME: Extract helpers for gt/display blocks to shrink the function a
bit and avoid duplication between bdw/chv (and other similar cases for
upstream).
Signed-off-by: NVille Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: NJani Nikula <jani.nikula@intel.com>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

Inspired by a review bikeshed from Jani.

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

CHV has the Gen8 master interrupt register, as well as Gen8
GT/PCU interrupt registers.

The display block is based on VLV, with the main difference
of adding pipe C.

v2: Rewrite the order of operations to make more sense
    Don't bail out if MASTER_CTL register doesn't show an interrupt,
    as display interrupts aren't reported there.

v3: Rebase on top of Egbert Eich's hpd irq handling rework by using
the relevant port hotplug logic like for vlv.

v4: Rebase on top of Ben's gt irq #define refactoring.

v5: Squash in gen8_gt_irq_handler refactoring from Zhao Yakui
<yakui.zhao@intel.com>

v6: Adapt to upstream changes, dev_priv->irq_received is gone.

v7: Enable 3 the commented-out 3 pipe support.

v8: Rebase on top of Paulo's irq setup rework, use the renamed macros from
upstream.

v9: Grab irq_lock around i915_enable_pipestat()

FIXME: There's probably some potential for more shared code between bdw and chv.

Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com> (v2)
Reviewed-by: NJani Nikula <jani.nikula@intel.com>
[danvet: Drop the unnecessary cast Jani spotted.]
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

For clients that submit large batch buffers the command parser has
a substantial impact on performance. On my HSW ULT system performance
drops as much as ~20% on some tests. Most of the time is spent in the
command lookup code. Converting that from the current naive search to
a hash table lookup reduces the performance drop to ~10%.

The choice of value for I915_CMD_HASH_ORDER allows all commands
currently used in the parser tables to hash to their own bucket (except
for one collision on the render ring). The tradeoff is that it wastes
memory. Because the opcodes for the commands in the tables are not
particularly well distributed, reducing the order still leaves many
buckets empty. The increased collisions don't seem to have a huge
impact on the performance gain, but for now anyhow, the parser trades
memory for performance.

NB: Ville noticed that the error paths through the ring init code
will leak memory. I've not addressed that here. We can do a follow
up pass to handle all of the leaks.

v2: improved comment describing selection of hash key mask (Damien)
replace a BUG_ON() with an error return (Tvrtko, Ville)
commit message improvements
Signed-off-by: NBrad Volkin <bradley.d.volkin@intel.com>
Reviewed-by: NDamien Lespiau <damien.lespiau@intel.com>
Reviewed-by: NTvrtko Ursulin <tvrtko.ursulin@intel.com>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

Use the same code for enabling/disabling planes on all platforms. Rename
the functions to reflect that they're no longer specific to any
platform.

For now we leave the plane enable/disable to ccur at the same old
position in the modeset sequence.
Signed-off-by: NVille Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: NChris Wilson <chris@chris-wilson.co.uk>
[danvet: Frob drm_vblank_on conflict.]
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

During the review of

commit 1f70999f
Author: Chris Wilson <chris@chris-wilson.co.uk>
Date:   Mon Jan 27 22:43:07 2014 +0000

    drm/i915: Prevent recursion by retiring requests when the ring is full

Ville raised the point that our interaction with request->tail was
likely to foul up other uses elsewhere (such as hang check comparing
ACTHD against requests).

However, we also need to restore the implicit retire requests that certain
test cases depend upon (e.g. igt/gem_exec_lut_handle), this raises the
spectre that the ppgtt will randomly call i915_gpu_idle() and recurse
back into intel_ring_begin().
Signed-off-by: NChris Wilson <chris@chris-wilson.co.uk>
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=78023Reviewed-by: NBrad Volkin <bradley.d.volkin@intel.com>
[danvet: Remove now unused 'tail' variable as spotted by Brad.]
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

A few improvements to the fallback method for waiting upon ring space:

1. Fix the start/end wait tracepoints to always be paired.
2. Increase responsiveness of checking
3. Mark the process as waiting upon io
4. Check for signal interruptions
Signed-off-by: NChris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: NBrad Volkin <bradley.d.volkin@intel.com>
[danvet: Drop the s/msleep/io_schedule_timeout/ change again since the
latter isn't exported.]
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

There is a good debate to be had about how best to fit the aliasing
PPGTT into the code. However, as it stands right now, getting aliasing
PPGTT bindings is a hack, and done through implicit arguments. To make
this absolutely clear, WARN and return an error if a driver writer tries
to do something they shouldn't.

I have no issue with an eventual revert of this patch. It makes sense
for what we have today.
Signed-off-by: NBen Widawsky <ben@bwidawsk.net>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

It was always the intention to do the topdown allocation for context
objects (Chris' idea originally). Unfortunately, I never managed to land
the patch, but someone else did, so now we can use it.

As a reminder, hardware contexts never need to be in the precious GTT
aperture space - which is what is what happens with the normal bottom up
allocation we do today. Doing a top down allocation increases the odds
that the HW contexts can get out of the way, especially with per FD
contexts as is done in full PPGTT
Signed-off-by: NBen Widawsky <ben@bwidawsk.net>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

Signed-off-by: NImre Deak <imre.deak@intel.com>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

Add runtime PM support for VLV, but leave it disabled. The next patch
enables it.

The suspend/resume sequence used is based on [1] and [2]. In practice we
depend on the GT RC6 mechanism to save the HW context depending on the
render and media power wells. By the time we run the runtime suspend
callback the display side is also off and the HW context for that is
managed by the display power domain framework.

Besides the above there are Gunit registers that depend on a system-wide
power well. This power well goes off once the device enters any of the
S0i[R123] states. To handle this scenario, save/restore these Gunit
registers. Note that this is not the complete register set dictated by
[2], to remove some overhead, registers that are known not to be used are
ignored. Also some registers are fully setup by initialization functions
called during resume, these are not saved either. The list of registers
can be further reduced, see the TODO note in the code.

[1] VLV_gfx_clocking_PM_reset_y12w21d3 / "Driver D3 entry/exit"
[2] VLV2_S0IXRegs

v2:
- unchanged
v3:
- fix s/GEN6_PMIIR/GEN6_PMIMR/ typo when saving/restoring registers
  (Ville)
v4:
- rebased on the previous patch fixing GEN register prefixes
Reviewed-by: NVille Syrjälä <ville.syrjala@linux.intel.com>
[ rebased (according to v4) ]
Signed-off-by: NImre Deak <imre.deak@intel.com>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>