The perfect solution for psr_exit is the hardware tracking the changes and
doing the psr exit by itself. This scenario works for HSW and BDW with some
environments like Gnome and Wayland.

However there are many other scenarios that this isn't true. Mainly one right
now is KDE users on HSW and BDW with PSR on. User would miss many screen
updates. For instances any key typed could be seen only when mouse cursor is
moved. So this patch introduces the ability of trigger PSR exit on kernel side
on some common cases that.

Most of the cases are coverred by psr_exit at set_domain. The remaining cases
are coverred by triggering it at set_domain, busy_ioctl, sw_finish and
mark_busy.

The downside here might be reducing the residency time on the cases this
already work very wall like Gnome environment. But so far let's get focused
on fixinge issues sio PSR couild be used for everybody and we could even
get it enabled by default. Later we can add some alternatives to choose the
level of PSR efficiency over boot flag of even over crtc property.

v2: remove exit from connector_dpms. Daniel pointed this is the wrong way and
also this isn't needed for BDW and HSW anyway.

Cc: Daniel Vetter <daniel.vetter@ffwll.ch>
Reviewed-by: NVijay Purushothaman <vijay.a.purushothaman@intel.com>
Signed-off-by: NRodrigo Vivi <rodrigo.vivi@intel.com>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

obj->framebuffer_references isn't an atomic_t so the decrement needs to
be protected by some lock. struct_mutex seems like the appropriate lock
here, and we may already take it for the obj unref anyway.
Signed-off-by: NVille Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

The DRM core will translate calls to legacy cursor ioctls into universal
cursor calls automatically, so there's no need to maintain the legacy
cursor support.  This greatly simplifies the transition since we don't
have to handle reference counting differently depending on which cursor
interface was called.

The aim here is to transition to the universal plane interface with
minimal code change.  There's a lot of cleanup that can be done (e.g.,
using state stored in crtc->cursor->fb rather than intel_crtc) that is
left to future patches.

v4:
 - Drop drm_gem_object_unreference() that is no longer needed now that
   we receive the GEM obj directly rather than looking up the ID.
v3:
 - Pass cursor obj to intel_crtc_cursor_set_obj() if cursor fb changes,
   even if 'visible' is false.  intel_crtc_cursor_set_obj() will notice
   that the cursor isn't visible and disable it properly, but we still
   need to get intel_crtc->cursor_addr set properly so that we behave
   properly if the cursor becomes visible again in the future without
   changing the cursor buffer (noted by Chris Wilson and verified
   via i-g-t kms_cursor_crc).
 - s/drm_plane_init/drm_universal_plane_init/.  Due to type
   compatibility between enum and bool, everything actually works
   correctly with the wrong init call, except for the type of plane that
   gets exposed to userspace (it shows up as type 'primary' rather than
   type 'cursor').
v2:
 - Remove duplicate dimension checks on cursor
 - Drop explicit cursor disable from crtc destroy (fb & plane
   destruction will take care of that now)
 - Use DRM plane helper to check update parameters

Cc: intel-gfx@lists.freedesktop.org
Signed-off-by: NMatt Roper <matthew.d.roper@intel.com>
Reviewed-by: Pallavi G<pallavi.g@intel.com>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

Refactor cursor buffer setting such that the code to actually update the
cursor lives in a new function, intel_crtc_cursor_set_obj(), and takes
a GEM object as a parameter.  The existing legacy cursor ioctl handler,
intel_crtc_cursor_set() will now perform the userspace handle lookup and
then call this new function.

This refactoring is in preparation for the universal plane cursor
support where we'll want to update the cursor with an actual GEM buffer
object (obtained via drm_framebuffer) rather than a userspace handle.

v2:  Drop obvious kerneldoc and replace with note about function's
     reference consumption
Reviewed-by: NDaniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: NMatt Roper <matthew.d.roper@intel.com>
Reviewed-by: Pallavi G<pallavi.g@intel.com>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

Keeping track of the power domains is a bit messy since crtc->active
is currently updated by the platform hooks, but we need to be aware of
which state transition exactly is going on. Maybe we simply need to
shovel all the power domain handling down into platform code to
simplify this. But doing that requires some more auditing since
currently the ->mode_set callbacks still read some random registers
(to e.g. figure out the reference clocks).

Also note that intel_crtc_update_dpms is always call first/last even
for encoders which have their own dpms functions. Hence we really only
need to update this place here.

Being a quick "does it blow up?" run not really tested yet.

v2: Don't do runtime PM in the DPMS hooks for HAS_DDI platforms since
that is stalled. Also add a comment to explain what's going on.
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

Intel hardware allows the primary plane to be disabled independently of
the CRTC.  Provide custom primary plane handling to allow this.

v8:
 - Pin/unpin properly when clipping causes the primary plane to be
   disabled when it has previously been enabled.
 - s/drm_primary_helper_check_update/drm_plane_helper_check_update/
v7:
 - Clip primary plane to invisible when crtc is disabled since
   intel_crtc->config.pipe_src_{w,h} may be garbage otherwise.
 - Unpin old fb before pinning new one in the "just pin and
   return" case that is used when the crtc is disabled.
 - Don't treat implicit disabling of the primary plane (caused by
   clipping) the same way as explicit disabling (caused by fb=0).
   For implicit disables, we should leave the fb set and pinned,
   whereas for explicit disables we need to unpin the fb before
   primary->fb is cleared.
v6:
 - Pass rectangles to primary helper check function and get plane
   visibility back.
 - Wait for pending pageflips on primary plane update/disable.
 - Allow primary plane to be updated while the crtc is disabled (changes
   will take effect when the crtc is re-enabled if modeset passes -1
   for the fb id).
 - Drop WARN() if we try to disable the primary plane when it's
   already been disabled.  This will happen if the crtc gets disabled
   after the primary plane has already been disabled independently.
v5:
 - Use new drm_primary_helper_check_update() helper function to check
   setplane parameter validity.
 - Swap primary plane's pipe for pre-gen4 FBC (caught by Ville Syrjälä)
 - Cleanup primary plane properly on crtc init failure
v4:
 - Don't add a primary_plane field to intel_crtc; that was left over
   from a much earlier iteration of this patch series, but is no longer
   needed/used now that the DRM core primary plane support has been
   merged.
v3:
 - Provide gen-specific primary plane format lists (suggested by Daniel
   Vetter).
 - If the primary plane is already enabled, go ahead and just call the
   primary plane helper to do the update (suggested by Daniel Vetter).
 - Don't try to disable the primary plane on destruction; the DRM layer
   should have already taken care of this for us.
v2:
 - Unpin fb properly on primary plane disable
 - Provide an Intel-specific set of primary plane formats
 - Additional sanity checks on setplane (in line with the checks
   currently being done by the DRM core primary plane helper)
Reviewed-by: NChon Ming Lee <chon.ming.lee@intel.com>
Signed-off-by: NMatt Roper <matthew.d.roper@intel.com>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

In a future patch, we'll allow the primary plane to be disabled by
userspace via the universal plane API.  If a modeset is requested while
the primary plane is disabled, crtc->primary->fb will be NULL which
generally triggers a full modeset (except in fastboot situations).  If
we detect that the crtc is active, but there's no primary plane fb,
we should still allow a simple plane update rather than a full modeset
if the mode isn't actually changing (after re-enabling the primary plane
of course).

v2:
 - Enable plane after set_base to avoid enabling the plane if set_base
   fails, and to make flip+enable atomic (suggested by Ville)
 - Drop BUG to WARN if we somehow enter the 'fb_changed' modeset case
   with the crtc disabled (suggested by Ville)
Reviewed-by: NChon Ming Lee <chon.ming.lee@intel.com>
Signed-off-by: NMatt Roper <matthew.d.roper@intel.com>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

This doesn't look possible but a little extra defense against the
improbable is worth it - an oops here could lockup the machine.
Signed-off-by: NChris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

Now that we forced the clock buffers on in .pre_pll_enable() we
should probably undo the damage after we've turned the PLL off.

We do the clock buffer force enable in the .pre_pll_enable() hook
as we need to know which port is going to be used, but in the disable
case we don't need the port since we just disable the clock buffers
to both channels. So we can do this in chv_disable_pll() instead
of having to add any kind of .post_pll_disable() hook.

v2: Improve the commit message
Signed-off-by: NVille Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: NMika Kuoppala <mika.kuoppala@intel.com>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

Because of the upcoming vblank interrupt driven watermark update
mechanism we will have use for vblank interrupts during plane
enabling/disabling. So don't call drm_vblank_off() until planes
are off, and call drm_vblank_on() just before we start to enable
the planes.

v2: Pimp commit message (Paulo)
Reviewed-by: NPaulo Zanoni <paulo.r.zanoni@intel.com>
Signed-off-by: NVille Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

Apparently it does more harm than good. Thomas Richter reports that
it helps his machine (Thinkpad X31) and there's another report from a
Fujitsu S6010. Also, we've nuked it on i845G already to make Chris'
machine happy.

Cc: Thomas Richter <richter@rus.uni-stuttgart.de>
References: http://mid.mail-archive.com/538C54E0.8090507@rus.uni-stuttgart.deSigned-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

It is possible for userspace to create a big object large enough for a
256x256, and then switch over to using it as a 64x64 cursor. This
requires the cursor update routines to check for a change in width on
every update, rather than just when the cursor is originally enabled.

This also fixes an issue with 845g/865g which cannot change the base
address of the cursor whilst it is active.
Signed-off-by: NChris Wilson <chris@chris-wilson.co.uk>
[Antti:rebased, adjusted macro names and moved some lines, no functional
changes]
Reviewed-by: NAntti Koskipaa <antti.koskipaa@linux.intel.com>
Tested-by: NAntti Koskipaa <antti.koskipaa@linux.intel.com>
Cc: stable@vger.kernel.org
Testcase: igt/kms_cursor_crc/cursor-size-change
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

Some platforms may not have it, and enumerating it is both confusing and
time consuming due to the hotplug and DDC probing.
Signed-off-by: NJesse Barnes <jbarnes@virtuousgeek.org>
Reviewed-by: NBen Widawsky <ben@bwidawsk.net>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

We need to do this anytime we power gate the DPIO common well.
Signed-off-by: NJesse Barnes <jbarnes@virtuousgeek.org>
Reviewed-by: NVille Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

This needs to be done before we power back on the CMN_BC well so the PHY
can calibrate properly.
Signed-off-by: NJesse Barnes <jbarnes@virtuousgeek.org>
Reviewed-by: NVille Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

This is a bit like the CMN reset de-assert we do in DPIO_CTL, except
that it resets the whole common lane section of the PHY.  This is
required on machines where the BIOS doesn't do this for us on boot or
resume to properly re-calibrate and get the PHY ready to transmit data.

Without this patch, such machines won't resume correctly much of the time,
with the symptom being a 'port ready' timeout and/or a link training
failure.

Note that simply asserting reset at suspend and de-asserting at resume
is not sufficient, nor is simply de-asserting at boot.  Both of these
cases have been tested and have still been found to have failures on
some configurations.

v2: extract simpler set_power_well function for use in reset_dpio (Imre)
    move to reset_dpio (Daniel & Ville)
v3: don't reset if DPIO reset is already de-asserted (Imre)
Signed-off-by: NJesse Barnes <jbarnes@virtuousgeek.org>
Reviewed-by: NVille Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

For atomic, it will be quite necessary to not need to care so much
about locking order.  And 'state' gives us a convenient place to stash a
ww_ctx for any sort of update that needs to grab multiple crtc locks.

Because we will want to eventually make locking even more fine grained
(giving locks to planes, connectors, etc), split out drm_modeset_lock
and drm_modeset_acquire_ctx to track acquired locks.

Atomic will use this to keep track of which locks have been acquired
in a transaction.

v1: original
v2: remove a few things not needed until atomic, for now
v3: update for v3 of connection_mutex patch..
v4: squash in docbook
v5: doc tweaks/fixes
Signed-off-by: NRob Clark <robdclark@gmail.com>
Reviewed-by: NDaniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: NDave Airlie <airlied@redhat.com>

After the split-out of crtc locks from the big mode_config.mutex
there's still two major areas it protects:
- Various connector probe states, like connector->status, EDID
  properties, probed mode lists and similar information.
- The links from connector->encoder and encoder->crtc and other
  modeset-relevant connector state (e.g. properties which control the
  panel fitter).

The later is used by modeset operations. But they don't really care
about the former since it's allowed to e.g. enable a disconnected VGA
output or with a mode not in the probed list.

Thus far this hasn't been a problem, but for the atomic modeset
conversion Rob Clark needs to convert all modeset relevant locks into
w/w locks. This is required because the order of acquisition is
determined by how userspace supplies the atomic modeset data. This has
run into troubles in the detect path since the i915 load detect code
needs _both_ protections offered by the mode_config.mutex: It updates
probe state and it needs to change the modeset configuration to enable
the temporary load detect pipe.

The big deal here is that for the probe/detect users of this lock a
plain mutex fits best, but for atomic modesets we really want a w/w
mutex. To fix this lets split out a new connection_mutex lock for the
modeset relevant parts.

For simplicity I've decided to only add one additional lock for all
connector/encoder links and modeset configuration states. We have
piles of different modeset objects in addition to those (like bridges
or panels), so adding per-object locks would be much more effort.

Also, we're guaranteed (at least for now) to do a full modeset if we
need to acquire this lock. Which means that fine-grained locking is
fairly irrelevant compared to the amount of time the full modeset will
take.

I've done a full audit, and there's just a few things that justify
special focus:
- Locking in drm_sysfs.c is almost completely absent. We should
  sprinkle mode_config.connection_mutex over this file a bit, but
  since it already lacks mode_config.mutex this patch wont make the
  situation any worse. This is material for a follow-up patch.

- omap has a omap_framebuffer_flush function which walks the
  connector->encoder->crtc links and is called from many contexts.
  Some look like they don't acquire mode_config.mutex, so this is
  already racy. Again fixing this is material for a separate patch.

- The radeon hot_plug function to retrain DP links looks at
  connector->dpms. Currently this happens without any locking, so is
  already racy. I think radeon_hotplug_work_func should gain
  mutex_lock/unlock calls for the mode_config.connection_mutex.

- Same applies to i915's intel_dp_hot_plug. But again, this is already
  racy.

- i915 load_detect code needs to acquire this lock. Which means the
  w/w dance due to Rob's work will be nicely contained to _just_ this
  function.

I've added fixme comments everywhere where it looks suspicious but in
the sysfs code. After a quick irc discussion with Dave Airlie it
sounds like the lack of locking in there is due to sysfs cleanup fun
at module unload.

v1: original (only compile tested)

v2: missing mutex_init(), etc (from Rob Clark)

v3: i915 needs more care in the conversion:
- Protect the edp pp logic with the connection_mutex.
- Use connection_mutex in the backlight code due to
  get_pipe_from_connector.
- Use drm_modeset_lock_all in suspend/resume paths.
- Update lock checks in the overlay code.

Cc: Alex Deucher <alexdeucher@gmail.com>
Cc: Rob Clark <robdclark@gmail.com>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>
Reviewed-by: NRob Clark <robdclark@gmail.com>

Generated using semantic patches:

@@
expression E;
@@

- drm_get_encoder_name(&E)
+ E.name

@@
expression E;
@@

- drm_get_encoder_name(E)
+ E->name

v2: Turn drm_get_encoder_name(&E) into E.name instead of &(E)->name.
Acked-by: NDavid Herrmann <dh.herrmann@gmail.com>
Signed-off-by: NJani Nikula <jani.nikula@intel.com>
Signed-off-by: NDave Airlie <airlied@redhat.com>

Generated using semantic patches:

@@
expression E;
@@

- drm_get_connector_name(&E)
+ E.name

@@
expression E;
@@

- drm_get_connector_name(E)
+ E->name

v2: Turn drm_get_connector_name(&E) into E.name instead of &(E)->name.
Acked-by: NDavid Herrmann <dh.herrmann@gmail.com>
Signed-off-by: NJani Nikula <jani.nikula@intel.com>
Signed-off-by: NDave Airlie <airlied@redhat.com>

A single object may be referenced by multiple registers fundamentally
breaking the static allotment of ids in the current design. When the
object is used the second time, the physical address of the first
assignment is relinquished and a second one granted. However, the
hardware is still reading (and possibly writing) to the old physical
address now returned to the system. Eventually hilarity will ensue, but
in the short term, it just means that cursors are broken when using more
than one pipe.

v2: Fix up leak of pci handle when handling an error during attachment,
and avoid a double kmap/kunmap. (Ville)
Rebase against -fixes.

v3: And fix the error handling added in v2 (Ville)

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=77351Signed-off-by: NChris Wilson <chris@chris-wilson.co.uk>
Cc: Ville Syrjälä <ville.syrjala@linux.intel.com>
Cc: Jani Nikula <jani.nikula@linux.intel.com>
Cc: stable@vger.kernel.org
Reviewed-by: NVille Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

In the upcoming patches we plan to break the correlation between
engine command streamers (a.k.a. rings) and ringbuffers, so it
makes sense to refactor the code and make the change obvious.

No functional changes.
Signed-off-by: NOscar Mateo <oscar.mateo@intel.com>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

On gen2 the scanline counter behaves a bit differently from the
later generations. Instead of adding one to the raw scanline
counter value, we must subtract one.

On HSW/BDW the scanline counter requires a +2 adjustment on HDMI
outputs. DP outputs on the on the other require the typical +1
adjustment.

As the fixup we must apply to the hardware scanline counter
depends on several factors, compute the desired offset at modeset
time and tuck it away for when it's needed.

v2: Clarify HSW+ situation
Signed-off-by: NVille Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: "Akash Goel <akash.goels@gmail.com>"
Reviewed-by: "Sourab Gupta <sourabgupta@gmail.com>"
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=78997Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

We have to write to the primary plane base address registrer when we
enable/disable the primary plane in response to sprite coverage. Those
writes will cause the flip counter to increment which could interfere
with the detection of CS flip completion. We could end up completing
CS flips before the CS has even executed the commands from the ring.

To avoid such issues, wait for CS flips to finish before we toggle the
primary plane on/off.

v2: Rebased due to atomic sprite update changes

Testcase: igt/kms_mmio_vs_cs_flip/setplane_vs_cs_flip
Signed-off-by: NVille Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by Rodrigo Vivi <rodrigo.vivi@gmail.com>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

Gen2 reports FIFO underruns whenever no planes are enabled on the pipe.
So in order to avoid false positives we must enable the FIFO underrun
reporting only when at least one plane is enabled on the pipe. For
now just move the underrun reporting enable/disable points to the
other side of the plane enable/disable point. That doesn't cover cases
when we turn off all the planes for the pipe but leave the pipe running
on purpose, but it's better than the current situation.

On gen4+ we can actually move the underrun reporting enable/disable to
the opposite ends of the crtc enable/disable hooks. I suppose in theory
we could leave the underrun reporting enabled all the time, except on
VLV where PIPESTAT stops working when the display power well is down.
If we ever get around to unifying the PIPESTAT irq handling for all
gmch platforms, we should still follow the VLV route for other platforms.
It would also micro-optimize the irq handler a bit since we could then
skip the PIPESTAT reads for all disabled pipes.

Gen3 is still a mystery, but for now I'm going to assume it behaves
like gen4+.
Signed-off-by: NVille Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: NThomas Wood <thomas.wood@intel.com>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

FIFO underruns don't generate interrupts on gmch platforms, so
if we want to know whether a modeset triggered FIFO underruns we
need to explicitly check for them.

As a modeset on one pipe could cause underruns on other pipes,
check for underruns on all pipes.
Signed-off-by: NVille Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: NThomas Wood <thomas.wood@intel.com>
[danvet: Fix up merge error, kudos to Ville for noticing it.]
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

If a pipe is already active when we init/resume there might not be a
full modeset afterwards so drm_vblank_on() may not get called. In such
a case if someone is holding a vblank reference across a suspend/resume
cycle drm_vblank_get() called after resuming won't re-enable the vblank
interrupts.

So in order to make sure vblank interrupts get re-enabled post-resume,
call drm_vblank_on() in intel_sanitize_crtc() if the crtc is already
active.

v2: Also drm_vblank_off() if the pipe got disabled magically
Signed-off-by: NVille Syrjälä <ville.syrjala@linux.intel.com>
Testecase: igt/kms_flip/vblank-vs-suspend
Tested-by: NDaniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

We don't have hardware based disable bits on gmch platforms, so need
to block spurious underrun reports in software. Which means that we
_must_ start out with fifo underrun reporting disabled everywhere.

This is in big contrast to ilk/hsw/cpt where there's only _one_
disable bit for all platforms and hence we must allow underrun
reporting on disabled pipes. Otherwise nothing really works,
especially the CRC support since that's key'ed off the same irq
disable bit.

This allows us to ditch the fifo underrun reporting hack from the vlv
runtime pm code and unexport the internal function from i915_irq.c
again. Yay!

v2: Keep the display irq disabling, spotted by Imre.

Cc: Imre Deak <imre.deak@intel.com>
Reviewed-by: NVille Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

Only the low-level irq handling functions still use integer crtc
indices with this. But fixing that will require a lot more sugery
and some good ideas for backwards compat with old ums userspace.
Both in drivers and in the drm core.
Reviewed-by: NVille Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

Originally these functions have been for user modesetting drivers to
ensure vblank processing doesn't fall over completely around modeset
changes. This has been carried over ever since then.

Now that Ville cleaned our vblank handling with an explicit
drm_vblank_off/on braket when disabling/enabling crtcs. So this seems
to be unnecessary now. The most important side effect was that due to
the delayed vblank disabling we have been pretty much guaranteed to
receive a vblank interrupt soonish after a crtc was enabled.

Note that our vblank handling across modeset is still fairly decent
fubar - we don't actually handle vblank counter all to well.
drm_update_vblank_count will make sure that the frame counter always
rolls forward, but userspace isn't really all to ready to cope with
the big jumps this causes.

This isn't a big mostly because the hardware retains the frame
counter. But with runtime pm and also across suspend/resume we fall
over.

Fixing this is a lot more involved and also needs som i-g-ts. So
material for another patch series.
Reviewed-by: NVille Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

All of the .queue_flip() callbacks duplicate the same code to pin the
buffers and calculate the gtt_offset. Move that code to
intel_crtc_page_flip(). In order to do that we must also move the ring
selection logic there.
Signed-off-by: NVille Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: NChris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

Now that we've plugged the mmio vs. ring flip race, we shouldn't need
these vblank waits in the modeset codepaths anymore. So get rid of
them.

v2: gen2 needs to wait for planes to turn off before disabling pipe
Signed-off-by: NVille Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: NRodrigo Vivi <rodrigo.vivi@gmail.com>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

Now that the vblank wait is gone from intel_enable_primary_plane(),
hsw_enable_ips() needs to do the vblank wait itself.
Signed-off-by: NVille Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: NRodrigo Vivi <rodrigo.vivi@gmail.com>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

Starting from ILK, mmio flips also cause a flip done interrupt to be
signalled. This means if we first do a set_base and follow it
immediately with the CS flip, we might mistake the flip done interrupt
caused by the set_base as the flip done interrupt caused by the CS
flip.

The hardware has a flip counter which increments every time a mmio or
CS flip is issued. It basically counts the number of DSPSURF register
writes. This means we can sample the counter before we put the CS
flip into the ring, and then when we get a flip done interrupt we can
check whether the CS flip has actually performed the surface address
update, or if the interrupt was caused by a previous but yet
unfinished mmio flip.

Even with the flip counter we still have a race condition of the CS flip
base address update happens after the mmio flip done interrupt was
raised but not yet processed by the driver. When the interrupt is
eventually processed, the flip counter will already indicate that the
CS flip has been executed, but it would not actually complete until the
next start of vblank. We can use the DSPSURFLIVE register to check
whether the hardware is actually scanning out of the buffer we expect,
or if we managed hit this race window.

This covers all the cases where the CS flip actually changes the base
address. If the base address remains unchanged, we might still complete
the CS flip before it has actually completed. But since the address
didn't change anyway, the premature flip completion can't result in
userspace overwriting data that's still being scanned out.

CTG already has the flip counter and DSPSURFLIVE registers, and
although the flip done interrupt is still limited to CS flips alone,
the code now also checks the flip counter on CTG as well.

v2: s/dspsurf/gtt_offset/ (Chris)

Testcase: igt/kms_mmio_vs_cs_flip/setcrtc_vs_cs_flip
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=73027Signed-off-by: NVille Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: NRodrigo Vivi <rodrigo.vivi@gmail.com>
[danvet: Add g4x_ prefix to flip_count_after_eq.]
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

drm_vblank_off() will turn off vblank interrupts, but as long as the
refcount is elevated drm_vblank_get() will not re-enable them. This
is a problem is someone is holding a vblank reference while a modeset is
happening, and the driver requires vblank interrupt to work during that
time.

Add drm_vblank_on() as a counterpart to drm_vblank_off() which will
re-enabled vblank interrupts if the refcount is already elevated. This
will allow drivers to choose the specific places in the modeset sequence
at which vblank interrupts get disabled and enabled.

Testcase: igt/kms_flip/*-vs-suspend
Signed-off-by: NVille Syrjälä <ville.syrjala@linux.intel.com>
[danvet: Add Testcase tag for the igt I've written.]
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

Seems like we shouldn't leave the data lane resert deasserted when
the port if disabled. So propagate the reset the data lanes in
the encoder .post_disable() hook.
Signed-off-by: NVille Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: NMika Kuoppala <mika.kuoppala@intel.com>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

During the enable sequence we first enable the dclkp output to the
display controller, and then enable the PLL. Do the opposite during
the disable sequence.
Signed-off-by: NVille Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: NMika Kuoppala <mika.kuoppala@intel.com>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

We need to pick the correct data lanes based on the port not the
pipe, so move the data lane deassert into the encoder .pre_enable()
hook from the chv_enable_pll().
Signed-off-by: NVille Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: NMika Kuoppala <mika.kuoppala@intel.com>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

Setup the pipe config dpll state correctly for CHV. Also add
a assert_pipe_disabled() to chv_disable_pll(), and program the
DPLL_MD registers in chv_enable_pll().
Signed-off-by: NVille Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: NMika Kuoppala <mika.kuoppala@intel.com>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

Signed-off-by: NVille Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: NAntti Koskipää <antti.koskipaa@linux.intel.com>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>