Because we merged the PC8 and runtime PM features, so calling
intel_runtime_pm_get now has the same meaning, and we plan to just
remove hsw_disable_package_c8 for this exact reason.

My first patch tried to completely kill
intel_aux_display_runtime_get/put, because I was assuming that whoever
needed more than just runtime PM would have to get the appropriate
power domain instead of that, but it seems some people still want the
intel_aux_display_runtime_get abstraction, so keep it until someone
else tries to replace it with the more-standard power domain calls.
Signed-off-by: NPaulo Zanoni <paulo.r.zanoni@intel.com>
Reviewed-by: NImre Deak <imre.deak@intel.com>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

Since after the latest patches it's only being used to prevent
getting/putting the runtime PM refcount.
Reviewed-by: NJesse Barnes <jbarnes@virtuousgeek.org>
Signed-off-by: NPaulo Zanoni <paulo.r.zanoni@intel.com>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

The requirements_met variable was used to track two things: enabled
CRTCs and the power well. After the latest chagnes, we get a runtime
PM reference whenever we get any of the power domains, and we get
power domains when we enable CRTCs or the power well, so we should
already be covered, not needing this specific tracking.

v2: - Rebase.
v3: - Rebase.
Signed-off-by: NPaulo Zanoni <paulo.r.zanoni@intel.com>
Reviewed-by: NImre Deak <imre.deak@intel.com>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

Any power domain will require the HW to be in PCI D0 state, so just do
the simple thing.

Dear maintainer: since intel_display_power_put() and
intel_display_power_get() are almost identical, git-am has failed
apply the patch on my local machine once: it added both chunks to
put(), instead of one chunk to get() and another to put(). When you
apply this patch to your tree, please check if it is correct.

v2: - Add the warning above.
v3: - Rebase.
Reviewed-by: NImre Deak <imre.deak@intel.com>
Signed-off-by: NPaulo Zanoni <paulo.r.zanoni@intel.com>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

Currently, when our driver becomes idle for i915.pc8_timeout (default:
5s) we enable PC8, so we save some power, but not everything we can.
Then, while PC8 is enabled, if we stay idle for more
autosuspend_delay_ms (default: 10s) we'll enter runtime PM and put the
graphics device in D3 state, saving even more power. The two features
are separate things with increasing levels of power savings, but if we
disable PC8 we'll never get into D3.

While from the modularity point of view it would be nice to keep these
features as separate, we have reasons to merge them:
 - We are not aware of anybody wanting a "PC8 without D3" environment.
 - If we keep both features as separate, we'll have to to test both
   PC8 and PC8+D3 code paths. We're already having a major pain to
   make QA do automated testing of just one thing, testing both paths
   will cost even more.
 - Only Haswell+ supports PC8, so if we want to add runtime PM support
   to, for example, IVB, we'll have to copy some code from the PC8
   feature to runtime PM, so merging both features as a single thing
   will make it easier for enabling runtime PM on other platforms.

This patch only does the very basic steps required to have PC8 and
runtime PM merged on a single feature: the next patches will take care
of cleaning up everything.

v2: - Rebase.
v3: - Rebase.
    - Fully remove the deprecated i915 params since Daniel doesn't
      consider them as part of the ABI.
v4: - Rebase.
    - Fix typo in the commit message.
v5: - Rebase, again.
    - Add a huge comment explaining the different forcewake usage
      (Chris, Daniel).
    - Use open-coded forcewake functions (Daniel).
Signed-off-by: NPaulo Zanoni <paulo.r.zanoni@intel.com>
Reviewed-by: NImre Deak <imre.deak@intel.com>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

This function is only used on ILK+, so rename it accordingly.
Signed-off-by: NDamien Lespiau <damien.lespiau@intel.com>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

Make sure the line_time_us isn't zero in the gmch watermarks code as
that would cause a div by zero. This can be triggered by specifying
a very fast pixel clock for the mode.

At some point we should probably just switch over to using the same
math we use on PCH platforms which avoids such intermediate rounded
results.

Also we should verify the user provided mode much more rigorously.
At the moment we accept pretty much anything.

Note that "very fast mode" here means above 74.25 GHz.
Signed-off-by: NVille Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: NChris Wilson <chris@chris-wilson.co.uk>
[danvet: Add Ville's clarification of what "very fast" means.]
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

Based on an early draft from Jesse.

Add support for powering on/off the dynamic power wells on VLV by
registering its display and dpio dynamic power wells with the power
domain framework.

For now power on all PHY TX lanes regardless of the actual lane
configuration. Later this can be optimized when the PHY side setup
enables only the required lanes. Atm, it enables all lanes in all
cases.

v2:
- undef function local COND macro after its last use (Ville)
- Take dev_priv->irq_lock around the whole sequence of
  intel_set_cpu_fifo_underrun_reporting_nolock() and
  valleyview_disable_display_irqs(). They are short and releasing
  the lock in between only makes proving correctness more difficult.
- sanitize local var names in vlv_power_well_enabled()
v3:
- rebase on latest -nightly
Signed-off-by: NImre Deak <imre.deak@intel.com>
Reviewed-by: NJesse Barnes <jbarnes@virtuousgeek.org>
[danvet: Resolve conflict due to my changes in the previous patch.
Also throw in an assert_spin_locked for safety. And finally appease
checkpatch.]
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

Suggested by Daniel.

v2:
- sanitize the state checking condition, the original was rather
  confusing (partly due to the unfortunate naming of
  i915.disable_power_well) (Ville)
- simpler message+backtrace generation by using WARN instead of WARN_ON
  (Ville)
- check if always-on power wells are truly on all the time
Signed-off-by: NImre Deak <imre.deak@intel.com>
Reviewed-by: NJesse Barnes <jbarnes@virtuousgeek.org>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

We need to do the same for other platforms in upcoming patches.

v2:
- s/p/pipe (Ville)
- Call the new helper with the vbl_lock already held. The part it
  protects is short, so releasing it between pipes only makes proving
  correctness more difficult.
Signed-off-by: NImre Deak <imre.deak@intel.com>
Reviewed-by: NJesse Barnes <jbarnes@virtuousgeek.org>
[danvet: Resolve conflict with Damien's s/p/pipe/ change.]
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

Parts that poke port specific HW blocks like the encoder HW state
readout or connector hotplug detect code need a way to check whether
required power domains are on or enable/disable these. For this purpose
add a set of power domains that refer to the port HW blocks. Get the
proper port power domains during modeset.

For now when requesting the power domain for a DDI port get it for a 4
lane configuration. This can be optimized later to request only the 2
lane power domain, when proper support is added on the VLV PHY side for
this. Atm, the PHY setup code assumes a 4 lane config in all cases.
Signed-off-by: NImre Deak <imre.deak@intel.com>
Reviewed-by: NJesse Barnes <jbarnes@virtuousgeek.org>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

Reading code free of special cases wins over the small overhead of
calling a noop handler. Suggested by Jesse.
Signed-off-by: NImre Deak <imre.deak@intel.com>
Reviewed-by: NJesse Barnes <jbarnes@virtuousgeek.org>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

Split the 'set' power well handler into an 'enable', 'disable' and
'sync_hw' handler. This maps more conveniently to higher level
operations, for example it allows us to push the hsw package c8 handling
into the corresponding hsw/bdw enable/disable handlers and the hsw BIOS
hand-over setting into the hsw/bdw sync_hw handler.

No functional change.
Signed-off-by: NImre Deak <imre.deak@intel.com>
Reviewed-by: NJesse Barnes <jbarnes@virtuousgeek.org>
[danvet: Appease checkpatch's whitespace complaints.]
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

Whenever we request a power domain it has to guarantee that all HW
resources are enabled that are needed to access a HW register associated
with that power domain. In case a register is on an always-on power well
this won't result in turning on a power well, but it may require
enabling some other HW resource. One such resource is the HSW/BDW device
D0 state that is required for all register accesses and thus for all
power wells/power domains.

So far the init power domain (guaranteeing access to all HW registers)
was part of the default i9xx always-on power well, but not the HSW/BDW
always-on power wells. Add the domain to the latter power wells too.

Atm, all the always-on power wells have noop handlers, so this doesn't
change the functionality.

v2:
- clarify semantics of always-on power wells (Paulo)
Signed-off-by: NImre Deak <imre.deak@intel.com>
Reviewed-by: NJesse Barnes <jbarnes@virtuousgeek.org>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

These macros are used only locally, so move them to the .c file.

No functional change.

v2:
- add init power domain to always-on power wells in the following
  - separate - patch (Paulo)
Signed-off-by: NImre Deak <imre.deak@intel.com>
Reviewed-by: NJesse Barnes <jbarnes@virtuousgeek.org>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

These functions are used only by a single call site and are simple
enough to just fold them in.

Note that in later patches the parts folded in here are further
simplified as we'll remove hsw_{disable,enable}_package_c8 and the NULL
check of the power well enable/disable handlers. All this means that at
the end intel_display_power_get/put() becomes more understandable as we
don't need to jump between two functions when reading the code.

No functional change.

v2:
- clarify the rational for the change (Chris)
Signed-off-by: NImre Deak <imre.deak@intel.com>
Reviewed-by: NJesse Barnes <jbarnes@virtuousgeek.org>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

We have two names for the same register CHICKEN_PIPESL_1 and
HSW_PIPE_SLICE_CHICKEN_1. Unify it to just one.

Also rename the FBCQ disable bit to resemble the name we've
given to a similar bit on earlier platforms.
Signed-off-by: NVille Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: NDamien Lespiau <damien.lespiau@intel.com>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

gen7_enable_fbc() may write to some registers which we've already
touched, so use RMW so that we don't undo any previous updates.

Also note that we implemnt WaFbcAsynchFlipDisableFbcQueue:bdw.
Signed-off-by: NVille Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: NDamien Lespiau <damien.lespiau@intel.com>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

Misplaced parens cause us to totally clobber the CHICKEN_PIPESL_1
registers with 0xffffffff. Move the parens to the correct place
to avoid this.

In particular this caused bit 30 of said registers to be set, which
caused the sprite CSC to produce incorrect results.

Cc: stable@vger.kernel.org
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=72220Signed-off-by: NVille Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: NDamien Lespiau <damien.lespiau@intel.com>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

Consistency throughout the code base is good and remove some room for
mistakes (as explained in the "drm/i915: Use a pipe variable to cycle
through the pipes" commit)

So, let's replace the for_each_pipe(i) occurences by for_each_pipe(pipe)
when it's reasonable and practical to do so (eg. when there isn't another
pipe variable already).
Suggested-by: NChris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: NDamien Lespiau <damien.lespiau@intel.com>
Reviewed-by: NChris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

Signed-off-by: NVille Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: NBen Widawsky <ben@bwidawsk.net>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

According to BSpec we need to always set this magic bit in ring buffer
mode.
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>

Ben and I believe this will be necessary on production hardware.
Signed-off-by: NKenneth Graunke <kenneth@whitecape.org>
[danvet: Shuffle lines to group all ROW_CHICKEN writes and add a
cautious comment that this might not be needed on production hw.]
Reviewed-by: NVille Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: NBen Widawsky <ben@bwidawsk.net>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

I believe this will be necessary on production hardware.
Signed-off-by: NKenneth Graunke <kenneth@whitecape.org>
Reviewed-by: NVille Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: NBen Widawsky <ben@bwidawsk.net>
[danvet: Fix whitespace fail spotted by checkpatch. Also add missing
:bdw w/a tag that Ville spotted.]
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

Since the addition of dev_priv->mm.busy, there's no more need for
dev_priv->pc8.gpu_idle, so kill it.

Notice that when you remove gpu_idle, hsw_package_c8_gpu_idle and
hsw_package_c8_gpu_busy become identical to hsw_enable_package_c8 and
hsw_disable_package_c8, so just use them.

Also, when we boot the machine, dev_priv->mm.busy initially considers
the machine as idle. This is opposed to dev_priv->pc8.gpu_idle, which
considered it busy. So dev_priv->pc8.disable_count has to be
initalized to 1 now.
Signed-off-by: NPaulo Zanoni <paulo.r.zanoni@intel.com>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

This got lost when we shuffled around our internal branch and
GEN7_FEATURES macro. There were no HW changes to support FBC, so we just
need to set the flag.

v2: Don't allow FBC for any pipe but A on platforms with DDI. (Paulo)

Cc: Daisy Sun <daisy.sun@intel.com>
Signed-off-by: NBen Widawsky <ben@bwidawsk.net>
Reviewed-by: NPaulo Zanoni <paulo.r.zanoni@intel.com>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

Signed-off-by: NImre Deak <imre.deak@intel.com>
Reviewed-by: NJesse Barnes <jbarnes@virtuousgeek.org>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

The power domains framework is internal to the i915 driver, so pass
drm_i915_private instead of drm_device to its functions.

Also remove a dangling intel_set_power_well() declaration.

No functional change.
Signed-off-by: NImre Deak <imre.deak@intel.com>
Reviewed-by: NJesse Barnes <jbarnes@virtuousgeek.org>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

This reverts commit 116f2b6d.

This optimization causes widespread corruption in games, and even in
glxgears, on my ivb:gt1. The corruption appears like z-fighting of
overlapping polygons in the HiZ buffer.

The observation ties in very closely with the description of the
optimization disabled by default on IVB:

"The Hierarchical Z RAW Stall Optimization allows non-overlapping
polygons in the same 8x4 pixel/sample area to be processed without
stalling waiting for the earlier ones to write to Hierarchical Z
buffer."

No reason is given for why it is disabled by default, usually for such
optimizations it is that it is incomplete. However, there is no
indication whether this a gt1 only issue either. Before considering
reenabling this optimization, I would first suggest reproducing the
corruption in piglit.

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=75623Signed-off-by: NChris Wilson <chris@chris-wilson.co.uk>
Cc: Chia-I Wu <olv@lunarg.com>
Cc: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

V2:  edit the commit message to contain more info
The W/A spreadsheet says this is still required, but the b-spec says
it's not for BYT-T.  So the documentation is not clear.  However,
our experience with the other SKUs of BYT-I/M on Android and Linux
suggests that setting this bit actually causes GPU hang for certain
OGL benchmark applications.

Removing this bit completely resolves the GPU hangs.
Signed-off-by: NSinclair Yeh <sinclair.yeh@intel.com>
Acked-by: NJesse Barnes <jbarnes@virtuousgeek.org>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

Add a comment next to our WIZ hashing setup to remind people about the
link between WIZ hashing disable bit and PS/WM thread counts.
Suggested-by: NChris Wilson <chris@chris-wilson.co.uk>
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>

BSpec recommends using 8x4 hashing mode when MSAA is used. But in
practice 16x4 seems to have a slight edge in performance (on IVB and
HSW at least). So just use 16x4.
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>

BSpec recommends using 8x4 hashing mode when MSAA is used. But in
practice 16x4 seems to have a slight edge in performance (on IVB and
HSW at least). So just use 16x4.
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>

BSpec recommends using 8x4 hashing mode when MSAA is used. But in
practice 16x4 seems to have a slight edge in performance (on IVB and
HSW at least). So just use 16x4.
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>

The need to set all of the mask bits for 3D_CHICKEN3 was required
only for pre-production hardware.
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>

Based on the name, the workaround we implement is
WaStripsFansDisableFastClipPerformanceFix. Unfortunately there's no
description in the w/a database, so this is just a guess.
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>

On SNB we set up WaSetupGtModeTdRowDispatch:snb early in
gen6_init_clock_gating(). That sets a bit in the GEN6_GT_MODE register.
However later we go and disable all the bits in the same register. And
then we go on to set some other bit. So apparently we never actually
implemented this workaround since the "disable all bits" part was there
already before the w/a got supposedly implemented.

These are the relevant commits:

 commit 6547fbdb
 Author: Daniel Vetter <daniel.vetter@ffwll.ch>
 Date:   Fri Dec 14 23:38:29 2012 +0100

    drm/i915: Implement WaSetupGtModeTdRowDispatch

 commit f8f2ac9a
 Author: Ben Widawsky <ben@bwidawsk.net>
 Date:   Wed Oct 3 19:34:24 2012 -0700

    drm/i915: Fix GT_MODE default value

So, let's drop the "disable all bits" part, move both writes to
closer proxomity to each other, and name the WIZ hashing bits
appropriately. BSpec is still a bit confused how the bits should
actually be interpreted, but I took the the description for the
high bit since the low bit part only lists values for a single bit.

Also add a comment about our choice of WIZ hashing mode.
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>

We reserve the space for the power context in stolen memory at a fixed
address from a delayed work. This races with the subsequent driver
init/resume code which could allocate something at that address, so the
reservation for the power context fails. Reserve the space up-front, so
this can't happen. This also adds a missing struct_mutex lock around the
stolen allocation, which wasn't taken in the delayed work path.
Signed-off-by: NImre Deak <imre.deak@intel.com>
Reviewed-by: NJesse Barnes <jbarnes@virtuousgeek.org>
Cc: stable@vger.kernel.org
Signed-off-by: NJani Nikula <jani.nikula@intel.com>

Only the hardware really access them, so no need to have cpu
gtt access available.

Split out from Chris vma-bind rework.

Note that this is only possible due to the split-up of the mappable
pin flag into PIN_GLOBAL and PIN_MAPPABLE.

Cc: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Ben Widawsky <benjamin.widawsky@intel.com>
Reviewed-by: NChris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

Anything more than just one bool parameter is just a pain to read,
symbolic constants are much better.

Split out from Chris' vma-binding rework patch.

v2: Undo the behaviour change in object_pin that Chris spotted.

v3: Split out misplaced hunk to handle set_cache_level errors,
spotted by Jani.

v4: Keep the current over-zealous binding logic in the execbuffer code
working with a quick hack while the overall binding code gets shuffled
around.

v5: Reorder the PIN_ flags for more natural patch splitup.

v6: Pull out the PIN_GLOBAL split-up again.

Cc: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Ben Widawsky <benjamin.widawsky@intel.com>
Reviewed-by: NChris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>