drm/i915: vlv: fix switcheroo/legacy suspend/resume

During switcheroo/legacy suspend we don't call the suspend_late handler
but when resuming afterwards we call resume_early. This happened to work
so far, since suspend_late only disabled the PCI device. This changed in

commit 016970beb05da6285c2f3ed2bee1c676cb75972e
Author: Sagar Kamble <sagar.a.kamble@intel.com>
Date:   Wed Aug 13 23:07:06 2014 +0530

    drm/i915: Sharing platform specific sequence between runtime and system susp

after which we also saved/restored the VLV Gunit HW state in
suspend_late/resume_early. So now since we don't save the state during
suspend a following resume will restore a corrupted state.

Fix this by calling the suspend_late handler during both switcheroo and
legacy suspend.

CC: Sagar Kamble <sagar.a.kamble@intel.com>
Signed-off-by: Imre Deak <imre.deak@intel.com>
Reviewed-by: Sagar Kamble <sagar.a.kamble@intel.com>
Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915: propagate error from legacy resume handler

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

drm/i915: unify legacy S3 suspend and S4 freeze handlers

i915_suspend() is called from the DRM legacy S3 suspend/S4 freeze paths
and the switcheroo suspend path. For switcheroo we only ever need to
perform a full suspend (PM_EVENT_SUSPEND) and for the DRM legacy path
we can handle the S4 freeze (PM_EVENT_FREEZE) the same way as S3
suspend. The only difference atm between suspend and freeze is that
during freeze we don't disable the PCI device, but there is no reason
why we can't do so. So unify the two cases to reduce complexity.

Note that for the DRM legacy case the thaw event is not handled, so
we disable the display before creating the hibernation image and it
won't get re-enabled until reboot. We could fix this leaving the
display enabled for the image creation/writing (if we care enough
about UMS), but this can be done as a follow-up.

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

drm/i915: factor out i915_drm_suspend_late

This is needed by an upcoming patch fixing the switcheroo/legacy suspend
paths.

No functional change.

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

drm/i915: Emit even number of dwords when emitting LRIs

The number of DWords should be even when doing ring emits as
command sequences require QWord alignment.

There was some discussion about the maximum length of the MI_LRI
command. Quoting Mika

"I did some test with bdw:

"The maximum is 128 writes, resulting the 8 bit length
field of the command being 0xff, thus following the spec.
The 128'th write went through.

"Perhaps the max command length is then less in older gens?

"Perhaps WARN_ON(x > 128) in MI_LOAD_REGISTER_IMM would be in place
but one needs minor tweak to command parser a bit also then.

#define I915_MAX_WA_REGS 16

keeps us safe for now atleast."

Ville commented that on pre-gen6 the length field seems to be
restricted to 0x3f though. So for all cases we should be ok.

v2: user LRI variant that can write multiple regs in one go (Damien).
We can simply insert one NOP at the end instead of one per register write.

Cc: Mika Kuoppala <mika.kuoppala@intel.com>
Signed-off-by: Arun Siluvery <arun.siluvery@linux.intel.com>
Reviewed-by: Damien Lespiau <damien.lespiau@intel.com>
[danvet: Add a summary of the MI_LRI length discussion.]
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915: Add rotation support for cursor plane (v5)

The cursor plane also supports 180 degree rotation. Add a new
"cursor-rotation" property on the crtc which controls this.

Unlike sprites, the cursor has a fixed size, so if you have a small
cursor image with the rest of the bo filled by transparent pixels,
simply flipping the rotation property will cause the visible part
of the cursor to shift. This is something to keep in mind when
using cursor rotation.

v2: Fix gen4/vlv by offsetting the base address appropriately

v3: Removing cursor-rotation property and using rotation property on cursor
plane.
v4: Changing the author name back to Ville.

v5 (by Matt Roper): Slight tweaking to apply against latest di-nightly
codebase.

Cc: Sagar Kamble <sagar.a.kamble@intel.com>
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Sonika Jindal <sonika.jindal@intel.com>
Reviewed-by: Matt Roper <matthew.d.roper@intel.com>
Tested-by (IVB): Matt Roper <matthew.d.roper@intel.com>
Signed-off-by: Matt Roper <matthew.d.roper@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915: Correctly reject invalid flags for wait_ioctl

Not having checks for this isn't good.

I've checked igt and libdrm and they all already clear flags properly.
So we're lucky and should be able to sneak this ABI clarification in.

Testcase: igt/gem_wait/invalid-flags
Signed-off-by: Daniel Vetter <daniel.vetter@intel.com>
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=85280
Reviewed-by: Damien Lespiau <damien.lespiau@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915: use macros to assign mmio access functions

This is beautification prep work since vgt will add even more special
cases. With these macros it's much easier to see what's going on
really.

Signed-off-by: Yu Zhang <yu.c.zhang@linux.intel.com>
[danvet: #undef the temporary macros after the function again. And
write a commit message.]
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915: only run hsw_power_well_post_enable when really needed

Only run it after we actually enable the power well. When we're
booting the machine there are cases where we run
hsw_power_well_post_enable without really needing, and even though
this is not causing any real bugs, it is unneeded and causes confusion
to people debugging interrupts.

Signed-off-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915/chv: Use 16 and 32 for low and high drain latency precision.

Current chv spec teels we can only use either 16 or 32 bits as precision.

Although in the past VLV went from 16/32 to 32/64 and spec might not be updated,
these precision values brings stability and fixes some issues Wayne was facing.

Cc: Wayne Boyer <wayne.boyer@intel.com>
Cc: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Tested-by: Wayne Boyer <wayne.boyer@intel.com>
Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
[danvet: Sprinkle const as requested by Ville.]
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915: Fold in intel_mst_port_dp_detect

The indirection here seems to serve no purpose. Probably leftovers
from earlier revisions. Spotted while trying to review some mst
patches.

Cc: Dave Airlie <airlied@gmail.com>
Signed-off-by: Daniel Vetter <daniel.vetter@intel.com>

drm/i915: add runtime PM get/put call in i915_execlists

Otherwise we will get WARNs when we read context status registers and
the machine is suspended.

Testcase: igt/pm_rpm/debugfs-read
Signed-off-by: Michel Thierry <michel.thierry@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915: Fix chv PCS DW11 register defines

I managed to fumble the per spline PCS DW11 register defines in:

commit 570e2a747bc06cd8620662c5125ec2dc964c511b
Author: Ville Syrjälä <ville.syrjala@linux.intel.com>
Date:   Mon Aug 18 14:42:46 2014 +0300

    drm/i915: Clear TX FIFO reset master override bits on chv

Fortunately the bit in DW0 that was cleared due to this didn't have
any effect as long as the bit we meant to clear was already zero.

Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Jani Nikula <jani.nikula@intel.com>
[danvet: Fix commit ref as pointed out by Jani.]
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915: Don't claim that we're resetting PCH ADPA register

intel_crt_reset() resets the ADPA register on all gen5+ platforms.
However the debug message claims it's touching the PCH ADPA register
which is clearly not what it does on VLV. Drop the PCH part from
the debug message.

Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Jani Nikula <jani.nikula@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915: disable IPS while getting the pipe CRCs.

For some yet-undiscovered reason, when IPS gets enabled, the pipe CRC
changes. Since hsw_enable_ips() doesn't really guarantees to enable
IPS (it depends on package C-states), we can't really predict if IPS
is enabled or disabled while running our CRC tests, so let's just
completely disable IPS while pipe CRCs are being used.

If we find a way to make IPS not change the pipe CRC result, we may
want to fix IPS and then revert this patch. While this doesn't happen,
let's merge this patch, so every IGT test relying on the CRCs can
work on pipe A.

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=72864
Testcase: igt/kms_cursor_crc (and others)
Signed-off-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915: call drm_vblank_cleanup() earlier at unload

In its current place, it just segfaults while trying to access the
CRTC structures:

[ 9132.421681] Call Trace:
[ 9132.421707]  [<ffffffffa01130d8>] i915_get_crtc_scanoutpos+0x1e8/0x220 [i915]
[ 9132.421727]  [<ffffffffa001da34>] drm_calc_vbltimestamp_from_scanoutpos+0x94/0x330 [drm]
[ 9132.421744]  [<ffffffffa001d240>] ?vblank_disable_and_save+0x40/0x1e0 [drm]
[ 9132.421769]  [<ffffffffa0114328>] i915_get_vblank_timestamp+0x68/0xb0 [i915]
[ 9132.421786]  [<ffffffffa001d094>] drm_get_last_vbltimestamp+0x44/0x80 [drm]
[ 9132.421801]  [<ffffffffa001d3a6>] vblank_disable_and_save+0x1a6/0x1e0 [drm]
[ 9132.421817]  [<ffffffffa001eac1>] drm_vblank_cleanup+0x61/0xa0 [drm]
[ 9132.421849]  [<ffffffffa0177a5e>] i915_driver_unload+0xde/0x290 [i915]
[ 9132.421867]  [<ffffffffa0020264>] drm_dev_unregister+0x24/0xb0 [drm]
[ 9132.421884]  [<ffffffffa002090e>] drm_put_dev+0x1e/0x70 [drm]
[ 9132.421901]  [<ffffffffa00e01e0>] i915_pci_remove+0x10/0x20 [i915]
[ 9132.421910]  [<ffffffff81347556>] pci_device_remove+0x36/0xb0
[ 9132.421920]  [<ffffffff8140084a>] __device_release_driver+0x7a/0xf0
[ 9132.421928]  [<ffffffff81400fc8>] driver_detach+0xb8/0xc0
[ 9132.421936]  [<ffffffff8140054a>] bus_remove_driver+0x4a/0xb0
[ 9132.421944]  [<ffffffff81401717>] driver_unregister+0x27/0x50
[ 9132.421953]  [<ffffffff81346f65>] pci_unregister_driver+0x25/0x70
[ 9132.421971]  [<ffffffffa00229c8>] drm_pci_exit+0x78/0xa0 [drm]
[ 9132.422000]  [<ffffffffa017a6d2>] i915_exit+0x20/0x94e [i915]
[ 9132.422009]  [<ffffffff810fb9dc>] SyS_delete_module+0x13c/0x1f0
[ 9132.422019]  [<ffffffff8131c5fb>] ?
trace_hardirqs_on_thunk+0x3a/0x3f
[ 9132.422028]  [<ffffffff816f7792>] system_call_fastpath+0x16/0x1b

This means it has to be before intel_modeset_cleanup, which cleans the
CRTC structures. But if we move it to before intel_fbdev_fini(), we
get WARNs because intel_fbdev_fini() still tries to use the vblanks,
so the only acceptable point for drm_vblank_cleanup() seems to be this
place.

Related commit:

    commit cbb47d179fb345c579cd8cd884693903fceed26a
    Author: Chris Wilson <chris@chris-wilson.co.uk>
    Date:   Mon Sep 23 17:33:20 2013 -0300
        drm/i915: Add some missing steps to i915_driver_load error path

Testsuite: igt/drv_module_reload
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=77511
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=83484
Signed-off-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915: spt does not have pch backlight override bit

SPT is always in the PCH override mode, and the bit MBZ. Only set
override on LPT.

v2: check for PCH version (Ville)

Signed-off-by: Jani Nikula <jani.nikula@intel.com>
Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915: preserve swizzle settings if necessary v4

Some machines (like MBAs) might use a tiled framebuffer but not enable
display swizzling at boot time.  We want to preserve that configuration
if possible to prevent a boot time mode set.  On IVB+ it shouldn't
affect performance anyway since the memory controller does internal
swizzling anyway.

For most other configs we'll be able to enable swizzling at boot time,
since the initial framebuffer won't be tiled, thus we won't see any
corruption when we enable it.

v2: preserve swizzling if BIOS had it set (Daniel)
v3: preserve swizzling only if we inherited a tiled framebuffer (Daniel)
    check display swizzle setting in detect_bit_6_swizzle (Daniel)
    use gen6 as cutoff point (Daniel)
v4: fixup swizzle preserve again, had wrong init order (Daniel)

Reported-by: Kristian Høgsberg <hoegsberg@gmail.com>
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915: vlv: fix gunit HW state corruption during S4 suspend

During S4 freeze we don't call intel_suspend_complete(), which would
save the gunit HW state, but during S4 thaw/restore events we call
intel_resume_prepare() which restores it, thus ending up in a corrupted
HW state.

Fix this by calling intel_suspend_complete() from the corresponding
freeze_late event handler.

The issue was introduced in
commit 016970beb05da6285c2f3ed2bee1c676cb75972e
Author: Sagar Kamble <sagar.a.kamble@intel.com>
Date:   Wed Aug 13 23:07:06 2014 +0530

CC: Sagar Kamble <sagar.a.kamble@intel.com>
Signed-off-by: Imre Deak <imre.deak@intel.com>
Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915: Check workaround status on dfs read time

As the workaround list has the value as initialization time
constant, we can do the simple checking on the go without
negleting igt.

Signed-off-by: Mika Kuoppala <mika.kuoppala@intel.com>
Reviewed-by: Arun Siluvery <arun.siluvery@linux.intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915: Build workaround list in ring initialization

If we build the workaround list in ring initialization
and decouple it from the actual writing of values, we
gain the ability to decide where and how we want to apply
the values.

The advantage of this will become more clear when
we need to initialize workarounds on older gens where
it is not possible to write all the registers through ring
LRIs.

v2: rebase on newest bdw workarounds

Cc: Arun Siluvery <arun.siluvery@linux.intel.com>
Cc: Damien Lespiau <damien.lespiau@intel.com>
Signed-off-by: Mika Kuoppala <mika.kuoppala@intel.com>
Reviewed-by: Arun Siluvery <arun.siluvery@linux.intel.com>
[danvet: Resolve tiny conflict in comments and ocd alignments a bit.]
[danvet2: Remove bogus force_wake_get call spotted by Paulo and QA.]
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915: remove redundant #ifdef CONFIG_COMPAT

The whole file is only built with CONFIG_COMPAT=y.

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

drm/i915: remove dead code from legacy suspend handler

The legacy DRM suspend logic (effective in UMS) doesn't handle any S4 thaw
events so we don't need to care about it either. Only S3 suspend and S4
freeze events are handled. Leave an assert behind to be sure.

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

drm/i915: Make intel_pipe_has_type() and some callers take intel_crtc

For consistency, since that's the rule followed for internal functions.

Signed-off-by: Ander Conselvan de Oliveira <ander.conselvan.de.oliveira@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915: Make *_crtc_mode_set() take an intel_crtc insted of drm_crtc

For consistency, since that's the rule followed for internal functions.

Signed-off-by: Ander Conselvan de Oliveira <ander.conselvan.de.oliveira@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915: Make *_find_best_dpll() take an intel_crtc insted of drm_crtc

For consistency, since that's the rule followed for internal functions.

Signed-off-by: Ander Conselvan de Oliveira <ander.conselvan.de.oliveira@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915: Replace some loop through encoders with intel_pipe_has_type()

In the ironlake mode set code, there was two instances of a loop through
encoders to find out if one of them has INTEL_OUTPUT_LVDS type. Simplify
the code by deleting some lines and use intel_pipe_has_type() instead.

Signed-off-by: Ander Conselvan de Oliveira <ander.conselvan.de.oliveira@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915: Document that mmap forwarding is discouraged

Too many new drm driver writers seem to look at i915 for inspiration.
But we have two ways to do mmap, so discourage readers from the old,
ugly version. In a new driver we'd just expose two mmap offsets per
object, one for the gtt map and the other for the cpu map.

v2: Make it clear that i915 does cpu mmaps this way for past
cluelessness^W^W historical reasons. Asked for by Jani.

Cc: "Cheng, Yao" <yao.cheng@intel.com>
Cc: David Herrmann <dh.herrmann@gmail.com>
Reviewed-by: David Herrmann <dh.herrmann@gmail.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915/bdw: Remove BDW preproduction W/As until C stepping.

Let's clean this a bit

v2: Rebase after other Mika's patch that removed some BDW production workarounds.
v3: Removed stepping info.

Reviewed-by: Mika Kuoppala <mika.kuoppala@intel.com>
Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915: Convert a couple more INTEL_INFO-esque macros to be pointer agnostic

Just a couple more macros that assume that they were being passed a
struct drm_device when they want a struct drm_i915_private. Use our
magic macro to ease transitioning over to using drm_i915_privates

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Jani Nikula <jani.nikula@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915/skl: Add 180 degree HW rotation support

Add support for 180 degree rotation for primary and sprite planes

Signed-off-by: Sonika Jindal <sonika.jindal@intel.com>
Reviewed-by: Damien Lespiau <damien.lespiau@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915: Suppress no action noise from oom shrinker

If we are not able to free anything (the shrinker leaves nothing on the
global object lists), do not log anything. This is useful when other
subsystems are being stress-tested for their oom behaviour and i915.ko
is shouting into the logs about doing nothing.

Reported-by: Dave Jones <davej@redhat.com>
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915: Report the current number of bytes freed during oom

The shrinker reports the number of pages freed, but we try to log the
number of bytes - which leads to some nonsense values being reportedly
freed during oom.

Reported-by: Maarten Lankhorst <maarten.lankhorst@canonical.com>
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915: Add missing '\n' to cdclk debug message

Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915: Do not export RC6p and RC6pp if they don't exist

Avoid to expose RC6 and RC6pp to the platforms that doesn't support it.
So powertop can be changed to show RC6p and RC6pp only on the platforms
they are available.

v2: Simplify by merging RC6p and RC6pp groups and respect the spec that
mentions deep and deepest RC6 on SNB and IVB although they keep disabled
by default.

v3: Remove unecessary space.

v4: RC6p and RC6pp is only for SNB and IVB; unify debug msg and use
    has_rc6p() on sanitize options instead of is gen 6 and ivb.

v5: yet another fix on has_rc6p macro. final is_gen6 or is_ivb! To make sure
    we are excluding hsw and baytrail.

References: https://bugs.freedesktop.org/show_bug.cgi?id=84524
Cc: Josh Triplett <josh.triplett@intel.com>
Cc: Paulo Zanoni <paulo.r.zanoni@intel.com>
Cc: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Reviewed-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915: Fix not checking cursor and object sizes

Even if the fb is the same we should still check if the sizes are
valid to be set.

Signed-off-by: Gustavo Padovan <gustavo.padovan@collabora.co.uk>
Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915: move check of intel_crtc_cursor_set_obj() out

Move check inside intel_crtc_cursor_set_obj() to
intel_check_cursor_plane(), we only use it there so move them out to
make the merge of intel_crtc_cursor_set_obj() into
intel_check_cursor_plane() easier.

This is another step toward the atomic modesetting support and unification
of plane operations such pin/unpin of fb objects on i915.

v2: take Ville's comment: move crtc_{w,h} assignment a bit down in the
code
v3: take Ville's comment: kept only the restructuring changes, the rest of
the code was moved to a separated patch since it is a bug fix (we weren't
checking sizes when the fb was the same)

Signed-off-by: Gustavo Padovan <gustavo.padovan@collabora.co.uk>
[danvet: Fixup commit message mixup.]
Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915: remove leftover from pre-universal planes days

Now that universal planes are in place we don't need this plane unref on
failures.

Suggested-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Gustavo Padovan <gustavo.padovan@collabora.co.uk>
Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915: Merge of visible and !visible paths for primary planes

Fold intel_pipe_set_base() in the update primary plane path merging
pieces of code that are common to both paths.

Basically the the pin/unpin procedures are the same for both paths
and some checks can also be shared (some of the were moved to the
check() stage)

v2: take Ville's comments:
- remove unnecessary plane check
- move mutex lock to inside the conditional
- make the pin fail message a debug one
- add a fixme for the fastboot hack
- call intel_frontbuffer_flip() after FBC update

v3: take more Ville's comments:
- fold update code under if (intel_crtc->active), and do the
visible/!visible split inside.
- check ret inside the same conditional we assign it

v4: don't use intel_enable_primary_hw_plane(), the primary_enabled
check inside will break page flips

v5: take more Ville's comments:
- set primary_enabled to true and add BDW hack
- unify if (old_fb) and if (old_fb != fb)

v6: take more Ville's comments:
- make was_primary bool and fix its check
- add the BDW vblank wait comment

Suggested-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Gustavo Padovan <gustavo.padovan@collabora.co.uk>
Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915: Remove IS_ULT()

As stated in the few previous commits, IS_ULT/ULX() is better
per-platform as it has different consequences depending on the platform.

We now can get rid of it.

Signed-off-by: Damien Lespiau <damien.lespiau@intel.com>
Reviewed-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915/skl: Don't check for ULT/ULX when detecting the PCH

IS_ULT() wasn't taking into account SKL so we had a warn with SPT-LP.

We don't realy need those checks here, and as we don't need to introduce
IS_SKL_ULT/ULX() at the moment, let's just drop them.

Signed-off-by: Damien Lespiau <damien.lespiau@intel.com>
Reviewed-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915: Use IS_HSW_ULT() in HSW CDCLK clock read-out

hsw_get_cdclk_freq() is really just HSW, so we can use IS_HSW_ULT()
instead of IS_ULT() there.

Signed-off-by: Damien Lespiau <damien.lespiau@intel.com>
Reviewed-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915: Spell out IS_HSW/BDW_ULT() in intel_crt_present()

The quality of being a ULT or ULX package doesn't tell anything across
generations and so a global IS_ULT() macro doesn't make much sense, esp.
as we're adding new products.

So, spell out which ULT/ULX SKUs we are talking about here, namely HSW
and BDW.

Signed-off-by: Damien Lespiau <damien.lespiau@intel.com>
Reviewed-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915: Use IS_HSW_ULT() in HAS_IPS()

HAS_IPS() has a '|| IS_BROADWELL()', no need to check for IS_BDW_ULT().

Signed-off-by: Damien Lespiau <damien.lespiau@intel.com>
Reviewed-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915: Use IS_HSW_ULT() in a HSW specific code path

No need to add the BDW pci ULT/ULX checks inside a if (IS_HASWELL(dev))
code path.

Signed-off-by: Damien Lespiau <damien.lespiau@intel.com>
Reviewed-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915: kerneldoc for intel_fifo_underrun.c

v2: Fix spelling fail.

Reviewed-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@intel.com>

drm/i915: Filter gmch fifo underruns in the shared handler

This simplifies the code in the vlv irq handler. Also this now
means that we correctly filter underruns on gen2-4.

And as the real upshot I need to document one less function for
the fifo underrun code.

v2: Shorten one long line.

Reviewed-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915: Add wrappers to handle fifo underrun interrupts

Way too much copypasta all over. And this also clarifies a bit what's
going on since it separates the "do we have an underrun irq" from the
"should we report the underrun" check.

v2: Fix excessively long lines.

Reviewed-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@intel.com>

drm/i915: Use dev_priv in public intel_fifo_underrun.c functions

It's the new rule!

Reviewed-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915: Extract intel_fifo_underrun.c

Prep work for some nice documentation. Requires that we export the
display irq enable/disable functions on ilk/ibx. But we already export
them for vlv/i915. So not more inconsistency.

v2: Rebase on top of skl stage 1.

Reviewed-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@intel.com>

Revert "drm/i915: Enable full PPGTT on gen7"

This reverts commit 8c50f10d73b50139dcfe48bc22f2c8c7822c1983.

It's not yet solid and Dave objected to pulling the tree in its
current state.

Cc: Michel Thierry <michel.thierry@intel.com>
Cc: Dave Airlie <airlied@gmail.com>
Cc: Chris Wilson <chris@chris-wilson.co.uk>
References: http://mid.mail-archive.com/CAPM=9ty2r1MLE=wzC-_vNSUzXVqAyXiGgocpSV9qOp0gzpK3xA@mail.gmail.com
References: http://lists.freedesktop.org/archives/intel-gfx/2014-October/053926.html
Acked-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Daniel Vetter <daniel.vetter@intel.com>

Merge branch 'drm-intel-next-fixes' into drm-intel-next

So I've sent the first pull request to Dave and I expect his request
for a merge tree any second now ;-)

More seriously I have some pending patches for 3.19 that depend upon
both trees, hence backmerge. Conflicts are all trivial.

Conflicts:
drivers/gpu/drm/i915/i915_irq.c
drivers/gpu/drm/i915/intel_display.c

v2: Of course I've forgotten the fixup script for the silent conflict.

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

drm/i915: fix short vs. long hpd detection

Fix short vs. long hpd detection for non-g4x and non-pch split
platforms.

Broken since introduction in
commit 13cf550448b58abf8f44f5d6a560f2d20871c965
Author: Dave Airlie <airlied@redhat.com>
Date:   Wed Jun 18 11:29:35 2014 +1000

    drm/i915: rework digital port IRQ handling (v2)

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=83175
Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Jani Nikula <jani.nikula@intel.com>

drm/i915: Don't trust the DP_DETECT bit for eDP ports on CHV

On CHV the display DDC pins may be muxed to an alternate function if
there's no need for DDC on a specific port, which is the case for eDP
ports since there's no way to plug in a DP++ HDMI dongle.

This causes problems when trying to determine if the port is present
since the the DP_DETECTED bit is the latched state of the DDC SDA pin
at boot. If the DDC pins are muxed to an alternate function the bit
may indicate that the port isn't present.

To work around this look at the VBT as well as the DP_DETECTED bit
to determine if we should attempt registering an eDP port. Do this
only for ports B and C since port D doesn't support eDP (no PPS/BLC).

In theory someone could also wire up a normal DP port w/o DDC lines.
That would just mean that simple DP++ HDMI dongles wouldn't work
on such a port. With this change we would still fail to register
such DP ports. But let's hope no one wires their board in such a way,
and if they do we can extend the VBT checks to cover normal DP ports
as well.

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=84265
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Damien Lespiau <damien.lespiau@intel.com>
Signed-off-by: Jani Nikula <jani.nikula@intel.com>

drm/i915: properly reenable gen8 pipe IRQs

We were missing the pipe B/C vblank bits! Take a look at
gen8_de_irq_postinstall for a comparison.

This should fix a bunch of IGT tests.

There are a few more things we could improve on this code, but this
should be the minimal fix to unblock us.

v2: s/extra_iir/extra_ier/ because IIR doesn't make sense (Ville)

Bugzilla:https://bugs.freedesktop.org/show_bug.cgi?id=83640
Testcase: igt/*
Cc: stable@vger.kernel.org
Signed-off-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Jani Nikula <jani.nikula@intel.com>

drm/i915: Move DIV_ROUND_CLOSEST_ULL macro to header

Move the duplicated DIV_ROUND_CLOSEST_ULL macro into the intel_drv.h
header file so that it can be shared between intel_display.c
and intel_panel.c.

Signed-off-by: U. Artie Eoff <ullysses.a.eoff@intel.com>
Reviewed-By: Joe Konno <joe.konno@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915: intel_backlight scale() math WA

Improper truncated integer division in the scale() function causes
actual_brightness != brightness. This (partial) work-around should be
sufficient for a majority of use-cases, but it is by no means a complete
solution.

TODO: Determine how best to scale "user" values to "hw" values, and
vice-versa, when the ranges are of different sizes. That would be a
buggy scenario even with this work-around.

The issue was introduced in the following (v3.17-rc1) commit:

    6dda730 drm/i915: respect the VBT minimum backlight brightness

Note that for easier backporting this commit adds a duplicated macro.
A follow-up cleanup patch rectifies this for 3.18+

v2: (thanks to Chris Wilson) clarify commit message, use rounded division
macro

v3: -DIV_ROUND_CLOSEST() fails to build with CONFIG_X86_32=y. (Jani)
    -Use DIV_ROUND_CLOSEST_ULL() instead. (Damien)
    -v1 and v2 originally authored by Joe Konno.

Signed-off-by: U. Artie Eoff <ullysses.a.eoff@intel.com>
Cc: stable@vger.kernel.org
Reviewed-By: Joe Konno <joe.konno@intel.com>
[danvet: Add backporting note.]
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915: Update DRIVER_DATE to 20141003

Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915: Remove the duplicated logic between the two shrink phases

We can use the same logic to walk the different bound/unbound lists
during shrinker (as the unbound list is a degenerate case of the bound
list), slightly compacting the code.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915: kerneldoc for interrupt enable/disable functions

Just start with the basics for now.

Since there's a lot of different functionality in i915_irq.c I've
decided to split it into different sections and pull in just the
relevant functions. Splitting into different files looks like a lot
more work since the interrupt handlers do an awful lot of reuse all
over.

v2: Rebase onto changed function names.

Signed-off-by: Daniel Vetter <daniel.vetter@intel.com>
Reviewed-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915: Use dev_priv instead of dev in irq setup functions

It's the new world order!

Not going full monty on these here and rolling this out throughout the
subsequent call chains since this is just for the kerneldoc. Later on
we can go more crazy, especially once we've embedded drm_device
correctly.

v2: Also frob the runtime_pm functions ...

Signed-off-by: Daniel Vetter <daniel.vetter@intel.com>
Reviewed-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915: s/pm._irqs_disabled/pm.irqs_enabled/

Double negations just parse harder. Also this allows us to ditch some
init code since clearing to 0 dtrt. Also ditch the assignment in
intel_pm_setup, that's not redundant since we do the assignement now
while setting up interrupts.

While at it do engage in a bit of OCD and wrap up the few lines of
setup/teardown code into little helper functions: intel_irq_fini for
cleanup and intel_irq_init_hw for hw setup.

v2: Use _install/_uninstall for the new wrapper function names as
Paulo suggested.

Signed-off-by: Daniel Vetter <daniel.vetter@intel.com>
Reviewed-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915: Clear TX FIFO reset master override bits on chv

Clear the override bits to make sure the hardware manages
the TX FIFO reset master on its own.

v2: Squash with the earlier attempt at forcing the override bits

Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Mika Kuoppala <mika.kuoppala@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915: Make sure hardware uses the correct swing margin/deemph bits on chv

The register can house two different swing marging/deemph settings at
once. However only one gets used based on some other bits. Make sure we
set those bits correctly to make the hardware use the settings we
provided.

Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Mika Kuoppala <mika.kuoppala@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915: make sink_crc return -EIO on aux read/write failure

Even though it's unliky, we should check each aux transaction not just
the first one. Also

commit ce31d9f4fc05964f6c0dd3a8661dc1a1d843a1e2
Author: Rodrigo Vivi <rodrigo.vivi@intel.com>
Date:   Mon Sep 29 18:29:52 2014 -0400

    drm/i915: preserve other DP_TEST_SINK bits.

added a new aux transaction before the one which was checked. Fix
this.

Cc: Daniel Vetter <daniel.vetter@ffwll.ch>
Cc: Todd Previte <tprevite@gmail.com>
Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915: Constify send buffer for intel_dp_aux_ch

Inspired by Ville constifying the send buffer for pach_aux.

Cc: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@intel.com>

drm/i915: De-magic the PSR AUX message

Use pack_aux() to construct the PSR exit DPMS D0 AUX message,
and use the defines from dp_dp_helper.h to populate the message
contents.

v2: Use sizeof() for message size (Jani)
    Use a generic loop to write EDP_PSR_AUX_DATA registers

Cc: Jani Nikula <jani.nikula@linux.intel.com>
Cc: Rodrigo Vivi <rodrigo.vivi@gmail.com>
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915: Reinstate error level message for non-simulated gpu hangs

This seems to have been accidentally lost in

commit be62acb4cce1389a28296852737e3917d9cc5b25
Author: Mika Kuoppala <mika.kuoppala@linux.intel.com>
Date:   Fri Aug 30 16:19:28 2013 +0300

    drm/i915: ban badly behaving contexts

Without this real gpu hangs only log output at info level, which gets
filtered away by piglit's testrunner.

v2: Tune down to notice level. Note that we need to add drm/i915 so
that at least the automatic igt dmesg filtering still picks it up.

v3: git add and lack of coffee don't mix well.

v4: Message is in between hw and sw reset, so switch verb to
continuous form.

v5: Use i915_stop_rings_allow_warn for consistency. For Chris' case of
injecting lots of hangs I guess we need to revamp this all anyway when
merging. For now this should plug the regression for piglit testing
mesa.

v6: Make it compile (Mika).

Cc: Mika Kuoppala <mika.kuoppala@intel.com>
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Kenneth Graunke <kenneth@whitecape.org>
Reported-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Daniel Vetter <daniel.vetter@intel.com>

drm/i915: Kerneldoc for intel_runtime_pm.c

I've decided not to document the functions exported to the audio
driver since really, they shouldn't exist ...

v2: Improvements from Imre's review plus a few more spelling fixes
I've spotted.

Reviewed-by: Imre Deak <imre.deak@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915: Call runtime_pm_disable directly

Allows us to mark it static and so forgoe the kerneldoc for it.

Note that intel_power_domains_fini is also called from failure paths
in the driver load sequence. But the call to runtime_pm_disable for
that is harmless since by default runtime pm is already disabled.

v2: Augment the commit message as discussed with Imre on irc.

Reviewed-by: Imre Deak <imre.deak@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915: Move intel_display_set_init_power to intel_runtime_pm.c

I've decided to not move intel_display_port_power_domain because
that's just a hack in our design ...

Reviewed-by: Imre Deak <imre.deak@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915: Bikeshed rpm functions name a bit.

- fini goes with init, so call it intel_power_domains_fini. While
  at it shovel some of the fini code that leaked out of it back in.

- give power_enabled functions the verb _is_ to make the meaning clearer.
  Also use a __ prefix instead of _unlocked to really discourage users.

- rename runtime_pm_init/fini to enable/disable since that's what they do.

Reviewed-by: Imre Deak <imre.deak@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915: Extract intel_runtime_pm.c

Geez is the audio hack ugly.

Signed-off-by: Daniel Vetter <daniel.vetter@intel.com>
Reviewed-by: Imre Deak <imre.deak@intel.com>
[danvet: Rebased on top of the skl patches.]
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915: Remove intel_modeset_suspend_hw

Another layer of indirection for just an lpt-only w/a is a bit
excessive. Reduce it.

This was added in

commit 7d708ee40a6b9ca1112a322e554c887df105b025
Author: Imre Deak <imre.deak@intel.com>
Date:   Wed Apr 17 14:04:50 2013 +0300

    drm/i915: HSW: allow PCH clock gating for suspend

Cc: Imre Deak <imre.deak@intel.com>
Cc: Paulo Zanoni <paulo.r.zanoni@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@intel.com>
Reviewed-by: Imre Deak <imre.deak@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915: spelling fixes for frontbuffer tracking kerneldoc

Oh well.

v2: Fix one more spelling fail Paulo spotted.

Reviewed-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915: Tighting frontbuffer tracking around flips

So I think I've spotted a small gap in the frontbuffer tracking
while discussing the logic with Paulo on irc:

1. Userspace schedules gpu rendering to the current frontbuffer.
This gets tracked in dev_priv->fb_tracking.busy_bits.

2. We pageflip a fully rendered buffer before the frontbuffer
rendering completes.

3. The request retiring will never clear busy_bits (since at retire
time the old frontbuffer won't have obj->frontbuffer_bits set), so
these bits now are stuck until someone again does a bit of frontbuffer
tracking.

If we clear stale busy_bits in flip_prepare this gap is closed.

Reviewed-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915: Enable pixel replicated modes on BDW and HSW.

Haswell and later silicon has added a new pixel replication register
to the pipe timings for each transcoder. Now in addition to the
DPLL_A_MD register for the pixel clock double, we also need to write
to the TRANS_MULT_n (0x6002c) register to double the pixel data. Writing
to the DPLL only double the pixel clock.

ver2: Macro name change from MULTIPLY to PIPE_MULTI. (Daniel)
ver3: Do not set pixel multiplier if transcoder is eDP (Ville)
ver4: Macro name change to PIPE_MULT and default else pixel_multiplier

Cc: Ville =?iso-8859-1?Q?Syrj=E4l=E4?= <ville.syrjala@linux.intel.com>
Cc: Daniel Vetter <daniel.vetter@ffwll.ch>
Cc: Jani Nikula <jani.nikula@intel.com>
Signed-off-by: Clint Taylor <clinton.a.taylor@intel.com>
Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
[danvet: Appease checkpatch and move one hunk back into the right
place that git am misplace!?]
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

Merge branch 'topic/skl-stage1' into drm-intel-next-queued

SKL stage 1 patches still need polish so will likely miss the 3.18
merge window. We've decided to postpone to 3.19 so let's pull this in
to make patch merging and conflict handling easier.

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

drm/i915: preserve other DP_TEST_SINK bits.

Sink crc was implemented based on dp 1.1 spec that had all TEST_SINK bits
reserved reading all 0s. But when reviewing my latest changes on sink crc
Todd warned me that on new specs we have other valid bits on this reg that we
might want to preserve.

Cc: Todd Previte <tprevite@gmail.com>
Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Reviewed-by: Todd Previte <tprevite@gmail.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915/bdw: WaDisableFenceDestinationToSLM

This WA affect BDW GT3 pre-production steppings.

Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Reviewed-by: Mika Kuoppala <mika.kuoppala@intel.com>
[danvet: Don't mention steppings ...]
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915: Add IS_BDW_GT3 macro.

It will be usefull to specify w/a that affects only BDW GT3.

Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Reviewed-by: Jesse Barnes <jbarnes@virtuousgeek.org>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915: Fix Sink CRC

In some cases like when PSR just got enabled the panel need more vblank
times to calculate CRC. I figured that out with the new PSR test cases
facing some cases that I had a green screen but a blank CRC. Even with
2 vblank waits on kernel + 2 vblank waits on test case.

So let's give up to 6 vblank wait time. However we now check for
TEST_CRC_COUNT that shows when panel finished to calculate CRC and
has it ready.

v2: Jani pointed out attempts decrements was wrong and should never reach
the error condition. And Daniel pointed out that EIO is more appropriated than
EGAIN. Also I realized that I have to read test_crc_count after setting
test_sink

v3: Rebase and adding error message

Cc: Todd Previte <tprevite@gmail.com>
Cc: Daniel Vetter <daniel.vetter@ffwll.ch>
Cc: Jani Nikula <jani.nikula@intel.com>
Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Reviewed-by: Todd Previte <tprevite@gmail.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915: Don't spam dmesg with rps messages on vlv/chv

If the GPU frequency isn't going to change don't spam dmesg with
debug messages about it.

Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Jani Nikula <jani.nikula@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@intel.com>

drm/i915: Do not leak pages when freeing userptr objects

sg_alloc_table_from_pages() can build us a table with coalesced ranges which
means we need to iterate over pages and not sg table entries when releasing
page references.

Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Cc: "Barbalho, Rafael" <rafael.barbalho@intel.com>
Tested-by: Rafael Barbalho <rafael.barbalho@intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: stable@vger.kernel.org
[danvet: Remove unused local variable sg.]
Signed-off-by: Daniel Vetter <daniel.vetter@intel.com>

drm/i915: Do not store the error pointer for a failed userptr registration

If we fail to create our mmu notification, we report the error back and
currently store the error inside the i915_mm_struct. This not only causes
subsequent registerations of the same mm to fail (an issue if the first
was interrupted by a signal and needed to be restarted) but also causes
us to eventually try and free the error pointer.

[   73.419599] BUG: unable to handle kernel NULL pointer dereference at 000000000000004c
[   73.419831] IP: [<ffffffff8114af33>] mmu_notifier_unregister+0x23/0x130
[   73.420065] PGD 8650c067 PUD 870bb067 PMD 0
[   73.420319] Oops: 0000 [#1] SMP DEBUG_PAGEALLOC
[   73.420580] CPU: 0 PID: 42 Comm: kworker/0:1 Tainted: G        W      3.17.0-rc6+ #1561
[   73.420837] Hardware name: Intel Corporation SandyBridge Platform/LosLunas CRB, BIOS ASNBCPT1.86C.0075.P00.1106281639 06/28/2011
[   73.421405] Workqueue: events __i915_mm_struct_free__worker
[   73.421724] task: ffff880088a81220 ti: ffff880088168000 task.ti: ffff880088168000
[   73.422051] RIP: 0010:[<ffffffff8114af33>]  [<ffffffff8114af33>] mmu_notifier_unregister+0x23/0x130
[   73.422410] RSP: 0018:ffff88008816bd50  EFLAGS: 00010286
[   73.422765] RAX: 0000000000000003 RBX: ffff880086485400 RCX: 0000000000000000
[   73.423137] RDX: ffff88016d80ee90 RSI: ffff880086485400 RDI: 0000000000000044
[   73.423513] RBP: ffff88008816bd70 R08: 0000000000000001 R09: 0000000000000000
[   73.423895] R10: 0000000000000320 R11: 0000000000000001 R12: 0000000000000044
[   73.424282] R13: ffff880166e5f008 R14: ffff88016d815200 R15: ffff880166e5f040
[   73.424682] FS:  0000000000000000(0000) GS:ffff88016d800000(0000) knlGS:0000000000000000
[   73.425099] CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
[   73.425537] CR2: 000000000000004c CR3: 0000000087f5f000 CR4: 00000000000407f0
[   73.426157] Stack:
[   73.426597]  ffff880088a81248 ffff880166e5f038 fffffffffffffffc ffff880166e5f008
[   73.427096]  ffff88008816bd98 ffffffff814a75f2 ffff880166e5f038 ffff8800880f8a28
[   73.427603]  ffff88016d812ac0 ffff88008816be00 ffffffff8106321a ffffffff810631af
[   73.428119] Call Trace:
[   73.428606]  [<ffffffff814a75f2>] __i915_mm_struct_free__worker+0x42/0x80
[   73.429116]  [<ffffffff8106321a>] process_one_work+0x1ba/0x610
[   73.429632]  [<ffffffff810631af>] ? process_one_work+0x14f/0x610
[   73.430153]  [<ffffffff810636db>] worker_thread+0x6b/0x4a0
[   73.430671]  [<ffffffff8108d67d>] ? trace_hardirqs_on+0xd/0x10
[   73.431501]  [<ffffffff81063670>] ? process_one_work+0x610/0x610
[   73.432030]  [<ffffffff8106a206>] kthread+0xf6/0x110
[   73.432561]  [<ffffffff8106a110>] ? __kthread_parkme+0x80/0x80
[   73.433100]  [<ffffffff8169c22c>] ret_from_fork+0x7c/0xb0
[   73.433644]  [<ffffffff8106a110>] ? __kthread_parkme+0x80/0x80
[   73.434194] Code: 0f 1f 84 00 00 00 00 00 66 66 66 66 90 8b 46 4c 85 c0 0f 8e 10 01 00 00 55 48 89 e5 41 55 41 54 53 48 89 f3 49 89 fc 48 83 ec 08 <48> 83 7f 08 00 0f 84 b1 00 00 00 48 c7 c7 40 e6 ac 82 e8 26 65
[   73.435942] RIP  [<ffffffff8114af33>] mmu_notifier_unregister+0x23/0x130
[   73.437017]  RSP <ffff88008816bd50>
[   73.437704] CR2: 000000000000004c

Fixes regression from commit ad46cb533d586fdb256855437af876617c6cf609
Author: Chris Wilson <chris@chris-wilson.co.uk>
Date:   Thu Aug 7 14:20:40 2014 +0100

    drm/i915: Prevent recursive deadlock on releasing a busy userptr

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=84207
Testcase: igt/gem_render_copy_redux
Testcase: igt/gem_userptr_blits/create-destroy-sync
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Jacek Danecki <jacek.danecki@intel.com>
Cc: "Gong, Zhipeng" <zhipeng.gong@intel.com>
Cc: Jacek Danecki <jacek.danecki@intel.com>
Cc: "Ursulin, Tvrtko" <tvrtko.ursulin@intel.com>
Cc: stable@vger.kernel.org
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@intel.com>

Revert "drm/i915/bdw: BDW Software Turbo"

This reverts commit c76bb61a71083b2d90504cc6d0dda2047c5d63ca.

It's apparently too broken so that Rodrigo submitted a patch to add a
config option for it. Given that the design is also ... suboptimal and
that I've only merged this to get lead engineers and managers off my
back for one second let's just revert this.

/me puts on combat gear again

It was worth a shot ...

References: http://mid.mail-archive.com/1411686380-1953-1-git-send-email-rodrigo.vivi@intel.com
Cc: Jesse Barnes <jbarnes@virtuousgeek.org>
Cc: Daisy Sun <daisy.sun@intel.com>
Cc: Jesse Barnes <jbarnes@virtuousgeek.org>
Cc: Rodrigo Vivi <rodrigo.vivi@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@intel.com>

drm/i915: Broadwell DDI Buffer translation - more tuning

BDW display - DP buffer translation values changed to give better margin.

Further change to entry 6; set dword 0 bit 31=1.

Both changes were approved already but this one didn't landed BSpec yet
this is why it is in a separated patch. Making reviewer's life easier.
Also alowing separated tests and any future bisect that might be needed.

Reference: Predator r74080 / HSD 4394389

v2: Arthur noticed I was changing the wrong bit.

Cc: Arthur Runyan <arthur.j.runyan@intel.com>
Cc: Paulo Zanoni <paulo.r.zanoni@intel.com>
Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Reviewed-by: Arthur Runyan <arthur.j.runyan@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915: Broadwell DDI Buffer translation changed to give better margin.

Reference: Predator r73977 / HSD 4394389

Cc: Arthur Runyan <arthur.j.runyan@intel.com>
Cc: Paulo Zanoni <paulo.r.zanoni@intel.com>
Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Reviewed-by: Arthur Runyan <arthur.j.runyan@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915: Make sure PSR is ready for been re-enabled.

Let's make sure PSR is propperly disabled before to re-enabled it.

According to Spec, after disabled PSR CTL, the Idle state might occur
up to 24ms, that is one full frame time (1/refresh rate),
plus SRD exit training time (max of 6ms),
plus SRD aux channel handshake (max of 1.5ms).

So if something went wrong PSR will be disabled until next full
enable/disable setup.

v2: The 24ms above takes in account 16ms for refresh rate on 60Hz mode. However
on low frequency modes this can take longer. So let's use 50ms for safeness.

v3: Move wait out of psr.lock critical area.

Cc: Daniel Vetter <daniel.vetter@ffwll.ch>
Cc: Paulo Zanoni <paulo.r.zanoni@intel.com>
Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Reviewed-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915: Minimize the huge amount of unecessary fbc sw cache clean.

The sw cache clean on BDW is a tempoorary workaround because we cannot
set cache clean on blt ring with risk of hungs. So we are doing the cache clean on sw.
However we are doing much more than needed. Not only when using blt ring.
So, with this extra w/a we minimize the ammount of cache cleans and call it only
on same cases that it was being called on gen7.

The traditional FBC Cache clean happens over LRI on BLT ring when there is a
frontbuffer touch happening. frontbuffer tracking set fbc_dirty variable
to let BLT flush that it must clean FBC cache.

fbc.need_sw_cache_clean works in the opposite information direction
of ring->fbc_dirty telling software on frontbuffer tracking to perform
the cache clean on sw side.

v2: Clean it a little bit and fully check for Broadwell instead of gen8.

v3: Rebase after frontbuffer organization.

v4: Wiggle confused me. So fixing v3!

Cc: Daniel Vetter <daniel.vetter@ffwll.ch>
Cc: Paulo Zanoni <paulo.r.zanoni@intel.com>
Reviewed-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915: Avoid re-configure panel on every PSR re-enable.

The panel has to be reconfigured only when it really loose the power.
The traditional enable/disable sequence already take care of this so we can
minimize the time spend on every re-enable.

Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Reviewed-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915/skl: Move gen9 pm initialization into its own branch

Gen9 is different enough (for instance, fetching the memory latency
values is different from ILK+) to not take the HAS_PCH_SPLIT() branch,
so let's prefer a clean separation.

v2: Rebase on top of the broadwell_init_clock_gating() name change

Reviewed-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Signed-off-by: Damien Lespiau <damien.lespiau@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915/skl: Introduce intel_num_planes()

It can be handy to get the number of planes for this pipe, ie including
the primary plane to loop over them. Introduce a little function to do
so.

Reviewed-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Signed-off-by: Damien Lespiau <damien.lespiau@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915/skl: Introduce a I915_MAX_PLANES macro

This can be useful to declare structures around pipes and planes and
don't have to go back auditing the code if the next platorm bump that
number.

Reviewed-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Signed-off-by: Damien Lespiau <damien.lespiau@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915/skl: Adjust assert_sprites_disabled()

Let's put to good use the new PLANE_CTL macros.

Reviewed-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Signed-off-by: Damien Lespiau <damien.lespiau@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915/skl: Implement drm_plane vfuncs

SKL Uses the same hardware for all planes now, so called "universal"
planes. Ie both the primary planes and sprite planes share the same
logic. This patch implements the drm_plane vfuncs for "sprites" ie
planes that aren't the primary plane.

v2: Couple of fixes:
  - Actually enabled the planes and fix the plane number

Reviewed-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Signed-off-by: Damien Lespiau <damien.lespiau@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915/skl: Skylake has 2 "sprite" planes per pipe

On Skylake, we use plane1 as primary plane and plane2/3 as sprite
planes.

v2: Rebase on top of the for_each_pipe() change adding dev_priv as first
    argument.

Reviewed-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Signed-off-by: Damien Lespiau <damien.lespiau@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915/skl: Implement WaDisableDgMirrorFixInHalfSliceChicken5:skl

Reviewed-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Signed-off-by: Damien Lespiau <damien.lespiau@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915/skl: Implement Wa4x4STCOptimizationDisable:skl

Reviewed-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Signed-off-by: Damien Lespiau <damien.lespiau@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915/skl: Implement WaDisableSDEUnitClockGating:skl

Reviewed-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Signed-off-by: Damien Lespiau <damien.lespiau@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>