drm/i915/rpm: Enable runtime pm autosuspend by default

Let's enable runtime pm autosuspend by default everywhere.
So, we can allow D3hot and bigger power savings on idle scenarios.

But at this time let's not touch the autosuspend_delay time,
what caused some regression on our previous attempt.

Also, the latest identified issue on GuC PM has been fixed by
commit 1a52faed3131 ("drm/i915/guc: Take GT PM ref when deregistering
context")

v1: Enable runtime pm autosuspend by default for Gen12
and later versions.

v2: Enable runtime pm autosuspend by default for all
platforms(Syrjala Ville)

v3: Change commit message(Nikula Jani)

Signed-off-by: Tilak Tangudu <tilak.tangudu@intel.com>
Reviewed-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20211116155238.3226516-1-tilak.tangudu@intel.com

drm/i915/dg2: Implement WM0 cursor WA for DG2

Bug in the register unit which results in WM1 register
used when only WM0 is enabled on cursor.
Software workaround is when only WM0 enabled on cursor,
copy contents of CUR_WM_0[30:0] (exclude the enable bit)
into CUR_WM_1[30:0].

v2:  - s/dev_priv/i915/ (Ville Syrjälä)
     - Removed unneeded brackets (Ville Syrjälä)

HSDES: 14012656716

Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Stanislav Lisovskiy <stanislav.lisovskiy@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20211118093907.18510-1-stanislav.lisovskiy@intel.com

drm/i915/pxp: fix includes for headers in include/drm

Use <> not "" for including headers from include/drm.

Signed-off-by: Jani Nikula <jani.nikula@intel.com>
Reviewed-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20211116135813.19806-1-jani.nikula@intel.com

drm/i915: drop intel_display.h include from intel_display_power.h

Use forward declarations instead.

Cc: Imre Deak <imre.deak@intel.com>
Signed-off-by: Jani Nikula <jani.nikula@intel.com>
Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Reviewed-by: Imre Deak <imre.deak@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20211116171434.20516-5-jani.nikula@intel.com

drm/i915: move structs from intel_display_power.h to .c

Anything internal to the implementation should be hidden away. Move the
intel_display_power structs to the .c file.

Cc: Imre Deak <imre.deak@intel.com>
Signed-off-by: Jani Nikula <jani.nikula@intel.com>
Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Reviewed-by: Imre Deak <imre.deak@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20211116171434.20516-4-jani.nikula@intel.com

drm/i915/debugfs: move debug printing to intel_display_power.c

The debugfs should have no special privileges to look into the
implementation guts. Move the actual debug printing of power domains to
intel_display_power.c.

Cc: Imre Deak <imre.deak@intel.com>
Signed-off-by: Jani Nikula <jani.nikula@intel.com>
Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Reviewed-by: Imre Deak <imre.deak@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20211116171434.20516-3-jani.nikula@intel.com

drm/i915: drop intel_display.h include from intel_dpll_mgr.h

Use forward declarations instead.

Signed-off-by: Jani Nikula <jani.nikula@intel.com>
Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Reviewed-by: Imre Deak <imre.deak@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20211116171434.20516-2-jani.nikula@intel.com

drm/i915: drop intel_display.h include from intel_ddi.h

Use forward declarations instead.

Signed-off-by: Jani Nikula <jani.nikula@intel.com>
Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Reviewed-by: Imre Deak <imre.deak@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20211116171434.20516-1-jani.nikula@intel.com

drm/i915: Clean up CRC register defines

Use REG_BIT() & co. for the CRC registers.

Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20211112193813.8224-8-ville.syrjala@linux.intel.com
Reviewed-by: Rodrigo Vivi <rodrigo.vivi@intel.com>

drm/i915: Clean up DPINVGTT/VLV_DPFLIPSTAT bits

Use REG_BIT() & co. for DPINVTT/VLV_DPFLIPSTAT bits.

Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20211112193813.8224-9-ville.syrjala@linux.intel.com
Reviewed-by: Rodrigo Vivi <rodrigo.vivi@intel.com>

drm/i915: Clean up FPGA_DBG/CLAIM_ER bits

Use REG_BIT() & co. for FPGA_DBG/CLAIM_ER bits.

Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20211112193813.8224-10-ville.syrjala@linux.intel.com
Reviewed-by: Rodrigo Vivi <rodrigo.vivi@intel.com>

drm/i915/: Extend VRR platform support to Gen 11

VRR is supported on Gen 11 HW , hence extend the support
in the driver to enable this for Gen 11.

Cc: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Manasi Navare <manasi.d.navare@intel.com>
Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20211116231209.28621-1-manasi.d.navare@intel.com

drm/i915: Disable DSB usage for now

Turns out the DSB has trouble correctly loading the gamma LUT.
From a cursory look maybe like some entries do not load
properly, or they get loaded with some gibberish. Unfortunately
our current kms_color/etc. tests do not seem to catch this.

I had a brief look at the generated DSB batch and it looked
correct. Tried a few quick tricks like writing the index
register twice/etc. but didn't see any improvement.
Also tried switching to the 10bit gamma mode in case
there is yet another issue with the multi-segment mode, but
even the 10bit mode was showing issues.

Switching to mmio fixes all of it. I suppose one theory is that
maybe the DSB bangs on the LUT too quickly and it can't keep up
and instead some data either gets dropped or corrupted. To confirm
that someone should try to slow down the DSB's progress a bit.
Another thought was that maybe the LUT has crappy dual porting
and you get contention if you try to load it during active
scanout. But why then would the mmio path work, unless it's
just sufficiently slow?

Whatever the case, this is currently busted so let's disable
it until we get to the root of the problem.

Closes: https://gitlab.freedesktop.org/drm/intel/-/issues/3916
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20211014181856.17581-2-ville.syrjala@linux.intel.com
Reviewed-by: Uma Shankar <uma.shankar@intel.com>

drm/i915: Declare .(de)gamma_lut_tests for icl+

All interpolated gamma modes including the icl+ multi segment
mode require non-decreasing entries for the interpolation to
work correctly. For some reason we're forgetting to declare
that for icl+. Let us do so.

References: https://gitlab.freedesktop.org/drm/intel/-/issues/3916
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20211014181856.17581-1-ville.syrjala@linux.intel.com
Reviewed-by: Uma Shankar <uma.shankar@intel.com>

drm/i915: Fix framestart_delay commens in VRR code

Since I originally wrote these comments we decided to change our
definition of framestart_delay from 0-3 to 1-4. Adjust the comments
to match that new convention. The actual code was adjusted already.

Cc: Manasi Navare <manasi.d.navare@intel.com>
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20211117183103.27418-3-ville.syrjala@linux.intel.com
Reviewed-by: Manasi Navare <manasi.d.navare@intel.com>

drm/i915: Do vblank evasion correctly if vrr push has already been sent

Let's adjust the vblank evasion to account for the case where
a push has already been sent. In that case the vblank exit will start
at vmin vblank start (as opposed to vmax vblank start when no push
has been sent).

This should minimize the effects of the tiny race between sampling
the frame counter vs. intel_vrr_send_push() during the previous frame.
This will also be required if we want to do mailbox style updates with
vrr since then we'd definitely do multiple commits per frame. Currently
mailbox updates are only used by the legacy cursor, but we don't do
vrr push for those.

Cc: Manasi Navare <manasi.d.navare@intel.com>
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20211117183103.27418-2-ville.syrjala@linux.intel.com
Reviewed-by: Manasi Navare <manasi.d.navare@intel.com>

drm/i915: Move vrr push after the frame counter sampling again

Moving the vrr push to happen before sampling the frame counter
was wrong. If we are already in vblank when the push is sent
the vblank exit will start immediately which causes the sampled
frame counter to correspond to the next frame instead of the current
frame.

So put things back into the original order (except we should
keep the vrr push within the irq disable section to avoid
pointless irq related delays here).

We'll just have to accept the tiny race that exists between
sampling the frame counter vs. vrr push. And let's at least
document said race properly in a comment.

I suppose we could try to minimize the race by sampling the frame
counter just before sending the push, but that would require
changing drm_crtc_arm_vblank_event() to accept a caller provided
vblank counter value, so leave it be for now. Another thing we
could do is change the vblank evasion to account for the case
where a push was already sent. That would anyway be required
for mailbox style updates. Currently mailbox updates are only
used by the legacy cursor, but we don't do a vrr push for those.

Cc: Manasi Navare <manasi.d.navare@intel.com>
Fixes: 6f9976bd1310 ("drm/i915: Do vrr push before sampling the frame counter")
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20211117183103.27418-1-ville.syrjala@linux.intel.com
Reviewed-by: Manasi Navare <manasi.d.navare@intel.com>

drm/i915/vlv_dsi: Double pixelclock on read-back for dual-link panels

In intel_dsi_get_config() double the pclk returned by foo_dsi_get_pclk()
for dual-link panels. This fixes the following WARN triggering:

 i915 0000:00:02.0: [drm] *ERROR* [CRTC:51:pipe A] mismatch in pixel_rate (expected 235710, found 118056)
 i915 0000:00:02.0: [drm] *ERROR* [CRTC:51:pipe A] mismatch in hw.pipe_mode.crtc_clock (expected 235710, found 118056)
 i915 0000:00:02.0: [drm] *ERROR* [CRTC:51:pipe A] mismatch in hw.adjusted_mode.crtc_clock (expected 235710, found 118056)
 i915 0000:00:02.0: [drm] *ERROR* [CRTC:51:pipe A] mismatch in port_clock (expected 235710, found 118056)
 ------------[ cut here ]------------
 pipe state doesn't match!
 WARNING: CPU: 3 PID: 136 at drivers/gpu/drm/i915/display/intel_display.c:9125 intel_display_finish_reset+0x1bd3/0x2050 [i915]
 ...

This has been tested on a Xiaomi Mi Pad 2 (with CHT x5-Z8500 SoC) tablet,
with a 1536x2048 dual-link DSI panel.

Note this fix was taken from icl_dsi.c which does the same in
its get_config().

Cc: Tsuchiya Yuto <kitakar@gmail.com>
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Acked-by: Jani Nikula <jani.nikula@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20211024155020.126328-1-hdegoede@redhat.com

drm/i915: Fix fastsets on TypeC ports following a non-blocking modeset

After a non-blocking modeset on a TypeC port's CRTC - possibly blocked
later in drm_atomic_helper_wait_for_dependencies() - a fastset on the
same CRTC may copy the state of CRTC before this gets updated to reflect
the up-to-date DP-alt vs. TBT-alt TypeC mode DPLL used for the CRTC. In
this case after the first (non-blocking) commit completes enabling the
DPLL required for the up-to-date TypeC mode the following fastset will
update the CRTC state pointing to the wrong DPLL. A subsequent disabling
modeset will try to disable the wrong PLL, triggering a state checker
WARN (and leaving the DPLL which is actually used active for good).

Fix the above race by copying the DPLL state for fastset CRTCs from the
old CRTC state at the point where it's guaranteed to be up-to-date
already. This could be handled in the encoder's update_prepare() hook as
well, but that's a bigger change, which is better done as a follow-up.

v2: Copy dpll_hw_state as well. (Ville)

Testcase: igt/kms_busy/extended-modeset-hang-newfb-with-reset
Closes: https://gitlab.freedesktop.org/drm/intel/-/issues/4308
Cc: Ville Syrjala <ville.syrjala@linux.intel.com>
Signed-off-by: Imre Deak <imre.deak@intel.com>
Reviewed-by: Mika Kahola <mika.kahola@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20211115181121.156197-1-imre.deak@intel.com

drm/i915/guc: fix NULL vs IS_ERR() checking

The intel_engine_create_virtual() function does not return NULL.  It
returns error pointers.

Fixes: e5e32171a2cf ("drm/i915/guc: Connect UAPI to GuC multi-lrc interface")
Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Reviewed-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20211116114916.GB11936@kili

drm/i915/driver: add i915_driver_ prefix to functions

Add the i915_driver_ prefix to the switcheroo functions in
i915_driver.[ch].

Signed-off-by: Jani Nikula <jani.nikula@intel.com>
Acked-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Link: https://patchwork.freedesktop.org/patch/msgid/20211111101304.13094-3-jani.nikula@intel.com

drm/i915/driver: rename driver to i915_drm_driver

As a name, "driver" is too generic and short to be easily located in a
file this size. Rename it to i915_drm_driver.

Signed-off-by: Jani Nikula <jani.nikula@intel.com>
Acked-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Link: https://patchwork.freedesktop.org/patch/msgid/20211111101304.13094-2-jani.nikula@intel.com

drm/i915/driver: rename i915_drv.c to i915_driver.c

This is more about trimming i915_drv.h than the renamed
i915_driver.[ch]. Split out i915_driver.[ch] out of i915_drv.h as a
feasible thing to do.

Signed-off-by: Jani Nikula <jani.nikula@intel.com>
Acked-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Link: https://patchwork.freedesktop.org/patch/msgid/20211111101304.13094-1-jani.nikula@intel.com

drm/i915: Disable D3Cold in s2idle and runtime pm

s2idle and runtime pm puts the pci gfx device in D3Hot, ACPI runtime
monitors the pci tree,if it sees complete tree as D3Hot,it transitions
the device to D3Cold.But i915 do not have D3Cold support in S2idle or in
runtime pm. so disabling D3cold in above flows and its FIXME.

Added pci D3Cold enable/disable in s2idle and runtime suspend/resume
flows.

Signed-off-by: Tilak Tangudu <tilak.tangudu@intel.com>
Reviewed-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20211115154054.3220476-1-tilak.tangudu@intel.com

drm/i915/fbc: fix the FBC kernel-doc warnings

Fix the recently introduced 'make htmldocs' warnings:

$ make htmldocs 2>&1 > /dev/null | grep i915
./drivers/gpu/drm/i915/display/intel_fbc.c:635: warning: Excess function parameter 'i915' description in 'intel_fbc_is_active'
./drivers/gpu/drm/i915/display/intel_fbc.c:1638: warning: Excess function parameter 'i915' description in 'intel_fbc_handle_fifo_underrun_irq'
./drivers/gpu/drm/i915/display/intel_fbc.c:635: warning: Function parameter or member 'fbc' not described in 'intel_fbc_is_active'
./drivers/gpu/drm/i915/display/intel_fbc.c:635: warning: Excess function parameter 'i915' description in 'intel_fbc_is_active'
./drivers/gpu/drm/i915/display/intel_fbc.c:1638: warning: Function parameter or member 'fbc' not described in 'intel_fbc_handle_fifo_underrun_irq'
./drivers/gpu/drm/i915/display/intel_fbc.c:1638: warning: Excess function parameter 'i915' description in 'intel_fbc_handle_fifo_underrun_irq'

Fixes: e49a656b924e ("drm/i915/fbc: Start passing around intel_fbc")
Cc: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Jani Nikula <jani.nikula@intel.com>
Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20211115140549.27629-1-jani.nikula@intel.com

agp/intel-gtt: reduce intel-gtt dependencies more

Don't include stuff on behalf of users if they're not strictly necessary
for the header.

Cc: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Signed-off-by: Jani Nikula <jani.nikula@intel.com>
Acked-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Acked-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Reviewed-by: Lucas De Marchi <lucas.demarchi@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/7bcaa1684587b9b008d3c41468fb40e63c54fbc7.1636977089.git.jani.nikula@intel.com

agp/intel-gtt: Replace kernel.h with the necessary inclusions

When kernel.h is used in the headers it adds a lot into dependency hell,
especially when there are circular dependencies are involved.

Replace kernel.h inclusion with the list of what is really being used.

Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Signed-off-by: Jani Nikula <jani.nikula@intel.com>
Acked-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Reviewed-by: Lucas De Marchi <lucas.demarchi@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/4d6a976459547407979f4b4c05a52785523e6bd8.1636977089.git.jani.nikula@intel.com

drm/i915: include intel-gtt.h only where needed

Only intel_gt.c and intel_ggtt.c need the interface.

Cc: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Signed-off-by: Jani Nikula <jani.nikula@intel.com>
Acked-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Acked-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Reviewed-by: Lucas De Marchi <lucas.demarchi@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/034f57db24d6936ac2e4e6830261d791240cdd79.1636977089.git.jani.nikula@intel.com

drm/i915/dsi/xelpd: Disable DC states in Video mode

MIPI DSI transcoder cannot be in video mode to support any of the
display C states.

Bspec: 49195 (For DC*co DSI transcoders cannot be in video mode)
Bspec: 49193 (Hardware does not support DC5 or DC6 with MIPI DSI enabled)
Bspec: 49188 (desc of DSI_DCSTATE_CTL talks about cmd mode PM control

v2: Align to the power domain ordering (Jani)
    Add bspec references (Imre)

Signed-off-by: Vandita Kulkarni <vandita.kulkarni@intel.com>
Reviewed-by: Imre Deak <imre.deak@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20211019151435.20477-4-vandita.kulkarni@intel.com

drm/i915/dsi/xelpd: Add DSI transcoder support

Update ADL_P device info to support DSI0, DSI1

v2: Re-define cpu_transcoder_mask only (Jani)

Signed-off-by: Vandita Kulkarni <vandita.kulkarni@intel.com>
Reviewed-by: Jani Nikula <jani.nikula@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20211019151435.20477-3-vandita.kulkarni@intel.com

drm/i915/dsi/xelpd: Fix the bit mask for wakeup GB

v2: Fix the typo, move out the hardcoding from
    macro(Jani, Ville)

Fixes: f87c46c43175 ("drm/i915/dsi/xelpd: Add WA to program LP to HS wakeup guardband")
Signed-off-by: Vandita Kulkarni <vandita.kulkarni@intel.com>
Reviewed-by: Jani Nikula <jani.nikula@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20211019151435.20477-2-vandita.kulkarni@intel.com

drm/i915: make array states static const

Don't populate the read-only array states on the stack but instead it
static. Also makes the object code smaller.

Signed-off-by: Colin Ian King <colin.king@canonical.com>
Signed-off-by: Jani Nikula <jani.nikula@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20210915112702.12783-1-colin.king@canonical.com

drm/i915/psr: Fix PSR2 handling of multiplanar format

When a plane with a multiplanar format is added to the state by
drm_atomic_add_affected_planes(), only the UV plane is
added, so a intel_atomic_get_new_plane_state() call to get the Y
plane state can return a null pointer.
To fix this, intel_atomic_get_plane_state() should be called and
the return needs to be checked for errors, as it could return a EAGAIN
as other atomic state could be holding the lock for the Y plane.

Other issue with the patch being fixed is that the Y plane is not
being committed to hardware because the corresponded plane bit is not
set in update_planes when UV and Y planes are added to the state by
drm_atomic_add_affected_planes().

Cc: Jouni Högander <jouni.hogander@intel.com>
Cc: Mika Kahola <mika.kahola@intel.com>
Fixes: 3809991ff5f4 ("drm/i915/display: Add initial selective fetch support for biplanar formats")
Signed-off-by: José Roberto de Souza <jose.souza@intel.com>
Reviewed-by: Jouni Högander <jouni.hogander@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20211108213807.39865-1-jose.souza@intel.com

drm/1915/fbc: Replace plane->has_fbc with a pointer to the fbc instance

With multiple fbc instances we need to find the right one for each
plane. Rather than going looking for the right instance every time
let's just replace the has_fbc boolean with a pointer that gets us
there straight away.

Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20211104144520.22605-18-ville.syrjala@linux.intel.com
Acked-by: Jani Nikula <jani.nikula@intel.com>
Reviewed-by: Mika Kahola <mika.kahola@intel.com>

drm/i915/fbc: Start passing around intel_fbc

In preparation for multiple FBC instances start passing around
intel_fbc pointers rather than i915 pointers. And once there are
multiple of these we can't rely on container_of() to get back to
the i915, so we toss in a fbc->i915 pointer already.

Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20211104144520.22605-17-ville.syrjala@linux.intel.com
Acked-by: Jani Nikula <jani.nikula@intel.com>
Reviewed-by: Mika Kahola <mika.kahola@intel.com>

drm/i915/fbc: s/dev_priv/i915/

Rename 'dev_priv' to 'i915' to match modern style.

Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20211104144520.22605-16-ville.syrjala@linux.intel.com
Acked-by: Jani Nikula <jani.nikula@intel.com>
Reviewed-by: Mika Kahola <mika.kahola@intel.com>

drm/i915: Relocate FBC_LLC_READ_CTRL

In the case of FBC_LLC_READ_CTRL the "FBC" stands for
frame buffer _caching_, not frame buffer compression.
Move the register definition out from the middle of the
frame buffer compression register definitions. Let's
just stick it somewhere with similar looking register
offsets.

And while at it switch it over to REG_BIT().

Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20211104144520.22605-15-ville.syrjala@linux.intel.com
Acked-by: Jani Nikula <jani.nikula@intel.com>
Reviewed-by: Mika Kahola <mika.kahola@intel.com>

drm/i915/fbc: Finish polishing FBC1 registers

Switch all the FBC1 registers over to REG_BTT()/etc.
And while at it add a few more registers/bits that
escaped the net previously.

Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20211104144520.22605-14-ville.syrjala@linux.intel.com
Acked-by: Jani Nikula <jani.nikula@intel.com>
Reviewed-by: Mika Kahola <mika.kahola@intel.com>

drm/i915/fbc: Clean up all register defines

The FBC register defines are a mess:
- namespace changes between DPFC_, FBC_, and some platform
  specific prefix at a whim
- ilk+ reuses most g4x bits but still has some separate bit
  defines elsewhere
- it's not clear from the defines that the bit defines are
  shared

So let's clean it up:
- both g4x and ilk register share the same defines now
- only defines which conflict have a _PLATFORM suffix, everyone
  else just gets comments to indicate which platforms do what
- namespace is consistent DPFC_ now
- SNB system agent fence registers also get a consistent namespace
- REG_BIT() & co. for everything

Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20211104144520.22605-13-ville.syrjala@linux.intel.com
Acked-by: Jani Nikula <jani.nikula@intel.com>
Reviewed-by: Mika Kahola <mika.kahola@intel.com>

drm/i915/fbc: Nuke BDW_FBC_COMP_SEG_MASK

Just use a same mask for ivb/hsw as for bdw+. The extra bit
in the bdw mask is mbz on ivb/hsw anyway so this is just
pointless complexity.

Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20211104144520.22605-12-ville.syrjala@linux.intel.com
Acked-by: Jani Nikula <jani.nikula@intel.com>
Reviewed-by: Mika Kahola <mika.kahola@intel.com>

drm/i915/fbc: Introduce intel_fbc_set_false_color()

Pull the direct FBC register frobbing out from the debugfs code
into the fbc code. Also add a vfunc for this so we don't need
extra platforms checks.

Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20211104144520.22605-11-ville.syrjala@linux.intel.com
Acked-by: Jani Nikula <jani.nikula@intel.com>
Reviewed-by: Mika Kahola <mika.kahola@intel.com>

drm/i915/fbc: Introduce .program_cfb() vfunc

Eliminate the last if-ladder by pulling the CFB/LLB programming
into a vfunc as well.

Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20211104144520.22605-10-ville.syrjala@linux.intel.com
Acked-by: Jani Nikula <jani.nikula@intel.com>
Reviewed-by: Mika Kahola <mika.kahola@intel.com>

drm/i915/fbc: s/gen7/ivb/

"gen7" in display code is not really sensible. We shall call
these things "ivb".

Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20211104144520.22605-9-ville.syrjala@linux.intel.com
Acked-by: Jani Nikula <jani.nikula@intel.com>
Reviewed-by: Mika Kahola <mika.kahola@intel.com>

drm/i915/fbc: Introduce .nuke() vfunc

Eliminate yet another if-ladder by adding .nuke() vfunc.

We also rename all *_recompress() stuff to *_nuke() since
that's the terminology the spec uses. Also "recompress"
is a bit confusing by perhaps implying that this triggers
an immediate recompression. Depending on the hardware that
may definitely not be the case, and in general we don't
specifically know when the hardware decides to compress.
So all we do is "nuke" the current compressed framebuffer
and leave it up to the hardware to recompress later if it
so chooses.

Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20211104144520.22605-8-ville.syrjala@linux.intel.com
Acked-by: Jani Nikula <jani.nikula@intel.com>
Reviewed-by: Mika Kahola <mika.kahola@intel.com>

drm/i915/fbc: Introduce intel_fbc_funcs

Replace the "if-ladders everywhere" approach with vfuncs.

Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20211104144520.22605-7-ville.syrjala@linux.intel.com
Acked-by: Jani Nikula <jani.nikula@intel.com>
Reviewed-by: Mika Kahola <mika.kahola@intel.com>

drm/i915/fbc: Extract helpers to compute FBC control register values

Declutter the *_fbc_activate() functions by pulling all the
control register value computations into helpers.

I left the enable bit in *_fbc_activate() in the hopes of maybe
using the helpers in the *_fbc_deactivate() paths as well instead
of the current rmw approach. That won't be possible at least
quite yet since we clobber the fbc->params before deactivating
FBC so we could end up changing some of the values live, which
given FBC's lack of/poor double buffering would likely not go
so well.

Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20211104144520.22605-6-ville.syrjala@linux.intel.com
Acked-by: Jani Nikula <jani.nikula@intel.com>
Reviewed-by: Mika Kahola <mika.kahola@intel.com>

drm/i915/fbc: Introduce intel_fbc_is_compressing()

Move the direct FBC status register reads from the debugfs code
behind an abstract api.

Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20211104144520.22605-5-ville.syrjala@linux.intel.com
Acked-by: Jani Nikula <jani.nikula@intel.com>
Reviewed-by: Mika Kahola <mika.kahola@intel.com>

drm/i915/fbc: Just use params->fence_y_offset always

No need to tiptoe around programming DPFC_FENCE_YOFF with
params->fence_y_offset vs. 0. If the fence is not enabled
it doesn't even matter what we program here.

Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20211104144520.22605-4-ville.syrjala@linux.intel.com
Acked-by: Jani Nikula <jani.nikula@intel.com>
Reviewed-by: Mika Kahola <mika.kahola@intel.com>

drm/i915/fbc: Extract {skl,glk}_fbc_program_cfb_stride()

Declutter gen7_fbc_activate() by sucking the override
stride programming stuff into helpers.

Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20211104144520.22605-3-ville.syrjala@linux.intel.com
Acked-by: Jani Nikula <jani.nikula@intel.com>
Reviewed-by: Mika Kahola <mika.kahola@intel.com>

drm/i915/fbc: Extract snb_fbc_program_fence()

We have two identical copies of the snb+ system agent
CPU fence programming code. Extract into a helper.

Also there's no real point in insisting that we
program 0 into DPFC_CPU_FENCE_OFFSET when the fence is
disabled. So just always stick the computed Y offset there
whether or not the fence is actually used or not.

Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20211104144520.22605-2-ville.syrjala@linux.intel.com
Acked-by: Jani Nikula <jani.nikula@intel.com>
Reviewed-by: Mika Kahola <mika.kahola@intel.com>

drm/i915/dsi: transmit brightness command in HS state

In Video Mode, if DSI transcoder is set to transmit packets
in LP Escape mode, screen flickering would be obseved when
brightness commands are continuously and quickly transmitted
to a panel.

The problem may be resolved by changing the mode to transmit
packets from Low Power to HS.

Cc: Ville Syrjala <ville.syrjala@linux.intel.com>
Cc: Jani Nikula <jani.nikula@linux.intel.com>
Cc: Vandita Kulkarni <vandita.kulkarni@intel.com>
Cc: Lee Shawn C <shawn.c.lee@intel.com>
Cc: Cooper Chiou <cooper.chiou@intel.com>
Signed-off-by: William Tseng <william.tseng@intel.com>
Signed-off-by: Jani Nikula <jani.nikula@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20211110010217.26759-1-william.tseng@intel.com

drm/i915/dp: For PCON TMDS mode set only the relavant bits in config DPCD

Currently we reset the whole PCON linkConfig DPCD to set the TMDS mode.
This also resets the Source control bit and HDMI link enable bit and
goes to autonomous mode of operation, which is seen to spoil the PCONs
internal state.

This patch avoids resetting the PCON link config register and sets only
the source control bit, with FRL Enable bit set to 0 (TMDS mode) in the
configuration DPCD. It then enables the HDMI Link Enable bit.

v2: Removed the redundant resetting of the bits as the buffer is already
initialized to 0. (Uma)
Updated comments and commit message.

v3: Rebase

Signed-off-by: Ankit Nautiyal <ankit.k.nautiyal@intel.com>
Reviewed-by: Uma Shankar <uma.shankar@intel.com>
Signed-off-by: Uma Shankar <uma.shankar@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20211110072947.171659-3-ankit.k.nautiyal@intel.com

drm/i915/dp: Optimize the FRL configuration for HDMI2.1 PCON

Currently the HDMI2.1 PCON's frl link config DPCD registers are
reset and configured even if they are already configured.
Also the HDMI Link Mode does not settle to FRL MODE immediately after
HDMI Link Status is active.

This patch:
-Checks if the PCON is already configured for FRL.
-Include HDMI Link Mode in wait for loop along with HDMI Link status DPCD.

v2: Rebase

Signed-off-by: Ankit Nautiyal <ankit.k.nautiyal@intel.com>
Reviewed-by: Uma Shankar <uma.shankar@intel.com>
Signed-off-by: Uma Shankar <uma.shankar@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20211110072947.171659-2-ankit.k.nautiyal@intel.com

Revert "drm/i915/tgl/dsi: Gate the ddi clocks after pll mapping"

This reverts commit 991d9557b0c4 ("drm/i915/tgl/dsi: Gate the ddi clocks
after pll mapping"). The Bspec was updated recently with the pll ungate
sequence similar to that of icl dsi enable sequence. Hence reverting.

Bspec: 49187
Fixes: 991d9557b0c4 ("drm/i915/tgl/dsi: Gate the ddi clocks after pll mapping")
Cc: <stable@vger.kernel.org> # v5.4+
Signed-off-by: Vandita Kulkarni <vandita.kulkarni@intel.com>
Signed-off-by: Jani Nikula <jani.nikula@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20211109120428.15211-1-vandita.kulkarni@intel.com

drm/i915: pin: delete duplicate check in intel_pin_and_fence_fb_obj()

The "ret" variable is checked on the previous line so we know it's
zero.  No need to check again.

Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Reviewed-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20211109114850.GB16587@kili

drm/i915: Call intel_update_active_dpll() for both bigjoiner pipes

Currently we're only calling intel_update_active_dpll() for the
bigjoiner master pipe but not for the slave. With TC ports this
leads to the two pipes end up trying to use different PLLs
(TC vs. TBT). What's worse we're enabling the PLL that didn't get
intel_update_active_dpll() called on it at the spot where we
need the clocks turned on. So we turn on the wrong PLL and the
DDI is now trying to source its clock from the other PLL which is
still disabled. Naturally that doesn't end so well and the DDI
fails to start up.

The state checker also gets a bit unhappy (which is a good thing)
when it notices that one of the pipes was using the wrong PLL.

Let's fix this by remembering to call intel_update_active_dpll()
for both pipes. That should get the correct PLL turned on when
we need it, and the state checker should also be happy.

Cc: Imre Deak <imre.deak@intel.com>
Cc: Manasi Navare <manasi.d.navare@intel.com>
Closes: https://gitlab.freedesktop.org/drm/intel/-/issues/4434
Fixes: e12d6218fda2 ("drm/i915: Reduce bigjoiner special casing")
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20211105212156.5697-1-ville.syrjala@linux.intel.com
Reviewed-by: Imre Deak <imre.deak@intel.com>

drm/i915: Use unlocked register accesses for LUT loads

We have to bash in a lot of registers to load the higher
precision LUT modes. The locking overhead is significant, especially
as we have to get this done as quickly as possible during vblank.
So let's switch to unlocked accesses for these. Fortunately the LUT
registers are mostly spread around such that two pipes do not have
any registers on the same cacheline. So as long as commits on the
same pipe are serialized (which they are) we should get away with
this without angering the hardware.

The only exceptions are the PREC_PIPEGCMAX registers on ilk/snb which
we don't use atm as they are only used in the 12bit gamma mode. If/when
we add support for that we may need to remember to still serialize
those registers, though I'm not sure ilk/snb are actually affected
by the same cacheline issue. I think ivb/hsw at least were, but they
use a different set of registers for the precision LUT.

I have a test case which is updating the LUTs on two pipes from a
single atomic commit. Running that in a loop for a minute I get the
following worst case with the locks in place:
 intel_crtc_vblank_work_start: pipe B, frame=10037, scanline=1081
 intel_crtc_vblank_work_start: pipe A, frame=12274, scanline=769
 intel_crtc_vblank_work_end: pipe A, frame=12274, scanline=58
 intel_crtc_vblank_work_end: pipe B, frame=10037, scanline=74

And here's the worst case with the locks removed:
 intel_crtc_vblank_work_start: pipe B, frame=5869, scanline=1081
 intel_crtc_vblank_work_start: pipe A, frame=7616, scanline=769
 intel_crtc_vblank_work_end: pipe B, frame=5869, scanline=1096
 intel_crtc_vblank_work_end: pipe A, frame=7616, scanline=777

The test was done on a snb using the 10bit 1024 entry LUT mode.
The vtotals for the two displays are 793 and 1125. So we can
see that with the locks ripped out the LUT updates are pretty
nicely confined within the vblank, whereas with the locks in
place we're routinely blasting past the vblank end which causes
visual artifacts near the top of the screen.

Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20211020223339.669-5-ville.syrjala@linux.intel.com
Reviewed-by: Uma Shankar <uma.shankar@intel.com>

drm/i915: Use vblank workers for gamma updates

The pipe gamma registers are single buffered so they should only
be updated during the vblank to avoid screen tearing. In fact they
really should only be updated between start of vblank and frame
start because that is the only time the pipe is guaranteed to be
empty. Already at frame start the pipe begins to fill up with
data for the next frame.

Unfortunately frame start happens ~1 scanline after the start
of vblank which in practice doesn't always leave us enough time to
finish the gamma update in time (gamma LUTs can be several KiB of
data we have to bash into the registers). However we must try our
best and so we'll add a vblank work for each pipe from where we
can do the gamma update. Additionally we could consider pushing
frame start forward to the max of ~4 scanlines after start of
vblank. But not sure that's exactly a validated configuration.
As it stands the ~100 first pixels tend to make it through with
the old gamma values.

Even though the vblank worker is running on a high prority thread
we still have to contend with C-states. If the CPU happens be in
a deep C-state when the vblank interrupt arrives even the irq
handler gets delayed massively (I've observed dozens of scanlines
worth of latency). To avoid that problem we'll use the qos mechanism
to keep the CPU awake while the vblank work is scheduled.

With all this hooked up we can finally enjoy near atomic gamma
updates. It even works across several pipes from the same atomic
commit which previously was a total fail because we did the
gamma updates for each pipe serially after waiting for all
pipes to have latched the double buffered registers.

In the future the DSB should take over this responsibility
which will hopefully avoid some of these issues.

Kudos to Lyude for finishing the actual vblank workers.
Works like the proverbial train toilet.

v2: Add missing intel_atomic_state fwd declaration
v3: Clean up properly when not scheduling the worker
v4: Clean up the rest and add tracepoints
v5: s/intel_wait_for_vblank_works/intel_wait_for_vblank_workers/ (Jani,Uma)

CC: Lyude Paul <lyude@redhat.com>
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20211020223339.669-4-ville.syrjala@linux.intel.com
Reviewed-by: Uma Shankar <uma.shankar@intel.com>

drm/i915: Do vrr push before sampling the frame counter

Do the vrr push before we sample the frame counter to
know when the commit has been latched. Doing these in the
wrong order could lead us to complete the flip before it
has actually happened.

Cc: Manasi Navare <manasi.d.navare@intel.com>
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20211020223339.669-3-ville.syrjala@linux.intel.com
Reviewed-by: Uma Shankar <uma.shankar@intel.com>

drm/i915/dsi: disable lpdt if it is not enabled

Avoid setting LP_DATA_TRANSFER when enable_lpdt is false

Cc: Ville Syrjala <ville.syrjala@linux.intel.com>
Cc: Jani Nikula <jani.nikula@linux.intel.com>
Cc: Vandita Kulkarni <vandita.kulkarni@intel.com>
Cc: Lee Shawn C <shawn.c.lee@intel.com>
Cc: Cooper Chiou <cooper.chiou@intel.com>
Signed-off-by: William Tseng <william.tseng@intel.com>
Signed-off-by: Jani Nikula <jani.nikula@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20211109034125.11291-1-william.tseng@intel.com

drm/i915: Fix Memory BW formulae for ADL-P

The earlier update to BW formulae broke ADL-P. Include
display 13 to use TGL path for BW parameters.

Fixes: c64a9a7c05be ("drm/i915: Update memory bandwidth formulae")
Cc: Matt Roper <matthew.d.roper@intel.com>
Reported-by: José Roberto de Souza <jose.souza@intel.com>
Signed-off-by: Radhakrishna Sripada <radhakrishna.sripada@intel.com>
Reviewed-by: Caz Yokoyama <caz.yokoyama@intel.com>
Reviewed-by: José Roberto de Souza <jose.souza@intel.com>
Signed-off-by: José Roberto de Souza <jose.souza@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20211106003714.17894-1-radhakrishna.sripada@intel.com

drm/i915/display/adlp: Disable underrun recovery

It was also defeatured for ADL-P and other platforms.

BSpec: 55424
Signed-off-by: José Roberto de Souza <jose.souza@intel.com>
Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20211104010858.43559-1-jose.souza@intel.com

drm/i915/audio: rename intel_init_audio_hooks to intel_audio_hooks_init

Follow the filename based prefix naming.

Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Jani Nikula <jani.nikula@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20211104161858.21786-6-jani.nikula@intel.com

drm/i915/audio: move intel_audio_funcs internal to intel_audio.c

It's all internal to intel_audio.c.

Cc: Dave Airlie <airlied@redhat.com>
Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Lucas De Marchi <lucas.demarchi@intel.com>
Signed-off-by: Jani Nikula <jani.nikula@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20211104161858.21786-4-jani.nikula@intel.com

drm/i915/audio: define the audio struct separately from drm_i915_private

Add a standalone definition of struct intel_audio_private, and note that
all of it is private to intel_audio.c.

v2: Rebase

Cc: Dave Airlie <airlied@redhat.com>
Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Lucas De Marchi <lucas.demarchi@intel.com>
Signed-off-by: Jani Nikula <jani.nikula@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20211104161858.21786-3-jani.nikula@intel.com

drm/i915/audio: name the audio sub-struct in drm_i915_private

Add name to the audio sub-struct in drm_i915_private, and remove the
tautologies and other inconsistencies in the member names.

v2: Call the mutex member mutex, not lock. (Ville)

Cc: Dave Airlie <airlied@redhat.com>
Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Lucas De Marchi <lucas.demarchi@intel.com>
Signed-off-by: Jani Nikula <jani.nikula@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20211104161858.21786-2-jani.nikula@intel.com

drm/i915/audio: group audio under anonymous struct in drm_i915_private

With an anonymous struct, this can be pure hierarchical organization
without code changes. We'll follow up with adding a name to the
sub-struct separately.

Cc: Dave Airlie <airlied@redhat.com>
Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Lucas De Marchi <lucas.demarchi@intel.com>
Signed-off-by: Jani Nikula <jani.nikula@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20211104161858.21786-1-jani.nikula@intel.com

drm/i915: Update memory bandwidth formulae

The formulae has been updated to include more variables. Make
sure the code carries the same.

Bspec: 64631, 54023

v2: Make GEN11 follow the default route and fix calculation of
    maxdebw(RK)
v3: Fix div by zero on default case
    Correct indent for fallthrough(Jani)
v4: Fix div by zero on gen11.
v5: Fix 0 max_numchannels case
v6:
    - Split gen11/gen12 algorithms
    - Fix RKL deburst value
    - Fix difference b/ween ICL and TGL algorithms
    - Protect deinterleave from being 0
    - Warn when numchannels exceeds max_numchannels
    - Fix scaling of clk_max from different units
    - s/deinterleave/channelwidth/ in calculating peakbw
    - Fix off by one for num_planes TGL+
    - Fix SAGV check
v7: Fix div by zero error on gen11
v8: Even though the algorithm for gen11 says that we need to return
    derated bw for a qgv point whose planes are less than no of active
    planes, we return 0 for deratedbw when only one plane is allowed.
    We modify the algorithm to accommodate the case where no of active
    planes are same as the min no of planes supported by a qgv point.
v9: Fix dclk scaling for dg1

Cc: Ville Syrjälä <ville.syrjala@linux.intel.com>
Suggested-by: Matt Roper <matthew.d.roper@intel.com>
Signed-off-by: Radhakrishna Sripada <radhakrishna.sripada@intel.com>
Reviewed-by: Anusha Srivatsa <anusha.srivatsa@intel.com>
Signed-off-by: Matt Roper <matthew.d.roper@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20211015210041.16858-1-radhakrishna.sripada@intel.com

drm/i915: Split vlv/chv sprite plane update into noarm+arm pair

Chop vlv_sprite_update() into two halves. Fist half becomes
the _noarm() variant, second part the _arm() variant.

Fortunately I have already previously grouped the register
writes into roughtly the correct order, so the split looks
surprisingly clean.

Looks like most of the hardware logic was copied from the
pre-ctg sprite C, so SPSTRIDE/POS/SIZE are armed by SPSURF,
while the rest are self arming. SPCONSTALPHA is the one
entirely new register that didn't exist in the old sprite C,
and looks like that one is self arming. The CHV pipe B CSC
is also self arming, like the rest of the CHV pipe B
additions.

I didn't have time to capture i915_update_info numbers for
these, but since all the other platforms generally showed
improvements, and crucially no regression, I am fairly
confident this should behave similarly.

Cc: Stanislav Lisovskiy <stanislav.lisovskiy@intel.com>
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20211018115030.3547-10-ville.syrjala@linux.intel.com
Reviewed-by: Stanislav Lisovskiy <stanislav.lisovskiy@intel.com>

drm/i915: Split ivb+ sprite plane update into noarm+arm pair

Chop ivb_sprite_update() into two halves. Fist half becomes
the _noarm() variant, second part the _arm() variant.

Fortunately I have already previously grouped the register
writes into roughtly the correct order, so the split looks
surprisingly clean.

Didn't bother with i915_update_info numbers for this one.
I expect the results to be pretty much identical to the snb
numbers from the corresponding g4x+ sprite modification.

Cc: Stanislav Lisovskiy <stanislav.lisovskiy@intel.com>
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20211018115030.3547-9-ville.syrjala@linux.intel.com
Reviewed-by: Stanislav Lisovskiy <stanislav.lisovskiy@intel.com>

drm/i915: Split g4x+ sprite plane update into noarm+arm pair

Chop g4x_sprite_update() into two halves. Fist half becomes
the _noarm() variant, second part the _arm() variant.

Fortunately I have already previously grouped the register
writes into roughtly the correct order, so the split looks
surprisingly clean.

Not much of a change in i915_update_info on these older
platforms that don't have so many planes or registers to
begin with. Here are the numbers from snb (totally unpatched
vs. both primary plane and sprite patched applied) running
kms_atomic_transition --r plane-all-transition --extended:
w/o patch                           w/ patch
Updates: 5404     Updates: 5405
       |         |
   1us |******        1us |******
       |*********         |*********
   4us |***********        4us |***********
       |**********         |**********
  16us |**       16us |**
       |         |
  66us |       66us |
       |         |
 262us |      262us |
       |         |
   1ms |        1ms |
       |         |
   4ms |        4ms |
       |         |
  17ms |       17ms |
       |         |
Min update: 1400ns     Min update: 1307ns
Max update: 19809ns     Max update: 20194ns
Average update: 6957ns     Average update: 6432ns
Overruns > 100us: 0     Overruns > 100us: 0

But there seems to be a slight improvement with
lockdep enabled:
w/o patch                           w/ patch
Updates: 17612     Updates: 16364
       |         |
   1us |        1us |
       |******         |******
   4us |**********        4us |**********
       |************         |*************
  16us |*************       16us |************
       |***         |*
  66us |       66us |
       |         |
 262us |      262us |
       |         |
   1ms |        1ms |
       |         |
   4ms |        4ms |
       |         |
  17ms |       17ms |
       |         |
Min update: 3141ns     Min update: 3562ns
Max update: 126450ns     Max update: 73354ns
Average update: 16373ns     Average update: 15153ns
Overruns > 250us: 0     Overruns > 250us: 0

Cc: Stanislav Lisovskiy <stanislav.lisovskiy@intel.com>
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20211018115030.3547-8-ville.syrjala@linux.intel.com
Reviewed-by: Stanislav Lisovskiy <stanislav.lisovskiy@intel.com>

drm/i915: Split pre-skl primary plane update into noarm+arm pair

Chop i9xx_plane_update() into two halves. Fist half becomes
the _noarm() variant, second part the _arm() variant.

Fortunately I have already previously grouped the register
writes into roughtly the correct order, so the split looks
surprisingly clean.

One slightly surprising fact was that the CHV pipe B PRIMPOS/SIZE
registers are self arming unlike their pre-ctg DSPPOS/SIZE
counterparts. In fact all the new CHV pipe B registers are
self arming.

Also we must remind ourselves that i830/i845 are a bit borked
in that all of their plane registers are self-arming.

I didn't do any i915_update_info measurements for this one
alone. I'll get total numbers with the corrsponding sprite
plane changes.

v2: Don't break my precious i830/i845

Cc: Stanislav Lisovskiy <stanislav.lisovskiy@intel.com>
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20211020212757.13517-1-ville.syrjala@linux.intel.com
Reviewed-by: Stanislav Lisovskiy <stanislav.lisovskiy@intel.com>

drm/i915: Split skl+ plane update into noarm+arm pair

Chop skl_program_plane() into two halves. Fist half becomes
the _noarm() variant, second part the _arm() variant.

Fortunately I have already previously grouped the register
writes into roughtly the correct order, so the split looks
surprisingly clean.

A few notable oddities I did not realize were self arming
are AUX_DIST and COLOR_CTL.

i915_update_info doesn't look too terrible on my cfl running
kms_atomic_transition --r plane-all-transition --extended:
w/o patch                           w/ patch
Updates: 2178                       Updates: 2018
       |                                   |
   1us |                               1us |
       |                                   |
   4us |                               4us |*****
       |*********                          |**********
  16us |**********                    16us |*******
       |***                                |
  66us |                              66us |
       |                                   |
 262us |                             262us |
       |                                   |
   1ms |                               1ms |
       |                                   |
   4ms |                               4ms |
       |                                   |
  17ms |                              17ms |
       |                                   |
Min update: 8332ns                  Min update: 6164ns
Max update: 48758ns                 Max update: 31808ns
Average update: 19959ns             Average update: 13159ns
Overruns > 100us: 0                 Overruns > 100us: 0

And with lockdep enabled:
w/o patch                           w/ patch
Updates: 2177     Updates: 2172
       |         |
   1us |        1us |
       |         |
   4us |        4us |
       |*******         |*********
  16us |**********       16us |**********
       |*******         |*
  66us |       66us |
       |         |
 262us |      262us |
       |         |
   1ms |        1ms |
       |         |
   4ms |        4ms |
       |         |
  17ms |       17ms |
       |         |
Min update: 12645ns     Min update: 9980ns
Max update: 50153ns     Max update: 33533ns
Average update: 25337ns     Average update: 18245ns
Overruns > 250us: 0     Overruns > 250us: 0

TODO: On icl+ everything seems to be armed by PLANE_SURF, so we
      can optimize this even further on modern platforms. But I
      think there's a bit of refactoring to be done first to
      figure out the best way to go about it (eg. just reusing
      the current skl+ functions, or doing a lower level split).

TODO: Split scaler programming as well, but IIRC the scaler
      has some oddball double buffering behaviour on some
      platforms, so needs proper reverse engineering

Cc: Stanislav Lisovskiy <stanislav.lisovskiy@intel.com>
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20211018115030.3547-6-ville.syrjala@linux.intel.com
Reviewed-by: Stanislav Lisovskiy <stanislav.lisovskiy@intel.com>

drm/i915: Split update_plane() into update_noarm() + update_arm()

The amount of plane registers we have to write has been steadily
increasing, putting more pressure on the vblank evasion mechanism
and forcing us to increase its time budget. Let's try to take some
of the pressure off by splitting plane updates into two parts:
1) write all non-self arming plane registers, ie. the registers
   where the write actually does nothing until a separate arming
   register is also written which will cause the hardware to latch
   the new register values at the next start of vblank
2) write all self arming plane registers, ie. registers which always
   just latch at the next start of vblank, and registers which also
   arm other registers to do so

Here we just provide the mechanism, but don't actually implement
the split on any platform yet. so everything stays now in the _arm()
hooks. Subsequently we can move a whole bunch of stuff into the
_noarm() part, especially in more modern platforms where the number
of registers we have to write is also the greatest. On older
platforms this is less beneficial probably, but no real reason
to deviate from a common behaviour.

And let's sprinkle some TODOs around the areas that will need
adapting.

Cc: Stanislav Lisovskiy <stanislav.lisovskiy@intel.com>
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20211018115030.3547-5-ville.syrjala@linux.intel.com
Reviewed-by: Stanislav Lisovskiy <stanislav.lisovskiy@intel.com>

drm/i915: Fix up the sprite namespacing

Give all sprite exclusive functions/etc. a proper namespace.

Cc: Stanislav Lisovskiy <stanislav.lisovskiy@intel.com>
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20211018115030.3547-4-ville.syrjala@linux.intel.com
Reviewed-by: Stanislav Lisovskiy <stanislav.lisovskiy@intel.com>

drm/i915: Fix async flip with decryption and/or DPT

We're currently forgetting to set the PLANE_SURF_DECRYPT
flag in the async flip path. So if the hardware were to
latch that bit despite this being an async flip we'd start
scanning out garbage. And if it doesn't latch it then I
guess we'd just end up with a weird register value that
doesn't actually match the hardware state, which isn't
great for anyone staring at register dumps.

Similarly the async flip path also forgets to call
skl_surf_address() which means the DPT address space to
GGTT address space downshift is not being applied to
the offset. Which means we are pointing PLANE_SURF
at some random location in GGTT instead of the correct
DPT page.

So let's fix two birds with one stone and extract the
PLANE_SURF calculation from skl_program_plane() into
a small helper and use it in the async flip path as well.

Cc: Anshuman Gupta <anshuman.gupta@intel.com>
Cc: Daniele Ceraolo Spurio <daniele.ceraolospurio@intel.com>
Cc: Juston Li <juston.li@intel.com>
Cc: Rodrigo Vivi <rodrigo.vivi@intel.com>
Cc: Uma Shankar <uma.shankar@intel.com>
Cc: Karthik B S <karthik.b.s@intel.com>
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20211018115030.3547-3-ville.syrjala@linux.intel.com
Reviewed-by: Stanislav Lisovskiy <stanislav.lisovskiy@intel.com>

drm/i915: Reject planar formats when doing async flips

Async flips are only capable of changing PLANE_SURF, hence we
they can't easily be used with planar formats.

Older platforms could require updating AUX_DIST as well, which
is not possible. We'd have to make sure AUX_DIST doesn't change
before allowing the async flip through. If we could get async
flips with CCS then that might be interesting, but since the hw
doesn't allow async flips with CCS I don't see much point in
allowing this for planar formats either. No one renders their
game content in YUV anyway.

icl+ could in theory do this I suppose since each color plane
has its own PLANE_SURF register, but I don't know if there is
some magic to guarantee that both the Y and UV plane would
async flip synchronously if you will. Ie. beyond just a clean
tear we'd potentially get some kind of weird tear with some
random mix of luma and chroma from the old and new frames.

So let's just say no to async flips when scanning out planar
formats.

Cc: Karthik B S <karthik.b.s@intel.com>
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20211018115030.3547-2-ville.syrjala@linux.intel.com
Reviewed-by: Stanislav Lisovskiy <stanislav.lisovskiy@intel.com>

drm/i915: Use intel_de_rmw() for icl combo phy programming

Streamline the code by using intel_de_rmw().

Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20211006204937.30774-16-ville.syrjala@linux.intel.com
Reviewed-by: José Roberto de Souza <jose.souza@intel.com>

drm/i915: Use intel_de_rmw() for icl mg phy programming

Streamline the code by using intel_de_rmw().

Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20211006204937.30774-15-ville.syrjala@linux.intel.com
Reviewed-by: José Roberto de Souza <jose.souza@intel.com>

drm/i915: Use intel_de_rmw() for tgl dkl phy programming

Streamline the code by using intel_de_rmw().

Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20211006204937.30774-14-ville.syrjala@linux.intel.com
Reviewed-by: José Roberto de Souza <jose.souza@intel.com>

drm/i915: Enable per-lane drive settings for icl+

Now that the link buf_trans, link training, and the
combo/mg/dkl/snps phy programming are all fixed up we can
allow per-lane DP drive settings on icl+. Make it so.

Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20211006204937.30774-13-ville.syrjala@linux.intel.com
Reviewed-by: José Roberto de Souza <jose.souza@intel.com>

drm/i915: Query the vswing levels per-lane for snps phy

Prepare for per-lane drive settings by querying the desired vswing
level per-lane.

Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20211006204937.30774-12-ville.syrjala@linux.intel.com
Reviewed-by: José Roberto de Souza <jose.souza@intel.com>

drm/i915: Query the vswing levels per-lane for tgl dkl phy

Prepare for per-lane drive settings by querying the desired vswing
level per-lane.

Note that the code only does two loops, with each one writing the
levels for two TX lanes. The register offsets also look a bit funny
because each time through the loop we write to the exact same
register offsets. The crucial bit is the HIP_INDEX_REG
write that steers the same mmio window into different places.

Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20211006204937.30774-11-ville.syrjala@linux.intel.com
Reviewed-by: José Roberto de Souza <jose.souza@intel.com>

drm/i915: Query the vswing levels per-lane for icl mg phy

Prepare for per-lane drive settings by querying the desired vswing
level per-lane.

Note that the code only does two loops, with each one writing the
levels for two TX lanes.

Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20211006204937.30774-10-ville.syrjala@linux.intel.com
Reviewed-by: José Roberto de Souza <jose.souza@intel.com>

drm/i915: Query the vswing levels per-lane for icl combo phy

Prepare for per-lane drive settings by querying the desired vswing
level per-lane.

Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20211006204937.30774-9-ville.syrjala@linux.intel.com
Reviewed-by: José Roberto de Souza <jose.souza@intel.com>

drm/i915: Stop using group access when progrmming icl combo phy TX

Program each TX lane individually so that we can start to use per-lane
drive settings.

Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20211006204937.30774-8-ville.syrjala@linux.intel.com
Reviewed-by: José Roberto de Souza <jose.souza@intel.com>

drm/i915: Fix comment about modeset parameters

The comment mentions that the KMS is enabled by default unless either the
i915.modeset module parameter or vga_text_mode_force boot option are used.

But the latter does not exist and instead the nomodeset option was meant.

Signed-off-by: Javier Martinez Canillas <javierm@redhat.com>
Signed-off-by: Jani Nikula <jani.nikula@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20211103122809.1040754-2-javierm@redhat.com

drm/i915: Restore memory mapping for DPT FBs across system suspend/resume

At least during hibernation the DPT mappings are lost with all stolen
memory content, so suspend/resume these mappings similarly to GGTT
mappings.

This fixes a problem where the restoring modeset during system resume fails
with pipe faults if a tiled framebuffer was active before suspend.

v2: Clarify the way restore works in intel_dpt_resume()'s Docbook entry.
    (Chris)

Cc: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Ville Syrjälä <ville.syrjala@linux.intel.com>
Cc: Vunny Sodhi <vunny.sodhi@intel.com>
Reported-and-tested-by: Vunny Sodhi <vunny.sodhi@intel.com>
Signed-off-by: Imre Deak <imre.deak@intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20211101183551.3580546-2-imre.deak@intel.com

drm/i915: Factor out i915_ggtt_suspend_vm/i915_ggtt_resume_vm()

Factor out functions that are needed by the next patch to suspend/resume
the memory mappings for DPT FBs.

No functional change, except reordering during suspend the
ggtt->invalidate(ggtt) call wrt. atomic_set(&ggtt->vm.open, open) and
mutex_unlock(&ggtt->vm.mutex). This shouldn't matter due to the i915
suspend sequence being single threaded.

Cc: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Imre Deak <imre.deak@intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20211101183551.3580546-1-imre.deak@intel.com

drm/i915/adlp/fb: Remove restriction on CCS AUX plane strides

As opposed to other GEN12 platforms ADLP provides a way to program the
stride of CCS surfaces independently of the main surface stride (within
the corresponding limit of the preceding and succeeding power-of-two
values of the main surface stride). Using this HW feature we can remove
the POT stride restriction on CCS surfaces, making the ADLP CCS FB uAPI
(FB modifiers) identical to that of TGL.

The HW makes the CCS stride flexible programming possible by deriving
the stride from the value programmed to the PLANE_STRIDE register. After
that the HW rounds up this value to the next power-of-two value and uses
this for walking the pages of the main surface mapped to GTT/DPT.

To align with the above scheme, introduce a scanout_stride view
parameter which will be programmed to the PLANE_STRIDE register and use
the mapping_stride view param to store the POT aligned value of the
same. By requiring userspace to pass in FBs with a CCS stride that
aligns with the main surface stride (matching the requirement of all
GEN12 platforms), the scanout_stride will be the userspace main surface
stride and the mapping_stride will be the POT rounded value of the same.

Cc: Ville Syrjälä <ville.syrjala@linux.intel.com>
Cc: Juha-Pekka Heikkila <juhapekka.heikkila@gmail.com>
Cc: Nanley G Chery <nanley.g.chery@intel.com>
Cc: Sameer Lattannavar <sameer.lattannavar@intel.com>
Signed-off-by: Imre Deak <imre.deak@intel.com>
Reviewed-by: Juha-Pekka Heikkila <juhapekka.heikkila@gmail.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20211026225105.2783797-8-imre.deak@intel.com

drm/i915/adlp/fb: Remove restriction on semiplanar UV plane offset

Since the surfaces of tiled FBs on ADLP are remapped it's pointless to
require an alignment in the allocated object. The necessary tile-row
alignment (to be programmed to the surface start register) will be
ensured later when flipping to the FB.

Signed-off-by: Imre Deak <imre.deak@intel.com>
Reviewed-by: Juha-Pekka Heikkila <juhapekka.heikkila@gmail.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20211026225105.2783797-7-imre.deak@intel.com

drm/i915/fb: Rename i915_color_plane_view::stride to mapping_stride

The next patch needs to distinguish between a view's mapping and scanout
stride. Rename the current stride parameter to mapping_stride with the
script below. mapping_stride will keep the same meaning as stride had
on all platforms so far, while the meaning of it will change on ADLP.

No functional changes.

@@
identifier intel_fb_view;
identifier i915_color_plane_view;
identifier color_plane;
expression e;
type T;
@@
struct intel_fb_view {
...
struct i915_color_plane_view {
...
- T stride;
+ T mapping_stride;
...
} color_plane[e];
...
};

@@
struct i915_color_plane_view pv;
@@
  pv.
-    stride
+    mapping_stride

@@
struct i915_color_plane_view *pvp;
@@
  pvp->
-     stride
+     mapping_stride

Signed-off-by: Imre Deak <imre.deak@intel.com>
Reviewed-by: Juha-Pekka Heikkila <juhapekka.heikkila@gmail.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20211026225105.2783797-6-imre.deak@intel.com

drm/i915/adlp/fb: Fix remapping of linear CCS AUX surfaces

During remapping CCS FBs the CCS AUX surface mapped size and offset->x,y
coordinate calculations assumed a tiled layout. This works as long as
the CCS surface height is aligned to 64 lines (ensuring a 4k bytes CCS
surface tile layout).  However this alignment is not required by the HW
(and the driver doesn't enforces it either).

Add the remapping logic required to remap the pages of CCS surfaces
without the above alignment, assuming the natural linear layout of the
CCS surface (vs. tiled main surface layout).

Cc: Juha-Pekka Heikkila <juhapekka.heikkila@gmail.com>
Cc: Ville Syrjälä <ville.syrjala@linux.intel.com>
Fixes: 3d1adc3d64cf ("drm/i915/adlp: Add support for remapping CCS FBs")
Signed-off-by: Imre Deak <imre.deak@intel.com>
Reviewed-by: Juha-Pekka Heikkila <juhapekka.heikkila@gmail.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20211026225105.2783797-5-imre.deak@intel.com

drm/i915/fb: Factor out functions to remap contiguous FB obj pages

Factor out functions needed to map contiguous FB obj pages to a GTT/DPT
VMA view in the next patch.

While at it s/4096/I915_GTT_PAGE_SIZE/ in add_padding_pages().

No functional changes.

v2: s/4096/I915_GTT_PAGE_SIZE/ (Matthew)

Signed-off-by: Imre Deak <imre.deak@intel.com>
Reviewed-by: Matthew Auld <matthew.auld@intel.com>
Reviewed-by: Juha-Pekka Heikkila <juhapekka.heikkila@gmail.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20211026225105.2783797-4-imre.deak@intel.com

drm/i915/adlp/fb: Prevent the mapping of redundant trailing padding NULL pages

So far the remapped view size in GTT/DPT was padded to the next aligned
offset unnecessarily after the last color plane with an unaligned size.
Remove the unnecessary padding.

Cc: Juha-Pekka Heikkila <juhapekka.heikkila@gmail.com>
Fixes: 3d1adc3d64cf ("drm/i915/adlp: Add support for remapping CCS FBs")
Signed-off-by: Imre Deak <imre.deak@intel.com>
Reviewed-by: Juha-Pekka Heikkila <juhapekka.heikkila@gmail.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20211026225105.2783797-3-imre.deak@intel.com

drm/i915/fb: Fix rounding error in subsampled plane size calculation

For NV12 FBs with odd main surface tile-row height the CCS surface
height was incorrectly calculated 1 less than the actual value. Fix this
by rounding up the result of divison. For consistency do the same for
the CCS surface width calculation.

Fixes: b3e57bccd68a ("drm/i915/tgl: Gen-12 render decompression")
Signed-off-by: Imre Deak <imre.deak@intel.com>
Reviewed-by: Juha-Pekka Heikkila <juhapekka.heikkila@gmail.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20211026225105.2783797-2-imre.deak@intel.com

drm/i915/display: program audio CDCLK-TS for keepalives

XE_LPD display adds support for display audio codec keepalive feature.
This feature works also when display codec is in D3 state and the audio
link is off (BCLK off). To enable this functionality, display driver
must update the AUD_TS_CDCLK_M/N registers whenever CDCLK is changed.
Actual timestamps are generated only when the audio codec driver
specifically enables the KeepAlive (KAE) feature.

This patch adds new hooks to intel_set_cdclk() in order to inform
display audio driver when CDCLK change is started and when it is
complete.

Bspec: 53679
Signed-off-by: Kai Vehmanen <kai.vehmanen@linux.intel.com>
Reviewed-by: Uma Shankar <uma.shankar@intel.com>
Acked-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Uma Shankar <uma.shankar@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20211021105915.4128635-1-kai.vehmanen@linux.intel.com

drm/i915/display: Check async flip state of every crtc and plane once

For every crtc in state, intel_atomic_check_async() was checking all
the crtc and plane states again.

v2: comparing pipe ids instead of crtc pointers when iterating over
planes

Cc: Karthik B S <karthik.b.s@intel.com>
Cc: Vandita Kulkarni <vandita.kulkarni@intel.com>
Cc: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: José Roberto de Souza <jose.souza@intel.com>
Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20211029202432.140745-1-jose.souza@intel.com

drm/i915/hdmi: Turn DP++ TMDS output buffers back on in encoder->shutdown()

Looks like our VBIOS/GOP generally fail to turn the DP dual mode adater
TMDS output buffers back on after a reboot. This leads to a black screen
after reboot if we turned the TMDS output buffers off prior to reboot.
And if i915 decides to do a fastboot the black screen will persist even
after i915 takes over.

Apparently this has been a problem ever since commit b2ccb822d376 ("drm/i915:
Enable/disable TMDS output buffers in DP++ adaptor as needed") if one
rebooted while the display was turned off. And things became worse with
commit fe0f1e3bfdfe ("drm/i915: Shut down displays gracefully on reboot")
since now we always turn the display off before a reboot.

This was reported on a RKL, but I confirmed the same behaviour on my
SNB as well. So looks pretty universal.

Let's fix this by explicitly turning the TMDS output buffers back on
in the encoder->shutdown() hook. Note that this gets called after irqs
have been disabled, so the i2c communication with the DP dual mode
adapter has to be performed via polling (which the gmbus code is
perfectly happy to do for us).

We also need a bit of care in handling DDI encoders which may or may
not be set up for HDMI output. Specifically ddc_pin will not be
populated for a DP only DDI encoder, in which case we don't want to
call intel_gmbus_get_adapter(). We can handle that by simply doing
the dual mode adapter type check before calling
intel_gmbus_get_adapter().

Cc: <stable@vger.kernel.org> # v5.11+
Fixes: fe0f1e3bfdfe ("drm/i915: Shut down displays gracefully on reboot")
Closes: https://gitlab.freedesktop.org/drm/intel/-/issues/4371
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20211029191802.18448-2-ville.syrjala@linux.intel.com
Reviewed-by: Stanislav Lisovskiy <stanislav.lisovskiy@intel.com>

drm/i915: Don't request GMBUS to generate irqs when called while irqs are off

We will need to do some i2c poking from the encoder->shutdown() hook.
Currently that gets called after irqs have been turned off. We still
poll the gmbus status bits even if the interrupt never arrives so
things will work just fine. But seems like asking gmbus to generate
interrupts we will never see is a bit pointless, so don't.

Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20211029191802.18448-1-ville.syrjala@linux.intel.com
Reviewed-by: Stanislav Lisovskiy <stanislav.lisovskiy@intel.com>