drm/i915/dsb: Don't use indexed register writes needlessly

Turns out the DSB indexed register write command has
rather significant initial overhead compared to the normal
MMIO write command. Based on some quick experiments on TGL
you have to write the register at least ~5 times for the
indexed write command to come out ahead. If you write the
register less times than that the MMIO write is faster.

So it seems my automagic indexed write logic was a bit
misguided. Go back to the original approach only use
indexed writes for the cases we know will benefit from
it (indexed LUT register updates).

Currently we shouldn't have any cases where this truly
matters (just some rare double writes to the precision
LUT index registers), but we will need to switch the
legacy LUT updates to write each LUT register twice (to
avoid some palette anti-collision logic troubles).
This would be close to the worst case for using indexed
writes (two writes per register, and 256 separate registers).
Using the MMIO write command should shave off around 30%
of the execution time compared to using the indexed write
command.

Cc: stable@vger.kernel.org
Fixes: 34d8311f4a1c ("drm/i915/dsb: Re-instate DSB for LUT updates")
Fixes: 25ea3411bd23 ("drm/i915/dsb: Use non-posted register writes for legacy LUT")
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20241120164123.12706-2-ville.syrjala@linux.intel.com
Reviewed-by: Uma Shankar <uma.shankar@intel.com>

Revert "drm/i915/dpt: Make DPT object unshrinkable"

This reverts commit 51064d471c53dcc8eddd2333c3f1c1d9131ba36c.

Now that we forcefully evict all DPT VMAs during suspend
there should be no problem allowing the shrinker to eat
the DPT objects.

Cc: Brian Geffon <bgeffon@google.com>
Cc: Vidya Srinivas <vidya.srinivas@intel.com>
Closes: https://gitlab.freedesktop.org/drm/i915/kernel/-/issues/12965
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20241127061117.25622-5-ville.syrjala@linux.intel.com
Acked-by: Brian Geffon <bgeffon@google.com>
Reviewed-by: Vidya Srinivas <vidya.srinivas@intel.com>
Tested-by: Vidya Srinivas <vidya.srinivas@intel.com>

drm/i915/dpt: Evict all DPT VMAs on suspend

Currently intel_dpt_resume() tries to blindly rewrite all the
PTEs for currently bound DPT VMAs. That is problematic because
the CPU mapping for the DPT is only really guaranteed to exist
while the DPT object has been pinned. In the past we worked
around this issue by making DPT objects unshrinkable, but that
is undesirable as it'll waste physical RAM.

Let's instead forcefully evict all the DPT VMAs on suspend,
thus guaranteeing that intel_dpt_resume() has nothing to do.
To guarantee that all the DPT VMAs are evictable by
intel_dpt_suspend() we need to flush the cleanup workqueue
after the display output has been shut down.

And for good measure throw in a few extra WARNs to catch
any mistakes.

Cc: Brian Geffon <bgeffon@google.com>
Cc: Vidya Srinivas <vidya.srinivas@intel.com>
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20241127061117.25622-4-ville.syrjala@linux.intel.com
Reviewed-by: Vidya Srinivas <vidya.srinivas@intel.com>
Tested-by: Vidya Srinivas <vidya.srinivas@intel.com>

drm/i915: Intruduce display.wq.cleanup

Introduce a dedicated workqueue for the commit cleanup work.
In the future we'll need this to guarantee all the cleanup
works have finished at a specific point during suspend.

Cc: Brian Geffon <bgeffon@google.com>
Cc: Vidya Srinivas <vidya.srinivas@intel.com>
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20241127061117.25622-3-ville.syrjala@linux.intel.com
Reviewed-by: Vidya Srinivas <vidya.srinivas@intel.com>
Tested-by: Vidya Srinivas <vidya.srinivas@intel.com>

drm/i915: Don't reuse commit_work for the cleanup

Currently we reuse the commit_work for a later cleanup step.
Let's not do that so that atomic ioctl handler won't accidentally
wait for the cleanup work when it really wants to just wait on the
commit_tail() part. We'll just add another work struct for the
cleanup.

Cc: Brian Geffon <bgeffon@google.com>
Cc: Vidya Srinivas <vidya.srinivas@intel.com>
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20241127061117.25622-2-ville.syrjala@linux.intel.com
Reviewed-by: Vidya Srinivas <vidya.srinivas@intel.com>
Tested-by: Vidya Srinivas <vidya.srinivas@intel.com>

drm/i915/dislay: convert i9xx_display_sr.[ch] to struct intel_display

Going forward, struct intel_display is the main device data structure
for display. Switch to it.

Cc: Rodrigo Vivi <rodrigo.vivi@intel.com>
Reviewed-by: Mika Kahola <mika.kahola@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20241126101222.2671224-2-jani.nikula@intel.com
Signed-off-by: Jani Nikula <jani.nikula@intel.com>

drm/i915/display: prefer DISPLAY_VER over GRAPHICS_VER in display s/r

Use display version checks for display scratch registers, not graphics
version. And for the older platforms it's the same thing anyway.

Cc: Rodrigo Vivi <rodrigo.vivi@intel.com>
Reviewed-by: Mika Kahola <mika.kahola@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20241126101222.2671224-1-jani.nikula@intel.com
Signed-off-by: Jani Nikula <jani.nikula@intel.com>

drm/i915/ddi: clarify intel_ddi_connector_get_hw_state() for DP MST

encoder->get_hw_state() returns false for DP MST, and currently always
interprets 128b/132b as MST. Therefore the DDI MST mode checks in
intel_ddi_connector_get_hw_state() are redundant.

Prepare for future, and handle 128b/132b SST and warn on 8b/10b MST.

Cc: Imre Deak <imre.deak@intel.com>
Reviewed-by: Imre Deak <imre.deak@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20241125120959.2366419-1-jani.nikula@intel.com
Signed-off-by: Jani Nikula <jani.nikula@intel.com>

drm/i915: Fixed an typo in i915_gem_gtt.c

in function `i915_gem_gtt_reserve` @node comment,
i915_vma has no `mode` member, `i915_vma.node` is the correct name

Reviewed-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Signed-off-by: Zhang He <zhanghe9702@163.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20241120123245.71101-1-zhanghe9702@163.com
Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>

drm/i915/irq: emphasize display_irqs_enabled is only about VLV/CHV

Use display_irqs_enabled only on VLV/CHV where it's relevant. Rename to
vlv_display_irqs_enabled, to emphasize it's really only about VLV/CHV.

Reviewed-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/f60104ea59687cb8c65b18b4f9ddd832a643407d.1732102179.git.jani.nikula@intel.com
Signed-off-by: Jani Nikula <jani.nikula@intel.com>

drm/i915/irq: hide display_irqs_enabled access

Move the check for display_irqs_enabled within vlv_display_irq_reset()
and vlv_display_irq_postinstall() to avoid looking at struct
intel_display members within i915 core irq code.

Within display irq code, vlv_display_irq_reset() may need to be called
with !display_irqs_enabled, so add a small wrapper.

Reviewed-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/ef43e26ebab7f84768391f5053c0eba44b647c89.1732102179.git.jani.nikula@intel.com
Signed-off-by: Jani Nikula <jani.nikula@intel.com>

drm/i915/plane: convert initial plane setup to struct intel_display

struct intel_display replaces struct drm_i915_private as the main
display device pointer. Convert initial plane setup to it, as much as
possible.

Reviewed-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/9e370d8e90235165539f81ca2d00fdd2e883397f.1732102179.git.jani.nikula@intel.com
Signed-off-by: Jani Nikula <jani.nikula@intel.com>

drm/i915/overlay: add intel_overlay_available() and use it

Avoid accessing struct intel_display members directly from
i915_getparam_ioctl(). Add intel_overlay_available() function to provide
the information for I915_PARAM_HAS_OVERLAY.

Reviewed-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/26041645168ce3e76cb8f73bcb4c747619117e06.1732102179.git.jani.nikula@intel.com
Signed-off-by: Jani Nikula <jani.nikula@intel.com>

drm/i915/overlay: convert to struct intel_display

struct intel_display replaces struct drm_i915_private as the main
display device pointer. Convert overlay to it, as much as possible.

Reviewed-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/3680586c05e82fd01d173cfb4f8df015d6db663c.1732102179.git.jani.nikula@intel.com
Signed-off-by: Jani Nikula <jani.nikula@intel.com>

drm/i915/ddi: simplify intel_ddi_get_encoder_pipes() slightly

Use a temporary variable for DDI mode to simplify the conditions. This
is in line with the other places that read DDI mode.

Reviewed-by: Imre Deak <imre.deak@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/84892d31807bd8118474dd873e73c4d459f61448.1732106557.git.jani.nikula@intel.com
Signed-off-by: Jani Nikula <jani.nikula@intel.com>

drm/i915/ddi: refactor intel_ddi_connector_get_hw_state()

Refactor the switch-case into an if-ladder similar to
intel_ddi_read_func_ctl() for clarity.

This highlights how TRANS_DDI_MODE_SELECT_FDI_OR_128B132B works on
different platforms.

Reviewed-by: Imre Deak <imre.deak@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/191f0210d720f3113a092e1ef0c7996a7dee85a0.1732106557.git.jani.nikula@intel.com
Signed-off-by: Jani Nikula <jani.nikula@intel.com>

drm/i915/ddi: split up intel_ddi_read_func_ctl() by output type

The function has become quite long, and the switch-case statement quite
complex with the fallthrougs. Simplify by splitting to individual
functions and an if-ladder.

This highlights how TRANS_DDI_MODE_SELECT_FDI_OR_128B132B works on
different platforms.

Reviewed-by: Imre Deak <imre.deak@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/2621df6e6b0b7ac75159cfb112755c35b30ce906.1732106557.git.jani.nikula@intel.com
Signed-off-by: Jani Nikula <jani.nikula@intel.com>

drm/i915/ddi: rename temp to ddi_func_ctl in intel_ddi_read_func_ctl()

The temp name is a bit vague for something used so much in the function.

Reviewed-by: Imre Deak <imre.deak@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/29d21b8f829e8139cc8ad857a86d3fc967f2ac07.1732106557.git.jani.nikula@intel.com
Signed-off-by: Jani Nikula <jani.nikula@intel.com>

drm/i915/dp: refactor clear/wait for act sent

Move clear_act_sent() and wait_for_act_sent() to intel_ddi.[ch] and make
independent of DP MST. They'll be needed for 128b/132b SST
operation. Rename accordingly.

Reviewed-by: Imre Deak <imre.deak@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/ef05f5bc222e8ba48d84f75a9ea5dd29667055d2.1732106557.git.jani.nikula@intel.com
Signed-off-by: Jani Nikula <jani.nikula@intel.com>

drm/i915/reg: convert DP_TP_CTL/DP_TP_STATUS to REG_BIT() and friends

Use the modern style for defining register contents. Expand the status
register contents a bit.

TODO: There are more VC payload mapping fields, spanning more registers,
and have more bits on more recent platforms.

v2:
- Fix DP_TP_STATUS_STREAMS_ENABLED_MASK mask (Imre)
- Drop status VC3 payload mapping for now

Cc: Imre Deak <imre.deak@intel.com>
Reviewed-by: Imre Deak <imre.deak@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/1ef15e6bb58ca847f89c9b39cbc9771cb57db408.1732106557.git.jani.nikula@intel.com
Signed-off-by: Jani Nikula <jani.nikula@intel.com>

drm/i915/ddi: use intel_ddi_* naming convention for encoder enable/disable

All the other encoder hooks are named intel_ddi_*, follow suit with
intel_ddi_enable() and intel_ddi_disable(), and the dp/hdmi variants.

Reviewed-by: Imre Deak <imre.deak@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/9533cf61773f2cab3a6a29acf9e6ecfc00b6e8fd.1732106557.git.jani.nikula@intel.com
Signed-off-by: Jani Nikula <jani.nikula@intel.com>

drm/i915/mst: unify MST topology callback naming to mst_topology_*

Make it easier to keep track of what belongs where.

Reviewed-by: Imre Deak <imre.deak@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/e8efc8fddc3ca93e1256a1dad13bc4eca07c3382.1732106557.git.jani.nikula@intel.com
Signed-off-by: Jani Nikula <jani.nikula@intel.com>

drm/i915/mst: simplify mst_connector_get_hw_state()

Use a variable for the encoder to simplify.

Reviewed-by: Imre Deak <imre.deak@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/eb8595ae36c6330cce1615059bd2c89a7db79668.1732106557.git.jani.nikula@intel.com
Signed-off-by: Jani Nikula <jani.nikula@intel.com>

drm/i915/mst: unify MST connector function naming to mst_connector_*

Similar to commit 5674e700d43e ("drm/i915/mst: change naming from fake
encoders to MST stream encoders"), name all MST connector related
functions to mst_connector_*.

Reviewed-by: Imre Deak <imre.deak@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/3d4814a94b97fcff88722e0effd2fb5893b256af.1732106557.git.jani.nikula@intel.com
Signed-off-by: Jani Nikula <jani.nikula@intel.com>

drm/i915/mst: pass intel_dp around in mst stream helpers

All of the functions in question operate on the primary encoder, and
more specifically the primary intel_dp, so pass it around instead of the
encoder for less ambiguity.

Suggested-by: Imre Deak <imre.deak@intel.com>
Reviewed-by: Imre Deak <imre.deak@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/165ee8b723c42bcdeb3adf6a1b34ea09371e5d64.1732106557.git.jani.nikula@intel.com
Signed-off-by: Jani Nikula <jani.nikula@intel.com>

drm/i915/display/xe3lpd: Avoid setting YUV420_MODE in PIPE_MISC

For Xe3_LPD the PIPE_MISC YUV420 Enable (bit 27), already implies enabling
full blend YUV420 mode and YUV420 Mode (bit 26) is removed.
Therefore, avoid setting YUV420 Mode for Xe3_LPD+ while programming
PIPE_MISC for YCbCr420 output format.

Bspec: 69749
Signed-off-by: Ankit Nautiyal <ankit.k.nautiyal@intel.com>
Reviewed-by: Juha-Pekka Heikkila <juhapekka.heikkila@gmail.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20241113115531.3394962-1-ankit.k.nautiyal@intel.com

drm/i915/psr: Disable psr1 if setup_time > vblank

Issue is seen when PSR enabled with setup frames and when try to disable
PSR at SRDONACK State (0x1). PSR FSM is stuck at SRDONACK(0x1) for more
than 5 seconds. Issue not seen with Setup frames disabled. Currently
disable psr1 if setuptime > vblank to workaround the above issue.

HSD: 16024594674
WA: 18037818876

v1: Initial version
v2: Add debug log and some cosmetic changes. [Jouni, Jani, Nemesa]

Signed-off-by: Animesh Manna <animesh.manna@intel.com>
Reviewed-by: Jouni Högander <jouni.hogander@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20241105103916.1857731-1-animesh.manna@intel.com

drm/i915/dp_mst: Fix connector initialization in intel_dp_add_mst_connector()

The connector initialization in intel_dp_add_mst_connector() depends on
the device pointer in connector to be valid, at least by connector
debug printing. The device pointer is initialized by drm_connector_init(),
however that function also exposes the connector to in-kernel users,
which can't be done before the connector is fully initialized. For now
make sure the device pointer is valid before it's used, until a
follow-up change moving this to DRM core.

This issue was revealed by the commit in the Fixes: line below, before
which the above debug printing checked and handled a NULL device pointer
gracefully in DRM core.

Cc: Jani Nikula <jani.nikula@intel.com>
Fixes: 529798bd786a ("drm/i915/mst: convert to struct intel_display")
Closes: https://gitlab.freedesktop.org/drm/i915/kernel/-/issues/12799
Reviewed-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Signed-off-by: Imre Deak <imre.deak@intel.com>
Reviewed-by: Jani Nikula <jani.nikula@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20241118131044.1278028-1-imre.deak@intel.com
Signed-off-by: Jani Nikula <jani.nikula@intel.com>

drm/i915/dp: Include the time taken by AUX Tx for timeout

As per DP spec the timeout for LANE_CHANNEL_EQ_DONE is 400ms. But this
timeout value is exclusively for the Aux RD Interval and excludes the
time consumed for the AUX Tx (i.e reading/writing FFE presets). Add
another 50ms for these AUX Tx to the 400ms timeout.
Ref: "Figure 3-52: 128b132b DP DPTC LANEx_CHANNEL_EQ_DONE Sequence" of
DP2.1a spec.

Co-developed-by: Srikanth V NagaVenkata <nagavenkata.srikanth.v@intel.com>
Signed-off-by: Arun R Murthy <arun.r.murthy@intel.com>
Reviewed-by: Suraj Kandpal <suraj.kandpal@intel.com>
Acked-by: Jani Nikula <jani.nikula@intel.com>
Signed-off-by: Suraj Kandpal <suraj.kandpal@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20240925034432.1777029-4-arun.r.murthy@intel.com

drm/i915/dp: read Aux RD interval just before setting the FFE preset

Figure 3-52: 128b132b DP DPTC LANEx_CHANNEL_EQ_DONE Sequence of
DP2.1a spec.
After reading LANEx_CHANNEL_EQ_DONE, read the FFE presets.
AUX_RD_INTERVAL and then write the new FFE presets.

v4: Read AUX_RD_INTERVAL before get/set TX FFE preset (Jani)

Co-developed-by: Srikanth V NagaVenkata <nagavenkata.srikanth.v@intel.com>
Signed-off-by: Arun R Murthy <arun.r.murthy@intel.com>
Reviewed-by: Suraj Kandpal <suraj.kandpal@intel.com>
Signed-off-by: Suraj Kandpal <suraj.kandpal@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20241011045825.2629469-1-arun.r.murthy@intel.com

drm/i915/dp: use fsleep instead of usleep_range for LT

Sleeping for < 10us use udelay, for 10us to 20ms use usleep_range() and
for > 10ms use msleep. flseep() will call the particular API based on
the above condition. (Documentation/timers/timers-howto.rst)
Aux RD Interval value depends on the value read from the dpcd register
which is updated from the sink device, hence use flseep.

Co-developed-by: Srikanth V NagaVenkata <nagavenkata.srikanth.v@intel.com>
Signed-off-by: Arun R Murthy <arun.r.murthy@intel.com>
Reviewed-by: Jani Nikula <jani.nikula@intel.com>
Signed-off-by: Suraj Kandpal <suraj.kandpal@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20240925034432.1777029-2-arun.r.murthy@intel.com

drm/{i915, xe}: Move power_domains suspend/resume to display_power

Move intel_power_domains_{suspend,resume} to inside
intel_display_power_{suspend_late, resume_early}.

With this also change the VLV suspend failure to call
the intel_display_power_resume_early. In the end, the only
function executed there for VLV is the intel_power_domains_resume.
Besides make the code more consistency give the call that was
immediately before: intel_display_power_suspend_late.

Reviewed-by: Jonathan Cavitt <jonathan.cavitt@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20241113225016.208673-7-rodrigo.vivi@intel.com
Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>

drm/xe/display: Delay dsm handler registration

Bring some consistency to register/unregister order at the
same time it aligns with i915 sequence order.

Reviewed-by: Jonathan Cavitt <jonathan.cavitt@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20241113225016.208673-6-rodrigo.vivi@intel.com
Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>

drm/xe/display: Delay hpd_init resume

Align with i915 and only initialize hotplugs after the display driver
access has been resumed.

Reviewed-by: Jonathan Cavitt <jonathan.cavitt@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20241113225016.208673-5-rodrigo.vivi@intel.com
Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>

drm/{i915, xe}/display: Move DP MST calls to display_driver

Move dp_mst suspend/resume functions from the drivers towards
intel_display_driver to continue with the unification.

Reviewed-by: Jonathan Cavitt <jonathan.cavitt@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20241113225016.208673-4-rodrigo.vivi@intel.com
Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>

drm/i915/display: Move regfile registers intel_display.restore

The intel_display struct already has a place for save/restore
stuff. Move the i915's regfile there since they are only
related to display.

Reviewed-by: Jonathan Cavitt <jonathan.cavitt@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20241113225016.208673-3-rodrigo.vivi@intel.com
Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>

drm/i915/display: Convert i915_suspend into i9xx_display_sr

These save & restore functions inside i915_suspend are old display
functions to save and restore a bunch of display related registers.

Move it under display and rename accordantly. Just don't move it
entirely towards intel_display struct yet because it depends
on drm_i915_private for the IS_MOBILE.

While doing this conversion also update the MIT header using
the new SPDX ones.

v2: Fix Makefile and include (Jani)
    Removed vga and gmbus (Jani, Ville)

Cc: Jesse Barnes <jbarnes@virtuousgeek.org>
Cc: Jani Nikula <jani.nikula@intel.com>
Reviewed-by: Jani Nikula <jani.nikula@intel.com>
Reviewed-by: Jonathan Cavitt <jonathan.cavitt@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20241113225016.208673-2-rodrigo.vivi@intel.com
Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>

drm/i915: Remove vga and gmbus seq out of i915_restore_display

Restrict this function to only save and restore registers
functionality. Then, that can be moved out later to under
display with a proper name.

Cc: Jani Nikula <jani.nikula@intel.com>
Reviewed-by: Jonathan Cavitt <jonathan.cavitt@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20241113225016.208673-1-rodrigo.vivi@intel.com
Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>

drm/i915/dp: Limit max compressed bpp to 18 when forcing DSC

While forcing DSC when it is not actually required, the max compressed
bpp from the source gets picked for compression for some pipe bpps.
Apparently, when DSC is not required, forcing DSC to the maximum compressed
bpp seems to cause underruns.

To avoid this, limit the max compressed bpp to 18 while forcing DSC. This
value works across platforms with different pipe bpps and prevents
underruns.

Signed-off-by: Ankit Nautiyal <ankit.k.nautiyal@intel.com>
Reviewed-by: Naga Venkata Srikanth V <nagavenkata.srikanth.v@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20241030084100.1361689-1-ankit.k.nautiyal@intel.com

drm/i915/dp: add comments about hooks called from MST stream encoders

It would be best to have self-explanatory code, but lacking that, add
some comments about the way the DDI encoder hooks get called from DP MST
stream encoders.

Reviewed-by: Imre Deak <imre.deak@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/5941b8ef48cac4d61de2e2806a80645de48e856c.1731409802.git.jani.nikula@intel.com
Signed-off-by: Jani Nikula <jani.nikula@intel.com>

drm/i915/mst: change naming from fake encoders to MST stream encoders

The fake encoders pretty much match individual MST streams. The encoders
remain as fake as ever, but change the naming to MST stream
encoders. Rename all the encoder hooks and related functions called from
them to mst_stream_* to clarify what type of encoders the hooks are
called on.

Reviewed-by: Imre Deak <imre.deak@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/37d9572e884ca2fdd84ea6722b4b7cefde04eed9.1731409802.git.jani.nikula@intel.com
Signed-off-by: Jani Nikula <jani.nikula@intel.com>

drm/i915/mst: convert to struct intel_display

struct intel_display will replace struct drm_i915_private as the main
display device data structure. Convert the DP MST code to struct
intel_display as much as possible.

Reviewed-by: Imre Deak <imre.deak@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/84c9923fccdd779f0ca713cab1d26a0c802b140c.1731409802.git.jani.nikula@intel.com
Signed-off-by: Jani Nikula <jani.nikula@intel.com>

drm/i915/display: make CHICKEN_TRANS() display version aware

Making register macros platform or display version aware is not exactly
something I want to promote widely, but in this case it's the lesser of
two evils. hsw_chicken_trans_reg() is not pretty, and it doesn't have a
suitable home.

v2: Rebase

Reviewed-by: Imre Deak <imre.deak@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/712c17ee22537b0628aa32695743bc017b3fe332.1731409802.git.jani.nikula@intel.com
Signed-off-by: Jani Nikula <jani.nikula@intel.com>

drm/i915/mst: use primary_encoder in fake mst encoder creation

Use a primary_encoder local variable in
intel_dp_create_fake_mst_encoder() for clarity.

Reviewed-by: Imre Deak <imre.deak@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/7c9ad1db97550b7eacb1f010521704f623b0c689.1731409802.git.jani.nikula@intel.com
Signed-off-by: Jani Nikula <jani.nikula@intel.com>

drm/i915/mst: introduce to_primary_encoder() and to_primary_dp()

Add helpers to_primary_encoder() and to_primary_dp() to convert fake MST
encoder pointers to primary encoder and DP pointers, respectively, and
use them.

The main point is to highlight the primary encoder and DP usage. Very
few places actually need the struct intel_dp_mst_encoder pointer, or the
primary struct intel_digital_port, so ditch them where possible for
clarity.

Reviewed-by: Imre Deak <imre.deak@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/ff8786c04bae053055783f720f89a4d9f9815c70.1731409802.git.jani.nikula@intel.com
Signed-off-by: Jani Nikula <jani.nikula@intel.com>

drm/i915/mst: rename intel_encoder to encoder

Switch to the modern style in fake mst encoder creation.

Reviewed-by: Imre Deak <imre.deak@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/4d29387fba64b925c6ab4c108782a7b5ac22628a.1731409802.git.jani.nikula@intel.com
Signed-off-by: Jani Nikula <jani.nikula@intel.com>

drm/i915/mst: pass primary encoder to primary encoder hooks

Pass the primary encoder to the primary encoder hooks. This is
pedantically correct, but intel_ddi_post_pll_disable() also works with
the fake encoder by coincidence.

Reviewed-by: Imre Deak <imre.deak@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/a0151fddfdd5ec11a26345232cdd3ae59c8cf56a.1731409802.git.jani.nikula@intel.com
Signed-off-by: Jani Nikula <jani.nikula@intel.com>

drm/i915/display: add mobile platform group

Identify mobile platforms separately in display, using the platform
group mechanism. This enables dropping the dependency on i915_drv.h
IS_MOBILE() from display code.

v2: Make snb_display static (kernel test robot)

Reviewed-by: Vinod Govindapillai <vinod.govindapillai@intel.com>
Reviewed-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/2389fd58446e83a79ec57bc0b037679662037b9e.1731321183.git.jani.nikula@intel.com
Signed-off-by: Jani Nikula <jani.nikula@intel.com>

drm/i915/display: pass struct pci_dev * to intel_display_device_probe()

Convert intel_display_device_probe() to accept struct pci_dev * instead
of struct drm_i915_private *. Return struct intel_display * in
preparation of allocating the memory of it later.

Reviewed-by: Vinod Govindapillai <vinod.govindapillai@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/ab4e960e3fff46cbeba185882b1e554f0ccd5877.1731321183.git.jani.nikula@intel.com
Signed-off-by: Jani Nikula <jani.nikula@intel.com>

drm/i915/display: convert display device identification to struct intel_display

Convert intel_display_device.[ch] to struct intel_display, including
callers, but excluding intel_display_device_probe() which will be
handled in follow-up.

v2: fix display->drm = display->drm goof-up

Reviewed-by: Vinod Govindapillai <vinod.govindapillai@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/865b27b66f599e707081d46fca9f679e19a4e8aa.1731321183.git.jani.nikula@intel.com
Signed-off-by: Jani Nikula <jani.nikula@intel.com>

drm/i915/display: rename i915 parameter to __display in feature helpers

All the feature check helpers now support (and some require) passing
struct intel_display. Rename the parameter to __display to reflect the
fact.

v2: Rebase

Reviewed-by: Vinod Govindapillai <vinod.govindapillai@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/4933ded2efa9f333c560b6e634f60d4808614030.1731321183.git.jani.nikula@intel.com
Signed-off-by: Jani Nikula <jani.nikula@intel.com>

drm/i915/display: convert HAS_ULTRAJOINER() to struct intel_display

Convert HAS_ULTRAJOINER() to struct intel_display. Do minimal drive-by
conversions to struct intel_display in the callers while at it.

Reviewed-by: Vinod Govindapillai <vinod.govindapillai@intel.com>
Reviewed-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/f562d35c94cf7ef18d261a9867d959a58f165f1e.1731321183.git.jani.nikula@intel.com
Signed-off-by: Jani Nikula <jani.nikula@intel.com>

drm/i915/display: convert HAS_HW_SAGV_WM() to struct intel_display

Convert HAS_HW_SAGV_WM() to struct intel_display. Do minimal drive-by
conversions to struct intel_display in the callers while at it.

v2: Rebase

Reviewed-by: Vinod Govindapillai <vinod.govindapillai@intel.com> # v1
Reviewed-by: Rodrigo Vivi <rodrigo.vivi@intel.com> # v1
Link: https://patchwork.freedesktop.org/patch/msgid/8babfd6f09df054d33d604a02e213200a3783737.1731321183.git.jani.nikula@intel.com
Signed-off-by: Jani Nikula <jani.nikula@intel.com>

drm/i915/display: convert HAS_SAGV() to struct intel_display

Convert HAS_SAGV() to struct intel_display. Do minimal drive-by
conversions to struct intel_display in the callers while at it.

Reviewed-by: Vinod Govindapillai <vinod.govindapillai@intel.com>
Reviewed-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/c0f65566d08c4ab5296740367979beb1670d1c06.1731321183.git.jani.nikula@intel.com
Signed-off-by: Jani Nikula <jani.nikula@intel.com>

drm/i915/display: convert HAS_MBUS_JOINING() to struct intel_display

Convert HAS_MBUS_JOINING() to struct intel_display. Do minimal drive-by
conversions to struct intel_display in the callers while at it.

v2: Rebase

Reviewed-by: Vinod Govindapillai <vinod.govindapillai@intel.com> # v1
Reviewed-by: Rodrigo Vivi <rodrigo.vivi@intel.com> # v1
Link: https://patchwork.freedesktop.org/patch/msgid/c198dd5a0ab0a5f37edce3d18ddb4991cf3d4792.1731321183.git.jani.nikula@intel.com
Signed-off-by: Jani Nikula <jani.nikula@intel.com>

drm/i915/display: convert HAS_IPS() to struct intel_display

Convert HAS_IPS() to struct intel_display. Do minimal drive-by
conversions to struct intel_display in the callers while at it.

Reviewed-by: Vinod Govindapillai <vinod.govindapillai@intel.com>
Reviewed-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/55fd6063b30ca227483a479cff0012e5a915196d.1731321183.git.jani.nikula@intel.com
Signed-off-by: Jani Nikula <jani.nikula@intel.com>

drm/i915/display: convert HAS_GMBUS_BURST_READ() to struct intel_display

Convert HAS_GMBUS_BURST_READ() to struct intel_display. Do minimal
drive-by conversions to struct intel_display in the callers while at it.

Reviewed-by: Vinod Govindapillai <vinod.govindapillai@intel.com>
Reviewed-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/34dfafeccf6f8ddd5437c79e18978151bfd96c24.1731321183.git.jani.nikula@intel.com
Signed-off-by: Jani Nikula <jani.nikula@intel.com>

drm/i915/display: convert HAS_DP20() to struct intel_display

Convert HAS_DP20() to struct intel_display. Do minimal drive-by
conversions to struct intel_display in the callers while at it.

v2: Rebase

Reviewed-by: Vinod Govindapillai <vinod.govindapillai@intel.com> # v1
Reviewed-by: Rodrigo Vivi <rodrigo.vivi@intel.com> # v1
Link: https://patchwork.freedesktop.org/patch/msgid/757d91d2a57052b11cbd9f4f40d836d9245ca926.1731321183.git.jani.nikula@intel.com
Signed-off-by: Jani Nikula <jani.nikula@intel.com>

drm/i915/display: convert HAS_DOUBLE_BUFFERED_M_N() to struct intel_display

Convert HAS_DOUBLE_BUFFERED_M_N() to struct intel_display. Do minimal
drive-by conversions to struct intel_display in the callers while at it.

v2: Rebase

Reviewed-by: Vinod Govindapillai <vinod.govindapillai@intel.com>
Reviewed-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/d313d32ae411b86eedb86c4a4949dc84588362df.1731321183.git.jani.nikula@intel.com
Signed-off-by: Jani Nikula <jani.nikula@intel.com>

drm/i915/display: convert HAS_4TILE() to struct intel_display

Convert HAS_4TILE() to struct intel_display. Do minimal drive-by
conversions to struct intel_display in the callers while at it.

Reviewed-by: Vinod Govindapillai <vinod.govindapillai@intel.com>
Reviewed-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/123799e7c96d03d8eea7d9bea3d450e82c88ae5d.1731321183.git.jani.nikula@intel.com
Signed-off-by: Jani Nikula <jani.nikula@intel.com>

drm/i915/display: convert HAS_D12_PLANE_MINIMIZATION() to struct intel_display

Convert HAS_D12_PLANE_MINIMIZATION() to struct intel_display. Do minimal
drive-by conversions to struct intel_display in the callers while at it.

Reviewed-by: Vinod Govindapillai <vinod.govindapillai@intel.com>
Reviewed-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/ed9402eb83ab301f3c1d60884376d715b2190fd3.1731321183.git.jani.nikula@intel.com
Signed-off-by: Jani Nikula <jani.nikula@intel.com>

drm/i915/display: identify discrete graphics

Identify discrete graphics separately in display, using the platform
group mechanism. This enables dropping the dependency on i915_drv.h
IS_DGFX() from display code.

Start grouping platform groups separately in INTEL_DISPLAY_PLATFORMS()
in anticipation of more groups to come.

Reviewed-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/c02230d26cc0d9dbd7ddcc064661b2ad03739b6a.1731321183.git.jani.nikula@intel.com
Signed-off-by: Jani Nikula <jani.nikula@intel.com>

drm/i915/hdcp: Fix when the first read and write are retried

Make sure that the first read/write in hdcp2_authentication_key_exchange
are only retried when we have either DP/DPMST encoder connected,
since we do this to give docks and dp encoders some extra time to
get their HDCP DPCD registers ready only for DP/DPMST encoders as
this issue is only observed here no need to burden other encoders
with extra retries as this causes the HDMI connector to have some
other timing issue and fails HDCP authentication.

--v2
-Add intent of patch [Chaitanya]
-Add reasoning for loop [Jani]
-Make sure we forfiet the 50ms wait for non DP/DPMST encoders.

--v3
-Remove the is_dp_encoder check [Jani/Chaitanya]
-Make the commit message more clearer [Jani]

Fixes: 9d5a05f86d2f ("drm/i915/hdcp: Retry first read and writes to downstream")
Signed-off-by: Suraj Kandpal <suraj.kandpal@intel.com>
Reviewed-by: Chaitanya Kumar Borah <chaitanya.kumar.borah@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20241104035951.517837-1-suraj.kandpal@intel.com

drm/i915/scaler: Extract intel_allocate_scaler()

intel_atomic_setup_scaler() is a mess. Make it a bit less so
by moving the scaler allocation loop into its own function.

Reviewed-by: Luca Coelho <luciano.coelho@intel.com>
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20241107122658.21901-9-ville.syrjala@linux.intel.com

drm/i915/scaler: Make scaler in_use a bool

Make scaler in_use a boolean since that's how it's used.

Reviewed-by: Luca Coelho <luciano.coelho@intel.com>
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20241107122658.21901-8-ville.syrjala@linux.intel.com

drm/i915/scaler: Carve up intel_atomic_setup_scalers()

Declutter intel_atomic_setup_scalers() a bit by splitting
the crtc scaling/pfit vs. plane scaling cases into their
own functions.

Reviewed-by: Luca Coelho <luciano.coelho@intel.com>
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20241107122658.21901-7-ville.syrjala@linux.intel.com

drm/i915/scaler: Convert the scaler code to intel_display

struct intel_display will replace struct drm_i915_private as
the main thing for display code. Convert the scaler code to
use it (as much as possible at this stage).

Reviewed-by: Jani Nikula <jani.nikula@intel.com>
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20241107122658.21901-6-ville.syrjala@linux.intel.com

drm/i915/scaler: Clean up intel_atomic_setup_scalers() a bit

intel_atomic_setup_scalers() is currently messing around with
the internals of the atomic states. Stop doing that and instead
use the regular interfaces so that we don't need to know any ugly
implementation details.

Reviewed-by: Luca Coelho <luciano.coelho@intel.com>
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20241107122658.21901-5-ville.syrjala@linux.intel.com

drm/i915/scaler: Pass the whole atomic state into intel_atomic_setup_scalers()

intel_atomic_setup_scalers() currently digs out the full atomic
state from the crtc state. Flip that on its head so that we instead
pass in the full atomic state and dig out the crtc state (and whatever
else we need). This is generallte the better approach as it works
in all phases of the atomic commit, whereas the other apporoach only
really works during .atomic_check().

Reviewed-by: Jani Nikula <jani.nikula@intel.com>
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20241107122658.21901-4-ville.syrjala@linux.intel.com

drm/i915/scaler: Remove redudant junk from skl_scaler.h

Nuke some redundant includes and forward declarations from
skl_scaler.h.

Reviewed-by: Jani Nikula <jani.nikula@intel.com>
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20241107122658.21901-3-ville.syrjala@linux.intel.com

drm/i915/scaler: s/intel_crtc/crtc/ etc.

Rename some variables from the intel_foo to just foo, to match
the more modern style used throughout the driver.

Reviewed-by: Jani Nikula <jani.nikula@intel.com>
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20241107122658.21901-2-ville.syrjala@linux.intel.com

drm/i915/crt: Nuke unused crt->connector

crt->connector is never used, nuke it.

Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20241107161123.16269-11-ville.syrjala@linux.intel.com
Reviewed-by: Jani Nikula <jani.nikula@intel.com>

drm/i915/crt: Rename some variables

Rename a bunch of local variables to the preferred
encoder/connector from intel_encoder/intel_connector.

Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20241107161123.16269-10-ville.syrjala@linux.intel.com
Reviewed-by: Jani Nikula <jani.nikula@intel.com>

drm/i915/crt: Drop pointless drm_device variables

Remove a bunch of pointless 'struct drm_device *dev' local variables.

Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20241107161123.16269-9-ville.syrjala@linux.intel.com
Reviewed-by: Jani Nikula <jani.nikula@intel.com>

drm/i915/crt: s/pipe_config/crtc_state/

Call the crtc state 'crtc_state' rather than 'pipe_config',
as is the modern style.

Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20241107161123.16269-8-ville.syrjala@linux.intel.com
Reviewed-by: Jani Nikula <jani.nikula@intel.com>

drm/i915/crt: Extract intel_crt_regs.h

Move the analog port register definitions into their
own file.

Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20241107161123.16269-7-ville.syrjala@linux.intel.com
Reviewed-by: Jani Nikula <jani.nikula@intel.com>

drm/i915/crt: Clean up ADPA_HOTPLUG_BITS definitions

ADPA_HOTPLUG_BITS is defined in terms of the individual
register bits and is defined in intel_crt.c, whereas the
counterpart mask (ADPA_CRT_HOTPLUG_MASK) is just defined
as a raw hex constant and lives in i915_reg.h. Just define
both the same way (with unified name to boot) and move both
to intel_crt.c since they are more an implementation detail
rather than anything to do with the actual register definitions.

Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20241107161123.16269-6-ville.syrjala@linux.intel.com
Reviewed-by: Jani Nikula <jani.nikula@intel.com>

drm/i915/crt: Use REG_BIT() & co.

Follow the modern style and use REG_BIT() & co. for the analog
port register definitions.

Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20241107161123.16269-5-ville.syrjala@linux.intel.com
Reviewed-by: Jani Nikula <jani.nikula@intel.com>

drm/i915/crt: Drop the unused ADPA_DPMS bit definitions

The ADPA_DPMS bit definitions are just an alias for the
sync disable bits, and unused one at that. Drop the
pointless definitions.

Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20241107161123.16269-4-ville.syrjala@linux.intel.com
Reviewed-by: Jani Nikula <jani.nikula@intel.com>

drm/i915/crt: Split long line

Split an overly long line in the CRT code.

Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20241107161123.16269-3-ville.syrjala@linux.intel.com
Reviewed-by: Jani Nikula <jani.nikula@intel.com>

drm/i915: Grab intel_display from the encoder to avoid potential oopsies

Grab the intel_display from 'encoder' rather than 'state'
in the encoder hooks to avoid the massive footgun that is
intel_sanitize_encoder(), which passes NULL as the 'state'
argument to encoder .disable() and .post_disable().

TODO: figure out how to actually fix intel_sanitize_encoder()...

Fixes: 40eb34c3f491 ("drm/i915/crt: convert to struct intel_display")
Fixes: ab0b0eb5c85c ("drm/i915/tv: convert to struct intel_display")
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20241107161123.16269-2-ville.syrjala@linux.intel.com
Reviewed-by: Jani Nikula <jani.nikula@intel.com>

drm/i915/psr: stop using bitwise OR with booleans in wm_optimization_wa()

Bitwise OR for booleans is a bit of an abuse. We can trivially drop the
usage.

Cc: Jouni Högander <jouni.hogander@intel.com>
Cc: Suraj Kandpal <suraj.kandpal@intel.com>
Reviewed-by: Jouni Högander <jouni.hogander@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/f7ea5854c9a306f56f4142f8d37d567ee2f768a7.1730201504.git.jani.nikula@intel.com
Signed-off-by: Jani Nikula <jani.nikula@intel.com>

drm/i915/psr: add LATENCY_REPORTING_REMOVED() register bit helper

Drop the wa_16013835468_bit_get() function in favour of the register
macro. It doesn't have to be so complicated, and we don't have to use
the workaround name in everything that's related to it.

Cc: Jouni Högander <jouni.hogander@intel.com>
Cc: Suraj Kandpal <suraj.kandpal@intel.com>
Reviewed-by: Jouni Högander <jouni.hogander@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/22934fee1ea37c777c35e4b520d5f11b6cd953d0.1730201504.git.jani.nikula@intel.com
Signed-off-by: Jani Nikula <jani.nikula@intel.com>

drm/i915/dp: demote source OUI read/write failure logging to debug

Commit 1f12d63a14d7 ("drm/i915/dp: Write the source OUI for non-eDP
sinks as well") started writing source OUI for non-eDP sinks as well,
increasing the possibilities of hitting read/write failures either due
to the sink behaviour or hotplug or whatever.

Even before that, commit 3fb0501f0c07 ("drm/i915/display/dp: Reduce log
level for SOURCE OUI write failures") already reduced write failures to
info level when source OUI was just for eDP.

Further reduce the log level to just debug. Switch to struct
intel_display while at it.

Closes: https://gitlab.freedesktop.org/drm/xe/kernel/-/issues/3372
Cc: Clint Taylor <clinton.a.taylor@intel.com>
Cc: Imre Deak <imre.deak@intel.com>
Reviewed-by: Matt Roper <matthew.d.roper@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20241106162325.4065078-1-jani.nikula@intel.com
Signed-off-by: Jani Nikula <jani.nikula@intel.com>

drm/i915/xe3lpd: Use DMC wakelock by default

Although Bspec doesn't explicitly mentions that, as of Xe3_LPD, using
DMC wakelock is the officially recommended way of accessing registers
that would be off during DC5/DC6 and the legacy method (where the DMC
intercepts MMIO to wake up the hardware) is to be avoided.

As such, update the driver to use the DMC wakelock by default starting
with Xe3_LPD. Since the feature is somewhat new to the driver, also
allow disabling it via a module parameter for debugging purposes.

For that, make the existing parameter allow values -1 (per-chip
default), 0 (disabled) and 1 (enabled), similarly to what is done for
other parameters.

v2:
  - Describe -1 in the same area where 0 and 1 are described. (Luca)

Reviewed-by: Luca Coelho <luciano.coelho@intel.com>
Signed-off-by: Gustavo Sousa <gustavo.sousa@intel.com>
Signed-off-by: Matt Roper <matthew.d.roper@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20241108130218.24125-16-gustavo.sousa@intel.com

drm/i915/dmc_wl: Sanitize enable_dmc_wl according to hardware support

Instead of checking for HAS_DMC_WAKELOCK() multiple times, let's use it
to sanitize the enable_dmc_wl parameter and use that variable when
necessary.

Reviewed-by: Luca Coelho <luciano.coelho@intel.com>
Signed-off-by: Gustavo Sousa <gustavo.sousa@intel.com>
Signed-off-by: Matt Roper <matthew.d.roper@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20241108130218.24125-15-gustavo.sousa@intel.com

drm/i915/dmc_wl: Add and use HAS_DMC_WAKELOCK()

A HAS_DMC_WAKELOCK() macro gives more semantic than openly checking the
display version. Define it and use it where appropriate.

v2:
  - Make this patch contain only the non-functional refactor. Functional
    changes related to including HAS_DMC() in the macro are done in
    upcoming changes. (Jani)

Cc: Jani Nikula <jani.nikula@intel.com>
Reviewed-by: Luca Coelho <luciano.coelho@intel.com>
Signed-off-by: Gustavo Sousa <gustavo.sousa@intel.com>
Signed-off-by: Matt Roper <matthew.d.roper@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20241108130218.24125-14-gustavo.sousa@intel.com

drm/i915/dmc_wl: Couple enable/disable with dynamic DC states

Enabling and disabling the DMC wakelock should be done as part of
enabling and disabling of dynamic DC states, respectively. We should not
enable or disable DMC wakelock independently of DC states, otherwise we
would risk ending up with an inconsistent state where dynamic DC states
are enabled and the DMC wakelock is disabled, going against current
recommendations and making MMIO transactions potentially slower. In
future display IPs that could have a worse outcome if DMC trap
implementation is completely removed.

So, let's make things safer by tying stuff together, removing the
independent calls, and also put warnings in place to detect inconsistent
calls.

Reviewed-by: Luca Coelho <luciano.coelho@intel.com>
Signed-off-by: Gustavo Sousa <gustavo.sousa@intel.com>
Signed-off-by: Matt Roper <matthew.d.roper@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20241108130218.24125-13-gustavo.sousa@intel.com

drm/i915/dmc_wl: Deal with existing references when disabling

It is possible that there are active wakelock references at the time we
are disabling the DMC wakelock mechanism. We need to deal with that in
two ways:

(A) Implement the missing step from Bspec:

    The Bspec instructs us to clear any existing wakelock request bit
    after disabling the mechanism. That gives a clue that it is okay to
    disable while there are locks held and we do not need to wait for
    them. However, since the spec is not explicit about it, we need
    still to get confirmation with the hardware team. Let's thus
    implement the spec and add a TODO note.

(B) Ensure a consistent driver state:

    The enable/disable logic would be problematic if the following
    sequence of events would happen:

    1. Function A calls intel_dmc_wl_get();
    2. Some function calls intel_dmc_wl_disable();
    3. Some function calls intel_dmc_wl_enable();
    4. Function A is done and calls intel_dmc_wl_put().

    At (2), the refcount becomes zero and then (4) causes an invalid
    decrement to the refcount. That would cause some issues:

        - At the time between (3) and (4), function A would think that
          the hardware lock is held but it could not be really held
          until intel_dmc_wl_get() is called by something else.
        - The call made to (4) could cause the refcount to become zero
          and consequently the hardware lock to be released while there
          could be innocent paths trusting they still have the lock.

    To fix that, we need to keep the refcount correctly in sync with
    intel_dmc_wl_{get,put}() calls and retake the hardware lock when
    enabling the DMC wakelock with a non-zero refcount.

    One missing piece left to be handled here is the following scenario:

    1. Function A calls intel_dmc_wl_get();
    2. Some function calls intel_dmc_wl_disable();
    3. Some function calls intel_dmc_wl_enable();
    4. Concurrently with (3), function A performs the MMIO in between
       setting DMC_WAKELOCK_CFG_ENABLE and asserting the lock with
       __intel_dmc_wl_take().

    I'm mostly sure this would cause issues future display IPs if DMC
    trap implementation was completely removed. We need to check with
    the hardware team whether it would be safe to assert the hardware
    lock before setting DMC_WAKELOCK_CFG_ENABLE to avoid this scenario.
    If not, then we would have to deal with that via software
    synchronization.

Reviewed-by: Luca Coelho <luciano.coelho@intel.com>
Signed-off-by: Gustavo Sousa <gustavo.sousa@intel.com>
Signed-off-by: Matt Roper <matthew.d.roper@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20241108130218.24125-12-gustavo.sousa@intel.com

drm/i915/dmc_wl: Allow simpler syntax for single reg in range tables

Allow simpler syntax for defining entries for single registers in range
tables. That makes them easier to type as well as to read, allowing one
to quickly tell whether a range actually refers to a single register or
a "true range".

Reviewed-by: Luca Coelho <luciano.coelho@intel.com>
Signed-off-by: Gustavo Sousa <gustavo.sousa@intel.com>
Signed-off-by: Matt Roper <matthew.d.roper@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20241108130218.24125-11-gustavo.sousa@intel.com

drm/i915/dmc_wl: Track registers touched by the DMC

There are extra registers that require the DMC wakelock when specific
dynamic DC states are in place. Those are registers that are touched by
the DMC and require DC exit for proper access. Add the range tables for
them and use the correct one depending on the enabled DC state.

v2:
  - Do not look into power domains guts (i.e.
    display->power.domains.dc_state). (Jani)
  - Come up with better names for variables containing register ranges.
    (Luca)
  - Keep a copy of dc_state in struct intel_dmc_wl.
  - Update commit message for a clearer explanation for the need of
    these new tables.

Bspec: 71583
Cc: Jani Nikula <jani.nikula@intel.com>
Reviewed-by: Luca Coelho <luciano.coelho@intel.com>
Signed-off-by: Gustavo Sousa <gustavo.sousa@intel.com>
Signed-off-by: Matt Roper <matthew.d.roper@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20241108130218.24125-10-gustavo.sousa@intel.com

drm/i915/dmc_wl: Rename lnl_wl_range to powered_off_ranges

In an upcoming change, we will add extra range tables for registers that
are touched by the DMC during DC states. The range table that we are
currently using is meant for registers that are powered off during DC
states. As such, let's rename the table to powered_off_ranges and also
add a comment regarding its purpose in the function that uses it.

Reviewed-by: Luca Coelho <luciano.coelho@intel.com>
Signed-off-by: Gustavo Sousa <gustavo.sousa@intel.com>
Signed-off-by: Matt Roper <matthew.d.roper@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20241108130218.24125-9-gustavo.sousa@intel.com

drm/i915/dmc_wl: Extract intel_dmc_wl_reg_in_range()

We will be using more than one range table in
intel_dmc_wl_check_range(). As such, move the logic to a new function
and name it intel_dmc_wl_reg_in_range().

Reviewed-by: Luca Coelho <luciano.coelho@intel.com>
Signed-off-by: Gustavo Sousa <gustavo.sousa@intel.com>
Signed-off-by: Matt Roper <matthew.d.roper@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20241108130218.24125-8-gustavo.sousa@intel.com

drm/i915/dmc_wl: Use sentinel item for range tables

We are currently using ARRAY_SIZE() to iterate address ranges in
intel_dmc_wl_check_range(). In upcoming changes, we will be using more
than a single table and will extract the range checking logic into a
dedicated function that takes a range table as argument. As we will not
able to use ARRAY_SIZE() then, let's make range tables contain a
sentinel item at the end and use that instead of having to pass the size
as parameter in this future function.

Reviewed-by: Luca Coelho <luciano.coelho@intel.com>
Signed-off-by: Gustavo Sousa <gustavo.sousa@intel.com>
Signed-off-by: Matt Roper <matthew.d.roper@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20241108130218.24125-7-gustavo.sousa@intel.com

drm/i915/dmc_wl: Get wakelock when disabling dynamic DC states

Bspec says that disabling dynamic DC states require taking the DMC
wakelock to cause an DC exit before writing to DC_STATE_EN. Implement
that.

In fact, testing on PTL revealed we end up failing to exit DC5/6 without
this step.

Bspec: 71583
Reviewed-by: Luca Coelho <luciano.coelho@intel.com>
Signed-off-by: Gustavo Sousa <gustavo.sousa@intel.com>
Signed-off-by: Matt Roper <matthew.d.roper@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20241108130218.24125-6-gustavo.sousa@intel.com

drm/i915/dmc_wl: Check for non-zero refcount in release work

When the DMC wakelock refcount reaches zero, we know that there are no
users and that we can do the actual release operation on the hardware,
which is queued with a delayed work. The idea of the delayed work is to
avoid performing the release if a new lock user appears (i.e. refcount
gets incremented) in a very short period of time.

Based on the above, the release work should bail out if refcount is
non-zero (meaning new lock users appeared in the meantime), but our
current code actually does the opposite: it bails when refcount is zero.
That means that the wakelock is not released when it should be; and
that, when the work is not canceled in time, it ends up being releasing
when it should not.

Fix that by inverting the condition.

Reviewed-by: Luca Coelho <luciano.coelho@intel.com>
Signed-off-by: Gustavo Sousa <gustavo.sousa@intel.com>
Signed-off-by: Matt Roper <matthew.d.roper@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20241108130218.24125-5-gustavo.sousa@intel.com

drm/i915/dmc_wl: Use non-sleeping variant of MMIO wait

Some display MMIO transactions for offsets in the range that requires
the DMC wakelock happen in atomic context (this has been confirmed
during tests on PTL). That means that we need to use a non-sleeping
variant of MMIO waiting function.

Implement __intel_de_wait_for_register_atomic_nowl() and use it when
waiting for acknowledgment of acquire/release.

v2:
  - No __intel_de_wait_for_register_atomic_nowl() wrapper to convert
    i915 to display. (Jani)
  - Add a quick explanation why DMC_WAKELOCK_CTL_TIMEOUT_US is defined
    in microseconds. (Luca)

Cc: Jani Nikula <jani.nikula@intel.com>
Reviewed-by: Luca Coelho <luciano.coelho@intel.com>
Signed-off-by: Gustavo Sousa <gustavo.sousa@intel.com>
Signed-off-by: Matt Roper <matthew.d.roper@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20241108130218.24125-4-gustavo.sousa@intel.com

drm/xe: Mimic i915 behavior for non-sleeping MMIO wait

In upcoming display changes, we will modify the DMC wakelock MMIO
waiting code to choose a non-sleeping variant implementation, because
the wakelock is also taking in atomic context.

While xe provides an explicit parameter (namely "atomic") to prevent
xe_mmio_wait32() from sleeping, i915 does not and implements that
behavior when slow_timeout_ms is zero.

So, for now, let's mimic what i915 does to allow for display to use
non-sleeping MMIO wait. In the future, we should come up with a better
and explicit interface for this behavior in i915, at least while display
code is not an independent entity with proper interfaces between xe and
i915.

v2:
  - Make the tone in comment the comment added in
    __intel_wait_for_register() more explanatory than a FIXME-like text.
    (Luca)

Reviewed-by: Luca Coelho <luciano.coelho@intel.com>
Signed-off-by: Gustavo Sousa <gustavo.sousa@intel.com>
Signed-off-by: Matt Roper <matthew.d.roper@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20241108130218.24125-3-gustavo.sousa@intel.com

drm/i915/dmc_wl: Use i915_mmio_reg_offset() instead of reg.reg

The macro i915_mmio_reg_offset() is the proper interface to get a
register's offset. Use that instead of looking directly at reg.reg.

Reviewed-by: Jani Nikula <jani.nikula@intel.com>
Reviewed-by: Luca Coelho <luciano.coelho@intel.com>
Signed-off-by: Gustavo Sousa <gustavo.sousa@intel.com>
Signed-off-by: Matt Roper <matthew.d.roper@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20241108130218.24125-2-gustavo.sousa@intel.com

drm/i915/psr: Disable Panel Replay as well if VRR is enabled

Having Panel Replay enabled together with VRR is causing following errors:

xe 0000:00:02.0: [drm] *ERROR* Timed out waiting PSR idle state
xe 0000:00:02.0: [drm] *ERROR* Fault errors on pipe A: 0x00000080
xe 0000:00:02.0: [drm] *ERROR* CPU pipe A FIFO underrun
xe 0000:00:02.0: [drm] *ERROR* Fault errors on pipe A: 0x00040080
xe 0000:00:02.0: [drm] *ERROR* Fault errors on pipe A: 0x00040080

Let's disable Panel Replay as well if VRR is enabled.

Signed-off-by: Jouni Högander <jouni.hogander@intel.com>
Reviewed-by: Mitul Golani <mitulkumar.ajitkumar.golani@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20241031081025.3215966-1-jouni.hogander@intel.com