drm/amd/display: Only clear symclk otg flag for HDMI

[Description]
There is a corner case where the symclk otg flag is cleared
when disabling the phantom pipe for subvp (because the phantom
and main pipe share the same link). This is undesired because
we need the maintain the correct symclk otg flag state for
the main pipe.

For now only clear the flag only for HDMI signal type, since
it's only set for HDMI signal type (phantom is virtual). The
ideal solution is to not clear it if the stream is phantom but
currently there's a bug that doesn't allow us to do this. Once
this issue is fixed the proper fix can be implemented.

Reviewed-by: Samson Tam <samson.tam@amd.com>
Acked-by: Wayne Lin <wayne.lin@amd.com>
Signed-off-by: Alvin Lee <alvin.lee2@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Fix lightup regression with DP2 single display configs

[WHY]
Previous fix for multiple displays downstream of DP2 MST hub caused regression

[HOW]
Match sink IDs instead of sink struct addresses

Reviewed-by: Nicholas Kazlauskas <nicholas.kazlauskas@amd.com>
Reviewed-by: Charlene Liu <charlene.liu@amd.com>
Acked-by: Wayne Lin <wayne.lin@amd.com>
Signed-off-by: Michael Strauss <michael.strauss@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Deep copy dml2_context when copying dc_state

[WHY&HOW]
dml2_context should be deep copied from src to dst dc_state.

Reviewed-by: George Shen <george.shen@amd.com>
Acked-by: Wayne Lin <wayne.lin@amd.com>
Signed-off-by: Dillon Varone <dillon.varone@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Create dc_state after resource initialization

[WHY&HOW]
After refactoring dc_state, it is always constructed at the time of its
creation. Construction can only happen after dc resources are initialized, so
move creation to be after this.

Reviewed-by: George Shen <george.shen@amd.com>
Acked-by: Wayne Lin <wayne.lin@amd.com>
Signed-off-by: Dillon Varone <dillon.varone@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Fix null reference to state when getting subvp type

[WHY&HOW]
Need to provide valid pointer to dc_state when getting subvp pipe type.

Reviewed-by: Alvin Lee <alvin.lee2@amd.com>
Acked-by: Wayne Lin <wayne.lin@amd.com>
Signed-off-by: Dillon Varone <dillon.varone@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Refactor phantom resource allocation

[WHY?]
Phantom streams and planes were previously not referenced explcitly on creation.

[HOW?]
To reduce memory management complexity, add an additional phantom streams and planes
reference into dc_state, and move mall_stream_config to stream_status inside
the state to make it safe to modify in shallow copies. Also consildates any logic
that is affected by this change to dc_state.

Reviewed-by: Nicholas Kazlauskas <nicholas.kazlauskas@amd.com>
Reviewed-by: Jun Lei <jun.lei@amd.com>
Acked-by: Wayne Lin <wayne.lin@amd.com>
Signed-off-by: Dillon Varone <dillon.varone@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Refactor dc_state interface

[WHY?]
Part of the dc_state interface that deals with adding streams and planes should
remain public, while others that deal with internal status' and subvp should be
private to DC.

[HOW?]
Move and rename the public functions to dc_state.h and private functions to
dc_state_priv.h. Also add some additional functions for extracting subvp meta
data from the state.

Reviewed-by: Nicholas Kazlauskas <nicholas.kazlauskas@amd.com>
Reviewed-by: Jun Lei <jun.lei@amd.com>
Acked-by: Wayne Lin <wayne.lin@amd.com>
Signed-off-by: Dillon Varone <dillon.varone@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Wake DMCUB before executing GPINT commands

[Why]
DMCUB can be in idle when we attempt to interface with the HW through
the GPINT mailbox resulting in a system hang.

[How]
Add dc_wake_and_execute_gpint() to wrap the wake, execute, sleep
sequence.

If the GPINT executes successfully then DMCUB will be put back into
sleep after the optional response is returned.

It functions similar to the inbox command interface.

Cc: Mario Limonciello <mario.limonciello@amd.com>
Cc: Alex Deucher <alexander.deucher@amd.com>
Cc: stable@vger.kernel.org
Reviewed-by: Hansen Dsouza <hansen.dsouza@amd.com>
Acked-by: Wayne Lin <wayne.lin@amd.com>
Signed-off-by: Nicholas Kazlauskas <nicholas.kazlauskas@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Wake DMCUB before sending a command

[Why]
We can hang in place trying to send commands when the DMCUB isn't
powered on.

[How]
For functions that execute within a DC context or DC lock we can
wrap the direct calls to dm_execute_dmub_cmd/list with code that
exits idle power optimizations and reallows once we're done with
the command submission on success.

For DM direct submissions the DM will need to manage the enter/exit
sequencing manually.

We cannot invoke a DMCUB command directly within the DM execution
helper or we can deadlock.

Cc: Mario Limonciello <mario.limonciello@amd.com>
Cc: Alex Deucher <alexander.deucher@amd.com>
Cc: stable@vger.kernel.org
Reviewed-by: Hansen Dsouza <hansen.dsouza@amd.com>
Acked-by: Wayne Lin <wayne.lin@amd.com>
Signed-off-by: Nicholas Kazlauskas <nicholas.kazlauskas@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Refactor DMCUB enter/exit idle interface

[Why]
We can hang in place trying to send commands when the DMCUB isn't
powered on.

[How]
We need to exit out of the idle state prior to sending a command,
but the process that performs the exit also invokes a command itself.

Fixing this issue involves the following:

1. Using a software state to track whether or not we need to start
   the process to exit idle or notify idle.

It's possible for the hardware to have exited an idle state without
driver knowledge, but entering one is always restricted to a driver
allow - which makes the SW state vs HW state mismatch issue purely one
of optimization, which should seldomly be hit, if at all.

2. Refactor any instances of exit/notify idle to use a single wrapper
   that maintains this SW state.

This works simialr to dc_allow_idle_optimizations, but works at the
DMCUB level and makes sure the state is marked prior to any notify/exit
idle so we don't enter an infinite loop.

3. Make sure we exit out of idle prior to sending any commands or
   waiting for DMCUB idle.

This patch takes care of 1/2. A future patch will take care of wrapping
DMCUB command submission with calls to this new interface.

Cc: Mario Limonciello <mario.limonciello@amd.com>
Cc: Alex Deucher <alexander.deucher@amd.com>
Cc: stable@vger.kernel.org
Reviewed-by: Hansen Dsouza <hansen.dsouza@amd.com>
Acked-by: Wayne Lin <wayne.lin@amd.com>
Signed-off-by: Nicholas Kazlauskas <nicholas.kazlauskas@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: fix usb-c connector_type

[why]
BIOS switches to use USB-C connector type 0x18, but VBIOS's
objectInfo table not supported yet. driver needs to patch it
based on enc_cap from system integration info table.

Reviewed-by: Charlene Liu <charlene.liu@amd.com>
Acked-by: Wayne Lin <wayne.lin@amd.com>
Signed-off-by: Allen Pan <allen.pan@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: add debug option for ExtendedVBlank DLG adjust

[why & how]
Add new option for debug usage

Reviewed-by: Charlene Liu <charlene.liu@amd.com>
Acked-by: Wayne Lin <wayne.lin@amd.com>
Signed-off-by: Muhammad Ahmed <ahmed.ahmed@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Set test_pattern_changed update flag on pipe enable

[Why]
In certain cases, ODM pipe split can occur while stream already has test
pattern enabled. The new pipe used in the ODM combine config must be
configured to output the test pattern in this case.

[How]
If the stream is configured to output test pattern, then set the
test_pattern_changed update flag for the new pipe when it gets enabled.

Reviewed-by: Alvin Lee <alvin.lee2@amd.com>
Acked-by: Wayne Lin <wayne.lin@amd.com>
Signed-off-by: George Shen <george.shen@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: dereference variable before checking for zero

[Why]
Driver incorrectly checks if pointer variable OutBpp is null instead of
if the value being pointed to is zero.

[How]
Dereference OutBpp before checking for a value of zero.

Reviewed-by: Chaitanya Dhere <chaitanya.dhere@amd.com>
Acked-by: Wayne Lin <wayne.lin@amd.com>
Signed-off-by: Josip Pavic <josip.pavic@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: get dprefclk ss info from integration info table

[why & how]
we have two SSC_En:
we get ssc_info from dce_info for MPLL_SSC_EN.
we used to call VBIOS cmdtbl's smu_info's SS persentage for DPRECLK SS info,
is used for DP AUDIO and VBIOS' smu_info table was from systemIntegrationInfoTable.

since dcn35 VBIOS removed smu_info, driver need to use integrationInfotable directly.

Reviewed-by: Nicholas Kazlauskas <nicholas.kazlauskas@amd.com>
Acked-by: Wayne Lin <wayne.lin@amd.com>
Signed-off-by: Charlene Liu <charlene.liu@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: skip error logging when DMUB is inactive from S3

[Why]
On resume from S3, while DMUB is inactive, DMUB queue and execute
calls will not work.  Skip reporting errors in these scenarios

[How]
Add new return code during DMUB queue and execute calls when DMUB
is in S3 state. Skip logging errors in these scenarios

Reviewed-by: Alvin Lee <alvin.lee2@amd.com>
Acked-by: Wayne Lin <wayne.lin@amd.com>
Signed-off-by: Samson Tam <samson.tam@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdkfd: Use partial hmm page walk during buffer validation in SVM

SVM uses hmm page walk to valid buffer before map to gpu vm. After have partial
migration/mapping do validation on same vm range as migration/map do instead of
whole svm range that can be very large. This change is expected to improve svm
code performance.

Signed-off-by: Xiaogang Chen <xiaogang.chen@amd.com>
Reviewed-by: Philip Yang <philip.yang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd: Add missing definitions for `SMU_MAX_LEVELS_VDDGFX`

It is reported that on a Topaz dGPU the kernel emits:
amdgpu: can't get the mac of 5

This is because there is no definition for max levels of VDDGFX
declared for SMU71 or SMU7. The correct definition is VDDC so
use this.

Link: https://gitlab.freedesktop.org/drm/amd/-/issues/3049
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Mario Limonciello <mario.limonciello@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

Documentation/amdgpu: Remove a spurious character

`/` wasn't meant to be in the Dragon Range line

Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Mario Limonciello <mario.limonciello@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

Documentation/amdgpu: Add Hawk Point processors

These have been announced so add them to the table.

Link: https://www.amd.com/en/product/13971
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Mario Limonciello <mario.limonciello@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: make flip_timestamp_in_us a 64-bit variable

[Why]
This variable currently overflows after about 71 minutes. This doesn't
cause any known functional issues but it does make debugging more
difficult.

[How]
Make it a 64-bit variable.

Reviewed-by: Aric Cyr <aric.cyr@amd.com>
Acked-by: Wayne Lin <wayne.lin@amd.com>
Signed-off-by: Josip Pavic <josip.pavic@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Add case for dcn35 to support usb4 dmub hpd event

[Why & how]
Refactor dc_is_dmub_outbox_supported() a bit and add case for dcn35 to
register dmub outbox notification irq to handle usb4 relevant hpd event.

Reviewed-by: Roman Li <roman.li@amd.com>
Reviewed-by: Jun Lei <jun.lei@amd.com>
Signed-off-by: Wayne Lin <wayne.lin@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: disable FPO and SubVP for older DMUB versions on DCN32x

There have recently been changes that break backwards compatibility,
that were introduced into DMUB firmware (for DCN32x) concerning FPO and
SubVP. So, since those are just power optimization features, we can just
disable them unless the user is using a new enough version of DMUB
firmware.

Cc: stable@vger.kernel.org
Link: https://gitlab.freedesktop.org/drm/amd/-/issues/2870
Fixes: ed6e2782e974 ("drm/amd/display: For cursor P-State allow for SubVP")
Reported-by: Mikhail Gavrilov <mikhail.v.gavrilov@gmail.com>
Closes: https://lore.kernel.org/r/CABXGCsNRb0QbF2pKLJMDhVOKxyGD6-E+8p-4QO6FOWa6zp22_A@mail.gmail.com/
Reviewed-by: Harry Wentland <harry.wentland@amd.com>
Signed-off-by: Hamza Mahfooz <hamza.mahfooz@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Add a new DC debug mask for PSR-SU

Some issues have been raised that appear to be tied to PSR-SU.
To allow users to confirm they're tied to PSR-SU without turning off
PSR entirely introduce a new debug mask:

amdgpu.dcdebugmask=0x200

Reviewed-by: Harry Wentland <harry.wentland@amd.com>
Signed-off-by: Mario Limonciello <mario.limonciello@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Add function for dumping clk registers

[why]
Allow devs to check raw clk register values by dumping them on the log

[how]
Add clk register dump implementation

Reviewed-by: Charlene Liu <charlene.liu@amd.com>
Acked-by: Wayne Lin <wayne.lin@amd.com>
Signed-off-by: Johnson Chen <johnson.chen@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Unify optimize_required flags and VRR adjustments

[why]
There is only a single call to dc_post_update_surfaces_to_stream so
there is no need to have two flags to control it. Unifying this to a
single flag allows dc_stream_adjust_vmin_vmax to skip actual
programming when there is no change required.

[how]
Remove wm_optimze_required flag and set only optimize_required in its
place. Then in dc_stream_adjust_vmin_vmax, check that the stream timing
range matches the requested one and skip programming if they are equal.

Reviewed-by: Jun Lei <jun.lei@amd.com>
Acked-by: Wayne Lin <wayne.lin@amd.com>
Signed-off-by: Aric Cyr <aric.cyr@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Revert " drm/amd/display: Use channel_width = 2 for vram table 3.0"

[Description]
Revert commit fec05adc40c2 ("drm/amd/display: Use channel_width = 2 for vram table 3.0")
Because the issue is being fixed from VBIOS side.

Reviewed-by: Samson Tam <samson.tam@amd.com>
Acked-by: Wayne Lin <wayne.lin@amd.com>
Signed-off-by: Alvin Lee <alvin.lee2@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: remove HPO PG in driver side

[why & how]
Add config to make HPO PG handled in dmubfw ips entry/exit

Reviewed-by: Charlene Liu <charlene.liu@amd.com>
Acked-by: Wayne Lin <wayne.lin@amd.com>
Signed-off-by: Muhammad Ahmed <ahmed.ahmed@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: do not send commands to DMUB if DMUB is inactive from S3

[Why]
On resume from S3, may get apply_idle_optimizations call while DMUB
is inactive which will just time out.

[How]
Set and track power state in dmub_srv and check power state before
sending commands to DMUB.  Add interface in both dmub_srv and
dc_dmub_srv

Reviewed-by: Nicholas Kazlauskas <nicholas.kazlauskas@amd.com>
Acked-by: Wayne Lin <wayne.lin@amd.com>
Signed-off-by: Samson Tam <samson.tam@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/swsmu: remove duplicate definition of smu v14_0_0 driver if version

There is a repeated define of smu v14_0_0 driver if version, so delete
one in driver if header.

Signed-off-by: Li Ma <li.ma@amd.com>
Reviewed-by: Kenneth Feng <kenneth.feng@amd.com>
Reviewed-by: Yifan Zhang <yifan1.zhang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: make an improvement on amdgpu_hmm_range_get_pages

Only schedule when hmm_range_fault returns error.

Signed-off-by: James Zhu <James.Zhu@amd.com>
Acked-by: Felix Kuehling <Felix.Kuehling@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: increase hmm range get pages timeout

When application tries to allocate all system memory and cause memory
to swap out. Needs more time for hmm_range_fault to validate the
remaining page for allocation. To be safe, increase timeout value to
1 second for 64MB range.

Signed-off-by: James Zhu <James.Zhu@amd.com>
Acked-by: Felix Kuehling <Felix.Kuehling@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdkfd: svm range always mapped flag not working on APU

On gfx943 APU there is no VRAM and page migration, queue CWSR area, svm
range with always mapped flag, is not mapped to GPU correctly. This
works fine if retry fault on CWSR area can be recovered, but could cause
deadlock if there is another retry fault recover waiting for CWSR to
finish.

Fix this by mapping svm range with always mapped flag to GPU with ACCESS
attribute if XNACK ON.

There is side effect, because all GPUs have ACCESS attribute by default
on new svm range with XNACK on, the CWSR area will be mapped to all GPUs
after this change. This side effect will be fixed with Thunk change to
set CWSR svm range with ACCESS_IN_PLACE attribute on the GPU that user
queue is created.

Signed-off-by: Philip Yang <Philip.Yang@amd.com>
Reviewed-by: Felix Kuehling <felix.kuehling@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdkfd: only flush mes process context if mes support is there

Fix up on mes process context flush to prevent non-mes devices from
spamming error messages or running into undefined behaviour during
process termination.

Fixes: bd33bb1409b4 ("drm/amdkfd: fix mes set shader debugger process management")
Signed-off-by: Jonathan Kim <jonathan.kim@amd.com>
Reviewed-by: Eric Huang <jinhuieric.huang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: fix documentation for dm_crtc_additional_color_mgmt()

warning: expecting prototype for drm_crtc_additional_color_mgmt().
Prototype was for dm_crtc_additional_color_mgmt() instead

Reported-by: kernel test robot <lkp@intel.com>
Closes: https://lore.kernel.org/oe-kbuild-all/202312141801.o9eBCxt9-lkp@intel.com/
Signed-off-by: Melissa Wen <mwen@igalia.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: fix documentation for amdgpu_dm_verify_lut3d_size()

It takes the plane state rather than the crtc state.

Fixes: aba8b76baabd ("drm/amd/display: add plane shaper LUT support")
Reported-by: Stephen Rothwell <sfr@canb.auug.org.au>
Reviewed-by: Melissa Wen <mwen@igalia.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Cc: Melissa Wen <mwen@igalia.com>
Cc: Harry.Wentland@amd.com

drm/amd/pm: fix a double-free in amdgpu_parse_extended_power_table

The amdgpu_free_extended_power_table is called in every error-handling
paths of amdgpu_parse_extended_power_table. However, after the following
call chain of returning:

amdgpu_parse_extended_power_table
  |-> kv_dpm_init / si_dpm_init
      (the only two caller of amdgpu_parse_extended_power_table)
        |-> kv_dpm_sw_init / si_dpm_sw_init
            (the only caller of kv_dpm_init / si_dpm_init, accordingly)
              |-> kv_dpm_fini / si_dpm_fini
                  (goto dpm_failed in xx_dpm_sw_init)
                    |-> amdgpu_free_extended_power_table

As above, the amdgpu_free_extended_power_table is called twice in this
returning chain and thus a double-free is triggered. Similarily, the
last kfree in amdgpu_parse_extended_power_table also cause a double free
with amdgpu_free_extended_power_table in kv_dpm_fini.

Fixes: 84176663e70d ("drm/amd/pm: create a new holder for those APIs used only by legacy ASICs(si/kv)")
Signed-off-by: Zhipeng Lu <alexious@zju.edu.cn>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

gpu/drm/radeon: fix two memleaks in radeon_vm_init

When radeon_bo_create and radeon_vm_clear_bo fail, the vm->page_tables
allocated before need to be freed. However, neither radeon_vm_init
itself nor its caller have done such deallocation.

Fixes: 6d2f2944e95e ("drm/radeon: use normal BOs for the page tables v4")
Signed-off-by: Zhipeng Lu <alexious@zju.edu.cn>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drivers/amd/pm: fix a use-after-free in kv_parse_power_table

When ps allocated by kzalloc equals to NULL, kv_parse_power_table
frees adev->pm.dpm.ps that allocated before. However, after the control
flow goes through the following call chains:

kv_parse_power_table
  |-> kv_dpm_init
        |-> kv_dpm_sw_init
      |-> kv_dpm_fini

The adev->pm.dpm.ps is used in the for loop of kv_dpm_fini after its
first free in kv_parse_power_table and causes a use-after-free bug.

Fixes: a2e73f56fa62 ("drm/amdgpu: Add support for CIK parts")
Signed-off-by: Zhipeng Lu <alexious@zju.edu.cn>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/pm: fix a double-free in si_dpm_init

When the allocation of
adev->pm.dpm.dyn_state.vddc_dependency_on_dispclk.entries fails,
amdgpu_free_extended_power_table is called to free some fields of adev.
However, when the control flow returns to si_dpm_sw_init, it goes to
label dpm_failed and calls si_dpm_fini, which calls
amdgpu_free_extended_power_table again and free those fields again. Thus
a double-free is triggered.

Fixes: 841686df9f7d ("drm/amdgpu: add SI DPM support (v4)")
Signed-off-by: Zhipeng Lu <alexious@zju.edu.cn>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/pm: Remove unneeded semicolon

./drivers/gpu/drm/amd/pm/swsmu/amdgpu_smu.c:1418:2-3: Unneeded semicolon

Reported-by: Abaci Robot <abaci@linux.alibaba.com>
Closes: https://bugzilla.openanolis.cn/show_bug.cgi?id=7743
Signed-off-by: Yang Li <yang.lee@linux.alibaba.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu/debugfs: fix error code when smc register accessors are NULL

Should be -EOPNOTSUPP.

Fixes: 5104fdf50d32 ("drm/amdgpu: Fix a null pointer access when the smc_rreg pointer is NULL")
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Add 3x4 CTM support for plane CTM

Create drm_color_ctm_3x4 to support 3x4-dimension plane CTM matrix and
convert DRM CTM to DC CSC float matrix.

v3:
- rename ctm2 to ctm_3x4 (Harry)

Reviewed-by: Harry Wentland <harry.wentland@amd.com>
Signed-off-by: Joshua Ashton <joshua@froggi.es>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: add plane CTM support

Map the plane CTM driver-specific property to DC plane, instead of DC
stream. The remaining steps to program DPP block are already implemented
on DC shared-code.

v3:
- fix comment about plane and CRTC CTMs priorities (Harry)

Reviewed-by: Harry Wentland <harry.wentland@amd.com>
Signed-off-by: Melissa Wen <mwen@igalia.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/pm: add power save mode workload for smu 13.0.10

add power save mode workload for smu 13.0.10, so that in compute mode,
pmfw will add margin since some applications requres higher margin.

Signed-off-by: Kenneth Feng <kenneth.feng@amd.com>
Reviewed-by: Likun Gao <Likun.Gao@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu/vpe: enable vpe dpm

enable vpe dpm

Signed-off-by: Peyton Lee <peytolee@amd.com>
Reviewed-by: Lang Yu <lang.yu@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: add plane CTM driver-specific property

Plane CTM for pre-blending color space conversion. Only enable
driver-specific plane CTM property on drivers that support both pre- and
post-blending gamut remap matrix, i.e., DCN3+ family. Otherwise it
conflits with DRM CRTC CTM property.

Reviewed-by: Harry Wentland <harry.wentland@amd.com>
Signed-off-by: Melissa Wen <mwen@igalia.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: copy 3D LUT settings from crtc state to stream_update

When commiting planes, we copy color mgmt resources to the stream state.
Do the same for shaper and 3D LUTs.

Reviewed-by: Harry Wentland <harry.wentland@amd.com>
Signed-off-by: Joshua Ashton <joshua@froggi.es>
Co-developed-by: Melissa Wen <mwen@igalia.com>
Signed-off-by: Melissa Wen <mwen@igalia.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/pm: support return vpe clock table

pm supports return vpe clock table and soc clock table

Signed-off-by: Peyton Lee <peytolee@amd.com>
Reviewed-by: Li Ma <li.ma@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: fix ftrace event amdgpu_bo_move always move on same heap

Issue: during evict or validate happened on amdgpu_bo, the 'from' and
'to' is always same in ftrace event of amdgpu_bo_move

where calling the 'trace_amdgpu_bo_move', the comment says move_notify
is called before move happens, but actually it is called after move
happens, here the new_mem is same as bo->resource

Fix: move trace_amdgpu_bo_move from move_notify to amdgpu_bo_move

Signed-off-by: Wang, Beyond <Wang.Beyond@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: allow newer DC hardware to use degamma ROM for PQ/HLG

Need to funnel the color caps through to these functions so it can check
that the hardware is capable.

v2:
- remove redundant color caps assignment on plane degamma map (Harry)
- pass color caps to degamma params

v3:
- remove unused color_caps parameter from set_color_properties (Harry)

Reviewed-by: Harry Wentland <harry.wentland@amd.com>
Signed-off-by: Joshua Ashton <joshua@froggi.es>
Signed-off-by: Melissa Wen <mwen@igalia.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: add plane blend LUT and TF support

Map plane blend properties to DPP blend gamma. Plane blend is a
post-3D LUT curve that linearizes color space for blending. It may be
defined by a user-blob LUT and/or predefined transfer function. As
hardcoded curve (ROM) is not supported on blend gamma, we use AMD color
module to fill parameters when setting non-linear TF with empty LUT.

v2:
- rename DRM TFs to AMDGPU TFs

Reviewed-by: Harry Wentland <harry.wentland@amd.com>
Signed-off-by: Joshua Ashton <joshua@froggi.es>
Signed-off-by: Melissa Wen <mwen@igalia.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: handle empty LUTs in __set_input_tf

Unlike degamma, blend gamma doesn't support hardcoded curve
(predefined/ROM), but we can use AMD color module to fill blend gamma
parameters when we have non-linear plane gamma TF without plane gamma
LUT. The regular degamma path doesn't hit this.

Reviewed-by: Harry Wentland <harry.wentland@amd.com>
Signed-off-by: Joshua Ashton <joshua@froggi.es>
Signed-off-by: Melissa Wen <mwen@igalia.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: add plane 3D LUT support

Wire up DC 3D LUT to DM plane color management (pre-blending). On AMD
display HW, 3D LUT comes after a shaper curve and we always have to
program a shaper curve to delinearize or normalize the color space
before applying a 3D LUT (since we have a reduced number of LUT
entries).

In this version, the default values of 3D LUT for size and bit_depth are
17x17x17 and 12-bit, but we already provide here a more generic
mechanisms to program other supported values (9x9x9 size and 10-bit).

v2:
- started with plane 3D LUT instead of CRTC 3D LUT support

v4:
- lut3d_size is the max dimension size instead of # of entries

Reviewed-by: Harry Wentland <harry.wentland@amd.com>
Signed-off-by: Melissa Wen <mwen@igalia.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: warn when there are still mappings when a BO is destroyed v2

This can only happen when there is a reference counting bug.

v2: fix typo

Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/radeon: Prevent multiple debug error lines on suspend

Fix to avoid multiple debug error lines printed on every suspend by Radeon driver's debugfs.

radeon_debugfs_init() calls debugfs_create_file() for every ring.

This results in printing multiple error lines to the screen and dmesg similar to this:

[   92.378726] debugfs: File 'radeon_ring_gfx' in directory '0000:00:01.0' already present!
[   92.378732] debugfs: File 'radeon_ring_cp1' in directory '0000:00:01.0' already present!
[   92.378734] debugfs: File 'radeon_ring_cp2' in directory '0000:00:01.0' already present!
[   92.378737] debugfs: File 'radeon_ring_dma1' in directory '0000:00:01.0' already present!
[   92.378739] debugfs: File 'radeon_ring_dma2' in directory '0000:00:01.0' already present!
[   92.380775] debugfs: File 'radeon_ring_uvd' in directory '0000:00:01.0' already present!
[   92.406620] debugfs: File 'radeon_ring_vce1' in directory '0000:00:01.0' already present!
[   92.406624] debugfs: File 'radeon_ring_vce2' in directory '0000:00:01.0' already present!

Patch v1: The fix was to run lookup() for the file before trying to (re)create that debug file.
Patch v2: Call the radeon_debugfs_init() only once when radeon ring is initialized (as suggested
by Christian K. - thanks)

Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Woody Suwalski <terraluna977@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: add plane shaper TF support

Enable usage of predefined transfer func in addition to shaper 1D LUT.
That means we can save some complexity by just setting a predefined
curve, instead of programming a custom curve when preparing color space
for applying 3D LUT.

Reviewed-by: Harry Wentland <harry.wentland@amd.com>
Signed-off-by: Melissa Wen <mwen@igalia.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: add plane shaper LUT support

Map DC shaper LUT to DM plane color management. Shaper LUT can be used
to delinearize and/or normalize the color space for computational
efficiency and achiving specific visual styles. If a plane degamma is
apply to linearize the color space, a custom shaper 1D LUT can be used
just before applying 3D LUT.

v2:
- use DPP color caps to verify plane 3D LUT support
- add debug message if shaper LUT programming fails

v4:
- remove helper to check 3D LUT color caps (Harry)
- update desc of lut3d-setup helper from MPC to DPP

v5:
- remove color_mgmt_changed check that prevents color updates (Joshua)

Reviewed-by: Harry Wentland <harry.wentland@amd.com>
Signed-off-by: Melissa Wen <mwen@igalia.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: add HDR multiplier support

With `dc_fixpt_from_s3132()` translation, we can just use it to set
hdr_mult.

Reviewed-by: Harry Wentland <harry.wentland@amd.com>
Signed-off-by: Joshua Ashton <joshua@froggi.es>
Signed-off-by: Melissa Wen <mwen@igalia.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: add dc_fixpt_from_s3132 helper

Detach value translation from CTM to reuse it for programming HDR
multiplier property.

Reviewed-by: Harry Wentland <harry.wentland@amd.com>
Signed-off-by: Joshua Ashton <joshua@froggi.es>
Signed-off-by: Melissa Wen <mwen@igalia.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: reject atomic commit if setting both plane and CRTC degamma

DC only has pre-blending degamma caps (plane/DPP) that is currently in
use for CRTC/post-blending degamma, so that we don't have HW caps to
perform plane and CRTC degamma at the same time. Reject atomic updates
when serspace sets both plane and CRTC degamma properties.

Reviewed-by: Harry Wentland <harry.wentland@amd.com>
Signed-off-by: Melissa Wen <mwen@igalia.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: add plane degamma TF and LUT support

Set DC plane with user degamma LUT or predefined TF from driver-specific
plane color properties. If plane and CRTC degamma are set in the same
time, plane degamma has priority.  That means, we only set CRTC degamma
if we don't have plane degamma LUT or TF to configure. We return -EINVAL
if we don't have plane degamma settings, so we can continue and check
CRTC degamma.

Reviewed-by: Harry Wentland <harry.wentland@amd.com>
Signed-off-by: Joshua Ashton <joshua@froggi.es>
Signed-off-by: Melissa Wen <mwen@igalia.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: fix tear down order in amdgpu_vm_pt_free

When freeing PD/PT with shadows it can happen that the shadow
destruction races with detaching the PD/PT from the VM causing a NULL
pointer dereference in the invalidation code.

Fix this by detaching the the PD/PT from the VM first and then
freeing the shadow instead.

Signed-off-by: Christian König <christian.koenig@amd.com>
Fixes: https://gitlab.freedesktop.org/drm/amd/-/issues/2867
Cc: <stable@vger.kernel.org>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: decouple steps for mapping CRTC degamma to DC plane

The next patch adds pre-blending degamma to AMD color mgmt pipeline, but
pre-blending degamma caps (DPP) is currently in use to provide DRM CRTC
atomic degamma or implict degamma on legacy gamma. Detach degamma usage
regarging CRTC color properties to manage plane and CRTC color
correction combinations.

Reviewed-by: Harry Wentland <harry.wentland@amd.com>
Signed-off-by: Melissa Wen <mwen@igalia.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: mark plane as needing reset if color props change

We should reset a plane state if at least one of the color management
properties differs from old and new state.

Reviewed-by: Harry Wentland <harry.wentland@amd.com>
Signed-off-by: Joshua Ashton <joshua@froggi.es>
Co-developed-by: Melissa Wen <mwen@igalia.com>
Signed-off-by: Melissa Wen <mwen@igalia.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: set sdr_ref_white_level to 80 for out_transfer_func

Otherwise this is just initialized to 0. This needs to actually have a
value so that compute_curve can work for PQ EOTF.

Reviewed-by: Harry Wentland <harry.wentland@amd.com>
Signed-off-by: Joshua Ashton <joshua@froggi.es>
Co-developed-by: Melissa Wen <mwen@igalia.com>
Signed-off-by: Melissa Wen <mwen@igalia.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd: include drm/drm_edid.h only where needed

Including drm_edid.h from amdgpu_mode.h causes the rebuild of literally
hundreds of files when drm_edid.h is modified, while there are only a
handful of files that actually need to include drm_edid.h.

Signed-off-by: Jani Nikula <jani.nikula@intel.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: add CRTC gamma TF support

Add predefined transfer function programming. There is no post-blending
out gamma ROM for hardcoded curves, but we can use AMD color modules to
program LUT parameters from pre-defined coefficients and an empty
regamma LUT (or bump up LUT parameters with pre-defined TF values).

v2:
- update crtc color mgmt if regamma TF differs between states (Joshua)
- map inverse EOTF to DC transfer function (Melissa)

v3:
- update AMDGPU TF list

v4:
- update comment regarding regamma behavior

Reviewed-by: Harry Wentland <harry.wentland@amd.com>
Signed-off-by: Joshua Ashton <joshua@froggi.es>
Co-developed-by: Melissa Wen <mwen@igalia.com>
Signed-off-by: Melissa Wen <mwen@igalia.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: encapsulate atomic regamma operation

We will wire up MPC 3D LUT to DM CRTC color pipeline in the next patch,
but so far, only for atomic interface. By checking
set_output_transfer_func in DC drivers with MPC 3D LUT support, we can
verify that regamma is only programmed when 3D LUT programming fails. As
a groundwork to introduce 3D LUT programming and better understand each
step, detach atomic regamma programming from the crtc colocr updating
code.

Reviewed-by: Harry Wentland <harry.wentland@amd.com>
Signed-off-by: Melissa Wen <mwen@igalia.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: add comments to describe DM crtc color mgmt behavior

Describe some expected behavior of the AMD DM color mgmt programming.

Reviewed-by: Harry Wentland <harry.wentland@amd.com>
Signed-off-by: Melissa Wen <mwen@igalia.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: add CRTC gamma TF driver-specific property

Add AMD pre-defined transfer function property to default DRM CRTC gamma
to convert to wire encoding with or without a user gamma LUT. There is
no post-blending regamma ROM for pre-defined TF. When setting Gamma TF
(!= Identity) and LUT at the same time, the color module will combine
the pre-defined TF and the custom LUT values into the LUT that's
actually programmed.

v2:
- enable CRTC prop in the end of driver-specific prop sequence
- define inverse EOTFs as supported regamma TFs
- reword driver-specific function doc to remove shaper/3D LUT

v3:
- spell out TF+LUT behavior in the commit and comments (Harry)

Reviewed-by: Harry Wentland <harry.wentland@amd.com>
Co-developed-by: Joshua Ashton <joshua@froggi.es>
Signed-off-by: Joshua Ashton <joshua@froggi.es>
Signed-off-by: Melissa Wen <mwen@igalia.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: add plane blend LUT and TF driver-specific properties

Blend 1D LUT or a pre-defined transfer function (TF) can be set to
linearize content before blending, so that it's positioned just before
blending planes in the AMD color mgmt pipeline, and after 3D LUT
(non-linear space). Shaper and Blend LUTs are 1D LUTs that sandwich 3D
LUT. Drivers should advertize blend properties according to HW caps.

There is no blend ROM for pre-defined TF. When setting blend TF (!=
Identity) and LUT at the same time, the color module will combine the
pre-defined TF and the custom LUT values into the LUT that's actually
programmed.

v3:
- spell out TF+LUT behavior in the commit and comments (Harry)

v5:
- get blend blob correctly

Reviewed-by: Harry Wentland <harry.wentland@amd.com>
Signed-off-by: Joshua Ashton <joshua@froggi.es>
Signed-off-by: Melissa Wen <mwen@igalia.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: add plane shaper LUT and TF driver-specific properties

On AMD HW, 3D LUT always assumes a preceding shaper 1D LUT used for
delinearizing and/or normalizing the color space before applying a 3D
LUT. Add pre-defined transfer function to enable delinearizing content
with or without shaper LUT, where AMD color module calculates the
resulted shaper curve. We apply an inverse EOTF to go from linear
values to encoded values. If we are already in a non-linear space and/or
don't need to normalize values, we can bypass shaper LUT with a linear
transfer function that is also the default TF value.

There is no shaper ROM. When setting shaper TF (!= Identity) and LUT at
the same time, the color module will combine the pre-defined TF and the
custom LUT values into the LUT that's actually programmed.

v2:
- squash commits for shaper LUT and shaper TF
- define inverse EOTF as supported shaper TFs

v3:
- spell out TF+LUT behavior in the commit and comments (Harry)
- replace BT709 EOTF by inv OETF

v5:
- get shaper blob correctly (Joshua)

Reviewed-by: Harry Wentland <harry.wentland@amd.com>
Signed-off-by: Melissa Wen <mwen@igalia.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdkfd: fix mes set shader debugger process management

MES provides the driver a call to explicitly flush stale process memory
within the MES to avoid a race condition that results in a fatal
memory violation.

When SET_SHADER_DEBUGGER is called, the driver passes a memory address
that represents a process context address MES uses to keep track of
future per-process calls.

Normally, MES will purge its process context list when the last queue
has been removed.  The driver, however, can call SET_SHADER_DEBUGGER
regardless of whether a queue has been added or not.

If SET_SHADER_DEBUGGER has been called with no queues as the last call
prior to process termination, the passed process context address will
still reside within MES.

On a new process call to SET_SHADER_DEBUGGER, the driver may end up
passing an identical process context address value (based on per-process
gpu memory address) to MES but is now pointing to a new allocated buffer
object during KFD process creation.  Since the MES is unaware of this,
access of the passed address points to the stale object within MES and
triggers a fatal memory violation.

The solution is for KFD to explicitly flush the process context address
from MES on process termination.

Note that the flush call and the MES debugger calls use the same MES
interface but are separated as KFD calls to avoid conflicting with each
other.

Signed-off-by: Jonathan Kim <jonathan.kim@amd.com>
Tested-by: Alice Wong <shiwei.wong@amd.com>
Reviewed-by: Eric Huang <jinhuieric.huang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd: Fix a probing order problem on SDMA 2.4

commit 751e293f2c99 ("drm/amd: Move microcode init from sw_init to
early_init for SDMA v2.4") made a fateful mistake in
`adev->sdma.num_instances` wasn't declared when sdma_v2_4_init_microcode()
was run. This caused probing to fail.

Move the declaration to right before sdma_v2_4_init_microcode().

Closes: https://gitlab.freedesktop.org/drm/amd/-/issues/3043
Fixes: 751e293f2c99 ("drm/amd: Move microcode init from sw_init to early_init for SDMA v2.4")
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Mario Limonciello <mario.limonciello@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: Switch to aca bank for xgmi pcs err cnt

Instead of software managed counters.

Signed-off-by: Hawking Zhang <Hawking.Zhang@amd.com>
Reviewed-by: Yang Wang <kevinyang.wang@amd.com>
Reviewed-by: Stanley.Yang <Stanley.Yang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: add plane 3D LUT driver-specific properties

Add 3D LUT property for plane color transformations using a 3D lookup
table. 3D LUT allows for highly accurate and complex color
transformations and is suitable to adjust the balance between color
channels. It's also more complex to manage and require more
computational resources.

Since a 3D LUT has a limited number of entries in each dimension we want
to use them in an optimal fashion. This means using the 3D LUT in a
colorspace that is optimized for human vision, such as sRGB, PQ, or
another non-linear space. Therefore, userpace may need one 1D LUT
(shaper) before it to delinearize content and another 1D LUT after 3D
LUT (blend) to linearize content again for blending. The next patches
add these 1D LUTs to the plane color mgmt pipeline.

v3:
- improve commit message about 3D LUT
- describe the 3D LUT entries and size (Harry)

v4:
- advertise 3D LUT max size as the size of a single-dimension

v5:
- get lut3d blob correctly (Joshua)
- fix doc about 3d-lut dimension size (Sebastian)

Signed-off-by: Melissa Wen <mwen@igalia.com>
Reviewed-by: Harry Wentland <harry.wentland@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/radeon: include drm/drm_edid.h only where needed

Including drm_edid.h from radeon_mode.h causes the rebuild of more than
a hundred files when drm_edid.h is modified, while there are only a
handful of files that actually need to include drm_edid.h.

Signed-off-by: Jani Nikula <jani.nikula@intel.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: Enable tunneling on high-priority compute queues

This improves latency if the GPU is already busy with other work.
This is useful for VR compositors that submit highly latency-sensitive
compositing work on high-priority compute queues while the GPU is busy
rendering the next frame.

Userspace merge request:
https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/26462

v2: bump driver version (Alex)

Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Signed-off-by: Friedrich Vock <friedrich.vock@gmx.de>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu/sdma5.2: add begin/end_use ring callbacks

Add begin/end_use ring callbacks to disallow GFXOFF when
SDMA work is submitted and allow it again afterward.

This should avoid corner cases where GFXOFF is erroneously
entered when SDMA is still active.  For now just allow/disallow
GFXOFF in the begin and end helpers until we root cause the
issue.  This should not impact power as SDMA usage is pretty
minimal and GFXOSS should not be active when SDMA is active
anyway, this just makes it explicit.

v2: move everything into sdma5.2 code.  No reason for this
to be generic at this point.
v3: Add comments in new code

Link: https://gitlab.freedesktop.org/drm/amd/-/issues/2220
Reviewed-by: Mario Limonciello <mario.limonciello@amd.com> (v1)
Tested-by: Mario Limonciello <mario.limonciello@amd.com> (v1)
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Cc: stable@vger.kernel.org # 5.15+

drm/amd/pm: enable Wifi RFI mitigation feature support for SMU13.0.7

Fulfill the SMU13.0.7 support for Wifi RFI mitigation feature.

--
v10->v11:
  - downgrade the prompt level on message failure(Lijo)
v13:
 - Fix the format issue (IIpo Jarvinen)
 - Remove duplicate code (IIpo Jarvinen)

Signed-off-by: Evan Quan <quanliangl@hotmail.com>
Reviewed-by: Mario Limonciello <mario.limonciello@amd.com>
Signed-off-by: Ma Jun <Jun.Ma2@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/pm: enable Wifi RFI mitigation feature support for SMU13.0.0

Fulfill the SMU13.0.0 support for Wifi RFI mitigation feature.

--
v10->v11:
 - downgrade the prompt level on message failure(Lijo)
v13:
 - Fix the format issue (IIpo Jarvinen)
 - Move function smu_v13_0_0_set_wbrf_exclusion_ranges to
smu_v13_0.c as a generic code for later use (IIpo Jarvinen)

Co-developed-by: Evan Quan <quanliangl@hotmail.com>
Signed-off-by: Evan Quan <quanliangl@hotmail.com>
Reviewed-by: Mario Limonciello <mario.limonciello@amd.com>
Signed-off-by: Ma Jun <Jun.Ma2@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/pm: add flood detection for wbrf events

To protect PMFW from being overloaded.

Signed-off-by: Evan Quan <quanliangl@hotmail.com>
Reviewed-by: Mario Limonciello <mario.limonciello@amd.com>
Signed-off-by: Ma Jun <Jun.Ma2@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/pm: setup the framework to support Wifi RFI mitigation feature

With WBRF feature supported, as a driver responding to the frequencies,
amdgpu driver is able to do shadow pstate switching to mitigate possible
interference(between its (G-)DDR memory clocks and local radio module
frequency bands used by Wifi 6/6e/7).

--
v1->v2:
  - update the prompt for feature support(Lijo)
v8->v9:
  - update parameter document for smu_wbrf_event_handler(Simon)
v9->v10:
v10->v11:
 - correct the logics for wbrf range sorting(Lijo)
v13:
 - Fix the format issue (IIpo Jarvinen)

Signed-off-by: Evan Quan <quanliangl@hotmail.com>
Reviewed-by: Mario Limonciello <mario.limonciello@amd.com>
Signed-off-by: Ma Jun <Jun.Ma2@amd.com>
Signed-off-by: Ma Jun <Jun.Ma2@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/pm: update driver_if and ppsmc headers for coming wbrf feature

Add those data structures to support Wifi RFI mitigation feature.

Signed-off-by: Evan Quan <quanliangl@hotmail.com>
Reviewed-by: Mario Limonciello <mario.limonciello@amd.com>
Signed-off-by: Ma Jun <Jun.Ma2@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

Merge tag 'platform-drivers-x86-amd-wbrf-v6.8-1' into amd-drm-next

Immutable branch between pdx86 amd wbrf branch and wifi / amdgpu due for the v6.8 merge window

platform-drivers-x86-amd-wbrf-v6.8-1: v6.7-rc1 + AMD WBRF support
for merging into the wifi subsys and amdgpu driver for 6.8.

Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdkfd: Import DMABufs for interop through DRM

Use drm_gem_prime_fd_to_handle to import DMABufs for interop. This
ensures that a GEM handle is created on import and that obj->dma_buf
will be set and remain set as long as the object is imported into KFD.

Signed-off-by: Felix Kuehling <Felix.Kuehling@amd.com>
Reviewed-by: Ramesh Errabolu <Ramesh.Errabolu@amd.com>
Reviewed-by: Xiaogang.Chen <Xiaogang.Chen@amd.com>
Acked-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdkfd: Export DMABufs from KFD using GEM handles

Create GEM handles for exporting DMABufs using GEM-Prime APIs. The GEM
handles are created in a drm_client_dev context to avoid exposing them
in user mode contexts through a DMABuf import.

Signed-off-by: Felix Kuehling <Felix.Kuehling@amd.com>
Reviewed-by: Ramesh Errabolu <Ramesh.Errabolu@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: xgmi_fill_topology_info

1. Use the mirrored topology info to fill links for VF.
The new solution is required to simplify and optimize host driver logic.
Only use the new solution for VFs that support full duplex and
extended_peer_link_info otherwise the info would be incomplete.

2. avoid calling extended_link_info on VF as its not supported

Signed-off-by: Vignesh Chander <Vignesh.Chander@amd.com>
Reviewed-by: Zhigang Luo <zhigang.luo@amd.com>
Reviewed-by: Jonathan Kim <jonathan.kim@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Fix memory leak in dm_set_writeback()

'wb_info' needs to be freed on error paths or it would leak the memory.

Smatch pointed this out.

Fixes: c81e13b929df ("drm/amd/display: Hande writeback request from userspace")
Signed-off-by: Harshit Mogalapalli <harshit.m.mogalapalli@oracle.com>
Reviewed-by: Alex Hung <alex.hung@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: add plane HDR multiplier driver-specific property

Multiplier to 'gain' the plane. When PQ is decoded using the fixed func
transfer function to the internal FP16 fb, 1.0 -> 80 nits (on AMD at
least) When sRGB is decoded, 1.0 -> 1.0.  Therefore, 1.0 multiplier = 80
nits for SDR content. So if you want, 203 nits for SDR content, pass in
(203.0 / 80.0).

v4:
- comment about the PQ TF need for L-to-NL (from Harry's review)

Reviewed-by: Harry Wentland <harry.wentland@amd.com>
Signed-off-by: Joshua Ashton <joshua@froggi.es>
Co-developed-by: Melissa Wen <mwen@igalia.com>
Signed-off-by: Melissa Wen <mwen@igalia.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: document AMDGPU pre-defined transfer functions

Brief documentation about pre-defined transfer function usage on AMD
display driver and standardized EOTFs and inverse EOTFs.

v3:
- Document BT709 OETF (Pekka)
- Fix description of sRGB and pure power funcs (Pekka)

v4:
- Add description of linear and non-linear forms (Harry)

Reviewed-by: Harry Wentland <harry.wentland@amd.com>
Co-developed-by: Harry Wentland <harry.wentland@amd.com>
Signed-off-by: Harry Wentland <harry.wentland@amd.com>
Signed-off-by: Melissa Wen <mwen@igalia.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: explicitly define EOTF and inverse EOTF

Instead of relying on color block names to get the transfer function
intention regarding encoding pixel's luminance, define supported
Electro-Optical Transfer Functions (EOTFs) and inverse EOTFs, that
includes pure gamma or standardized transfer functions.

v3:
- squash linear and unity TFs to identity (Pekka)
- define the right TFs for BT.709 (Pekka and Harry)
- add comment about AMD TF coefficients

Suggested-by: Harry Wentland <harry.wentland@amd.com>
Reviewed-by: Harry Wentland <harry.wentland@amd.com>
Signed-off-by: Melissa Wen <mwen@igalia.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: add plane degamma TF driver-specific property

Allow userspace to tell the kernel driver the input space and,
therefore, uses correct predefined transfer function (TF) to go from
encoded values to linear values.

v2:
- rename TF enum prefix from DRM_ to AMDGPU_ (Harry)
- remove HLG TF

Reviewed-by: Harry Wentland <harry.wentland@amd.com>
Signed-off-by: Joshua Ashton <joshua@froggi.es>
Co-developed-by: Melissa Wen <mwen@igalia.com>
Signed-off-by: Melissa Wen <mwen@igalia.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Disable PSR-SU on Parade 0803 TCON again

When screen brightness is rapidly changed and PSR-SU is enabled the
display hangs on panels with this TCON even on the latest DCN 3.1.4
microcode (0x8002a81 at this time).

This was disabled previously as commit 072030b17830 ("drm/amd: Disable
PSR-SU on Parade 0803 TCON") but reverted as commit 1e66a17ce546 ("Revert
"drm/amd: Disable PSR-SU on Parade 0803 TCON"") in favor of testing for
a new enough microcode (commit cd2e31a9ab93 ("drm/amd/display: Set minimum
requirement for using PSR-SU on Phoenix")).

As hangs are still happening specifically with this TCON, disable PSR-SU
again for it until it can be root caused.

Cc: stable@vger.kernel.org
Cc: aaron.ma@canonical.com
Cc: binli@gnome.org
Cc: Marc Rossi <Marc.Rossi@amd.com>
Cc: Hamza Mahfooz <Hamza.Mahfooz@amd.com>
Signed-off-by: Mario Limonciello <mario.limonciello@amd.com>
Link: https://bugs.launchpad.net/ubuntu/+source/linux/+bug/2046131
Acked-by: Alex Deucher <alexander.deucher@amd.com>
Reviewed-by: Harry Wentland <harry.wentland@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: 3.2.264

Summary:

Bug fixes for:
  * DCN35 power gating
  * P-state change, & prefetch logic
  * ABM
  * DP 2.1

Acked-by: Aurabindo Pillai <aurabindo.pillai@amd.com>
Signed-off-by: Aric Cyr <aric.cyr@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: fix HW block PG sequence

[why]
Power up and power down has reverted programming order.
also make sure disable root clock last.

Reviewed-by: Muhammad Ahmed <ahmed.ahmed@amd.com>
Acked-by: Aurabindo Pillai <aurabindo.pillai@amd.com>
Signed-off-by: Charlene Liu <charlene.liu@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Force p-state disallow if leaving no plane config

[Description]
- When we're in a no plane config, DCN is always asserting
  P-State allow
- This creates a scenario where the P-State blackout can start
  just as VUPDATE takes place and transitions the DCN config to
  a one where one or more HUBP's are active which can result in
  underflow
- To fix this issue, force p-state disallow and unforce after
  the transition from no planes case -> one or more planes active

Reviewed-by: Samson Tam <samson.tam@amd.com>
Acked-by: Aurabindo Pillai <aurabindo.pillai@amd.com>
Signed-off-by: Alvin Lee <alvin.lee2@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: For prefetch mode > 0, extend prefetch if possible

[Description]
For mode programming we want to extend the prefetch as much as possible
(up to oto, or as long as we can for equ) if we're not already applying
the 60us prefetch requirement. This is to avoid intermittent underflow
issues during prefetch.

The prefetch extension is applied under the following scenarios:
1. We're in prefetch mode 1 (i.e. we don't support MCLK switch in blank)
2. We're using subvp or drr methods of p-state switch, in which case we
   we don't care if prefetch takes up more of the blanking time

Mode programming typically chooses the smallest prefetch time possible
(i.e. highest bandwidth during prefetch) presumably to create margin between
p-states / c-states that happen in vblank and prefetch. Therefore we only
apply this prefetch extension when p-state in vblank is not required (UCLK
p-states take up the most vblank time).

Reviewed-by: Jun Lei <jun.lei@amd.com>
Acked-by: Aurabindo Pillai <aurabindo.pillai@amd.com>
Signed-off-by: Alvin Lee <alvin.lee2@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Exit from idle state before accessing HW data

[why & how]
User interface cannot guarantee system is in
idle state, so need to ensure we exit idle state
before accessing any HW data.

Reviewed-by: Nicholas Kazlauskas <nicholas.kazlauskas@amd.com>
Acked-by: Aurabindo Pillai <aurabindo.pillai@amd.com>
Signed-off-by: Sung Joon Kim <sungkim@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>