drm/amd/display: Fix uninitialized variable.

[Why]
Uninitialized variable causes diag compilation build failure.

[How]
- Ensure that variable in question is always initialized before being
used.
- The variable in question is the USB4 DP training pattern. In case an
unsupported training pattern has been requested, update status
accordingly and abort current link training attempt.

Reviewed-by: Meenakshikumar Somasundaram <Meenakshikumar.Somasundaram@amd.com>
Acked-by: Alan Liu <HaoPing.Liu@amd.com>
Signed-off-by: Jimmy Kizito <Jimmy.Kizito@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Move all linux includes into OS types

Move all linux includes into OS types.

Acked-by: Alan Liu <HaoPing.Liu@amd.com>
Signed-off-by: Harry Wentland <harry.wentland@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Fix null timing generator resource

[Why]
For some customer blending transition cases, the
available pipe for second stream is a pipe index that is
greater than the number of timing generators, which
can cause a problem in acquire_first_free_pipe since it
assumes same index for pipe and timing generator

[How]
Added logic to use last timing generator index
if the pipe index is greater than number of timing generators.

Acked-by: Alan Liu <HaoPing.Liu@amd.com>
Signed-off-by: Eric Bernstein <eric.bernstein@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Add flag to modify MST delay

[Why]
Some panels may require more MST delay on discovery

[How]
Add panel patch and debug mst delay flag

Reviewed-by: Charlene Liu <Charlene.Liu@amd.com>
Acked-by: Rodrigo Siqueira <Rodrigo.Siqueira@amd.com>
Signed-off-by: Duncan Ma <duncan.ma@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Disable TBT3 DSC work around by default.

[Why]
Some TBT3 docks have DPOAs which report USB4 capability and are expected
to support USB4 DPOA features such as FEC/DSC.

[How]
By default, do not override FEC/DSC capabilities reported by TBT3 docks.

Reviewed-by: Meenakshikumar Somasundaram <Meenakshikumar.Somasundaram@amd.com>
Acked-by: Rodrigo Siqueira <Rodrigo.Siqueira@amd.com>
Signed-off-by: Jimmy Kizito <Jimmy.Kizito@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Fix refresh rate issue on Club 3D

Recently we introduced a patch for fixing an MST issue, but it caused a
regression on Club 3D since we could not set a refresh rate higher than
60Hz. This commit fixes this issue by adding a proper check after
validating the stream.

Fixes: f04d275d94e164 ("drm/amd/display: add mst port output bw check")
Reviewed-by: Hersen Wu <hersenxs.wu@amd.com>
Acked-by: Rodrigo Siqueira <Rodrigo.Siqueira@amd.com>
Signed-off-by: Rodrigo Siqueira <Rodrigo.Siqueira@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Remove configuration option for dpia hpd delay

[Why]
DC debug option to configure dpia hpd processing delay is not required.

[How]
Remove dc debug option for dpia hpd delay and also added log for
querying dpia hpd state.

Reviewed-by: Mustapha Ghaddar <Mustapha.Ghaddar@amd.com>
Reviewed-by: Jun Lei <Jun.Lei@amd.com>
Acked-by: Rodrigo Siqueira <Rodrigo.Siqueira@amd.com>
Signed-off-by: Meenakshikumar Somasundaram <meenakshikumar.somasundaram@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Indicate stream change on ODM change

[Why]
With ODM policy 2 to 1, there exists a new use case
scenario where stream content is unchanged, but ODM
may be used.  When this happens, the stream needs
to be committed with a new pipe setting.
This did not happen due to stream change
detection logic not accounting for ODM.

[How]
Set ODM flag in stream and commit stream when change
in ODM has been detected due to policy change.

Reviewed-by: Samson Tam <Samson.Tam@amd.com>
Acked-by: Rodrigo Siqueira <Rodrigo.Siqueira@amd.com>
Signed-off-by: Chris Park <chris.park@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: disable otg toggle w/a on boot

This w/a has a bad interaction with seamless boot toggling an
active stream. Most panels recover, however some fail leading
to display corruption.

Reviewed-by: Charlene Liu <Charlene.Liu@amd.com>
Acked-by: Rodrigo Siqueira <Rodrigo.Siqueira@amd.com>
Signed-off-by: Dmytro Laktyushkin <Dmytro.Laktyushkin@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Fix dmub soft hang for PSR 1

[Why]
Unexpected change of aux hw mapping causes dmub soft hang when
initiate aux transation at wrong aux channel.

ddc_channel stands for hw dp aux index which is from vbios,
but link_index is pure software concept for link count depending on which link
is probed first. They are not interchangeable.

dmub aux transaction could pass if happens eDP link_index gets
the same value as vbios ddc_channel, e.g., ddc_channel = 1, link_index = 1
if they gets different, e.g., ddc_channel = 2, link_index = 0, overwrite
ddc_channel with link_index will have wrong ddc channel being used for aux
transaction in dmub PSR, cause aux transaction soft hang.

[How]
ddc_channel mapping to each link is determined by vbios and further
parsed in dc. Such info. should not be touched in any kind, otherwise
the mapping is screwed up leading to aux transaction timeout.

Reviewed-by: Aurabindo Jayamohanan Pillai <Aurabindo.Pillai@amd.com>
Acked-by: Rodrigo Siqueira <Rodrigo.Siqueira@amd.com>
Signed-off-by: Fangzhi Zuo <Jerry.Zuo@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Guard against NULL link encoder in log hw state

[Why & How]
Check lenc is not NULL since dynamic link encoder assignment could
end up assigning a NULL link encoder.

Reviewed-by: Michael Strauss <Michael.Strauss@amd.com>
Acked-by: Rodrigo Siqueira <Rodrigo.Siqueira@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: Remove incorrect ASSERT check for link_enc

[Why]
It's possible that we don't have a link encoder assignment if the
context is NULL but we're calling dc_add_stream_to_ctx from DM directly.

Link encoder assignment will happen later after global validation
runs with fast_validate = false.

[How]
Remove the ASSERTION. We already guard against NULL link_enc.

Reviewed-by: Michael Strauss <Michael.Strauss@amd.com>
Acked-by: Rodrigo Siqueira <Rodrigo.Siqueira@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: Guard against ddc_pin being NULL for AUX

[Why]
In the case where we don't support DMUB aux but we have DPIA links
in the configuration we might try to message AUX using the legacy
path - where DDC pin is NULL. This causes a NULL pointer dereference.

[How]
Guard against NULL DDC pin, return a failure for aux engine acquire.

Reviewed-by: Michael Strauss <Michael.Strauss@amd.com>
Acked-by: Rodrigo Siqueira <Rodrigo.Siqueira@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: Program ACP related register

- Setup the shift and mask of HDMI_ACP_SEND register
- Program the register in hdmi stream encoder
- Also update ACP register in azalia configuration

Reviewed-by: Harry Wentland <Harry.Wentland@amd.com>
Acked-by: Rodrigo Siqueira <Rodrigo.Siqueira@amd.com>
Signed-off-by: Alan Liu <HaoPing.Liu@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: add an option to skip wait for HPD when powering on eDP panel

[HOW/WHY]
Add an option to skip edp_wait_for_hpd_ready when necessary

Reviewed-by: Jayendran Ramani <Jayendran.Ramani@amd.com>
Acked-by: Rodrigo Siqueira <Rodrigo.Siqueira@amd.com>
Signed-off-by: Evgenii Krasnikov <Evgenii.Krasnikov@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Don't set dram clock change requirement for SubVP

[Description]
In general cases we want to keep the dram clock change requirement (we
prefer configs that support MCLK switch). Only override to false for
SubVP.

Reviewed-by: Jun Lei <Jun.Lei@amd.com>
Acked-by: Rodrigo Siqueira <Rodrigo.Siqueira@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: Extend soc BB capabilitiy

[why]
Some parts are consuming dangerously close to maximum number of states
supported when updating the BB (i.e. 8).

[how]
Change maximum stages from 9 to 20.

Acked-by: Rodrigo Siqueira <Rodrigo.Siqueira@amd.com>
Signed-off-by: Jun Lei <jun.lei@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Maintain consistent mode of operation during encoder assignment

[Why]
While applying a state to hardware, there is a transition period where
the back-end is reset using the old state; then enabled using the new
state.

Generally, the link encoder configuration module queries
stream-to-encoder assignments in either the new or old state based on a
mode variable. During the transition there is a need to query both
states, however toggling this mode variable can lead to incorrect
programming of encoders.

[How]
- Add new function to explicity query stream-to-encoder assignment
in the current state rather than intermittently switch the mode
of operation of the link encoder assignment module.
- Add additional checks for encoder assignment defects.
- Explicitly reset the mode of operation if application of state
to hardware ends prematurely.

Acked-by: Rodrigo Siqueira <Rodrigo.Siqueira@amd.com>
Signed-off-by: Jimmy Kizito <Jimmy.Kizito@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: guard for virtual calling destroy_link_encoders

[Why]:
On power down, virtual dal may try to delete link_encoders by
referencing uninitialized res_pool.

[How]:
Added guard against empty res_pool.

Acked-by: Rodrigo Siqueira <Rodrigo.Siqueira@amd.com>
Signed-off-by: Martin Leung <martin.leung@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: disable timing sync b/w odm halves

Fix for a bug where we would try to timing sync 2 odm halves.

Acked-by: Rodrigo Siqueira <Rodrigo.Siqueira@amd.com>
Signed-off-by: Dmytro Laktyushkin <Dmytro.Laktyushkin@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Add minimal pipe split transition state

[WHY?]
When adding/removing a plane to some configurations, unsupported pipe
programming can occur when moving to a new plane.  Such cases include pipe
split on multi-display, with MPO, and/or ODM.

[HOW?]
Add a safe transistion state that minimizes pipe usage before programming
new configuration. When adding a plane, the current state has the least
pipes required so it is applied without splitting.  This must be applied
prior to updating the plane_state for seamless transition.  When removing a
plane, the new state has the least pieps required so it is applied without
splitting.

Signed-off-by: Rodrigo Siqueira <Rodrigo.Siqueira@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Add SubVP control lock

Signed-off-by: Rodrigo Siqueira <Rodrigo.Siqueira@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Fix stream->link_enc unassigned during stream removal

[Why]
Found when running igt@kms_atomic.

Userspace attempts to do a TEST_COMMIT when 0 streams which calls
dc_remove_stream_from_ctx. This in turn calls link_enc_unassign which
ends up modifying stream->link = NULL directly, causing the global
link_enc to be removed preventing further link activity and future link
validation from passing.

[How]
We take care of link_enc unassignment at the start of
link_enc_cfg_link_encs_assign so this call is no longer necessary.

Fixes global state from being modified while unlocked.

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: Add basic infrastructure for enabling FAMS

We want to enable Firmware Assisted Memory (FAMS) Switching, but first,
we need to add the required code infrastructure in DC before allowing it
in amdgpu_dm.

Signed-off-by: Rodrigo Siqueira <Rodrigo.Siqueira@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Make OPTC3 function accessible to other DCN

[Why]
Newer DCN should use optc3

[How]
Declare optc3 vmin/vmax function in header.

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: Update gpuvm_max_page_table_levels IP param

After some experimental tests, we noticed that we need to set
gpuvm_max_page_table_levels to '4' to meet the hardware requirements.

Signed-off-by: Eric Bernstein <eric.bernstein@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Switch to correct DTO on HDMI

[Why]
For Pixel Rate control, when on HDMI, HDMI DTO should be selected
instead of DP DTO.

[How]
Pass HDMI parameter for HDMI stream, and select correct DTO.

Signed-off-by: Chris Park <chris.park@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Change DET policy for MPO cases

For MPO we want to allocate less than maximum DET for MPO pipes because
we need enogh buffer to move DET back to other pipes when removing an
MPO plane. Also update regular DET allocation to use DET override (DCN32
has an internal policy which driver does not want to use)

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: Use two pixel per container for k1/k2 div

Currently, we check if pixel_encoding is equal to
PIXEL_ENCODING_YCBCR422 to get the k1/k2 div parameters. This commit
changes this logic slightly by checking if two pixels per container are
used.

Signed-off-by: Rodrigo Siqueira <Rodrigo.Siqueira@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Apply ODM 2:1 policy for single display configuration

[Why]
Most of the time, a single display uses the ODM combine. When using
multi-display, we use ODM combine only if it is necessary. These cases
are not flexible enough for us, and we can improve them to take
advantage of our hardware. We want to have more control over the ODM
policy.

[How]
This commit add a new debug flag named
enable_single_display_2to1_odm_policy to control the ODM policy and
another flag named enable_dp_dig_pixel_rate_div_policy to fine control
the ODM combine. This is possible by adding a new "pipe.dest" parameter
that can be set to ODM 2:1 combined if we use a single display. For
dynamic ODM combine, when using DP-DIG, DCN applies K2=2 settings for
ODM combine. Note that this feature affects the following registers:

- timing.pix_clk_100khz -> DP_VID_M, DP_VID_N
- requested_pix_clk_100hz -> DP_DTOn_PHASE
- OTGn_PIXEL_RATE_DIVK2
- DP_PIXEL_PER_CYCLE_PROCESSING_MODE
- DIG_FIFO_OUTPUT_PIXEL_MODE
- DP_VID_N_MUL

Acked-by: Rodrigo Siqueira <Rodrigo.Siqueira@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/display: Add function to set pixels per cycle

Add function to set pixels per cycle in DIG stream encoder

Acked-by: Rodrigo Siqueira <Rodrigo.Siqueira@amd.com>
Signed-off-by: Eric Bernstein <eric.bernstein@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Prepare for new interfaces

[WHY]:
Lut pipeline will be hooked up differently in some asics
need to add new interfaces and missing registers.

[HOW]:
Add missing registers and hook up programming from DPP for pre-blend
lut.

Acked-by: Rodrigo Siqueira <Rodrigo.Siqueira@amd.com>
Signed-off-by: Martin Leung <Martin.Leung@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Add SubVP required code

This commit enables the SubVP feature. To achieve that, we need to:

- Don't force p-state disallow on SubVP (can't block dummy p-state)
- Send calculated watermark to DMCUB for SubVP
- Adjust CAB mode message to PMFW
- Add a proper locking sequence for SubVP
- Various fixes to SubVP static analysis and determining SubVP config
- Currently SubVP not supported with pipe split so merge all pipes
  before setting up SubVp

Reviewed-by: Jun Lei <Jun.Lei@amd.com>
Acked-by: Rodrigo Siqueira <Rodrigo.Siqueira@amd.com>
Acked-by: Alan Liu <HaoPing.Liu@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: Add missing registers for ACP

We are missing some ACP registers/mask value for some specific ASICs.
This commit includes it to those ASICs that support it.

Signed-off-by: Rodrigo Siqueira <Rodrigo.Siqueira@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: skip whole ras bad page framework on sriov

It should not init whole ras bad page framework on sriov guest side
due to it is handled on host side.

Signed-off-by: Stanley.Yang <Stanley.Yang@amd.com>
Reviewed-by: Hawking Zhang <Hawking.Zhang@amd.com>
Reviewed-by: Tao Zhou <tao.zhou1@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: Only send ras feature for gfx block

GFX is the only IP block that RAS TA needs to program
the hardware when receiving enable_feature command.

Changed from V1:
    remove amdgpu_ras_need_send_ras_feature inline function,
    use GFX RAS block check directly.

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

drm/amd/pm: update SMU 13.0.0 driver_if header

And bump the version to 0x2A.

Signed-off-by: Evan Quan <evan.quan@amd.com>
Acked-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdkfd: simplify vm_validate_pt_pd_bos

We don't need to validate and map root PD specially here,
it would be validated and mapped by amdgpu_vm_validate_pt_bos
if it is evicted.

The special case is when turning a GFX VM to a compute VM,
if vm_update_mode changed, we should make sure root PD gets
mapped. So just map root PD after updating vm->update_funcs
in amdgpu_vm_make_compute whether the vm_update_mode changed
or not.

v3:
 - Add some comments suggested by Christian.

v2:
 - Don't rename vm_validate_pt_pd_bos and make it public.

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

drm/amd: Add debug mask for subviewport mclk switch

[Why&How]
Expose a new dc debug mask enum to force a subviewport memory clock switch
to facilitate easy testing.

Signed-off-by: Aurabindo Pillai <aurabindo.pillai@amd.com>
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/amdgpu/display: add missing FP_START/END checks dcn32_clk_mgr.c

Properly handle FP code in dcn32_clk_mgr.c.

Fixes: 265280b99822 ("drm/amd/display: add CLKMGR changes for DCN32/321")
Reviewed-by: Rodrigo Siqueira <Rodrigo.Siqueira@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Fix __nedf2 undefined for 32 bit compilation

When we tried to compile DCN32/321 for 32-bit architecture, we got this
error message:

ERROR: modpost: "__nedf2" [drivers/gpu/drm/amd/amdgpu/amdgpu.ko] undefined!

This commit fixes this issue by rewriting a small part of the
dcn32_build_wm_range_table.

Cc: Aurabindo Pillai <aurabindo.pillai@amd.com>
Cc: Harry Wentland <harry.wentland@amd.com>
Cc: Alex Deucher <alexander.deucher@amd.com>
Cc: Randy Dunlap <rdunlap@infradead.org>
Fixes: 265280b99822 ("drm/amd/display: add CLKMGR changes for DCN32/321")
Acked-by: Randy Dunlap <rdunlap@infradead.org>
Tested-by: Randy Dunlap <rdunlap@infradead.org> # build-tested
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Rodrigo Siqueira <Rodrigo.Siqueira@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Fix __muldf3 undefined for 32 bit compilation

Sometimes when trying to enable some feature, we have to define some
values with educated guesses, but we mark those values as TBD, which
means "To Be Determined". However, the correct way to approach it is by
loading that information from the firmware. Anyway, some of the values
that we were experimenting with caused this issue:

ERROR: modpost: "__muldf3" [drivers/gpu/drm/amd/amdgpu/amdgpu.ko] undefined!

This commit fixes this issue by removing the division by two since it is
harmless in this case.

Cc: Aurabindo Pillai <aurabindo.pillai@amd.com>
Cc: Harry Wentland <harry.wentland@amd.com>
Cc: Alex Deucher <alexander.deucher@amd.com>
Cc: Randy Dunlap <rdunlap@infradead.org>
Fixes: 265280b99822 ("drm/amd/display: add CLKMGR changes for DCN32/321")
Acked-by: Randy Dunlap <rdunlap@infradead.org>
Tested-by: Randy Dunlap <rdunlap@infradead.org> # build-tested
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Rodrigo Siqueira <Rodrigo.Siqueira@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Fix __floatunsidf undefined for 32 bit compilation

When we tried to compile DCN32/321 for 32-bit architecture, we got this
error message:

ERROR: modpost: "__floatunsidf" [drivers/gpu/drm/amd/amdgpu/amdgpu.ko] undefined!

This was caused because we were trying to assign an unsigned int to a
double value which causes issues for 32-bit architecture. This issue can
be fixed by changing the value type.

Cc: Aurabindo Pillai <aurabindo.pillai@amd.com>
Cc: Harry Wentland <harry.wentland@amd.com>
Cc: Alex Deucher <alexander.deucher@amd.com>
Cc: Randy Dunlap <rdunlap@infradead.org>
Fixes: 265280b99822 ("drm/amd/display: add CLKMGR changes for DCN32/321")
Reported-by: Stephen Rothwell <sfr@canb.auug.org.au>
Acked-by: Randy Dunlap <rdunlap@infradead.org>
Tested-by: Randy Dunlap <rdunlap@infradead.org> # build-tested
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Rodrigo Siqueira <Rodrigo.Siqueira@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Fix __umoddi3 undefined for 32 bit compilation

While we tried to build amdgpu on i386, we got this error:

ERROR: modpost: "__umoddi3" [drivers/gpu/drm/amd/amdgpu/amdgpu.ko] undefined!

This commit fixes this issue by replacing the standard module operator
with div_u64_rem.

Cc: Aurabindo Pillai <aurabindo.pillai@amd.com>
Cc: Harry Wentland <harry.wentland@amd.com>
Cc: Alex Deucher <alexander.deucher@amd.com>
Cc: Randy Dunlap <rdunlap@infradead.org>
Fixes: d3dfceb58de5 ("drm/amd/display: Add dependant changes for DCN32/321")
Reported-by: Stephen Rothwell <sfr@canb.auug.org.au>
Acked-by: Randy Dunlap <rdunlap@infradead.org>
Tested-by: Randy Dunlap <rdunlap@infradead.org> # build-tested
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Rodrigo Siqueira <Rodrigo.Siqueira@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdkfd: Bump KFD API version for SMI profiling event

Indicate SMI profiling events available.

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: Asynchronously free smi_client

The synchronize_rcu may take several ms, which noticeably slows down
applications close SMI event handle. Use call_rcu to free client->fifo
and client asynchronously and eliminate the synchronize_rcu call in the
user thread.

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: Add unmap from GPU SMI event

SVM range unmapped from GPUs when range is unmapped from CPU, or with
xnack on from MMU notifier when range is evicted or migrated.

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: Add user queue eviction restore SMI event

Output user queue eviction and restore event. User queue eviction may be
triggered by svm or userptr MMU notifier, TTM eviction, device suspend
and CRIU checkpoint and restore.

User queue restore may be rescheduled if eviction happens again while
restore.

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: Add migration SMI event

For migration start and end event, output timestamp when migration
starts, ends, svm range address and size, GPU id of migration source and
destination and svm range attributes,

Migration trigger could be prefetch, CPU or GPU page fault and TTM
eviction.

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: Add GPU recoverable fault SMI event

Use ktime_get_boottime_ns() as timestamp to correlate with other
APIs. Output timestamp when GPU recoverable fault starts and ends to
recover the fault, if migration happened or only GPU page table is
updated to recover, fault address, if read or write fault.

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: Enable per process SMI event

Process receive event from same process by default. Add a flag to be
able to receive event from all processes, this requires super user
permission.

Event using pid 0 to send the event to all processes, to keep the
default behavior of existing SMI events.

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: Add KFD SMI event IDs and triggers

Define new system management interface event IDs for migration, GPU
recoverable page fault, user queues eviction, restore and unmap from
GPU events and corresponding event triggers, those will be implemented
in the following patches.

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>

Revert "drm/amdgpu/gmc11: avoid cpu accessing registers to flush VM"

This reverts commit 8748de873fedf4d55bdd99bbb738ee7ddf329792
since drv enabled mes to access registers.

Signed-off-by: Jack Xiao <Jack.Xiao@amd.com>
Reviewed-by: Hawking Zhang <Hawking.Zhang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: enable mes to access registers v2

Enable mes to access registers.

v2: squash mes sched ring enablement flag

Signed-off-by: Jack Xiao <Jack.Xiao@amd.com>
Reviewed-by: Hawking Zhang <Hawking.Zhang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu/mes: add mes register access interface

Add mes register access routines:
1. read register
2. write register
3. wait register
4. write and wait register

Signed-off-by: Jack Xiao <Jack.Xiao@amd.com>
Reviewed-by: Hawking Zhang <Hawking.Zhang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu/mes11: add mes11 misc op

Add misc op commands in mes11.

Signed-off-by: Jack Xiao <Jack.Xiao@amd.com>
Reviewed-by: Hawking Zhang <Hawking.Zhang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdkfd: fix cu mask for asics with wgps

GFX10 and up have work group processors (WGP) and WGP mode is the native
compile mode.

KFD and ROCr have no visibility into whether a dispatch is operating
in CU or WGP mode.

Enforce CU masking to be pairwise continguous in enablement and
round robin distribute CUs across the SEs in a pairwise manner to
assume WGP mode at all times.

Signed-off-by: Jonathan Kim <jonathan.kim@amd.com>
Reviewed-by: Felix Kuehling <felix.kuehling@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: add common interface for mes misc op

Add common interface for mes misc op, including accessing register
interface.

Signed-off-by: Jack Xiao <Jack.Xiao@amd.com>
Reviewed-by: Hawking Zhang <Hawking.Zhang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu/mes11: update mes interface for acessing registers

Update MES firmware api for accessing registers.

Signed-off-by: Jack Xiao <Jack.Xiao@amd.com>
Reviewed-by: Hawking Zhang <Hawking.Zhang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: fix documentation warning

Fixes this issue:
drivers/gpu/drm/amd/amdgpu/amdgpu_device.c:5094: warning: expecting prototype for amdgpu_device_gpu_recover_imp(). Prototype was for amdgpu_device_gpu_recover() instead

Fixes: cf727044144d ("drm/amdgpu: Rename amdgpu_device_gpu_recover_imp back to amdgpu_device_gpu_recover")
Reviewed-by: Kent Russell <kent.russell@amd.com>
Reported-by: Stephen Rothwell <sfr@canb.auug.org.au>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Re-org and cleanup the redundant code

[Why]
Redundant if-else cases for repeater and non-repeater checks

[How]
Without changing the core logic, rearranged the code by removing
redundant checks

Signed-off-by: Chandan Vurdigere Nataraj <chandan.vurdigerenataraj@amd.com>
Reviewed-by: Leo Li <sunpeng.li@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: expose additional modifier for DCN32/321

[Why&How]
Some userspace expect a backwards compatible modifier on DCN32/321. For
hardware with num_pipes more than 16, we expose the most efficient
modifier first. As a fall back method, we need to expose slightly inefficient
modifier AMD_FMT_MOD_TILE_GFX9_64K_R_X after the best option.

Also set the number of packers to fixed value as required per hardware
documentation. This value is cached during hardware initialization and
can be read through the base driver.

Signed-off-by: Aurabindo Pillai <aurabindo.pillai@amd.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd: Load TA firmware for DCN321/DCN32

[Why&How]
TA firmware is needed to enable HDCP.

Changes in v2:

Load separate firmware for PSP 13.0.0

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

drm/amd/amd_shared.h: Add missing doc for PP_GFX_DCS_MASK

This symbol is missing documentation:

drivers/gpu/drm/amd/include/amd_shared.h:224: warning: Enum value 'PP_GFX_DCS_MASK' not described in enum 'PP_FEATURE_MASK'

Document it.

Fixes: 680602d6c2d6 ("drm/amd/pm: enable DCS")
Signed-off-by: Mauro Carvalho Chehab <mchehab@kernel.org>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu/amdgpu_dm: fix kernel-doc markups

There are 4 undocumented fields at struct amdgpu_display_manager.

Add documentation for them, fixing those warnings:

drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.h:544: warning: Function parameter or member 'dmub_outbox_params' not described in 'amdgpu_display_manager'
drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.h:544: warning: Function parameter or member 'num_of_edps' not described in 'amdgpu_display_manager'
drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.h:544: warning: Function parameter or member 'disable_hpd_irq' not described in 'amdgpu_display_manager'
drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.h:544: warning: Function parameter or member 'dmub_aux_transfer_done' not described in 'amdgpu_display_manager'
drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.h:544: warning: Function parameter or member 'delayed_hpd_wq' not described in 'amdgpu_display_manager'

Signed-off-by: Mauro Carvalho Chehab <mchehab@kernel.org>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: change to_dal_irq_source_dnc32() storage class specifier to static

sparse reports
drivers/gpu/drm/amd/amdgpu/../display/dc/irq/dcn32/irq_service_dcn32.c:39:20: warning: symbol 'to_dal_irq_source_dcn32' was not declared. Should it be static?

to_dal_irq_source_dnc32() is only referenced in irq_service_dnc32.c, so change its
storage class specifier to static.

Fixes: 0efd4374f6b4 ("drm/amd/display: add dcn32 IRQ changes")
Reviewed-by: Aurabindo Pillai <aurabindo.pillai@amd.com>
Signed-off-by: Tom Rix <trix@redhat.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Remove unused globals FORCE_RATE and FORCE_LANE_COUNT

sparse reports
drivers/gpu/drm/amd/amdgpu/../display/dc/core/dc_link_dp.c:3885:6: warning: symbol 'FORCE_RATE' was not declared. Should it be static?
drivers/gpu/drm/amd/amdgpu/../display/dc/core/dc_link_dp.c:3886:10: warning: symbol 'FORCE_LANE_COUNT' was not declared. Should it be static?

Neither of thse variables is used in dc_link_dp.c.  Reviewing the commit listed in
the fixes tag shows neither was used in the original patch.  So remove them.

Fixes: 265280b99822 ("drm/amd/display: add CLKMGR changes for DCN32/321")
Reviewed-by: Aurabindo Pillai <aurabindo.pillai@amd.com>
Signed-off-by: Tom Rix <trix@redhat.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu/display: drop set but unused variable

No longer used so drop it.

Fixes: ec457f837890 ("drm/amd/display: Drop unnecessary detect link code")
Reported-by: kernel test robot <lkp@intel.com>
Reviewed-by: Harry Wentland <harry.wentland@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: Fix typos in amdgpu_stop_pending_resets

Change amdggpu to amdgpu and pedning to pending

Signed-off-by: Kent Russell <kent.russell@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Removed unused variable ret

Kernel test robot throws below warning ->

drivers/gpu/drm/amd/amdgpu/../display/dc/core/dc_link.c:
In function 'dc_link_reduce_mst_payload':
   drivers/gpu/drm/amd/amdgpu/../display/dc/core/dc_link.c:3782:32:
warning: variable 'ret' set but not used [-Wunused-but-set-variable]
    3782 |         enum act_return_status ret;

Removed the unused ret variable.

Reported-by: kernel test robot <lkp@intel.com>
Signed-off-by: Souptick Joarder (HPE) <jrdr.linux@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

Revert "drm/amdkfd: Free queue after unmap queue success"

This reverts commit ab8529b0cdb271d9b222cbbddb2641f3fca5df8f.

This causes KFDTest KFDMemoryTest.MemoryRegister test failed on gfx9.

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

drm/amdgpu/display/dc: Fix null pointer exception

We observed hard hang due to NULL derefrence This issue is seen after
running system all the time after two or three days

struct dc *dc = plane_state->ctx->dc; Randomly in long run we found
plane_state or plane_state->ctx is found NULL which causes exception.

BUG: kernel NULL pointer dereference, address: 0000000000000000
PF: supervisor read access in kernel mode
PF: error_code(0x0000) - not-present page
PGD 1dc7f2067 P4D 1dc7f2067 PUD 222c75067 PMD 0
Oops: 0000 [#1] SMP NOPTI
CPU: 5 PID: 29855 Comm: kworker/u16:4 ...
...
Workqueue: events_unbound commit_work [drm_kms_helper]
RIP: 0010:dcn10_update_pending_status+0x1f/0xee [amdgpu]
Code: 41 5f c3 0f 1f 44 00 00 b0 01 c3 0f 1f 44 00 00 41 55 41 54 55 53 48 8b 1f 4c 8b af f8 00 00 00 48 8b 83 88 03 00 00 48 85 db <4c> 8b 20 0f 84 bf 00 00 00 48 89 fd 48 8b bf b8 00 00 00 48 8b 07
RSP: 0018:ffff942941997ab8 EFLAGS: 00010286
RAX: 0000000000000000 RBX: ffff8d7fd98d2000 RCX: 0000000000000000
RDX: 0000000000000000 RSI: ffff8d7e3e87c708 RDI: ffff8d7f2d8c0690
RBP: ffff8d7f2d8c0000 R08: ffff942941997a34 R09: 00000000ffffffff
R10: 0000000000005000 R11: 00000000000000f0 R12: ffff8d7f2d8c0690
R13: ffff8d8035a41680 R14: 00000000000186a0 R15: ffff8d7f2d8c1dd8
FS:  0000000000000000(0000) GS:ffff8d8037340000(0000) knlGS:0000000000000000
CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
CR2: 0000000000000000 CR3: 0000000148030000 CR4: 00000000003406e0
Call Trace:
 dc_commit_state+0x6a2/0x7f0 [amdgpu]
 amdgpu_dm_atomic_commit_tail+0x460/0x19bb [amdgpu]

Tested-by: Rodrigo Siqueira <Rodrigo.Siqueira@amd.com>
Reviewed-by: Harry Wentland <harry.wentland@amd.com>
Signed-off-by: Rahul Kumar <rahul.kumar1@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: Follow up change to previous drm scheduler change.

Align refcount behaviour for amdgpu_job embedded HW fence with
classic pointer style HW fences by increasing refcount each
time emit is called so amdgpu code doesn't need to make workarounds
using amdgpu_job.job_run_counter to keep the HW fence refcount balanced.

Also since in the previous patch we resumed setting s_fence->parent to NULL
in drm_sched_stop switch to directly checking if job->hw_fence is
signaled to short circuit reset if already signed.

Signed-off-by: Andrey Grodzovsky <andrey.grodzovsky@amd.com>
Tested-by: Yiqing Yao <yiqing.yao@amd.com>
Acked-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/sched: Partial revert of 'drm/sched: Keep s_fence->parent pointer'

Problem:
This patch caused negative refcount as described in [1] because
for that case parent fence did not signal by the time of drm_sched_stop and hence
kept in pending list the assumption was they will not signal and
so fence was put to account for the s_fence->parent refcount but for
amdgpu which has embedded HW fence (always same parent fence)
drm_sched_fence_release_scheduled was always called and would
still drop the count for parent fence once more. For jobs that
never signaled this imbalance was masked by refcount bug in
amdgpu_fence_driver_clear_job_fences that would not drop
refcount on the fences that were removed from fence drive
fences array (against prevois insertion into the array in
get in amdgpu_fence_emit).

Fix:
Revert this patch and by setting s_job->s_fence->parent to NULL
as before prevent the extra refcount drop in amdgpu when
drm_sched_fence_release_scheduled is called on job release.

Also - align behaviour in drm_sched_resubmit_jobs_ext with that of
drm_sched_main when submitting jobs - take a refcount for the
new parent fence pointer and drop refcount for original kref_init
for new HW fence creation (or fake new HW fence in amdgpu - see next patch).

[1] - https://lore.kernel.org/all/731b7ff1-3cc9-e314-df2a-7c51b76d4db0@amd.com/t/#r00c728fcc069b1276642c325bfa9d82bf8fa21a3

Signed-off-by: Andrey Grodzovsky <andrey.grodzovsky@amd.com>
Tested-by: Yiqing Yao <yiqing.yao@amd.com>
Acked-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: Prevent race between late signaled fences and GPU reset.

Problem:
After we start handling timed out jobs we assume there fences won't be
signaled but we cannot be sure and sometimes they fire late. We need
to prevent concurrent accesses to fence array from
amdgpu_fence_driver_clear_job_fences during GPU reset and amdgpu_fence_process
from a late EOP interrupt.

Fix:
Before accessing fence array in GPU disable EOP interrupt and flush
all pending interrupt handlers for amdgpu device's interrupt line.

v2: Switch from irq_get/put to full enable/disable_irq for amdgpu

Signed-off-by: Andrey Grodzovsky <andrey.grodzovsky@amd.com>
Acked-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: Add put fence in amdgpu_fence_driver_clear_job_fences

This function should drop the fence refcount when it extracts the
fence from the fence array, just as it's done in amdgpu_fence_process.

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

drm/amdgpu: Remove useless amdgpu_display_freesync_ioctl() declaration

Signed-off-by: Leslie Shi <Yuliang.Shi@amd.com>
Reviewed-by: Hawking Zhang <Hawking.Zhang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: add mc wptr addr support for mes

MES requires mc wptr address for usermode queues.
Export bo gart address for mc wptr address.

Signed-off-by: Jack Xiao <Jack.Xiao@amd.com>
Reviewed-by: Hawking Zhang <Hawking.Zhang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display : Log DP link training failure reason

[Why]
Existing logs doesn't print DP LT failure reason

[How]
Update the existing log with DP LT failure reason

Signed-off-by: Chandan Vurdigere Nataraj <chandan.vurdigerenataraj@amd.com>
Reviewed-by: Leo Li <sunpeng.li@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/pm: enable VR0 HOT support for SMU 13.0.0

Enable VR0 Hot support for SMU 13.0.0.

Signed-off-by: Evan Quan <evan.quan@amd.com>
Acked-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: update GFX11 cs settings

Update GFX11 cs related settings.

Signed-off-by: Evan Quan <evan.quan@amd.com>
Acked-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu/display: reduce stack size in dml32_ModeSupportAndSystemConfigurationFull()

Move more stack variable in to dummy vars structure on the heap.

Fixes stack frame size errors:
drivers/gpu/drm/amd/amdgpu/../display/dc/dml/dcn32/display_mode_vba_32.c: In function 'dml32_ModeSupportAndSystemConfigurationFull':
drivers/gpu/drm/amd/amdgpu/../display/dc/dml/dcn32/display_mode_vba_32.c:3833:1: error: the frame size of 2720 bytes is larger than 2048 bytes [-Werror=frame-larger-than=]
 3833 | } // ModeSupportAndSystemConfigurationFull
      | ^

Fixes: dda4fb85e433 ("drm/amd/display: DML changes for DCN32/321")
Cc: Stephen Rothwell <sfr@canb.auug.org.au>
Cc: Aurabindo Pillai <aurabindo.pillai@amd.com>
Cc: Rodrigo Siqueira Jordao <Rodrigo.Siqueira@amd.com>
Reviewed-by: Rodrigo Siqueira <Rodrigo.Siqueira@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

Revert "drm/amd/display: ignore modifiers when checking for format support"

This reverts commit 5089c4a8ebea3c3ad9eedf038dad7098ebc06131.

This breaks validation and enumeration of display capable modifiers.

The early return true means the rest of the validation code never gets
executed, and we need that to enumerate the right modifiers to userspace
for the format.

The modifiers that are in the initial list generated for a plane are the
superset for all formats and we need the proper checks in this function
to filter some of them out for formats with which they're invalid to be
used.

Furthermore, the safety contract here is that we validate the incoming
modifiers to ensure the kernel can handle them and the display hardware
can handle them. This includes e.g. rejecting multi-plane images with DCC.

Note that the legacy swizzle mechanism allows encoding more swizzles, and
at fb creation time we convert them to modifiers and reject those with
no corresponding modifiers. If we are seeing rejections I'm happy to
help define modifiers that correspond to those, or if absolutely needed
implement a fallback path to allow for less strict validation of the
legacy path.

However, I'd like to revert this patch, since any of these is going to
be a significant rework of the patch, and I'd rather not the regression
gets into a release or forgotten in the meantime.

Reviewed-by: Aurabindo Pillai <aurabindo.pillai@amd.com>
Signed-off-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu/display: Fix spelling mistake "supporing" -> "supporting"

There is a spelling mistake in a dml_print message. Fix it.

Reviewed-by: Harry Wentland <harry.wentland@amd.com>
Signed-off-by: Colin Ian King <colin.i.king@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdkfd: Fix spelling mistake "mechanim" -> "mechanism"

There is a spelling mistake in a pr_debug message. Fix it.

Signed-off-by: Colin Ian King <colin.i.king@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

Revert "drm/amdgpu/display: set vblank_disable_immediate for DC"

This reverts commit 92020e81ddbeac351ea4a19bcf01743f32b9c800.

This causes stuttering and timeouts with DMCUB for some users
so revert it until we understand why and safely enable it
to save power.

Bug: https://gitlab.freedesktop.org/drm/amd/-/issues/1887
Acked-by: Harry Wentland <harry.wentland@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Cc: Nicholas Kazlauskas <nicholas.kazlauskas@amd.com>

drm/amdgpu: drop unexpected word 'for' in comments

there is an unexpected word 'for' in the comments that need to be dropped

file - drivers/gpu/drm/amd/amdgpu/amdgpu_ih.c
line - 245

 * position and also advance the position for for Vega10

changed to:

 * position and also advance the position for Vega10

Signed-off-by: Jiang Jian <jiangjian@cdjrlc.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Fix indentation in dcn32_get_vco_frequency_from_reg()

Clang warns:

  drivers/gpu/drm/amd/amdgpu/../display/dc/clk_mgr/dcn32/dcn32_clk_mgr.c:549:4: warning: misleading indentation; statement is not part of the previous 'else' [-Wmisleading-indentation]
                          pll_req = dc_fixpt_from_int(pll_req_reg & clk_mgr->clk_mgr_mask->FbMult_int);
                          ^
  drivers/gpu/drm/amd/amdgpu/../display/dc/clk_mgr/dcn32/dcn32_clk_mgr.c:542:3: note: previous statement is here
                  else
                  ^
  1 warning generated.

Indent this statement to the left, as it was clearly intended to be
called unconditionally, which will fix the warning.

Link: https://github.com/ClangBuiltLinux/linux/issues/1655
Fixes: 3e838f7ccf64 ("drm/amd/display: Get VCO frequency from registers")
Signed-off-by: Nathan Chancellor <nathan@kernel.org>
Reviewed-by: Leo Li <sunpeng.li@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: Update mes_v11_api_def.h

Update MES API to support oversubscription without aggregated doorbell
for usermode queues.

v2: Change oversubscription_no_aggregated_en to is_kfd_process (align
with MES)

Signed-off-by: Graham Sider <Graham.Sider@amd.com>
Acked-by: Felix Kuehling <Felix.Kuehling@amd.com>
Reviewed-by: Jack Xiao <Jack.Xiao@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdkfd: Enable GFX11 usermode queue oversubscription

Starting with GFX11, MES requires wptr BOs to be GTT allocated/mapped to
GART for usermode queues in order to support oversubscription. In the
case that work is submitted to an unmapped queue, MES must have a GART
wptr address to determine whether the queue should be mapped.

This change is accompanied with changes in MES and is applicable for
MES_API_VERSION >= 2.

v3:
- Use amdgpu_vm_bo_lookup_mapping for wptr_bo mapping lookup
- Move wptr_bo refcount increment to amdgpu_amdkfd_map_gtt_bo_to_gart
- Remove list_del_init from amdgpu_amdkfd_map_gtt_bo_to_gart
- Cleanup/fix create_queue wptr_bo error handling
v4:
- Add MES version shift/mask defines to amdgpu_mes.h
- Change version check from MES_VERSION to MES_API_VERSION
- Add check in kfd_ioctl_create_queue before wptr bo pin/GART map to
ensure bo is a single page.

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

drm/amdgpu: Fetch MES scheduler/KIQ versions

Store MES scheduler and MES KIQ version numbers in amdgpu_mes for GFX11.

Signed-off-by: Graham Sider <Graham.Sider@amd.com>
Acked-by: Felix Kuehling <Felix.Kuehling@amd.com>
Reviewed-by: Jack Xiao <Jack.Xiao@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: To flush tlb for MMHUB of RAVEN series

amdgpu: [mmhub0] no-retry page fault (src_id:0 ring:40 vmid:8 pasid:32769, for process test_basic pid 3305 thread test_basic pid 3305)
amdgpu: in page starting at address 0x00007ff990003000 from IH client 0x12 (VMC)
amdgpu: VM_L2_PROTECTION_FAULT_STATUS:0x00840051
amdgpu: Faulty UTCL2 client ID: MP1 (0x0)
amdgpu: MORE_FAULTS: 0x1
amdgpu: WALKER_ERROR: 0x0
amdgpu: PERMISSION_FAULTS: 0x5
amdgpu: MAPPING_ERROR: 0x0
amdgpu: RW: 0x1

When memory is allocated by kfd, no one triggers the tlb flush for MMHUB0.
There is page fault from MMHUB0.

v2:fix indentation
v3:change subject and fix indentation

Signed-off-by: Ruili Ji <ruiliji2@amd.com>
Reviewed-by: Philip Yang <philip.yang@amd.com>
Reviewed-by: Aaron Liu <aaron.liu@amd.com>
Acked-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

gpu/drm/radeon: Fix typo in comments

Remove the repeated word 'and' from comments

Signed-off-by: Jiang Jian <jiangjian@cdjrlc.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/fourcc: fix integer type usage in uapi header

Kernel uapi headers are supposed to use __[us]{8,16,32,64} types defined
by <linux/types.h> as opposed to 'uint32_t' and similar. See [1] for the
relevant discussion about this topic. In this particular case, the usage
of 'uint64_t' escaped headers_check as these macros are not being called
here. However, the following program triggers a compilation error:

  #include <drm/drm_fourcc.h>

  int main()
  {
   unsigned long x = AMD_FMT_MOD_CLEAR(RB);
   return 0;
  }

gcc error:
  drm.c:5:27: error: ‘uint64_t’ undeclared (first use in this function)
      5 |         unsigned long x = AMD_FMT_MOD_CLEAR(RB);
        |                           ^~~~~~~~~~~~~~~~~

This patch changes AMD_FMT_MOD_{SET,CLEAR} macros to use the correct
integer types, which fixes the above issue.

  [1] https://lkml.org/lkml/2019/6/5/18

Fixes: 8ba16d599374 ("drm/fourcc: Add AMD DRM modifiers.")
Signed-off-by: Carlos Llamas <cmllamas@google.com>
Reviewed-by: Simon Ser <contact@emersion.fr>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu/vcn: fix no previous prototype warning

Declare 'static', as the function is not intended to be used
outside of this translation unit.

Fixes: 4ed49c954e35 ("drm/amdgpu/vcn: add unified queue ib test")
Reported-by: kernel test robot <lkp@intel.com>
Reviewed-by: James Zhu <James.Zhu@amd.com>
Signed-off-by: Ruijing Dong <ruijing.dong@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

amdgpu/pm: Fix possible array out-of-bounds if SCLK levels != 2

 [v2]
simplified fix after Lijo's feedback
 removed clocks.num_levels from calculation of loop count
   removed unsafe accesses to shim table freq_values
 retained corner case output only min,now if
   clocks.num_levels == 1 && now > min

 [v1]
added a check to populate and use SCLK shim table freq_values only
   if using dpm_level == AMD_DPM_FORCED_LEVEL_MANUAL or
                         AMD_DPM_FORCED_LEVEL_PERF_DETERMINISM
removed clocks.num_levels from calculation of shim table size
removed unsafe accesses to shim table freq_values
   output gfx_table values if using other dpm levels
added check for freq_match when using freq_values for when now == min_clk

== Test ==
LOGFILE=aldebaran-sclk.test.log
AMDGPU_PCI_ADDR=`lspci -nn | grep "VGA\|Display" | cut -d " " -f 1`
AMDGPU_HWMON=`ls -la /sys/class/hwmon | grep $AMDGPU_PCI_ADDR | awk '{print $9}'`
HWMON_DIR=/sys/class/hwmon/${AMDGPU_HWMON}

lspci -nn | grep "VGA\|Display"  > $LOGFILE
FILES="pp_od_clk_voltage
pp_dpm_sclk"

for f in $FILES
do
  echo === $f === >> $LOGFILE
  cat $HWMON_DIR/device/$f >> $LOGFILE
done
cat $LOGFILE

Signed-off-by: Darren Powell <darren.powell@amd.com>
Reviewed-by: Kenneth Feng <kenneth.feng@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

amdgpu/pm: Fix incorrect variable for size of clocks array

 [v2]
No Changes, added RB
 [v1]
Size of pp_clock_levels_with_latency is PP_MAX_CLOCK_LEVELS, not MAX_NUM_CLOCKS.
Both are currently defined as 16, modifying in case one value is modified in future
Changed code in both arcturus and aldabaran.

Also removed unneeded var count, and used min_t function

Signed-off-by: Darren Powell <darren.powell@amd.com>
Reviewed-by: Lijo Lazar <lijo.lazar@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdkfd: Free queue after unmap queue success

After queue unmap or remove from MES successfully, free queue sysfs
entries, doorbell and remove from queue list. Otherwise, application may
destroy queue again, cause below kernel warning or crash backtrace.

For outstanding queues, either application forget to destroy or failed
to destroy, kfd_process_notifier_release will remove queue sysfs
entries, kfd_process_wq_release will free queue doorbell.

v2: decrement_queue_count for MES queue

 refcount_t: underflow; use-after-free.
 WARNING: CPU: 7 PID: 3053 at lib/refcount.c:28
  Call Trace:
   kobject_put+0xd6/0x1a0
   kfd_procfs_del_queue+0x27/0x30 [amdgpu]
   pqm_destroy_queue+0xeb/0x240 [amdgpu]
   kfd_ioctl_destroy_queue+0x32/0x70 [amdgpu]
   kfd_ioctl+0x27d/0x500 [amdgpu]
   do_syscall_64+0x35/0x80

 WARNING: CPU: 2 PID: 3053 at drivers/gpu/drm/amd/amdgpu/../amdkfd/kfd_device_queue_manager.c:400
  Call Trace:
   deallocate_doorbell.isra.0+0x39/0x40 [amdgpu]
   destroy_queue_cpsch+0xb3/0x270 [amdgpu]
   pqm_destroy_queue+0x108/0x240 [amdgpu]
   kfd_ioctl_destroy_queue+0x32/0x70 [amdgpu]
   kfd_ioctl+0x27d/0x500 [amdgpu]

 general protection fault, probably for non-canonical address
0xdead000000000108:
 Call Trace:
  pqm_destroy_queue+0xf0/0x200 [amdgpu]
  kfd_ioctl_destroy_queue+0x2f/0x60 [amdgpu]
  kfd_ioctl+0x19b/0x600 [amdgpu]

Signed-off-by: Philip Yang <Philip.Yang@amd.com>
Reviewed-by: Graham Sider <Graham.Sider@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdkfd: Add queue to MES if it becomes active

We remove the user queue from MES scheduler to update queue properties.
If the queue becomes active after updating, add the user queue to MES
scheduler, to be able to handle command packet submission.

v2: don't break pqm_set_gws

Signed-off-by: Philip Yang <Philip.Yang@amd.com>
Reviewed-by: Graham Sider <Graham.Sider@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: fix incorrect comparison in DML

[Why&How]
GCC 12 catches the following incorrect comparison in the if arm

drivers/gpu/drm/amd/amdgpu/../dal-dev/dc/dml/dcn32/display_mode_vba_32.c: In function ‘dml32_ModeSupportAndSystemConfigurationFull’:
drivers/gpu/drm/amd/amdgpu/../dal-dev/dc/dml/dcn32/display_mode_vba_32.c:3740:33: error: the comparison will always evaluate as ‘true’ for the address of ‘USRRetrainingSupport’ will never be NULL [-Werror=address]
3740 | || &mode_lib->vba.USRRetrainingSupport[i][j])) {
| ^~
In file included from ./drivers/gpu/drm/amd/amdgpu/../dal-dev/dc/dml/display_mode_lib.h:32,
from ./drivers/gpu/drm/amd/amdgpu/../dal-dev/dc/dc.h:45,
from drivers/gpu/drm/amd/amdgpu/../dal-dev/dc/dml/dcn32/display_mode_vba_32.c:30:
./drivers/gpu/drm/amd/amdgpu/../dal-dev/dc/dml/display_mode_vba.h:1175:14: note: ‘USRRetrainingSupport’ declared here
1175 | bool USRRetrainingSupport[DC__VOLTAGE_STATES][2];
|

Fix this by remove preceding & so that value is compared instead of
address

Fixes: dda4fb85e433 ("drm/amd/display: DML changes for DCN32/321")
Acked-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Aurabindo Pillai <aurabindo.pillai@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>