drm/amdgpu: retire RREG32_IDX/WREG32_IDX

those are not needed anymore

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

drm/amdgpu: retire indirect mmio reg support from cgs

not needed anymore

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

drm/amdgpu: replace indirect mmio access in non-dc code path

all the mmCUR_CONTROL instances are in mmr range and
can be accessd directly by using RREG32/WREG32

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

drm/amdgpu: remove inproper workaround for vega10

the workaround is not needed for soc15 ASICs except
for vega10. it is even not needed with latest vega10
vbios.

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

drm/amd/powerplay: error out on forcing clock setting not supported

For Arcturus, forcing clock to some specific level is not supported
with 54.18 and onwards SMU firmware. As according to firmware team,
they adopt new gfx dpm tuned parameters which can cover all the use
case in a much smooth way. Thus setting through driver interface
is not needed and maybe do a disservice.

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

drm/amdgpu: fix gfx hang during suspend with video playback (v2)

The system will be hang up during S3 suspend because of SMU is pending
for GC not respose the register CP_HQD_ACTIVE access request.This issue
root cause of accessing the GC register under enter GFX CGGPG and can
be fixed by disable GFX CGPG before perform suspend.

v2: Use disable the GFX CGPG instead of RLC safe mode guard.

Signed-off-by: Prike Liang <Prike.Liang@amd.com>
Tested-by: Mengbing Wang <Mengbing.Wang@amd.com>
Reviewed-by: Huang Rui <ray.huang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: add HDCP caps debugfs

Add debugfs to get HDCP capability. This is also useful for
kms_content_protection igt test.

Use:
cat /sys/kernel/debug/dri/0/DP-1/hdcp_sink_capability
cat /sys/kernel/debug/dri/0/HDMI-A-1/hdcp_sink_capability

Signed-off-by: Bhawanpreet Lakha <Bhawanpreet.Lakha@amd.com>
Reviewed-by: Rodrigo Siqueira <Rodrigo.Siqueira@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: query hdcp capability during link detect

[Why]
Query the hdcp caps of a link, it is useful and can be reported to the user

[How]
Create a query function and call it during link detect

Signed-off-by: Bhawanpreet Lakha <Bhawanpreet.Lakha@amd.com>
Reviewed-by: Rodrigo Siqueira <Rodrigo.Siqueira@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: Re-enable FRU check for most models v5

There is at least 1 VG20 DID that does not have an FRU, and trying to read
that will cause a hang. For now, explicitly support reading the FRU for
Arcturus and for the WKS VG20 DIDs, and skip for everything else.
This re-enables serial number reporting for server cards

v2: Add ASIC check
v3: Don't default to true for pre-VG20
v4: Use DID instead of parsing the VBIOS
v5: Sqaush in overflow warning fix

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: 3.2.80

Signed-off-by: Aric Cyr <aric.cyr@amd.com>
Acked-by: Rodrigo Siqueira <Rodrigo.Siqueira@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Make DMCUB bss/data firmware blob optional

[Why]
By moving everything out of .data into the other regions we can drop
the requirement for the second blob and unify it all into the inst/const
blob.

[How]
We need to still support the blob being there and not being there for
backwards compatibility.

Look for the DMCUB metadata section in the end of the inst/const blob
instead of bss/data is missing.

Clear CW2 if we don't have the data blob so we don't hang when
transitioning between data blob/blobless firmwares.

Don't memcpy the blob into CW2 region if it doesn't exist.

Signed-off-by: Nicholas Kazlauskas <nicholas.kazlauskas@amd.com>
Reviewed-by: Tony Cheng <Tony.Cheng@amd.com>
Acked-by: Rodrigo Siqueira <Rodrigo.Siqueira@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: determine USB C DP2 mode only when USB DP Alt is enabled

[why]
When display is connected with a native DP port, DP2 mode register value
is a don't care. Driver mistakenly reduce max supported lane count to 2
lane based on the don't care value.

[how]
Add additional check only if USB C DP alt mode is enabled, we will
determine max lane count supported based on current mode.

Signed-off-by: Wenjing Liu <wenjing.liu@amd.com>
Reviewed-by: Aric Cyr <Aric.Cyr@amd.com>
Acked-by: Rodrigo Siqueira <Rodrigo.Siqueira@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Check for null fclk voltage when parsing clock table

[WHY]
In cases where a clock table is malformed such that fclk entries have
frequencies but not voltages listed, we don't catch the error and set
clocks to 0 instead of using hardcoded values as we should.

[HOW]
Add check for clock tables fclk entry's voltage as well

Signed-off-by: Michael Strauss <michael.strauss@amd.com>
Reviewed-by: Eric Yang <eric.yang2@amd.com>
Acked-by: Rodrigo Siqueira <Rodrigo.Siqueira@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Acknowledge wm_optimized_required

[Why]
If dc->clk_mgr->funcs->are_clock_states_equal is set, then
wm_optimized_required is never checked. In that case, when going from a
higher mode to a lower mode, wm_optimized_required remains true until
the next mode change.

[How]
 - move from else-if to unconditional or

Signed-off-by: Joshua Aberback <joshua.aberback@amd.com>
Reviewed-by: Jun Lei <Jun.Lei@amd.com>
Reviewed-by: Nicholas Kazlauskas <Nicholas.Kazlauskas@amd.com>
Reviewed-by: Yongqiang Sun <yongqiang.sun@amd.com>
Acked-by: Rodrigo Siqueira <Rodrigo.Siqueira@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Update DRAM watermark before checking to update TTU

[Why]
In most cases, DRAM watermark is large enough that the result of the
condition to increase TTU doesn't change after DRAM watermark is
increased. However, there is are cases where the condition fails and
becomes true after DRAM watermark is increased. This results in minTTU <
DRAM watermarks which leads to PSR hang since p-state is requested but
not allowed.

[How]
Check whether to update TTU after DRAM watermark is updated.

Signed-off-by: Jaehyun Chung <jaehyun.chung@amd.com>
Reviewed-by: Jun Lei <Jun.Lei@amd.com>
Acked-by: Rodrigo Siqueira <Rodrigo.Siqueira@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Avoid create MST prop after registration

[Why]
Prop are created at boot stage, and not allowed to create new prop
after device registration.

[How]
Reuse the connector property from SST if exist.

Signed-off-by: Jerry (Fangzhi) Zuo <Jerry.Zuo@amd.com>
Reviewed-by: Hersen Wu <hersenxs.wu@amd.com>
Acked-by: Rodrigo Siqueira <Rodrigo.Siqueira@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Make cursor source translation adjustment optional

[Why]
In some usecases, like tiled display, the stream and plane configuration
can be setup in a way where the caller expects DAL to perform the
clipping, eg:

P0:
src_rect(0, 0, w, h)
dst_rect(0, 0, w, h)

P1:
src_rect(w, 0, w, h)
dst_rect(0, 0, w, h)

Cursor is enabled on both streams with the same position.

This can result in double cursor on tiled display, even though this
behavior is technically correct from the DC interface point of view.

We need a mechanism to control this dynamically.

[How]
This is something that should live in the DM layer based on detection
of the specified configuration but it's not something that we really
have enough information to deal with today.

Add a flag to the cursor position state that specifies whether we
want DC to do the translation or not and make it opt-in and let
the DM decide when to do it.

Signed-off-by: Nicholas Kazlauskas <nicholas.kazlauskas@amd.com>
Reviewed-by: Tony Cheng <Tony.Cheng@amd.com>
Acked-by: Rodrigo Siqueira <Rodrigo.Siqueira@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Revert to old formula in set_vtg_params

[Why]
New formula + cursor change causing underflow
on certain configs

[How]
Rever to old formula

Signed-off-by: Alvin Lee <alvin.lee2@amd.com>
Reviewed-by: Yongqiang Sun <yongqiang.sun@amd.com>
Acked-by: Rodrigo Siqueira <Rodrigo.Siqueira@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Refactor color management to take dm plane state

[Why]
- In amdgpu_dm_update_plane_color_mgmt() it is inconsistent in taking in
dm_crtc_state and dc_plane_state.
- Makes supporting plane-level color management with proper guard more
complicated than necessary.

[How]
Pass in dm_plane_state in place of dc_plane_state in
amdgpu_dm_update_plane_color_mgmt().

Signed-off-by: Stylon Wang <stylon.wang@amd.com>
Reviewed-by: Nicholas Kazlauskas <Nicholas.Kazlauskas@amd.com>
Acked-by: Rodrigo Siqueira <Rodrigo.Siqueira@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: initialize get_max_link_cap

[why]
usb3->usb2 switch system hang.
driver needs to limit the max sink cap based on DP4 mode.

[how]
based on s_dpalt check and DP4 check:
limit the USB-C DPALT DP maximum supported lane count.

Signed-off-by: Charlene Liu <Charlene.Liu@amd.com>
Reviewed-by: Chris Park <Chris.Park@amd.com>
Acked-by: Rodrigo Siqueira <Rodrigo.Siqueira@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: fix dml pipe merge logic

Dml merges mpc/odm combine pipes to do calculations. This merge is
imperfect if there is a viewport overlap. This change saves pre overlap
viewport for dml use.

Signed-off-by: Dmytro Laktyushkin <Dmytro.Laktyushkin@amd.com>
Reviewed-by: Tony Cheng <Tony.Cheng@amd.com>
Reviewed-by: Wesley Chalmers <Wesley.Chalmers@amd.com>
Acked-by: Rodrigo Siqueira <Rodrigo.Siqueira@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Update stream adjust in dc_stream_adjust_vmin_vmax

[Why]
After v_total_min and max are updated in vrr structure, the changes are
not reflected in stream adjust. When these values are read from stream
adjust it does not reflect the actual state of the system.

[How]
Set stream adjust values equal to vrr adjust values after vrr adjust
values are updated.

Signed-off-by: Isabel Zhang <isabel.zhang@amd.com>
Reviewed-by: Alvin Lee <Alvin.Lee2@amd.com>
Acked-by: Rodrigo Siqueira <Rodrigo.Siqueira@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: 3.2.79

Signed-off-by: Aric Cyr <aric.cyr@amd.com>
Acked-by: Rodrigo Siqueira <Rodrigo.Siqueira@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Remove unused defines

[Why]
Defines aren't used. Remove them.

Signed-off-by: Wyatt Wood <wyatt.wood@amd.com>
Reviewed-by: Anthony Koo <Anthony.Koo@amd.com>
Acked-by: Rodrigo Siqueira <Rodrigo.Siqueira@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Calculate scaling ratios on every medium/full update

[Why]
If a plane isn't being actively enabled or disabled then DC won't
always recalculate scaling rects and ratios for the primary plane.

This results in only a partial or corrupted rect being displayed on
the screen instead of scaling to fit the screen.

[How]
Add back the logic to recalculate the scaling rects into
dc_commit_updates_for_stream since this is the expected place to
do it in DC.

This was previously removed a few years ago to fix an underscan issue
but underscan is still functional now with this change - and it should
be, since this is only updating to the latest plane state getting passed
in.

Signed-off-by: Nicholas Kazlauskas <nicholas.kazlauskas@amd.com>
Reviewed-by: Aric Cyr <Aric.Cyr@amd.com>
Acked-by: Rodrigo Siqueira <Rodrigo.Siqueira@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Program viewport when source pos changes for DCN20 hw seq

[Why]
For medium updates that change nothing but the source rect position
the viewport doesn't change on DCN20.

We're missing the check for the position update bit that was there in
the DCN10 hardware sequencer.

[How]
Check the position bit along with the scaling bit like we were doing
with DCN20.

We shouldn't actually hit a case where context != current_state in
our programming/commit model but guard against it anyway since it was
guarded for the other bits.

Signed-off-by: Nicholas Kazlauskas <nicholas.kazlauskas@amd.com>
Reviewed-by: Zhan Liu <Zhan.Liu@amd.com>
Acked-by: Rodrigo Siqueira <Rodrigo.Siqueira@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Fix incorrect cursor pos on scaled primary plane

[Why]
Cursor pos is correctly adjusted from DC side for source rect offset
on DCN ASIC, but only on the overlay.

This is because DM places offsets the cursor for primary planes only
to workaround missing code in DCE for the adjustment we're now correctly
doing in DC for DCN ASIC.

[How]
Drop the adjustment for source rect from the DM side of things and put
the code where it actually belongs - in DC on the pipe level.

This matches what we do for DCN now.

Signed-off-by: Nicholas Kazlauskas <nicholas.kazlauskas@amd.com>
Reviewed-by: Zhan Liu <Zhan.Liu@amd.com>
Acked-by: Rodrigo Siqueira <Rodrigo.Siqueira@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: change default pipe_split policy for DCN1

[Why]
Changing policy to dynamic will allow 4k multi display configs
to be supported at DPM0

Signed-off-by: Eric Yang <Eric.Yang2@amd.com>
Reviewed-by: Tony Cheng <Tony.Cheng@amd.com>
Acked-by: Rodrigo Siqueira <Rodrigo.Siqueira@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Translate cursor position by source rect

[Why]
Cursor is drawn as part of the framebuffer for a plane on AMD hardware.
The cursor position on the framebuffer does not change even if the
source rect viewport for the cursor does. This causes the cursor to be
clipped.

The following IGT tests fail as a result of this issue:

- kms_plane_cursor@pipe-*-viewport-size-*

[How]
Offset cursor position by plane source rect viewport. If the viewport
is unscaled then the cursor is now correctly positioned on any
plane - primary or overlay.

There is still a hardware limitation for dealing with the cursor size
being incorrectly scaled but that's not something we can address.

Add some documentation explaining some of this in the code while we're
at it.

Signed-off-by: Nicholas Kazlauskas <nicholas.kazlauskas@amd.com>
Reviewed-by: Zhan Liu <Zhan.Liu@amd.com>
Acked-by: Rodrigo Siqueira <Rodrigo.Siqueira@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Fix ABM config copy for dmcub

Decouple dmcub config copy from dmcu iram copy.

Signed-off-by: Wyatt Wood <wyatt.wood@amd.com>
Reviewed-by: Anthony Koo <Anthony.Koo@amd.com>
Acked-by: Rodrigo Siqueira <Rodrigo.Siqueira@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: only blank dp stream which will be powered off

[why]
blank all dp stream would impact edp

[how]
only blank the one which will be powered off

Signed-off-by: Xiaodong Yan <Xiaodong.Yan@amd.com>
Reviewed-by: Yongqiang Sun <yongqiang.sun@amd.com>
Acked-by: Rodrigo Siqueira <Rodrigo.Siqueira@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: fix typo

[why]
MOD_HDCP_STATUS_HDCP1_ENABLE_ENCRYPTION and
MOD_HDCP_STATUS_HDCP2_ENABLE_ENCRYPTION were supposed to be
MOD_HDCP_STATUS_HDCP1_ENABLE_ENCRYPTION_FAILURE and
MOD_HDCP_STATUS_HDCP2_ENABLE_ENCRYPTION_FAILURE. Because of this
it always seems like mod_hdcp_hdcp1_enable_encryption
and mod_hdcp_hdcp2_enable_encryption are always passing

[how]
rename the elements to what they were supposed to be called

Signed-off-by: Joseph Gravenor <joseph.gravenor@amd.com>
Reviewed-by: Wenjing Liu <Wenjing.Liu@amd.com>
Acked-by: Rodrigo Siqueira <Rodrigo.Siqueira@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Use config flag to disable dmcu obj creation

[Why]
When dmcub is the default we no longer wish to create the psr and dmcu
objects. Currently a dc debug flag is used to implement this, but these
flags aren't populated until after dcn21_resource_construct is called.
This means the dmcub objects will never be created. Therefore we must
use a dc config flag, which is populated before dc resource construct.

[How]
Add a dc config flag.

Signed-off-by: Wyatt Wood <wyatt.wood@amd.com>
Reviewed-by: Nicholas Kazlauskas <Nicholas.Kazlauskas@amd.com>
Acked-by: Rodrigo Siqueira <Rodrigo.Siqueira@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Check power_down functions exist before calling

[WHY]
The power_down() function was only defined for specific asics and will
crash the system if it is called by an asic with eDP connected that does
not have it defined.

[HOW]
Add a check for the function's existence before calling it.

Signed-off-by: Sung Lee <sung.lee@amd.com>
Reviewed-by: Yongqiang Sun <yongqiang.sun@amd.com>
Acked-by: Rodrigo Siqueira <Rodrigo.Siqueira@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Don't change mpcc tree for medium updates on DCN20 hwseq

[Why]
Overlay planes disappear when the plane's alpha blending mode or global
opacity is modified.

These are considered UPDATE_TYPE_MEDIUM and trigger the update_mpcc path
in the DCN hardware sequencer.

On DCN10 we have an "optimization" to avoid touching the blending tree
on these updates, but this is actually required behavior based on how
update_mpcc is structured.

For full updates we acquire a MPCC for the plane, remove it if it
already exists then reinsert it after with insert_plane.

The call to insert_plane can take an optional mpcc to insert the new one
above to preserve the current blending order. The update_mpcc hwseq
function doesn't do this so the overlay gets sent to the very bottom
of the tree.

[How]
Copy the check over from DCN10 to DCN20. The only time we need to
actually touch the tree really is the full update, so this is also
an optimization on top of the fix.

Fixing the logic for insert_plane is rather simple (cache the bot_mpcc
and pass it to insert_plane) but is a change that impacts most display
usecases.

For now stick with the optimization.

Signed-off-by: Nicholas Kazlauskas <nicholas.kazlauskas@amd.com>
Reviewed-by: Chris Park <Chris.Park@amd.com>
Acked-by: Rodrigo Siqueira <Rodrigo.Siqueira@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: blank dp stream before power off receiver

[why]
power off dp receiver directly cause garbage during hw init

[how]
blank dp stream and then power off receiver

Signed-off-by: Xiaodong Yan <Xiaodong.Yan@amd.com>
Reviewed-by: Tony Cheng <Tony.Cheng@amd.com>
Acked-by: Rodrigo Siqueira <Rodrigo.Siqueira@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Power down hw blocks on boot

[WHY]
On headless boot a DIG may be turned on by VBIOS on RN.  This leads to
display_count being non-zero in hybrid graphics cases leading to SMU
DISPLAY_OFF message not being sent.

[HOW]
Power down hardware on boot
if seamless boot is not occurring
(power_down_display_on_boot == 1)

Signed-off-by: Sung Lee <sung.lee@amd.com>
Reviewed-by: Yongqiang Sun <yongqiang.sun@amd.com>
Acked-by: Rodrigo Siqueira <Rodrigo.Siqueira@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Remove hdcp display state with mst fix

[Why]
Due to previous code changes, displays transition from active to active
and added state immediately, making it redundant to have both display
states. Previous change to fix this caused HDCP to get into a bad state
when monitor is connected to MST hub, this change fixes that issue.

[How]
Change code behavior so when a device is added successfully the state
remains as active and when addition is unsuccessful change state to
inactive. This removes need for added and active state.

Signed-off-by: Isabel Zhang <isabel.zhang@amd.com>
Reviewed-by: Wenjing Liu <Wenjing.Liu@amd.com>
Acked-by: Rodrigo Siqueira <Rodrigo.Siqueira@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Implement abm config table copy to dmcub

[Why]
Driver must pass abm config table to dmub fw. This provides various
parameters for abm functionality.

[How]
There is too much data to be passed in an inbox message, so we must pass
this data using an indirect buffer.  Copy the table to cw7 via x86,
driver copies to fw_state structure.

Signed-off-by: Wyatt Wood <wyatt.wood@amd.com>
Reviewed-by: Nicholas Kazlauskas <Nicholas.Kazlauskas@amd.com>
Acked-by: Rodrigo Siqueira <Rodrigo.Siqueira@amd.com>
Acked-by: Tony Cheng <Tony.Cheng@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Program DSC before enabling link

[why]
Link or DIG BE can't be exposed to a higher stream bandwidth than they
can handle. When DSC is required to fit the stream into the link
bandwidth, DSC has to be programmed before the link is enabled to ensure
this. Without it, intermittent issues such as black screen after S3 or a
hot-plug can be seen with DSC timings like 4k144Hz or 8k60Hz.

[how]
Move DSC programming from before enabling stream to before enabling link

Signed-off-by: Nikola Cornij <nikola.cornij@amd.com>
Reviewed-by: Jun Lei <Jun.Lei@amd.com>
Acked-by: Rodrigo Siqueira <Rodrigo.Siqueira@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: prevent loop from occuring in pipe list

[Why]
If no free pipes are available, acquire_first_split_pipe is called to
get a pipe to use. This call may alter the ordering of the pipes in the
list so that, for example, the tail pipe changes.

If acquire_first_split_pipe returns the tail pipe, we'll have free_pipe
== tail_pipe. What tail_pipe refers to is not the current tail_pipe, but
what was previously the tail pipe - i.e. prior to the call to
acquire_first_split_pipe

The logic that follows will link free_pipe to the tail pipe, referring to
the current tail pipe. However, since tail_pipe is cached from before the
call to acquire_first_split_pipe, the wrong tail pipe will be used, and
it will end up being linked to itself, creating a loop that, if traversed,
will result in a soft hang.

[How]
Do not cache the tail pipe. Instead, check the tail pipe after the call to
acquire_first_split_pipe is made.

Signed-off-by: Josip Pavic <Josip.Pavic@amd.com>
Reviewed-by: Aric Cyr <Aric.Cyr@amd.com>
Acked-by: Rodrigo Siqueira <Rodrigo.Siqueira@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: 3.2.78

Signed-off-by: Aric Cyr <aric.cyr@amd.com>
Acked-by: Rodrigo Siqueira <Rodrigo.Siqueira@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu/sriov add amdgpu_amdkfd_pre_reset in gpu reset

[PATCH 2/2]
kfd_pre_reset will free mem_objs allocated by kfd_gtt_sa_allocate

Without this change, sriov tdr code path will never free those
allocated memories and get memory leak.

Signed-off-by: Jack Zhang <Jack.Zhang1@amd.com>
Reviewed-by: Monk Liu <monk.liu@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdkfd Avoid destroy hqd when GPU is on reset

This reverts commit 5161bba4311f in order to split it into two
different patches, and this will make it easier to understand.

[PATCH 1/2]
porting to gfx10 from
commit 1b0bfcff463f390c40 ("drm/amdgpu: Avoid destroy hqd when GPU is on reset")

Originally, MEC is touched
without GPU initialized first.

Signed-off-by: Jack Zhang <Jack.Zhang1@amd.com>
Reviewed-by: Monk Liu <monk.liu@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: update RAS related dmesg print

prefix RAS error related dmesg print with pci device info

Reviewed-by: Hawking Zhang <Hawking.Zhang@amd.com>
Signed-off-by: John Clements <john.clements@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: resolve mGPU RAS query instability

upon receiving uncorrectable error, query every GPU node for ras errors

Reviewed-by: Hawking Zhang <Hawking.Zhang@amd.com>
Signed-off-by: John Clements <john.clements@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/amdgpu: Correct gfx10's CG sequence

Incorrect CG sequence will cause gfx timedout,
if we keep switching power profile mode
(enter profile mod such as PEAK will disable CG,
exit profile mode EXIT will enable CG)
when run Vulkan test case(case used for test: vkexample).

Signed-off-by: Chengming Gui <Jack.Gui@amd.com>
Reviewed-by: Kenneth Feng <kenneth.feng@amd.com>
Acked-by: Evan Quan <evan.quan@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: add SPM golden settings for Navi12

Add RLC_SPM golden settings

Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Tianci.Yin <tianci.yin@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: add SPM golden settings for Navi14

Add RLC_SPM golden settings

Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Tianci.Yin <tianci.yin@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: add SPM golden settings for Navi10(v2)

Add RLC_SPM golden settings

Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Reviewed-by: Feifei Xu <Feifei.Xu@amd.com>
Signed-off-by: Tianci.Yin <tianci.yin@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: Print UTCL2 client ID on a gpuvm fault

UTCL2 client ID is useful information to get which
UTCL2 client caused the gpuvm fault. Print it out
for debug purpose

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

drm/amd/powerplay: fix a typo

Util -> Until

Fixes: 567c8fc4a0d28b63f ("drm/amd/powerplay: implement the is_dpm_running()")
Signed-off-by: Nirmoy Das <nirmoy.das@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu/vcn: add shared memory restore after wake up from sleep.

VCN shared memory needs restore after wake up during S3 test.

v2: Allocate shared memory saved_bo at sw_init and free it in sw_fini.

Signed-off-by: James Zhu <James.Zhu@amd.com>
Reviewed-by: Feifei Xu <Feifei.Xu@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: Fix oops when pp_funcs is unset in ACPI event

On ARCTURUS and RENOIR, powerplay is not supported yet.
When plug in or unplug power jack, ACPI event will issue.
Then kernel NULL pointer BUG will be triggered.
Check for NULL pointers before calling.

Signed-off-by: Aaron Ma <aaron.ma@canonical.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu/psp: dont warn on missing optional TA's

Replace dev_warn() with dev_info() and note that they are
optional to avoid confusing users.

The RAS TAs only exist on server boards and the HDCP and DTM
TAs only exist on client boards.  They are optional either way.

Acked-by: Nirmoy Das <nirmoy.das@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: rework sched_list generation

Generate HW IP's sched_list in amdgpu_ring_init() instead of
amdgpu_ctx.c. This makes amdgpu_ctx_init_compute_sched(),
ring.has_high_prio and amdgpu_ctx_init_sched() unnecessary.
This patch also stores sched_list for all HW IPs in one big
array in struct amdgpu_device which makes amdgpu_ctx_init_entity()
much more leaner.

v2:
fix a coding style issue
do not use drm hw_ip const to populate amdgpu_ring_type enum

v3:
remove ctx reference and move sched array and num_sched to a struct
use num_scheds to detect uninitialized scheduler list

v4:
use array_index_nospec for user space controlled variables
fix possible checkpatch.pl warnings

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

drm/amdgpu: sync ring type and drm hw_ip type

Use AMDGPU_HW_IP_* to set amdgpu_ring_type enum values

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

drm/amdgpu/sriov add amdgpu_amdkfd_pre_reset in gpu reset

kfd_pre_reset will free mem_objs allocated by kfd_gtt_sa_allocate

Without this change, sriov tdr code path will never free those allocated
memories and get memory leak.

v2:add a bugfix for kiq ring test fail

Signed-off-by: Jack Zhang <Jack.Zhang1@amd.com>
Reviewed-by: Monk Liu <monk.liu@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: fix the broken logic in dc_link.c

Add missing braces.

Signed-off-by: Yifan Zhang <yifan1.zhang@amd.com>
Reviewed-by: Feifei Xu <Feifei.Xu@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/powerplay: avoid using pm_en before it is initialized

hwmgr->pm_en is initialized at hwmgr_hw_init.
during amdgpu_device_init, there is amdgpu_asic_reset that calls to
pp_get_asic_baco_capability, while hwmgr->pm_en has not yet been initialized.

so avoid using pm_en in pp_get_asic_baco_capability.

Reviewed-by: Emily Deng <Emily.Deng@amd.com>
Signed-off-by: Tiecheng Zhou <Tiecheng.Zhou@amd.com>
Signed-off-by: Yintian Tao <yttao@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/powerplay: implement the is_dpm_running()

As the pmfw hasn't exported the interface of SMU feature
mask to APU SKU so just force on all the features to driver
inquired interface at early initial stage.

Signed-off-by: Prike Liang <Prike.Liang@amd.com>
Reviewed-by: Huang Rui <ray.huang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: re-order asic declarations

Fixes build error of:
"use of undeclared identifier 'RENOIR_A0'"

To fix the same, this patch re-orders the
ASIC declarations accordingly.

Fixes: 41ef3dcd86443fa ("drm/amd/display: Fix RV2 Variant Detection")
Signed-off-by: Shirish S <shirish.s@amd.com>
Reviewed-by: Zhan Liu <zhan.liu@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu/powerplay: using the FCLK DPM table to set the MCLK

1.Using the FCLK DPM table to set the MCLK for DPM states consist of
three entities:
 FCLK
 UCLK
 MEMCLK
All these three clk change together, MEMCLK from FCLK, so use the fclk
frequency.
2.we should show the current working clock freqency from clock table metric

Signed-off-by: Yuxian Dai <Yuxian.Dai@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Reviewed-by: Huang Rui <ray.huang@amd.com>
Reviewed-by: Kevin Wang <Kevin1.Wang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

amdgpu/drm: remove psp access on navi10 for sriov

Navi ASICs don't require to access through PSP to osssys registers.
This on SR-IOV configuration.

Signed-off-by: Alex Sierra <alex.sierra@amd.com>
Reviewed-by: Felix Kuehling <Felix.Kuehling@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Guard calls to hdcp_ta and dtm_ta

[Why]
The buffer used when calling psp is a shared buffer. If we have multiple calls
at the same time we can overwrite the buffer.

[How]
Add mutex to guard the shared buffer.

Signed-off-by: Bhawanpreet Lakha <Bhawanpreet.Lakha@amd.com>
Reviewed-by: Nicholas Kazlauskas <nicholas.kazlauskas@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: remove mod_hdcp_hdcp2_get_link_encryption_status()

It is not being used, so remove it

Signed-off-by: Bhawanpreet Lakha <Bhawanpreet.Lakha@amd.com>
Reviewed-by: Nicholas Kazlauskas <nicholas.kazlauskas@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/dc: Kill dc_conn_log_hex_linux()

DRM already supports tracing DPCD transactions, there's no reason for
the existence of this function. Also, it prints one byte per-line which
is way too loud. So, just remove it.

Reviewed-by: Nicholas Kazlauskas <nicholas.kazlauskas@amd.com>
Signed-off-by: Lyude Paul <lyude@redhat.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/amdgpu_dm/mst: Remove useless sideband tracing

We already trace DPCD reads/writes on both MST and SST, there's no
reason to have this code here (plus, toggling these things with a
define at the top of the file isn't how we do things in the kernel).

Reviewed-by: Nicholas Kazlauskas <nicholas.kazlauskas@amd.com>
Signed-off-by: Lyude Paul <lyude@redhat.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu/vcn: fix spelling mistake "fimware" -> "firmware"

There is a spelling mistake in a dev_err error message. Fix it.

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

drm/amdgpu: fix and cleanup amdgpu_gem_object_close v4

The problem is that we can't add the clear fence to the BO
when there is an exclusive fence on it since we can't
guarantee the the clear fence will complete after the
exclusive one.

To fix this refactor the function and also add the exclusive
fence as shared to the resv object.

v2: fix warning
v3: add excl fence as shared instead
v4: squash in fix for fence handling in amdgpu_gem_object_close

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

drm/amd/display: code cleanup of dc_link file on func dc_link_construct

Removes codestyle issues in dc_link file, on dc_link_construct and
translate_encoder_to_transmitter as suggested by checkpatch.pl.

Types covered:

CHECK: Lines should not end with a '('
WARNING: Missing a blank line after declarations
CHECK: Alignment should match open parenthesis
CHECK: Comparison to NULL could be written
CHECK: Logical continuations should be on the previous line
CHECK: Blank lines aren't necessary after an open brace '{'

Signed-off-by: Melissa Wen <melissa.srw@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: code cleanup on dc_link from is_same_edid to get_ddc_line

Removes codestyle issues on the file dc_link between is_same_edid and
get_ddc_line as suggested by checkpatch.pl.

Types covered:

CHECK: Blank lines aren't necessary after an open brace '{'
CHECK: Blank lines aren't necessary before a close brace '}'
WARNING: braces {} are not necessary for single statement blocks
CHECK: Comparison to NULL could be written
CHECK: Lines should not end with a '('
CHECK: Alignment should match open parenthesis
CHECK: Using comparison to false is error prone
CHECK: Using comparison to true is error prone
WARNING: Avoid multiple line dereference - prefer 'link->dpcd_caps.sink_count.bits.SINK_COUNT'
CHECK: Unnecessary parentheses around
WARNING: Missing a blank line after declarations

Signed-off-by: Melissa Wen <melissa.srw@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: codestyle cleanup on dc_link file until detect_dp func

Removes codestyle issues on the file dc_link until detect_dp func as
suggested by checkpatch.pl.

Types covered:

CHECK: Please don't use multiple blank lines
CHECK: Comparison to NULL could be written
ERROR: space required before the open parenthesis '('
CHECK: Alignment should match open parenthesis
CHECK: Lines should not end with a '('
WARNING: please, no space before tabs
WARNING: Comparisons should place the constant on the right side of the test
WARNING: braces {} are not necessary for single statement blocks
CHECK: Please don't use multiple blank lines

Signed-off-by: Melissa Wen <melissa.srw@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: cleanup codestyle type BLOCK_COMMENT_STYLE on dc_link

Solve comments alignment problems on dc_link file

Signed-off-by: Melissa Wen <melissa.srw@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: enable VCN2.5 DPG mode for Arcturus

Enable VCN2.5 DPG mode for arcturus after below items are applied.
ASD: 0x21000023
SOS: 0x17003B
VCN firmware Version ENC: 1.1 DEC: 1 VEP: 0 Revision: 16
VBIOS: 23

Signed-off-by: James Zhu <James.Zhu@amd.com>
Reviewed-by: Leo Liu <leo.liu@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu/vcn2.5: Add firmware w/r ptr reset sync

Add firmware write/read point reset sync through shared memory

Signed-off-by: James Zhu <James.Zhu@amd.com>
Reviewed-by: Leo Liu <leo.liu@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu/vcn2.0: Add firmware w/r ptr reset sync

Add firmware write/read point reset sync through shared memory

Signed-off-by: James Zhu <James.Zhu@amd.com>
Reviewed-by: Leo Liu <leo.liu@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu/vcn: Add firmware share memory support

Added firmware share memory support for VCN. Current multiple
queue mode is enabled only.

Signed-off-by: James Zhu <James.Zhu@amd.com>
Reviewed-by: Leo Liu <leo.liu@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu/vcn2.5: stall DPG when WPTR/RPTR reset

Add vcn dpg harware synchronization to fix race condition
issue between vcn driver and hardware.

Signed-off-by: James Zhu <James.Zhu@amd.com>
Reviewed-by: Leo Liu <leo.liu@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu/vcn2.0: stall DPG when WPTR/RPTR reset

Add vcn dpg harware synchronization to fix race condition
issue between vcn driver and hardware.

Signed-off-by: James Zhu <James.Zhu@amd.com>
Reviewed-by: Leo Liu <leo.liu@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu/vcn: fix race condition issue for dpg unpause mode switch

Couldn't only rely on enc fence to decide switching to dpg unpaude mode.
Since a enc thread may not schedule a fence in time during multiple
threads running situation.

v3: 1. Rename enc_submission_cnt to dpg_enc_submission_cnt
    2. Add dpg_enc_submission_cnt check in idle_work_handler

v4:  Remove extra counter check, and reduce counter before idle
    work schedule

Signed-off-by: James Zhu <James.Zhu@amd.com>
Reviewed-by: Leo Liu <leo.liu@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu/vcn: fix race condition issue for vcn start

Fix race condition issue when multiple vcn starts are called.

v2: Removed checking the return value of cancel_delayed_work_sync()
to prevent possible races here.

v3: Add total_submission_cnt to avoid gate power unexpectedly.

v4: Remove extra counter check, and reduce counter before idle
work schedule

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

drm/amdgpu: skip access sdma_v5_0 registers under SRIOV (v2)

Due to the new L1.0b0c011b policy, many SDMA registers are blocked which raise
the violation warning. There are total 6 pair register needed to be skipped
when driver init and de-init.
mmSDMA0/1_CNTL
mmSDMA0/1_F32_CNTL
mmSDMA0/1_UTCL1_PAGE
mmSDMA0/1_UTCL1_CNTL
mmSDMA0/1_CHICKEN_BITS,
mmSDMA0/1_SEM_WAIT_FAIL_TIMER_CNTL

v2: squash in warning fix

Signed-off-by: Yintian Tao <yttao@amd.com>
Reviewed-by: Emily Deng <Emily.Deng@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: stop disable the scheduler during HW fini

When we stop the HW for example for GPU reset we should not stop the
front-end scheduler. Otherwise we run into intermediate failures during
command submission.

The scheduler should only be stopped in very few cases:
1. We can't get the hardware working in ring or IB test after a GPU reset.
2. The KIQ scheduler is not used in the front-end and should be disabled during GPU reset.
3. In amdgpu_ring_fini() when the driver unloads.

Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Acked-by: Nirmoy Das <nirmoy.das@amd.com>
Test-by: Dennis Li <dennis.li@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/powerplay: added mutex protection on msg issuing

This could avoid the possible race condition.

Signed-off-by: Evan Quan <evan.quan@amd.com>
Reviewed-by: Kenneth Feng <kenneth.feng@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/powerplay: unified interfaces for message issuing and response checking

This can avoid potential race condition between them.

Signed-off-by: Evan Quan <evan.quan@amd.com>
Reviewed-by: Kenneth Feng <kenneth.feng@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/powerplay: avoid calling Vega20 specific SMU message implemention

Prepare for coming lock protection for SMU message issuing.

Signed-off-by: Evan Quan <evan.quan@amd.com>
Reviewed-by: Kenneth Feng <kenneth.feng@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/powerplay: avoid calling SMU10 specific SMU message implemention

Prepare for coming lock protection for SMU message issuing.

Signed-off-by: Evan Quan <evan.quan@amd.com>
Reviewed-by: Kenneth Feng <kenneth.feng@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/powerplay: avoid calling SMU9 specific SMU message implemention

Prepare for coming lock protection for SMU message issuing.

Signed-off-by: Evan Quan <evan.quan@amd.com>
Reviewed-by: Kenneth Feng <kenneth.feng@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/powerplay: avoid calling SMU8 specific SMU message implemention

Prepare for coming lock protection for SMU message issuing.

Signed-off-by: Evan Quan <evan.quan@amd.com>
Reviewed-by: Kenneth Feng <kenneth.feng@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/powerpaly: drop unused APIs

Drop unused smu7 message APIs.

Signed-off-by: Evan Quan <evan.quan@amd.com>
Reviewed-by: Kenneth Feng <kenneth.feng@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/powerplay: avoid calling SMU7 specific SMU message implemention

Prepare for coming lock protection for SMU message issuing.

Signed-off-by: Evan Quan <evan.quan@amd.com>
Reviewed-by: Kenneth Feng <kenneth.feng@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/powerplay: avoid calling CI specific SMU message implemention

Prepare for coming lock protection for SMU message issuing.

Signed-off-by: Evan Quan <evan.quan@amd.com>
Reviewed-by: Kenneth Feng <kenneth.feng@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: reroute VMC and UMD to IH ring 1 for oss v5

[Why]
Due Page faults can easily overwhelm the interrupt handler.
So to make sure that we never lose valuable interrupts on the primary ring
we re-route page faults to IH ring 1.
It also facilitates the recovery page process, since it's already
running from a process context.
This is valid for Arcturus and future Navi generation GPUs.

[How]
Setting IH_CLIENT_CFG_DATA for VMC and UMD IH clients.

Signed-off-by: Alex Sierra <alex.sierra@amd.com>
Reviewed-by: Felix Kuehling <Felix.Kuehling@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: call psp to program ih cntl in SR-IOV for Navi

call psp to program ih cntl in SR-IOV if supported on Navi and Arcturus.

Signed-off-by: Alex Sierra <alex.sierra@amd.com>
Reviewed-by: Felix Kuehling <Felix.Kuehling@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: enable IH ring 1 and ring 2 for navi

Support added into IH to enable ring1 and ring2 for navi10_ih.

Signed-off-by: Alex Sierra <alex.sierra@amd.com>
Reviewed-by: Felix Kuehling <Felix.Kuehling@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: ih doorbell size of range changed for nbio v7.4

[Why]
nbio v7.4 size of ih doorbell range is 64 bit. This requires 2 DWords per register.

[How]
Change ih doorbell size from 2 to 4. This means two Dwords per ring.
Current configuration uses two ih rings.

Signed-off-by: Alex Sierra <alex.sierra@amd.com>
Acked-by: Alex Deucher <alexander.deucher@amd.com>
Reviewed-by: Felix Kuehling <Felix.Kuehling@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: infinite retries fix from UTLC1 RB SDMA

[Why]
Previously these registers were set to 0. This was causing an
infinite retry on the UTCL1 RB, preventing higher priority RB such as paging RB.

[How]
Set to one the SDMAx_UTLC1_TIMEOUT registers for all SDMAs on Vega10, Vega12,
Vega20 and Arcturus.

Signed-off-by: Alex Sierra <alex.sierra@amd.com>
Acked-by: Alex Deucher <alexander.deucher@amd.com>
Reviewed-by: Felix Kuehling <Felix.Kuehling@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: fix non-pointer dereference for non-RAS supported

Backtrace on gpu recover test on Navi10.

[ 1324.516681] RIP: 0010:amdgpu_ras_set_error_query_ready+0x15/0x20 [amdgpu]
[ 1324.523778] Code: 4c 89 f7 e8 cd a2 a0 d8 e9 99 fe ff ff 45 31 ff e9 91 fe ff ff 0f 1f 44 00 00 55 48 85 ff 48 89 e5 74 0e 48 8b 87 d8 2b 01 00 <40> 88 b0 38 01 00 00 5d c3 66 90 0f 1f 44 00 00 55 31 c0 48 85 ff
[ 1324.543452] RSP: 0018:ffffaa1040e4bd28 EFLAGS: 00010286
[ 1324.549025] RAX: 0000000000000000 RBX: ffff911198b20000 RCX: 0000000000000000
[ 1324.556217] RDX: 00000000000c0a01 RSI: 0000000000000000 RDI: ffff911198b20000
[ 1324.563514] RBP: ffffaa1040e4bd28 R08: 0000000000001000 R09: ffff91119d0028c0
[ 1324.570804] R10: ffffffff9a606b40 R11: 0000000000000000 R12: 0000000000000000
[ 1324.578413] R13: ffffaa1040e4bd70 R14: ffff911198b20000 R15: 0000000000000000
[ 1324.586464] FS:  00007f4441cbf540(0000) GS:ffff91119ed80000(0000) knlGS:0000000000000000
[ 1324.595434] CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
[ 1324.601345] CR2: 0000000000000138 CR3: 00000003fcdf8004 CR4: 00000000003606e0
[ 1324.608694] DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000
[ 1324.616303] DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 0000000000000400
[ 1324.623678] Call Trace:
[ 1324.626270]  amdgpu_device_gpu_recover+0x6e7/0xc50 [amdgpu]
[ 1324.632018]  ? seq_printf+0x4e/0x70
[ 1324.636652]  amdgpu_debugfs_gpu_recover+0x50/0x80 [amdgpu]
[ 1324.643371]  seq_read+0xda/0x420
[ 1324.647601]  full_proxy_read+0x5c/0x90
[ 1324.652426]  __vfs_read+0x1b/0x40
[ 1324.656734]  vfs_read+0x8e/0x130
[ 1324.660981]  ksys_read+0xa7/0xe0
[ 1324.665201]  __x64_sys_read+0x1a/0x20
[ 1324.669907]  do_syscall_64+0x57/0x1c0
[ 1324.674517]  entry_SYSCALL_64_after_hwframe+0x44/0xa9
[ 1324.680654] RIP: 0033:0x7f44417cf081

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

drm/amd/amdgpu: Include headers for PWR and SMUIO registers

Clean up the smu10, smu12, and gfx9 drivers to use headers for
registers instead of hardcoding in the C source files.

Signed-off-by: Tom St Denis <tom.stdenis@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>