drm/amd/display: Set enabled to false at start of audio disable

[Why]
In an effort to stop redundant calls to dce110_disable_audio_stream
the audio->enabled flag was added to the audio resource struct. While
this state probably shouldn't have been tracked on the audio struct
itself it still works fine for some sequences.

However, it does not work for cases where we're freeing the audio
resource (such as hotplugs) or when dynamic audio is enabled.

In these cases the pipe_ctx->stream_res.audio = NULL before we can
set audio->enabled = false. The next time we acquire the audio resource
such as on hotplug the audio will not be enabled for the stream since
DC thinks it's still enabled.

Audio state tracking should cover this sequence.

[How]
Set audio->enabled = false at the start as long as we have
pipe_ctx->stream_res.audio.

Signed-off-by: Nicholas Kazlauskas <nicholas.kazlauskas@amd.com>
Reviewed-by: Zhan Liu <Zhan.Liu@amd.com>
Acked-by: Leo Li <sunpeng.li@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu/smu: move fan rpm query into the asic specific code

On vega20, there is an SMU message to query it.  On navi, it's fetched
from the metrics table.

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

drm/amd/powerplay: custom peak clock freq for navi10

v2:
add function smu_default_set_performance_level as default dpm level handler.
change function name smu_set_performance_level to smu_asic_set_performance_level

v1:
1.NAVI10_PEAK_SCLK_XTX    1830 Mhz
2.NAVI10_PEAK_SCLK_XT     1755 Mhz
3.NAVI10_PEAK_SCLK_XL     1625 Mhz

Signed-off-by: Kevin Wang <kevin1.wang@amd.com>
Reviewed-by: Huang Rui <ray.huang@amd.com>
Reviewed-by: Kenneth Feng <kenneth.feng@amd.com>
Reviewed-by: Evan Quan <evan.quan@amd.com>
Reviewed-by: Jack Gui <Jack.Gui@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: handle active dongle port type is DP++ or DP case

[Why]:
Some active dongles have DP++ port and DP port at the same time. Current
code doesn't cover DP++ case and processes as default DVI case, in which
audio is disabled. Because of dual mode, DP case is also treat as DVI case
for the other port.

[How]:
According DP 1.4 spec, add DP++ procedure similar with HDMI case. Also
add None dongle type for DP case.

Signed-off-by: Dale Zhao <dale.zhao@amd.com>
Reviewed-by: Wenjing Liu <wenjing.liu@amd.com>
Acked-by: Leo Li <sunpeng.li@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: do not read link setting if edp not connected

[Why]
Previously assume eDP sink present if connector present. Do not
need to enforce this restriction. Fix issue where driver attempt
to read link setting even though no edp connected.

{How]
Only read link setting after reading connection status.

Signed-off-by: Eric Yang <Eric.Yang2@amd.com>
Reviewed-by: Yongqiang Sun <yongqiang.sun@amd.com>
Acked-by: Leo Li <sunpeng.li@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Increase size of audios array

[Why]
The audios array defined in "struct resource_pool" is only 6 (MAX_PIPES)
but the max number of audio devices (num_audio) is 7. In some projects,
it will run out of audios array.

[How]
Incraese the audios array size to 7.

Signed-off-by: Tai Man <taiman.wong@amd.com>
Reviewed-by: Joshua Aberback <Joshua.Aberback@amd.com>
Acked-by: Leo Li <sunpeng.li@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: drop ASSERT() if eDP panel is not connected

[Why]
For boards that support eDP but do not have a physical eDP
display connected an ASSERT will be thrown. This is not a
critical failure and shouldn't be treated as such.

[How]
Drop the assertion.

Signed-off-by: Zhan Liu <zhan.liu@amd.com>
Reviewed-by: Nicholas Kazlauskas <Nicholas.Kazlauskas@amd.com>
Acked-by: Leo Li <sunpeng.li@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Only enable audio if speaker allocation exists

[Why]

In dm_helpers_parse_edid_caps, there is a corner case where no speakers
can be allocated even though the audio mode count is greater than 0.
Enabling audio when no speaker allocations exists can cause issues in
the video stream.

[How]

Add a check to not enable audio unless one or more speaker allocations
exist (since doing this can cause issues in the video stream).

Signed-off-by: Alvin Lee <alvin.lee2@amd.com>
Reviewed-by: Jun Lei <Jun.Lei@amd.com>
Acked-by: Leo Li <sunpeng.li@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Fix dc_create failure handling and 666 color depths

[Why]
It is possible (but very unlikely) that constructing dc fails
before current_state is created.

We support 666 color depth in some scenarios, but this
isn't handled in get_norm_pix_clk. It uses exactly the
same pixel clock as the 888 case.

[How]
Check for non null current_state before destructing.

Add case for 666 color depth to get_norm_pix_clk to
avoid assertion.

Signed-off-by: Julian Parkin <julian.parkin@amd.com>
Reviewed-by: Charlene Liu <Charlene.Liu@amd.com>
Acked-by: Leo Li <sunpeng.li@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: allocate 4 ddc engines for RV2

[Why]
Driver will create 0, 1, and 2 ddc engines for RV2,
but some platforms used 0, 1, and 3.

[How]
Still allocate 4 ddc engines for RV2.

Signed-off-by: Derek Lai <Derek.Lai@amd.com>
Reviewed-by: Aric Cyr <Aric.Cyr@amd.com>
Acked-by: Leo Li <sunpeng.li@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: put back front end initialization sequence

[Why]
Seamless boot optimization removed proper front end power off sequence.
In driver disable enable case, this causes driver to power gate hubp
and dpp while there is still memory fetching going on, this can cause
invalid memory requests to be generated which will hang data fabric.

[How]
Put back proper front end power off sequence

Signed-off-by: Eric Yang <Eric.Yang2@amd.com>
Reviewed-by: Anthony Koo <Anthony.Koo@amd.com>
Acked-by: Leo Li <sunpeng.li@amd.com>
Acked-by: Tony Cheng <Tony.Cheng@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Wait for flip to complete

[why]
In pipe split issue occurs when we program immediate flip while vsync flip is pending

[how]
Don't program immediate flip until flip is no longer pending

Signed-off-by: Alvin Lee <alvin.lee2@amd.com>
Reviewed-by: Jaehyun Chung <Jaehyun.Chung@amd.com>
Acked-by: Leo Li <sunpeng.li@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Change min_h_sync_width from 8 to 4

[Why]
Some display's hsync width is lower than the minimum dcn20 is set
to support right now. This will cause optc1_validate_timing to fail which
eventually will result in wrong set mode. This was set to 8 as per
HW team's request for no valid reason.

[How]
Changing min_h_sync_width to 4 will let us validate timing for
preffered mode and light up the headset. This change was made
to Vega 10 before for a similar issue.

Signed-off-by: Fatemeh Darbehani <fatemeh.darbehani@amd.com>
Reviewed-by: Joshua Aberback <Joshua.Aberback@amd.com>
Acked-by: Aric Cyr <Aric.Cyr@amd.com>
Acked-by: Leo Li <sunpeng.li@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: use encoder's engine id to find matched free audio device

[Why]
On some platforms, the encoder id 3 is not populated. So the encoders
are not stored in right order as index (id: 0, 1, 2, 4, 5) at pool. This
would cause encoders id 4 & id 5 to fail when finding corresponding
audio device, defaulting to the first available audio device. As result,
we cannot stream audio into two DP ports with encoders id 4 & id 5.

[How]
It need to create enough audio device objects (0 - 5) to perform matching.
Then use encoder engine id to find matched audio device.

Signed-off-by: Tai Man <taiman.wong@amd.com>
Reviewed-by: Charlene Liu <Charlene.Liu@amd.com>
Acked-by: Leo Li <sunpeng.li@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: fix DMCU hang when going into Modern Standby

[why]
When the system is going into suspend, set_backlight gets called
after the eDP got blanked. Since smooth brightness is enabled,
the driver will make a call into the DMCU to ramp the brightness.
The DMCU would try to enable ABM to do so. But since the display is
blanked, this ends up causing ABM1_ACE_DBUF_REG_UPDATE_PENDING to
get stuck at 1, which results in a dead lock in the DMCU firmware.

[how]
Disable brightness ramping when the eDP display is blanked.

Signed-off-by: Zi Yu Liao <ziyu.liao@amd.com>
Reviewed-by: Eric Yang <eric.yang2@amd.com>
Acked-by: Anthony Koo <Anthony.Koo@amd.com>
Acked-by: Leo Li <sunpeng.li@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Disable Audio on reinitialize hardware

[Why]
When we recover from hang, we do not want to skip the audio enable call.

[How]
Disable audio in dc_reinitialize_hardware

Signed-off-by: Alvin Lee <alvin.lee2@amd.com>
Reviewed-by: Jun Lei <Jun.Lei@amd.com>
Acked-by: Leo Li <sunpeng.li@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Read max down spread

[Why]
When launch D10.2, driver will write DPCD 0x107 with 0x00

[How]
Read MAX_DOWNSPREAD (0x0003h) then keep in current
link settings

Signed-off-by: Derek Lai <Derek.Lai@amd.com>
Reviewed-by: Tony Cheng <Tony.Cheng@amd.com>
Acked-by: Leo Li <sunpeng.li@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Check for valid stream_encode

Before accessing it's vtable, check that stream_encoder is non-null.

Signed-off-by: Ilya Bakoulin <Ilya.Bakoulin@amd.com>
Reviewed-by: Eric Bernstein <Eric.Bernstein@amd.com>
Acked-by: Leo Li <sunpeng.li@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Wait for backlight programming completion in set backlight level

[WHY]
Currently we don't wait for blacklight programming completion in DMCU
when setting backlight level. Some sequences such as PSR static screen
event trigger reprogramming requires it to be complete.

[How]
Add generic wait for dmcu command completion in set backlight level.

Signed-off-by: SivapiriyanKumarasamy <sivapiriyan.kumarasamy@amd.com>
Reviewed-by: Anthony Koo <Anthony.Koo@amd.com>
Acked-by: Leo Li <sunpeng.li@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Poll for GPUVM context ready (v2)

[Why]
Hardware docs state that we must wait until the GPUVM context is ready
after programming it.

[How]
Poll until the valid bit of PAGE_TABLE_BASE_ADDR_LO32 is set to 1 after
programming it.

v2: fix include for udelay (Alex)

Signed-off-by: Julian Parkin <julian.parkin@amd.com>
Reviewed-by: Charlene Liu <Charlene.Liu@amd.com>
Acked-by: Leo Li <sunpeng.li@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: wait for the whole frame after global unlock

[why]
The current code will not wait for the entire frame
 after global unlock.
This causes dsc dynamic target bpp update corruption when
there is a surface update immediately happens after this.

[how]
Wait for the entire whole frame after unlock before continuing
the rest of stream and surface update.

Signed-off-by: Wenjing Liu <Wenjing.Liu@amd.com>
Reviewed-by: Jun Lei <Jun.Lei@amd.com>
Acked-by: Leo Li <sunpeng.li@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Copy max_clks_by_state after dce_clk_mgr_construct

[Why]
For DCE110, DCE112 and DCE120 the max_clks_by_state for the clk_mgr are
copied from their respective table before the call to
dce_clk_mgr_construct, but then dce_clk_mgr_construct overwrites
these with the dce80_max_clks_by_state.

[How]
Copy these after we call dce_clk_mgr_construct so we're using the
right tables.

Signed-off-by: Nicholas Kazlauskas <nicholas.kazlauskas@amd.com>
Reviewed-by: David Francis <David.Francis@amd.com>
Acked-by: Leo Li <sunpeng.li@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Clock does not lower in Updateplanes

[why]
We reset the optimized_required in atomic_plane_disable
flag immediately after it is set in atomic_plane_disconnect, causing us to
never have flag set during next flip in UpdatePlanes.

[how]
Optimize directly after each time plane is removed.

Signed-off-by: Murton Liu <murton.liu@amd.com>
Reviewed-by: Tony Cheng <Tony.Cheng@amd.com>
Acked-by: Leo Li <sunpeng.li@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: skip retrain in dc_link_set_preferred_link_settings() if using passive dongle

[Why]
Fixes issue when we have a display connected using a passive
dongle and then emulate over it using a DP connection at 1 x 1.62 Ghz.
System hangs because register bus returns back 0xFFFFFFFF for all
register reads after setting register DIG_BE_CNTL in
dcn10_link_encoder_connect_dig_be_to_fe().  Hang occurs later
when trying to do a register read.

[How]
At the start of the emulation, dc_link_set_preferred_link_settings()
and dp_retrain_link_dp_test() is called, even though it is connected
using a passive dongle.

Add an extra condition in dp_retrain_link_dp_test() to check for
link->dongle_max_pix_clk > 0.  This is the only way we know if the
connection is using passive dongle so we don't retrain DP.

Signed-off-by: Samson Tam <Samson.Tam@amd.com>
Reviewed-by: Jun Lei <Jun.Lei@amd.com>
Acked-by: Leo Li <sunpeng.li@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: swap system aperture high/low

[why]
Currently logical values are swapped in HW, causing
system aperture to be undefined, so VA and PA cannot co-exist

[how]
program values correctly

Signed-off-by: Jun Lei <Jun.Lei@amd.com>
Reviewed-by: Yongqiang Sun <yongqiang.sun@amd.com>
Acked-by: Leo Li <sunpeng.li@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Set one 4:2:0-related PPS field as recommended by DSC spec

[why]
'second_line_offset_adj' was mistakenly left at zero, even though DSC spec
v1.2a recommends setting this field to 512 for 4:2:0.

[how]
Set 'second_line_offset_adj' to 512 for 4:2:0 and leave at zero otherwise

Signed-off-by: Nikola Cornij <nikola.cornij@amd.com>
Reviewed-by: Eric Bernstein <Eric.Bernstein@amd.com>
Acked-by: Leo Li <sunpeng.li@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Set default block_size, even in unexpected cases

We're not expected to enter the default case, but not returning a
default value here is incorrect.

Signed-off-by: Dmytro Laktyushkin <Dmytro.Laktyushkin@amd.com>
Reviewed-by: Eric Bernstein <Eric.Bernstein@amd.com>
Acked-by: Leo Li <sunpeng.li@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: No audio endpoint for Dell MST display

[Why]
There are certain MST displays (i.e. Dell P2715Q)
that although have the MST feature set to off may still
report it is a branch device and a non-zero
value for downstream port present.
This can lead to us incorrectly classifying a
dp dongle connection as being active and
disabling the audio endpoint for the display.

[How]
Modified the placement and
condition used to assign
the is_branch_dev bit.

Signed-off-by: Harmanprit Tatla <harmanprit.tatla@amd.com>
Reviewed-by: Aric Cyr <aric.cyr@amd.com>
Acked-by: Anthony Koo <Anthony.Koo@amd.com>
Acked-by: Leo Li <sunpeng.li@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: cap DCFCLK hardmin to 507 for NV10

[why]
Due to limitation in SMU/PPLIB, it is not possible to know Fmax @ Vmin for DCFCLK.
This causes issues at high display configurations where extra headroom of DCFCLK
can enable P-state switching

[how]
Use existing override logic.  If override not defined, then force
min = 507

Signed-off-by: Jun Lei <Jun.Lei@amd.com>
Reviewed-by: Eric Yang <eric.yang2@amd.com>
Acked-by: Leo Li <sunpeng.li@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: fix dsc disable

A regression caused dsc to never get disabled in certain situations.

Signed-off-by: Dmytro Laktyushkin <Dmytro.Laktyushkin@amd.com>
Reviewed-by: Nikola Cornij <Nikola.Cornij@amd.com>
Acked-by: Leo Li <sunpeng.li@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: fix up HUBBUB hw programming for VM

[why]
Some values were not being converted or bit-shifted properly for
HW registers, causing black screen

[how]
Fix up the values before programming HW

Signed-off-by: Jun Lei <jun.lei@amd.com>
Reviewed-by: Anthony Koo <Anthony.Koo@amd.com>
Acked-by: Leo Li <sunpeng.li@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: initialize p_state to proper value

[why]
On some modes SMU will be in infinite loop state at boot, this is
because driver assumes p_state_support is false, but this is the
opposite of the assumed boot state by SMU.  we optimize away
notifying SMU about no pstate, and so they will get stuck

[how]
when we init clk manager, init pstate to true, so it matches driver load
assumption

Signed-off-by: Jun Lei <Jun.Lei@amd.com>
Reviewed-by: Tony Cheng <Tony.Cheng@amd.com>
Acked-by: Leo Li <sunpeng.li@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: use VCN firmware offset for cache window

Since we are using the signed FW now, and also using PSP firmware loading,
but it's still potential to break driver when loading FW directly
instead of PSP, so we should add offset.

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

drm/ttm: use the same attributes when freeing d_page->vaddr

In function __ttm_dma_alloc_page(), d_page->addr is allocated
by dma_alloc_attrs() but freed with use dma_free_coherent() in
__ttm_dma_free_page().
Use the correct dma_free_attrs() to free d_page->vaddr.

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

drm/amd/powerplay: change sysfs pp_dpm_xxx format for navi10

v2:
set average clock value on level 1 when current clock equal
min or max clock (fine grained dpm support).

the navi10 gfxclk (sclk) support fine grained DPM,
so use level 1 to show current dpm freq in sysfs pp_dpm_xxx

Signed-off-by: Kevin Wang <kevin1.wang@amd.com>
Reviewed-by: Kenneth Feng <kenneth.feng@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: drop ras self test

this function is not needed any more. error injection is
the only way to validate ras but it can't be executed in
amdgpu_ras_init, where gpu is even not initialized

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

drm/amdgpu: only allow error injection to UMC IP block

error injection to other IP blocks (except UMC) will be enabled
until RAS feature stablize on those IP blocks

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

drm/amdgpu: disable GFX RAS by default

GFX RAS has not been stablized yet. disable GFX ras until
it is fully funcitonal.

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

drm/amdgpu: do not create ras debugfs/sysfs node for ASICs that don't have ras ability

driver shouldn't init any ras debugfs/sysfs node for ASICs that don't have ras
hardware ability

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

drm/amd/powerplay: report bootup clock as max supported on dpm disabled

With gfxclk or uclk dpm disabled, it's reasonable to report bootup clock
as the max supported.

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: Default disable GDS for compute VMIDs

The GDS and GWS blocks default to allowing all VMIDs to
access all entries. Graphics VMIDs can handle setting
these limits when the driver launches work. However,
compute workloads under HWS control don't go through the
kernel driver. Instead, HWS firmware should set these
limits when a process is put into a VMID slot.

Disable access to these devices by default by turning off
all mask bits (for OA) and setting BASE=SIZE=0 (for GDS
and GWS) for all compute VMIDs. If a process wants to use
these resources, they can request this from the HWS
firmware (when such capabilities are enabled). HWS will
then handle setting the base and limit for the process when
it is assigned to a VMID.

This will also prevent user kernels from getting 'stuck' in
GWS by accident if they write GWS-using code but HWS
firmware is not set up to handle GWS reset. Until HWS is
enabled to handle GWS properly, all GWS accesses will
MEM_VIOL fault the kernel.

v2: Move initialization outside of SRBM mutex

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

drm/amd/display: init res_pool dccg_ref, dchub_ref with xtalin_freq

[WHY] dc sw clock implementation of navi10 and raven are not exact the
same. dcccg, dchub reference clock initialization is done after dc calls
vbios dispcontroller_init table. for raven family, before
dispcontroller_init is called by dc, the ref clk values are referred
by sw clock implementation and program asic register using wrong
values. this causes dchub pstate error. This need provide valid ref
clk values. for navi10, since dispcontroller_init is not called,
dchubbub_global_timer_enable = 0, hubbub2_get_dchub_ref_freq will
hit aeert. this need remove hubbub2_get_dchub_ref_freq from this
location and move to dcn20_init_hw.

[HOW] for all asic, initialize dccg, dchub ref clk with data from
vbios firmware table by default. for raven asic family, use these data
from vbios, for asic which support sw dccg component, like navi10,
read ref clk by sw dccg functions and update the ref clk.

Signed-off-by: hersen wu <hersenxs.wu@amd.com>
Reviewed-by: Jun Lei <Jun.Lei@amd.com>
Acked-by: Leo Li <sunpeng.li@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu/pm: remove check for pp funcs in freq sysfs handlers

The dpm sensor function already does this for us.  This fixes
the freq*_input files with the new SMU implementation.

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

drm/amd/display: Force uclk to max for every state

Workaround for now to avoid underflow.

The uclk switch time should really be bumped up to 404, but doing so
would expose p-state hang issues for higher bandwidth display
configurations.

Signed-off-by: Nicholas Kazlauskas <nicholas.kazlauskas@amd.com>
Signed-off-by: Leo Li <sunpeng.li@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdkfd: Remove GWS from process during uninit

If we shut down a process without having destroyed its GWS-using
queues, it is possible that GWS BO will still be in the process
BO list during the gpuvm destruction. This list should be empty
at that time, so we should remove the GWS allocation at the
process uninit point if it is still around.

Signed-off-by: Joseph Greathouse <Joseph.Greathouse@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/amd/amdgpu: Fix offset for vmid selection in debugfs interface

The register debugfs interface was using the wrong bitmask for vmid
selection for GFX_CNTL.

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

drm/amd/powerplay: update vega20 driver if to fit latest SMU firmware

Optimization for the socket power calculation is introduced.

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: maintain SMU FW backward compatibility

Do not halt driver loading on if_version mismatch. As our
driver and FWs are backward compatible.

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: correct smu_update_table usage

The interface was used in a confusing way. In profile mode scenario,
the 2nd parameter of the interface was used in a different way from
other scenarios.

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/amd/powerplay: fix deadlock around smu_handle_task V2

As the lock was already held on the entrance to smu_handle_task.

- V2: lock in small granularity

Signed-off-by: Evan Quan <evan.quan@amd.com>
Reviewed-by: Kevin Wang <kevin1.wang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/powerplay: avoid access before allocation

No access before allocation.

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

drm/amd/powerplay: fix memory allocation failure check V2

Fix memory allocation failure check.

- V2: fix one more similar error

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 silent amdgpu_bo_move failures

Under memory pressure, buffer moves between RAM to VRAM  can
fail when there is no GTT space available. In those cases
amdgpu_bo_move falls back to ttm_bo_move_memcpy, which seems to
succeed, although it doesn't really support non-contiguous or
invisible VRAM. This manifests as VM faults with corrupted page
table entries in KFD eviction stress tests.

Print some helpful messages when lack of GTT space is causing buffer
moves to fail. Check that source and destination memory regions are
supported by ttm_bo_move_memcpy before taking that fallback.

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

drm/amdgpu: drop dead header

Not used anymore.

Reviewed-by: Evan Quan <evan.quan@amd.com>
Acked-by: Christian König <christian.koenig@amd.com>
Noticed-by: Dave Airlie <airlied@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/powerplay: Use proper enums in vega20_print_clk_levels

clang warns:

drivers/gpu/drm/amd/amdgpu/../powerplay/vega20_ppt.c:995:39: warning:
implicit conversion from enumeration type 'PPCLK_e' to different
enumeration type 'enum smu_clk_type' [-Wenum-conversion]
                ret = smu_get_current_clk_freq(smu, PPCLK_SOCCLK, &now);
                      ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~^~~~~~~~~~~~~~~~~~~
drivers/gpu/drm/amd/amdgpu/../powerplay/vega20_ppt.c:1016:39: warning:
implicit conversion from enumeration type 'PPCLK_e' to different
enumeration type 'enum smu_clk_type' [-Wenum-conversion]
                ret = smu_get_current_clk_freq(smu, PPCLK_FCLK, &now);
                      ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~^~~~~~~~~~~~~~~~~
drivers/gpu/drm/amd/amdgpu/../powerplay/vega20_ppt.c:1031:39: warning:
implicit conversion from enumeration type 'PPCLK_e' to different
enumeration type 'enum smu_clk_type' [-Wenum-conversion]
                ret = smu_get_current_clk_freq(smu, PPCLK_DCEFCLK, &now);
                      ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~^~~~~~~~~~~~~~~~~~~~

The values are mapped one to one in vega20_get_smu_clk_index so just use
the proper enums here.

Fixes: 096761014227 ("drm/amd/powerplay: support sysfs to get socclk, fclk, dcefclk")
Link: https://github.com/ClangBuiltLinux/linux/issues/587
Reviewed-by: Evan Quan <evan.quan@amd.com>
Signed-off-by: Nathan Chancellor <natechancellor@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu/gfx10: set SH_MEM_CONFIG.INITIAL_INST_PREFETCH

Prefetch mode 0 is not supported and can lead to hangs with certain very
specific code patterns. Set a sound prefetch mode for all VMIDs rather
than forcing all shaders to set the prefetch mode at the beginning.

Reduce code duplication a bit while we're at it. Note that the 64-bit
address mode enum and the retry all enum are both 0, so the only
functional change is in the INITIAL_INST_PREFETCH field.

Signed-off-by: Nicolai Hähnle <nicolai.haehnle@amd.com>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Acked-by: Felix Kuehling <Felix.Kuehling@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/powerplay: enable fw ctf,apcc dfll and gfx ss

enable fw ctf, apcc dfll and gfx ss on navi10.
fw ctf: when the fw ctf is triggered, the gfx and soc power domain
are shut down. fan speed is boosted to the maximum.
gfx ss: hardware feature, sanity check has been done.
apcc dfll: can check the scoreboard in smu fw to confirm if it's enabled.
no need to do further check since the gfx hardware control the frequency once
a pcc signal comes.

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

drm/amdgpu: check kdb_bin_size to exclude kdb loading sequence

The legacy navi10 sos binary will not carry on kdb image. the kdb_start_addr
is actually the start address of sys_drv image and shouldn't be sent to psp
bootloader.

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

drm/amdgpu: Fix unaligned memory copies

When starting a new mm_node, the page_offset becomes 0.

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

drm/amdgpu: replace simple_strtol() by kstrtou32()

The simple_strtol() function is deprecated. kstrto[l,u32]() is
the correct replacement as it can properly handle overflows.

This patch replaces the deprecated simple_strtol() use introduced recently.
As clk is of type uint32_t, we are safe to use kstrtou32().

It is also safe to return zero on string parsing error,
similar to the case of returning zero if buf is empty in parse_clk().

Fixes: bb5a2bdf36a8 ("drm/amdgpu: support dpm level modification under virtualization v3")
Signed-off-by: Wang Xiayang <xywang.sjtu@sjtu.edu.cn>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: remove memset after kzalloc

kzalloc has already zeroed the memory during the allocation.
So memset is unneeded.

Reviewed-by: Emil Velikov <emil.velikov@collabora.com>
Signed-off-by: Fuqian Huang <huangfq.daxian@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/amdgpu: hide #warning for missing DC config

It is annoying to have #warnings that trigger in randconfig
builds like

drivers/gpu/drm/amd/amdgpu/soc15.c:653:3: error: "Enable CONFIG_DRM_AMD_DC for display support on SOC15."
drivers/gpu/drm/amd/amdgpu/nv.c:400:3: error: "Enable CONFIG_DRM_AMD_DC for display support on navi."

Remove these and rely on the users to turn these on.

Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: Fix Vega20 Perf counter for pcie_bw

The perf counter for Vega20 is 108, instead of 104 which it was on all
previous GPUs, so add a check to use the appropriate value.

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

drm/amd/amdgpu: Add missing select_me_pipe_q() for gfx10

The ability to select GFX GRBM me/pipe/queue/vmid was missing from
the gfx10 driver.  This patch adds it.  Used by the debugfs register
interface to select GFX resources when read/writing registers.

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>

drm/amd/amdgpu: Add VMID to SRBM debugfs bank selection

Add 5 bits to the offset for SRBM selection to handle VMIDs.  Also
update the select_me_pipe_q() callback to also select VMID.

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>

drm/amdgpu: extend AMDGPU_CTX_PRIORITY_NORMAL comment

Currently the AMDGPU_CTX_PRIORITY_* defines are used in both
drm_amdgpu_ctx_in::priority and drm_amdgpu_sched_in::priority.

Extend the comment to mention the CAP_SYS_NICE or DRM_MASTER requirement
is only applicable with the former.

Cc: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Cc: Christian König <christian.koenig@amd.com>
Cc: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Emil Velikov <emil.velikov@collabora.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/powerplay: add helper of smu_clk_dpm_is_enabled for smu

v2: change function name to smu_clk_dpm_is_enabled.
add this helper function to check dpm clk feature is enabled.

Signed-off-by: Kevin Wang <kevin1.wang@amd.com>
Reviewed-by: Evan Quan <evan.quan@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/powerplay: fix save dpm level error for smu

the save dpm level should be save previous dpm profile level,
should not modified by get dpm level function.
eg: default auto
1. auto -> standard ==> dpm_level = standard, save_dpm = auto.
2. standard -> auto ==> dpm_level = auto, save_dpm = standard.

Signed-off-by: Kevin Wang <kevin1.wang@amd.com>
Reviewed-by: Evan Quan <evan.quan@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/powerplay: avoid double check feature enabled

the unforce_dpm_levels doesn't need to check feature enablement.
because the smu_get_dpm_freq_range function has check feature logic.

Signed-off-by: Kevin Wang <kevin1.wang@amd.com>
Reviewed-by: Evan Quan <evan.quan@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/powerplay: add standard profile dpm support for smu

1. the standard dpm is not support before.
2. use auto profile to adapt standard profile.

Signed-off-by: Kevin Wang <kevin1.wang@amd.com>
Reviewed-by: Evan Quan <evan.quan@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/powerplay: add socclk profile dpm support.

1.miss socclk profile support when bringup.
2.add feature check for socclk.

Signed-off-by: Kevin Wang <kevin1.wang@amd.com>
Reviewed-by: Evan Quan <evan.quan@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdkfd: Consistently apply noretry setting

Apply the same setting to SH_MEM_CONFIG and VM_CONTEXT1_CNTL. This
makes the noretry param no longer KFD-specific. On GFX10 I'm not
changing SH_MEM_CONFIG in this commit because GFX10 has different
retry behaviour in the SQ and I don't have a way to test it at the
moment.

Suggested-by: Christian König <Christian.Koenig@amd.com>
CC: Philip Yang <Philip.Yang@amd.com>
Signed-off-by: Felix Kuehling <Felix.Kuehling@amd.com>
Reviewed-by : Shaoyun.liu < Shaoyun.liu@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: return 'NULL' instead of 'false' from dcn20_acquire_idle_pipe_for_layer

clang complains that 'false' is a not a pointer:

drivers/gpu/drm/amd/amdgpu/../display/dc/dcn20/dcn20_resource.c:2428:10: error: expression which evaluates to zero treated as a null pointer constant of type 'struct pipe_ctx *' [-Werror,-Wnon-literal-null-conversion]
                return false;

Changing it to 'NULL' looks like the right thing that will shut up
the warning and make it easier to read, while not changing behavior.

Fixes: 7ed4e6352c16 ("drm/amd/display: Add DCN2 HW Sequencer and Resource")
Reviewed-by: Nathan Chancellor <natechancellor@gmail.com>
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Support clang option for stack alignment

As previously fixed for dml in commit 4769278e5c7f ("amdgpu/dc/dml:
Support clang option for stack alignment") and calcs in commit
cc32ad8f559c ("amdgpu/dc/calcs: Support clang option for stack
alignment"), dcn20 uses an option that is not available with clang:

clang: error: unknown argument: '-mpreferred-stack-boundary=4'
scripts/Makefile.build:281: recipe for target 'drivers/gpu/drm/amd/amdgpu/../display/dc/dcn20/dcn20_resource.o' failed

Use the same trick that we have in the other two files.

Fixes: 7ed4e6352c16 ("drm/amd/display: Add DCN2 HW Sequencer and Resource")
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/powerplay: add pstate mclk(uclk) support for navi10

add pstate mclk(uclk) support.

Signed-off-by: Kevin Wang <kevin1.wang@amd.com>
Reviewed-by: Evan Quan <evan.quan@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/powerplay: fix smu clock type change miss error

in the smu module, use the smu_xxxclk type to identify the CLK type
use SMU_SCLK, SMU_MCLK to replace PP_SCLK, PP_MCLK.

Signed-off-by: Kevin Wang <kevin1.wang@amd.com>
Reviewed-by: Feifei Xu <Feifei.Xu@amd.com>
Reviewed-by: Evan Quan <evan.quan@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: support key database loading for navi10

Starting from navi10, driver should send Key Database Load command
to bootloader before loading sys_drv and sos

Signed-off-by: John Clements <John.Clements@amd.com>
Signed-off-by: Hawking Zhang <Hawking.Zhang@amd.com>
Reviewed-by: Xiaojie Yuan <xiaojie.yuan@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: switch to macro for psp bootloader command

The command will be sent to psp bootloader from driver
to ask psp bootloader to exerise tOS, sys_drv and kdb loading

Signed-off-by: Hawking Zhang <Hawking.Zhang@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Reviewed-by: Xiaojie Yuan <xiaojie.yuan@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: Print out voltage in DM_PPLIB

As the clock is already logged, also log the voltage.

Signed-off-by: Paul Menzel <pmenzel@molgen.mpg.de>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Add drm_audio_component support to amdgpu_dm

[Why]
The drm_audio_component can be used to give pin ELD notifications
directly to the sound driver. This fixes audio endpoints disappearing
due to missing unsolicited notifications.

[How]
Send the notification via the audio component whenever we enable or
disable audio state on a stream. This matches what i915 does with
their drm_audio_component and what Takashi Iwai's proposed hack for
radeon/amdpgu did.

This is a bit delayed in when the notification actually occurs, however.
We wait until after all the programming is complete rather than sending
the notification mid sequence.

Particular care is needed for the get ELD callback since it can happen
outside the locking and fencing DRM does for atomic commits.

Cc: Leo Li <sunpeng.li@amd.com>
Cc: Harry Wentland <harry.wentland@amd.com>
Signed-off-by: Nicholas Kazlauskas <nicholas.kazlauskas@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Reviewed-by: Takashi Iwai <tiwai@suse.de>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Expose audio inst from DC to DM

[Why]
In order to give pin notifications to the sound driver from DM we need
to know whether audio is enabled on a stream and what pin it's using
from DC.

[How]
Expose the instance via stream status if it's a mapped resource for
the stream. It will be -1 if there's no audio mapped.

Cc: Leo Li <sunpeng.li@amd.com>
Cc: Harry Wentland <harry.wentland@amd.com>
Signed-off-by: Nicholas Kazlauskas <nicholas.kazlauskas@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdkfd: fix cp hang in eviction

The cp hang occurs in OCL conformance test only on supermicro
platform which has 40 cores and the test generates 40 threads.
The root cause is race condition in non-protected flags.

The fix is to add flags of is_evicted and is_active(init_mqd())
into protected area.

Signed-off-by: Eric Huang <JinhuiEric.Huang@amd.com>
Reviewed-by: Felix Kuehling <Felix.Kuehling@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: enable IP discovery by default on navi

Use the IP discovery table rather than hardcoding the
settings in the driver.

Reviewed-by: Xiaojie Yuan <xiaojie.yuan@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu/discovery: fix DCE_HWIP mapping error in hw_id_map array

ID of DCE_HWIP from vbios is DMU_HWID,
mismatch cause null pointer crash in navi10 modprobe.

Reviewed-by: Feifei Xu <Feifei.Xu@amd.com>
Reviewed-by: Xiaojie Yuan <xiaojie.yuan@amd.com>
Signed-off-by: tiancyin <tianci.yin@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/powerplay: bug fix for sysfs

when we set profile_peak to sysfs:power_dpm_force_performance_level,
we gets the wrong socclk level and mclk level.this patch fix this issue.

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

drm/amd/powerplay: increase the SMU msg response waiting time

This is expected to fix some mode1 reset failures. And this
affects SMU part only as the timeout setting for other parts
is controlled by a different macro.

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: Fix potential integer overflows

With mm_nodes larger than 4GB, byte_count in amdgpu_fill_buffer would
overflow.

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

drm/amdgpu/psp: add a mutex to protect access to the psp ring

We need to serialize access to the psp ring if there are multiple
callers at runtime.

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

drm/amdgpu/navi10: add uclk activity sensor

Query the metrics table for the current uclk activity.

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

drm/amdgpu: properly guard the generic discovery code

It's only available on navi and newer.

Reviewed-by: Xiaojie Yuan <xiaojie.yuan@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: add missing documentation on new module parameters

New parameters added for navi lack documentation.

Reviewed-by: Xiaojie Yuan <xiaojie.yuan@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: don't invalidate caches in RELEASE_MEM, only do the writeback

This RELEASE_MEM use has the Release semantic, which means we should write
back but not invalidate. Invalidations only make sense with the Acquire
semantic (ACQUIRE_MEM), or when RELEASE_MEM is used to do the combined
Acquire-Release semantic, which is a barrier, not a fence.

The undesirable side effect of doing invalidations for the Release semantic
is that it invalidates caches while shaders are running, because the Release
can execute in the middle of the next IB.

UMDs should use ACQUIRE_MEM at the beginning of IBs. Doing cache
invalidations for a fence (like in this case) doesn't do anything
for correctness.

Signed-off-by: Marek Olšák <marek.olsak@amd.com>
Acked-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: avoid 64-bit division

On 32-bit architectures, dividing a 64-bit integer in the kernel
leads to a link error:

ERROR: "__udivdi3" [drivers/gpu/drm/amd/amdgpu/amdgpu.ko] undefined!
ERROR: "__divdi3" [drivers/gpu/drm/amd/amdgpu/amdgpu.ko] undefined!

Change the two recently introduced instances to a multiply+shift
operation that is also much cheaper on 32-bit architectures.
We can do that here, since both of them are really 32-bit numbers
that change a few percent.

Fixes: bedbbe6af4be ("drm/amd/display: Move link functions from dc to dc_link")
Fixes: f18bc4e53ad6 ("drm/amd/display: update calculated bounding box logic for NV")
Acked-by: Slava Abramov <slava.abramov@amd.com>
Tested-by: Slava Abramov <slava.abramov@amd.com>
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu/psp11: simplify the ucode register logic

Split it between navi10 and newer and everything before
navi10.

Reviewed-by: Xiaojie Yuan <xiaojie.yuan@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: properly guard DC support in navi code

Need to add appropriate ifdef.

Acked-by: Leo Li <sunpeng.li@amd.com>
Reviewed-by: Nicholas Kazlauskas <nicholas.kazlauskas@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/powerplay: vega20: fix uninitialized variable use

If smu_get_current_rpm() fails, we can't use the output,
as that may be uninitialized:

drivers/gpu/drm/amd/amdgpu/../powerplay/vega20_ppt.c:3023:8: error: variable 'current_rpm' is used uninitialized whenever '?:' condition is false [-Werror,-Wsometimes-uninitialized]
        ret = smu_get_current_rpm(smu, &current_rpm);
              ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
drivers/gpu/drm/amd/amdgpu/../powerplay/inc/amdgpu_smu.h:735:3: note: expanded from macro 'smu_get_current_rpm'
        ((smu)->funcs->get_current_rpm ? (smu)->funcs->get_current_rpm((smu), (speed)) : 0)
         ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~
drivers/gpu/drm/amd/amdgpu/../powerplay/vega20_ppt.c:3024:12: note: uninitialized use occurs here
        percent = current_rpm * 100 / pptable->FanMaximumRpm;
                  ^~~~~~~~~~~
drivers/gpu/drm/amd/amdgpu/../powerplay/vega20_ppt.c:3023:8: note: remove the '?:' if its condition is always true
        ret = smu_get_current_rpm(smu, &current_rpm);
              ^
drivers/gpu/drm/amd/amdgpu/../powerplay/inc/amdgpu_smu.h:735:3: note: expanded from macro 'smu_get_current_rpm'
        ((smu)->funcs->get_current_rpm ? (smu)->funcs->get_current_rpm((smu), (speed)) : 0)
         ^
drivers/gpu/drm/amd/amdgpu/../powerplay/vega20_ppt.c:3020:22: note: initialize the variable 'current_rpm' to silence this warning
        uint32_t current_rpm;

Propagate the error code in that case.

Fixes: ee0db82027ee ("drm/amd/powerplay: move PPTable_t uses into asic level")
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: dcn20: include linux/delay.h

Without this header, we get a compiler error in some configurations:

.../dc/dcn20/dcn20_hwseq.c: In function 'dcn20_hwss_wait_for_blank_complete':
.../dc/dcn20/dcn20_hwseq.c:1493:3: error: implicit declaration of function 'udelay' [-Werror=implicit-function-declaration]

Note: the use of udelay itself may be problematic, as can occupy
the CPU for 200ms in a busy-loop here.

Fixes: 7ed4e6352c16 ("drm/amd/display: Add DCN2 HW Sequencer and Resource")
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

amdgpu: make pmu support optional

When CONFIG_PERF_EVENTS is disabled, we cannot compile the pmu
portion of the amdgpu driver:

drivers/gpu/drm/amd/amdgpu/amdgpu_pmu.c:48:38: error: no member named 'hw' in 'struct perf_event'
        struct hw_perf_event *hwc = &event->hw;
                                     ~~~~~  ^
drivers/gpu/drm/amd/amdgpu/amdgpu_pmu.c:51:13: error: no member named 'attr' in 'struct perf_event'
        if (event->attr.type != event->pmu->type)
            ~~~~~  ^
...

Use conditional compilation for this file.

Fixes: 9c7c85f7ea1f ("drm/amdgpu: add pmu counters")
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/powerplay: Zero initialize current_rpm in vega20_get_fan_speed_percent

clang warns (trimmed for brevity):

drivers/gpu/drm/amd/amdgpu/../powerplay/vega20_ppt.c:3023:8: warning:
variable 'current_rpm' is used uninitialized whenever '?:' condition is
false [-Wsometimes-uninitialized]
        ret = smu_get_current_rpm(smu, &current_rpm);
              ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

smu_get_current_rpm expands to a ternary operator conditional on
smu->funcs->get_current_rpm being not NULL. When this is false,
current_rpm will be uninitialized. Zero initialize current_rpm to
avoid using random stack values if that ever happens.

Fixes: ee0db82027ee ("drm/amd/powerplay: move PPTable_t uses into asic level")
Link: https://github.com/ClangBuiltLinux/linux/issues/588
Signed-off-by: Nathan Chancellor <natechancellor@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/powerplay: Zero initialize freq in smu_v11_0_get_current_clk_freq

clang warns (trimmed for brevity):

drivers/gpu/drm/amd/amdgpu/../powerplay/smu_v11_0.c:1098:10: warning:
variable 'freq' is used uninitialized whenever '?:' condition is false
[-Wsometimes-uninitialized]
                ret =  smu_get_current_clk_freq_by_table(smu, clk_id, &freq);
                       ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

If get_current_clk_freq_by_table is ever NULL, freq will fail to be
properly initialized. Zero initialize it to avoid using uninitialized
stack values.

smu_get_current_clk_freq_by_table expands to a ternary operator
conditional on smu->funcs->get_current_clk_freq_by_table being not NULL.
When this is false, freq will be uninitialized. Zero initialize freq to
avoid using random stack values if that ever happens.

Fixes: e36182490dec ("drm/amd/powerplay: fix dpm freq unit error (10KHz -> Mhz)")
Link: https://github.com/ClangBuiltLinux/linux/issues/585
Signed-off-by: Nathan Chancellor <natechancellor@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>