drm/amd/display: fix dtm unloading

there was a type in the terminate command.

We should be calling psp_dtm_unload() instead of psp_hdcp_unload()

Fixes: 143f23053333 ("drm/amdgpu: psp DTM init")
Signed-off-by: Bhawanpreet Lakha <Bhawanpreet.Lakha@amd.com>
Reviewed-by: Feifei Xu <Feifei.Xu@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Fix message for encryption

-msg_in is not needed for enabling encryption.
-Use hdcp2_set_encryption instead of hdcp1_enable_encryption for hdcp2.2

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: fix backwards byte order in rx_caps.

We were using incorrect byte order after we started using the drm_defines
So fix it.

Fixes: 02837a91ae75 ("drm/amd/display: add and use defines from drm_hdcp.h")
Signed-off-by: JinZe.Xu <JinZe.Xu@amd.com>
Signed-off-by: Bhawanpreet Lakha <Bhawanpreet.Lakha@amd.com>
Reviewed-by: Wenjing Liu <Wenjing.Liu@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: update HDCP DTM immediately after hardware programming

[why]
HDCP DTM needs to be aware of the upto date display topology
information in order to validate hardware consistency.

[how]
update HDCP DTM on update_stream_config call.

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

drm/amd/display: no hdcp retry if bksv or ksv list is revoked

[why]
According to the specs when bksv or ksv list fails SRM check,
HDCP TX should abort hdcp immediately.
However with the current code HDCP will be reattampt upto 4 times.

[how]
Add the logic that stop HDCP retry if bksv or ksv list
is revoked.

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

drm/amd/display: Handle revoked receivers

[Why]
PSP added a new return code for revoked receivers (SRM). We need to
handle that so we don't retry hdcp

This is already being handled on windows

[How]
Add the enums to psp interface header and handle them.

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/amdgpu/runpm: enable runpm on baco capable VI+ asics

Seems to work reliably on VI+ except for a few so enable runpm barring
those where baco for runtime power management is not supported.

[rajneesh] Picked https://patchwork.freedesktop.org/patch/335402/ to
enable runtime pm with baco for kfd. Also fixed a checkpatch warning and
added extra checks for VEGA20 and ARCTURUS.

Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Rajneesh Bhardwaj <rajneesh.bhardwaj@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdkfd: refactor runtime pm for baco

So far the kfd driver implemented same routines for runtime and system
wide suspend and resume (s2idle or mem). During system wide suspend the
kfd aquires an atomic lock that prevents any more user processes to
create queues and interact with kfd driver and amd gpu. This mechanism
created problem when amdgpu device is runtime suspended with BACO
enabled. Any application that relies on kfd driver fails to load because
the driver reports a locked kfd device since gpu is runtime suspended.

However, in an ideal case, when gpu is runtime  suspended the kfd driver
should be able to:

 - auto resume amdgpu driver whenever a client requests compute service
 - prevent runtime suspend for amdgpu  while kfd is in use

This change refactors the amdgpu and amdkfd drivers to support BACO and
runtime power management.

Reviewed-by: Oak Zeng <oak.zeng@amd.com>
Reviewed-by: Felix Kuehling <felix.kuehling@amd.com>
Signed-off-by: Rajneesh Bhardwaj <rajneesh.bhardwaj@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdkfd: show warning when kfd is locked

During system suspend the kfd driver aquires a lock that prohibits
further kfd actions unless the gpu is resumed. This adds some info which
can be useful while debugging.

Reviewed-by: Oak Zeng <oak.zeng@amd.com>
Reviewed-by: Felix Kuehling <felix.kuehling@amd.com>
Signed-off-by: Rajneesh Bhardwaj <rajneesh.bhardwaj@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: Fix missing error check in suspend

amdgpu_device_suspend might return an error code since it can be called
from both runtime and system suspend flows. Add the missing return code
in case of a failure.

Reviewed-by: Oak Zeng <oak.zeng@amd.com>
Reviewed-by: Felix Kuehling <felix.kuehling@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Rajneesh Bhardwaj <rajneesh.bhardwaj@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Remove set but not unused variable 'stream_status'

drivers/gpu/drm/amd/amdgpu/../display/dc/dcn10/dcn10_hw_sequencer.c:
 In function dcn10_post_unlock_program_front_end:
drivers/gpu/drm/amd/amdgpu/../display/dc/dcn10/dcn10_hw_sequencer.c:2623:29:
 warning: variable stream_status set but not used [-Wunused-but-set-variable]

commit bbf5f6c3f83b ("drm/amd/display: Split program front end part that occur outside lock")
involved this unused variable.

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

drm/amdgpu/powerplay: fix baco check for vega20

We need to handle the runpm case as well as GPU reset.

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

drm/amdgpu/smu: properly handle runpm/suspend/reset

We need some special handling when using baco vs. other
things.

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

drm/amdgpu: add flag for runtime suspend

So we know whether we in are in runtime suspend or
system suspend.

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

drm/amdgpu:/navi10: use the ODCAP enum to index the caps array

Rather than the FEATURE_ID flags.  Avoids a possible reading past
the end of the array.

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

drm/amdgpu: update smu_v11_0_pptable.h

Update to the latest changes.

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

drm/amdgpu: Do not move root PT bo to relocated list

As root PD has no parent, we just need move its status to idle.

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

drm/amdgpu: correct comment to clear up the confusion

Former comment looks to be one intended behavior in code,
actually it's not. So correct it.

Suggested-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Guchun Chen <guchun.chen@amd.com>
Acked-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/powerplay: suppress nonsupport profile mode overrun message

SMU12 not support WORKLOAD_DEFAULT_BIT and WORKLOAD_PPLIB_POWER_SAVING_BIT.

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

drm/amd/powerplay: update smu11_driver_if_navi10.h

To pair the latest SMU firmwares.

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

drm/amd/display: DCN2.x Do not program DPPCLK if same value

[WHY]
Programming DPPCLK to the same value currently set may cause
underflow while playing video in certain conditions.

[HOW]
Only program DPPCLK if clock is not the same as the
previous value programmed.

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: Use fb_base/fb_offset if available for translation

[Why]
In the case where we do have these passed in then we should be using
these instead of the registers - those are only programmed when booting
the adapter as primary.

[How]
Check if fb_base/fb_offset are non-zero. The value of 0 is invalid
for fb_base but is valid for fb_offset on dGPU.

Signed-off-by: Nicholas Kazlauskas <nicholas.kazlauskas@amd.com>
Reviewed-by: Aric Cyr <Aric.Cyr@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: Add psr get_state call

[Why]
Driver needs to know PSR state.

[How]
Use GPINT to call to firmware and get psr state.

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: Don't map ATOM_ENABLE to ATOM_INIT

[Why]
In DCN hardware sequencer we do actually call ATOM_INIT correctly per
pipe. The workaround is not necessary for command table offloading.

[How]
Drop the workaround since it's not needed.

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: add monitor patch to disable SCDC read/write

[why]
customer issue: found that for their specific panel, EDID register space
being overwritten during SCDC read write

[how]
customer accepted HDMI 2 features not working - disabled SCDC read/write
as well as HDMI 2 in general based on monitor patch

Signed-off-by: Martin Leung <martin.leung@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: add odm split logic to scaling calculations

Currently odm scaling calculations are only done when adding initial
odm pipe. Any scaling re-calculations will mess up odm because of this.

This change resolves the problem by updating scaling split logic to
handle odm.

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

drm/amd/display: 3.2.72

Signed-off-by: Anthony Koo <Anthony.Koo@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: Drop unused field from dc_panel_patch

In the struct dc_panel_patch, we have a field named
manage_secondary_link which is not used by any part of the code. This
commit drops the unused field.

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

drm/amd/display: Update register defines

[Why]
Some register defines are redundant or updated

[How]
Update register defines
Remove some redundant defines

Signed-off-by: Anthony Koo <Anthony.Koo@amd.com>
Reviewed-by: Wyatt Wood <Wyatt.Wood@amd.com>
Acked-by: Rodrigo Siqueira <Rodrigo.Siqueira@amd.com>
Acked-by: Sun peng Li <Sunpeng.Li@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Toggle VSR button cause system crash

[Why]
in dcn10_apply_ctx_for_surface(), if we hit !top_pipe_to_program,
pipe_ctx->update_flags could not get cleared.

[How]
Moved clearing pipe_ctx->update_flags logic before if
(!top_pipe_to_program) to gurantee pipe_ctx->update_flags get cleared.

Signed-off-by: Peikang Zhang <peikang.zhang@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: Add backlight support via AUX

Several modern displays support HDR/OLED panel, and one of the
interesting changes provided by this technology is the new approach for
handling backlight. More precisely, in the traditional way, we use PWM
to manage display backlight; however, with HDR/OLED we begin to use DP
Aux channel for adjusting the backlight.

This patch, enable the backlight control in HDR/OLED display via DP Aux.

Signed-off-by: Rodrigo Siqueira <Rodrigo.Siqueira@amd.com>
Reviewed-by: Harry Wentland <Harry.Wentland@amd.com>
Acked-by: Rodrigo Siqueira <Rodrigo.Siqueira@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Add AUX backlight register

Introduce vendor-specific registers for handling backlight via AUX.

Signed-off-by: Rodrigo Siqueira <Rodrigo.Siqueira@amd.com>
Reviewed-by: Nicholas Kazlauskas <Nicholas.Kazlauskas@amd.com>
Acked-by: Harry Wentland <Harry.Wentland@amd.com>
Acked-by: Rodrigo Siqueira <Rodrigo.Siqueira@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Hookup psr set version call

[Why]
Set version is currently not called.

[How]
Add call to set version from copy settings. This
should be the only time this is called as the eDP
caps should never change.

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: Pass override OUI in to dc_init_data

[WHY]
Vendor dongle requires propietary OUI and handshake sequence.

[HOW]
Add a new structure to dc_init_data, to allow creator to pass
an override vendor_oui. This value will be written to DP_SOURCE_OUI
instead of AMD signature, when dpcd_set_source_specific_data is
called.

Signed-off-by: Calvin Hou <Calvin.Hou@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: Remove unused values from psr struct

[Why]
Taking up unnecessary space.

[How]
Remove unused values from psr struct.

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: Don't treat missing command table as failure

[Why]
With the fallback case in place for command table offloading via DMCUB
it's no longer an error for the command table to be missing specific
entries.

[How]
Drop the BREAK_TO_DEBUGGER();

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: always apply T7/T9 delay logic

[why]
before we exit early in edp_reciever_ready if we detect that panel
is not edp or below rev 1.2. This will skip the backlight/t7 delay panel
patch.

[how]
edit logic to ensure panel patch is applied regardless of edp rev.

Signed-off-by: Martin Leung <martin.leung@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/amdgpu/vcn2.5: fix warning

Fix warning during switching to dpg pause mode for
VCN firmware Version ENC: 1.1 DEC: 1 VEP: 0 Revision: 16

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

drm/amdgpu: limit GDS clearing workaround in cold boot sequence

GDS clear workaround will cause gfx failure in suspend/resume case.

[   98.679559] [drm:amdgpu_device_ip_late_init [amdgpu]] *ERROR* late_init of IP block <gfx_v9_0> failed -110
[   98.679561] PM: dpm_run_callback(): pci_pm_resume+0x0/0xa0 returns -110
[   98.679562] PM: Device 0000:03:00.0 failed to resume async: error -110

As this workaround is specific to the HW bug of GDS's ECC error
existing in cold boot up, so bypass this workaround in suspend/
resume case after booting up.

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

drm/amdgpu/sriov set driver_table address in VF

With the recent patch to unify VRAM address for driver
table(a83f82e). VF cannot dump table info any more because
SMU_MSG_SetDriverDramAddrHigh/Low were deleted in the
function of smu_update_table.

Therefore, VF also needs to set driver_table address
in smu_hw_init to fix this regression issue.

Signed-off-by: Jack Zhang <Jack.Zhang1@amd.com>
Acked-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: use amdgpu_device_vram_access in amdgpu_ttm_access_memory v2

Make use of the better performance here as well.

This patch is only compile tested!

v2: fix calculation bug pointed out by Felix

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

drm/amdgpu: use amdgpu_device_vram_access in amdgpu_ttm_vram_read

This speeds up the access quite a bit from 2.2 MB/s to
2.9 MB/s on 32bit and 12,8 MB/s on 64bit.

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

drm/amdgpu: use the BAR if possible in amdgpu_device_vram_access v2

This should speed up debugging VRAM access a lot.

v2: add HDP flush/invalidate

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

drm/amdgpu: optimize amdgpu_device_vram_access a bit.

Only write the _HI register when necessary.

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

drm/amdgpu: fix amdgpu pmu to use hwc->config instead of hwc->conf

hwc->conf was designated specifically for AMD APU IOMMU purposes.  This
could cause problems in performance and/or function since APU IOMMU
implementation is elsewhere.  Also hwc->conf and hwc->config are
different members of an anonymous union so hwc->conf aliases as
hw->last_tag.

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

amdgpu: Prevent build errors regarding soft/hard-float FP ABI tags

On PowerPC, the compiler will tag object files with whether they
use hard or soft float FP ABI and whether they use 64 or 128-bit
long double ABI. On systems with 64-bit long double ABI, a tag
will get emitted whenever a double is used, as on those systems
a long double is the same as a double. This will prevent linkage
as other files are being compiled with hard-float.

On ppc64, this code will never actually get used for the time
being, as the only currently existing hardware using it are the
Renoir APUs. Therefore, until this is testable and can be fixed
properly, at least make sure the build will not fail.

Signed-off-by: Daniel Kolesa <daniel@octaforge.org>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu/vcn2.5: fix DPG mode power off issue on instance 1

Support pause_state for multiple instance, and it will fix vcn2.5 DPG mode
power off issue on instance 1.

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/amd/powerplay: handle features disablement for baco reset in SMU FW

SMU FW will handle the features disablement for baco reset
on Arcturus.

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

drm/amdgpu/sriov Don't send msg when smu suspend

For sriov and pp_onevf_mode, do not send message to set smu
status, because smu doesn't support these messages under VF.

Besides, it should skip smu_suspend when pp_onevf_mode is disabled.

Signed-off-by: Jack Zhang <Jack.Zhang1@amd.com>
Acked-by: Alex Deucher <alexander.deucher@amd.com>
Acked-by: Evan Quan <evan.quan@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Fix wrongly passed static prefix

Currently, gcc spews a warning as:
  drivers/gpu/drm/amd/amdgpu/../display/dc/dcn10/dcn10_hubbub.c: In function ‘hubbub1_verify_allow_pstate_change_high’:
  ./include/drm/drm_print.h:316:2: warning: ‘debug_data’ may be used uninitialized in this function [-Wmaybe-uninitialized]

This is because the code checks against a static value although it's
basically a constant and guaranteed to be set.

This patch changes the type prefix from static to const for addressing
the compile warning above and also for letting the compiler optimize
better.

Fixes: 62d591a8e00c ("drm/amd/display: create new files for hubbub functions")
Reviewed-by: Harry Wentland <harry.wentland@amd.com>
Signed-off-by: Takashi Iwai <tiwai@suse.de>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdkfd: Add queue information to sysfs

Provide compute queues information in sysfs under /sys/class/kfd/kfd/proc.
The format is /sys/class/kfd/kfd/proc/<pid>/queues/<queue id>/XX where
XX are size, type, and gpuid three files to represent queue size, queue
type, and the GPU this queue uses. <queue id> folder and files underneath
are generated when a queue is created. They are removed when the queue is
destroyed.

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

drm/amd/display: turn off the mst hub before we do detection

[why]
not turning off the mst hub before detection on reboot
causes us to not be able to light up displays with mst hook

[how]
on hw init, see if any displays are lit up. if so, turn them off

Signed-off-by: Joseph Gravenor <joseph.gravenor@amd.com>
Reviewed-by: Eric Yang <eric.yang2@amd.com>
Acked-by: Bhawanpreet Lakha <Bhawanpreet.Lakha@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Add initialitions for PLL2 clock source

[Why]
Starting from 14nm, the PLL is built into the PHY and the PLL is mapped
to PHY on 1 to 1 basis. In the code, the DP port is mapped to a PLL that was not
initialized. This causes DP to HDMI dongle to not light up the display.

[How]
Initializations added for PLL2 when creating resources.

Signed-off-by: Isabel Zhang <isabel.zhang@amd.com>
Reviewed-by: Eric Yang <eric.yang2@amd.com>
Acked-by: Bhawanpreet Lakha <Bhawanpreet.Lakha@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: fix workaround for incorrect double buffer register for DLG ADL and TTU

[Why]
these registers should have been double buffered. SW workaround we will have SW program the more aggressive (lower) values
whenever we are upating this register, so we will not have underflow at expense of less optimzal request pattern.

[How]
there is a driver bug where we don't check for 0, which is uninitialzed HW default.  since 0 is smaller than any value we need to program,
driver end up with not programming these registers

Signed-off-by: Tony Cheng <tony.cheng@amd.com>
Reviewed-by: Yongqiang Sun <yongqiang.sun@amd.com>
Acked-by: Bhawanpreet Lakha <Bhawanpreet.Lakha@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: linux enable oled panel support dc part

[Why] old panel has been enabled for window driver but not linux.

[How] enable oled panel support for linux. this patch is dc part.

Signed-off-by: Hersen Wu <hersenxs.wu@amd.com>
Reviewed-by: Harry Wentland <Harry.Wentland@amd.com>
Reviewed-by: Hersen Wu <hersenxs.wu@amd.com>
Acked-by: Bhawanpreet Lakha <Bhawanpreet.Lakha@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: external monitor abm enabled in modern standby

[why]
Resume from modern standby, edp stream disabled
but abm keep enabled. External monitor select OTG
source 0 which ABM enabled.
[how]
Disable abm before disable crtc when reset path
mode not call core link disable stream.

Signed-off-by: Jing Zhou <Jing.Zhou@amd.com>
Reviewed-by: Anthony Koo <Anthony.Koo@amd.com>
Acked-by: Bhawanpreet Lakha <Bhawanpreet.Lakha@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Add DMUB tracebuffer debugfs

[Why]
The DMUB tracebuffer is useful for understanding DMCUB execution state.

[How]
Add a "show" attribute debugfs so we can loop through the buffer
and print the entries.

The structs for the entry format are defined in the debugfs since
the tracebuffer header no longer exists in the DMUB service.

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

drm/amd/display: 3.2.71

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

drm/amd/display: Limit minimum DPPCLK to 100MHz.

[Why]
Underflow is observed when plug in a 4K@60 monitor with
1366x768 eDP due to DPPCLK is too low.

[How]
Limit minimum DPPCLK to 100MHz.

Signed-off-by: Yongqiang Sun <yongqiang.sun@amd.com>
Reviewed-by: Eric Yang <eric.yang2@amd.com>
Acked-by: Bhawanpreet Lakha <Bhawanpreet.Lakha@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Use uint64_t logger_mask instead of uint32_t

[WHY]
enum dc_log_type has more than 32 entries. User cannot set larger entries,
like LOG_DSC. Logs from LOG_GAMMA_DEBUG where being printed even though
flag was not enabled, because dal_logger_should_log check erroneously
passed.

[HOW]
Change struct dal_logger.mask and struct dc_init_data.mask to uint64_t.

Signed-off-by: David Galiffi <David.Galiffi@amd.com>
Reviewed-by: Tony Cheng <Tony.Cheng@amd.com>
Acked-by: Bhawanpreet Lakha <Bhawanpreet.Lakha@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Add wm ranges to clk_mgr

[WHY & HOW]
Having watermark ranges saved inside clk_mgr to be
available for debug at all times would be useful.
Add it to the clk_mgr_internal struct for reference.
Only populated for Renoir, unused for other asics.

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

drm/amd/display: Check hyperV flag in DC.

[Why]
hyperV flag should be passed from dm to DC, and override the
nv12 flip workaround flag.

[How]
Add flag to phy address config struct and pass the value in dm.

Signed-off-by: Yongqiang Sun <yongqiang.sun@amd.com>
Reviewed-by: Tony Cheng <Tony.Cheng@amd.com>
Acked-by: Bhawanpreet Lakha <Bhawanpreet.Lakha@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: fix inputting clk lvl into dml for RN

[Why]
Previous logic is only good for 15W parts. Other configuration
need a smarter logic to match clk levels with pp table in the fuse.

[How]
Cache all 8 DPM level's clock data, find lvl that match each pstate
in the pp table and build input into DML base on that

Signed-off-by: Eric Yang <Eric.Yang2@amd.com>
Signed-off-by: Sung Lee <sung.lee@amd.com>
Reviewed-by: Tony Cheng <Tony.Cheng@amd.com>
Acked-by: Bhawanpreet Lakha <Bhawanpreet.Lakha@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: decouple global lock out of pipe control lock

[why]
hwss should not guess what type of pipe lock is needed.
The caller of the lock function should know
the right type of pipe lock.
Decouple the setup of global lock outside of pipe control lock
logic.

Signed-off-by: Wenjing Liu <Wenjing.Liu@amd.com>
Reviewed-by: Anthony Koo <Anthony.Koo@amd.com>
Acked-by: Bhawanpreet Lakha <Bhawanpreet.Lakha@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Check engine is not NULL before acquiring

[Why]
Engine can be NULL in some cases, so we must not acquire it.

[How]
Check for NULL engine before acquiring.

Signed-off-by: Aric Cyr <aric.cyr@amd.com>
Reviewed-by: Harry Wentland <harry.wentland@amd.com>
Acked-by: Bhawanpreet Lakha <Bhawanpreet.Lakha@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Use dcfclk to populate watermark ranges

[WHY & HOW]
Previously drain clk was unconstrained and fill clk was constrained on fclk.
We want to change it to fill clk unconstrained and drain clock constrained
to dcfclk.

Signed-off-by: Sung Lee <sung.lee@amd.com>
Reviewed-by: Tony Cheng <Tony.Cheng@amd.com>
Acked-by: Bhawanpreet Lakha <Bhawanpreet.Lakha@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: add stream_enc_inst for PSP HDCP inst use

[why]
new HW engine mapping requirment use in PSP
[how]
report stream_enc_inst

Signed-off-by: Charlene Liu <Charlene.Liu@amd.com>
Reviewed-by: Wenjing Liu <Wenjing.Liu@amd.com>
Acked-by: Bhawanpreet Lakha <Bhawanpreet.Lakha@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Update hubbub description comment

Description for DCHUBBUB_TEST_DEBUG_DATA is changed to avoid any future confusions.

Signed-off-by: Peikang Zhang <peikang.zhang@amd.com>
Reviewed-by: Aric Cyr <Aric.Cyr@amd.com>
Acked-by: Bhawanpreet Lakha <Bhawanpreet.Lakha@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: remove early break in interdependent_lock

[Why]
The break in apply_ctx_interdependent_lock() may potentially
lead to early break from the loop leaving update plane unlocked

[How]
Remove break

Signed-off-by: Roman Li <roman.li@amd.com>
Reviewed-by: Anthony Koo <Anthony.Koo@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: dc_get_vmid_use_vector() crashes when get called

[Why]
int i can go out of boundary which will cause crash

[How]
Fixed the maximum value of i to avoid i going out of boundary

Signed-off-by: Peikang Zhang <peikang.zhang@amd.com>
Reviewed-by: Jun Lei <Jun.Lei@amd.com>
Acked-by: Bhawanpreet Lakha <Bhawanpreet.Lakha@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: 3.2.70

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

drm/amd/display: remove unused variable

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

drm/amd/display: Fix GSL acquire

[Why]
After locking refactor GSL is not acquired properly
resulting in immediate flip issues.

[How]
Do not copy old GSL state anymore since GSL is acquired
earlier now.

Signed-off-by: Aric Cyr <aric.cyr@amd.com>
Reviewed-by: Anthony Koo <Anthony.Koo@amd.com>
Acked-by: Bhawanpreet Lakha <Bhawanpreet.Lakha@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Added locking for atomic update stream and update planes

[Why]
Screen flickering when HDR switches between FP16 and ARGB2101010

[How]
Moved pipe_control_lock so stream update and plane update occur atomically

Signed-off-by: Anthony Koo <Anthony.Koo@amd.com>
Signed-off-by: Lucy Li <lucy.li@amd.com>
Reviewed-by: Aric Cyr <Aric.Cyr@amd.com>
Acked-by: Bhawanpreet Lakha <Bhawanpreet.Lakha@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Wait for clean shutdown in DMCUB reset

[Why]
The DMCUB may be currently executing commands when the reset is
triggered.

Before issuing a reset we should first wait for the DMCUB to finish
its work.

[How]
Send the GPINT command for halting the firmware before reset.

Get the ack for the command then wait for the scratch register to
become the correct value.

We want this to take under ~40us or so at most before we force reset
to cover PHY delay sequence max time.

Each register read will be at least ~1-3us so don't bother using udelay.

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

drm/amd/display: Add GPINT handler interface

[Why]
The General Purpose Interrupt is used on the DMCUB to pass lightweight
commands via a register to the DMCUB.

This is limited to 32-bit command and 32-bit response.

This will be used for shutting down the firmware in a clean manner.

[How]
Add the command IDs and the data register to correctly format
the commands.

Add the interface functions to dmub_srv for sending and receiving the
commands.

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

drm/amd/display: Fix various issues found by compiler warning as errors

[Why]
Diagnostics team reported various issues found when enabling warnings as errors

[How]
Fix implicit conversions

Signed-off-by: Eric Bernstein <eric.bernstein@amd.com>
Reviewed-by: Aric Cyr <Aric.Cyr@amd.com>
Acked-by: Bhawanpreet Lakha <Bhawanpreet.Lakha@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Indicate dsc updates explicitly

[Why]
DSC updates only set type to FULL UPDATE, but doesn't
flag the change

[How]
Add DSC flag update flag

Signed-off-by: Anthony Koo <Anthony.Koo@amd.com>
Reviewed-by: Aric Cyr <Aric.Cyr@amd.com>
Acked-by: Bhawanpreet Lakha <Bhawanpreet.Lakha@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Split program front end part that occur outside lock

[Why]
Eventually want to lock at a higher level in stack.
To do this, we need to be able to isolate the parts that need to be done
after pipe unlock.

[How]
Split out programming that is done post unlock.

Signed-off-by: Anthony Koo <Anthony.Koo@amd.com>
Reviewed-by: Aric Cyr <Aric.Cyr@amd.com>
Acked-by: Bhawanpreet Lakha <Bhawanpreet.Lakha@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Do not set optimized_require to false after plane disable

[WHY]
The optimized_require flag is needed to set watermarks and clocks lower
in certain conditions. This flag is set to true and then set to false
while programming front end in dcn20.

[HOW]
Do not set the flag to false while disabling plane.

Signed-off-by: Sung Lee <sung.lee@amd.com>
Reviewed-by: Tony Cheng <Tony.Cheng@amd.com>
Acked-by: Bhawanpreet Lakha <Bhawanpreet.Lakha@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Fix psr static frames calculation

[Why]
Driver crash with psr feature enabled due to divide-by-zero error.
This is a regression after rework to calculate static screen frame
number entry time.

[How]
Correct order of operations to avoid divide-by-zero.

Signed-off-by: Roman Li <roman.li@amd.com>
Reviewed-by: Zhan Liu <Zhan.Liu@amd.com>
Acked-by: Bhawanpreet Lakha <Bhawanpreet.Lakha@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: remove invalid dc_is_hw_initialized function

[why/how]
We found out that the register we read actually gets reset by SMU
after we loose power, meaning this always returns true

Signed-off-by: Joseph Gravenor <joseph.gravenor@amd.com>
Reviewed-by: Eric Yang <eric.yang2@amd.com>
Acked-by: Bhawanpreet Lakha <Bhawanpreet.Lakha@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Move USB-C workaround to after parameter variables are set

[Why]
The call to dp_enable_link_phy are using default/invalid values for clock id
and link settings.

[How]
Move workaround code to after its parameter variables are determined.

Signed-off-by: George Shen <george.shen@amd.com>
Reviewed-by: Tony Cheng <Tony.Cheng@amd.com>
Acked-by: Bhawanpreet Lakha <Bhawanpreet.Lakha@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Add set psr version message

[Why]
Must know psr version during runtime.

[How]
Add set psr version message structures.

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

drm/amdgpu: move xgmi init/fini to xgmi_add/remove_device call (v2)

For sriov, psp ip block has to be initialized before
ih block for the dynamic register programming interface
that needed for vf ih ring buffer. On the other hand,
current psp ip block hw_init function will initialize
xgmi session which actaully depends on interrupt to
return session context. This results an empty xgmi ta
session id and later failures on all the xgmi ta cmd
invoked from vf. xgmi ta session initialization has to
be done after ih ip block hw_init call.

to unify xgmi session init/fini for both bare-metal
sriov virtualization use scenario, move xgmi ta init
to xgmi_add_device call, and accordingly terminate xgmi
ta session in xgmi_remove_device call.

The existing suspend/resume sequence will not be changed.

v2: squash in return fix from Nirmoy

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

drm/amd/display: call psp set/get interfaces

Call the cmd ids for set/get srm according to the sysfs call

v2: Use define for the magic number

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

drm/amd/display: Load srm before enabling HDCP

[Why]
we need to load SRM before we start HDCP. Because for S3 case the sysfs call will be
after we have already enabled HDCP, so we might not be using the latest SRM

[How]
Set srm before starting HDCP.

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

drm/amd/display: Add sysfs interface for set/get srm

[Why]
PSP doesn't have the ability to store SRM in a non-volatile memory.  And since
the kernel cannot write to the storage directly, we need usermode to facilitate
this

As per spec the SRM needs to be persistent so this interface is to be
called by the usermode anytime the system goes down/powers on

*boot/resume: load from storage
*shutdown/suspend: save to storage

[How]
Provide a sysfs interface so that the usermode can set/get srm at the right times

save to storage: call "cat /sys/class/drm/card0/device/hdcp_srm > file" after boot and resume
-driver calls psp_get_srm() to get the stored srm and outputs it

load from storage: call "cat file > /sys/class/drm/card0/device/hdcp_srm" before shutdown and suspend
-driver reads the file from sysfs and calls psp_set_srm() to send the SRM to PSP

v2:
-update commit description
-add comment about sysfs file handling in the code

v3:
- squash in use after free fix (Dan Carpenter)

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

drm/amd/display: update psp interface header

[Why]
We need to support SRM(System Renewability Message)
As per hdcp spec (5.Renewability) SRM needs to be storage in a non-volatile
memory.

PSP owns the checking of SRM but doesn't have the ability to store it in a
non-volatile memory. So we need the kernel driver to facilitate it using the
interface provided by PSP

[How]
Add the interface to the header file, so the driver can use them

v2: update commit description

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

drm/amd/display: Pass amdgpu_device instead of psp_context

[Why]
We need this to create sysfs (followup patch)

[How]
Change the parameter

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

drm/amdgpu: rework synchronization of VM updates v4

If provided we only sync to the BOs reservation
object and no longer to the root PD.

v2: update comment, cleanup amdgpu_bo_sync_wait_resv
v3: use correct reservation object while clearing
v4: fix typo in amdgpu_bo_sync_wait_resv

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

drm/amdgpu: simplify and fix amdgpu_sync_resv

No matter what we always need to sync to moves.

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

drm/amdgpu: allow higher level PD invalidations

Allow partial invalidation on unallocated PDs. This is useful when we
need to silence faults to stop interrupt floods on Vega.

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

drm/amdgpu: return EINVAL instead of ENOENT in the VM code

That we can't find a PD above the root is expected can only happen if
we try to update a larger range than actually managed by the VM.

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

drm/amdgpu: fix parentheses in amdgpu_vm_update_ptes

For the root PD mask can be 0xffffffff as well which would
overrun to 0 if we don't cast it before we add one.

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

drm/amdgpu: make sure to never allocate PDs/PTs for invalidations

Make sure that we never allocate a page table for an invalidation operation.

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

drm/amdgpu: drop unnecessary restriction for huge root PDEs

The root PD can also contain huge PDEs.

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

drm/amdgpu: stop using amdgpu_bo_gpu_offset in the VM backend

We need to update page tables without any lock held.

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

drm/amdgpu: rework job synchronization v2

For unlocked page table updates we need to be able
to sync to fences of a specific VM.

v2: use SYNC_ALWAYS in the UVD code

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

drm/amdgpu: use the VM as job owner

For HMM we need to rework how VM synchronization works, so instead of the filp use VM as job owner.

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