drm/amd/display: Add comment where CONFIG_DRM_AMD_DC_DCN macro ends

Trivial patch which adds a comment for macro
endif's in amdgpu_dm.c

Reviewed-by: Ariel Bernstein <Eric.Bernstein@amd.com>
Reviewed-by: Harry Wentland <Harry.Wentland@amd.com>
Acked-by: Anson Jacob <Anson.Jacob@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Anson Jacob <Anson.Jacob@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdkfd: Fix retry fault drain race conditions

The check for whether to drain retry faults must be under the mmap write
lock to serialize with munmap notifier callbacks.

We were also missing checks on child ranges. To fix that, simplify the
logic by using a flag rather than checking on each prange. That also
allows draining less freqeuntly when many ranges are unmapped at once.

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

drm/amdkfd: lower the VAs base offset to 8KB

The low 16MB of virtual address space are currently reserved for kernel
mode allocations mapped into user virtual address space. This causes
conflicts with HMM/SVM mappings at low virtual addresses. We tried to
move those kernel mode allocations to the upper half of the 64-bit
virtual address space for GFX9, which is naturally reserved for kernel
use. However, TBA (trap handler code) has problems to access addresses
in the high virtual space. We have decided to set this to 8KB of the
lower address space as a temporary fix, while investigate TBA address
problem. It is very unlikely for user space to map memory at this low
region.

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

drm/amd/display: fix exit from amdgpu_dm_atomic_check() abruptly

make action upon failure in "drm_atomic_add_affected_connectors()"
consistent with the rest of failures in amdgpu_dm_atomic_check().

Signed-off-by: Shirish S <shirish.s@amd.com>
Reviewed-by: Harry Wentland <harry.wentland@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/amdgpu: fix the kfd pre_reset sequence in sriov

The KFD pre_reset should be called before reset been executed, it will
hold the lock to prevent other rocm process to sent the packlage to hiq
during host execute the real reset on the HW

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

drm/amdgpu: fix uvd crash on Polaris12 during driver unloading

There was a change(below) target for such issue:
d82e2c249c8f ("drm/amdgpu: Fix crash on device remove/driver unload")
But the fix for VI ASICs was missing there. This is a supplement for
that.

Fixes: d82e2c249c8f ("drm/amdgpu: Fix crash on device remove/driver unload")

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 SI handling in amdgpu_device_asic_has_dc_support()

Properly handle SI DC support when CONFIG_DRM_AMD_DC_SI is not
set.

Fixes: f7f12b25823c0d ("drm/amdgpu: default to true in amdgpu_device_asic_has_dc_support")
Reviewed-by: Evan Quan <evan.quan@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: Fix dangling kfd_bo pointer for shared BOs

If a kfd_bo was shared (e.g. a dmabuf export), the original kfd_bo may be
freed when the amdgpu_bo still lives on. Free the kfd_bo struct in the
release_notify callback then the amdgpu_bo is freed.

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

drm/amd/amdkfd: Don't sent command to HWS on kfd reset

When kfd need to be reset, sent command to HWS might cause hang and get unnecessary timeout.
This change try not to touch HW in pre_reset and keep queues to be in the evicted state
when the reset is done, so they are not put back on the runlist. These queues will be destroied
on process termination.

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

drm/amdgpu: correctly toggle gfx on/off around RLC_SPM_* register access

As part of the ib padding process, accessing the RLC_SPM_* register may
trigger gfx hang. Since gfxoff may be already kicked during the whole period.
To address that, we manually toggle gfx on/off around the RLC_SPM_*
register access.

This can resolve the gfx hang issue observed on running Talos with RDP launched
in parallel.

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

drm/amdgpu: correct xgmi ras error count reset

The error count reset for xgmi3x16 pcs is missed.

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

drm/amd/pm: Correct DPMS disable IP version check

Previously there was a check based on chip # for chips that aligned to
>=CHIP_NAVI10 to have RLC stopped as part of DPMS check.  This was because
of gfxclk being controlled by RLC in the newer designs.

As part of IP version checking though, this got changed to match IP
version for SMU.  Because Renoir designs also include smu11 that meant
that even GFX9 started to stop RLC earlier.

Adjust to match GFX IP version instead of SMU IP version to restore the
previous behavior.

Fixes: a8967967f6a5 ("drm/amdgpu/amdgpu_smu: convert to IP version checking")
Signed-off-by: Mario Limonciello <mario.limonciello@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/amdgpu: Fix csb.bo pin_count leak on gfx 9

[Why]
csb bo is not unpinned in gfx 9. It will lead to pin_count leak on
driver unload.

[How]
Call bo_free_kernel corresponding to bo_create_kernel in
gfx_rlc_init_csb. This will also unify the code path with other gfx
versions.

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

drm/amd/amdgpu: Avoid writing GMC registers under sriov in gmc9

[Why]
For Vega10, disabling gart of gfxhub could mess up KIQ and PSP
under sriov mode, and lead to DMAR on host side.

[How]
Skip writing GMC registers under sriov.

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

drm/amdgpu/powerplay: fix sysfs_emit/sysfs_emit_at handling

sysfs_emit and sysfs_emit_at requrie a page boundary
aligned buf address. Make them happy!

v2: fix sysfs_emit -> sysfs_emit_at missed conversions

Cc: Lang Yu <lang.yu@amd.com>
Cc: Darren Powell <darren.powell@amd.com>
Fixes: 6db0c87a0a8e ("amdgpu/pm: Replace hwmgr smu usage of sprintf with sysfs_emit")
Bug: https://gitlab.freedesktop.org/drm/amd/-/issues/1774
Reviewed-by: Lang Yu <lang.yu@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: Make sure to reserve BOs before adding or removing

BOs need to be reserved before they are added or removed, so ensure that
they are reserved during kfd_mem_attach and kfd_mem_detach

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

drm/amdkfd: avoid recursive lock in migrations back to RAM

[Why]:
When we call hmm_range_fault to map memory after a migration, we don't
expect memory to be migrated again as a result of hmm_range_fault. The
driver ensures that all memory is in GPU-accessible locations so that
no migration should be needed. However, there is one corner case where
hmm_range_fault can unexpectedly cause a migration from DEVICE_PRIVATE
back to system memory due to a write-fault when a system memory page in
the same range was mapped read-only (e.g. COW). Ranges with individual
pages in different locations are usually the result of failed page
migrations (e.g. page lock contention). The unexpected migration back
to system memory causes a deadlock from recursive locking in our
driver.

[How]:
Creating a task reference new member under svm_range_list struct.
Setting this with "current" reference, right before the hmm_range_fault
is called. This member is checked against "current" reference at
svm_migrate_to_ram callback function. If equal, the migration will be
ignored.

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

drm/amd/display: Don't allow partial copy_from_user

There is no reason to allow for partial buffers from userspace in our
debugfs. In this particular case callers will zero out the wr_buf but if
callers in the future don't do that we might be looking at corrupt data.

Linus puts it better than I can in
https://lkml.org/lkml/2021/10/26/993

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

drm/amd/display: 3.2.160

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

drm/amd/display: [FW Promotion] Release 0.0.91

Acked-by: Aurabindo Pillai <aurabindo.pillai@amd.com>
Signed-off-by: Anthony Koo <Anthony.Koo@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: add condition check for dmub notification

[Why & How]
In order to have dc_enable_dmub_notifications() more precise, add
one more condition to check if dc->debug.dpia_debug.bits.disable_dpia
is false.

Signed-off-by: Meenakshikumar Somasundaram <meenakshikumar.somasundaram@amd.com>
Signed-off-by: Wayne Lin <Wayne.Lin@amd.com>
Signed-off-by: Aurabindo Pillai <aurabindo.pillai@amd.com>
Reviewed-by: Nicholas Kazlauskas <Nicholas.Kazlauskas@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Added new DMUB boot option for power optimization

[Why]
During Z10, root clock gating and memory low power registers needs to
to be restored if optimization is enabled in driver.

[How]
Added new DMUB boot option for root clock gating and memory low power.

Acked-by: Aurabindo Pillai <aurabindo.pillai@amd.com>
Signed-off-by: Jake Wang <haonan.wang2@amd.com>
Reviewed-by: Nicholas Kazlauskas <Nicholas.Kazlauskas@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Add MPC meory shutdown support

[Why & How]
The MPC memory clocks should be powered down when not in use.

Acked-by: Aurabindo Pillai <aurabindo.pillai@amd.com>
Signed-off-by: Jake Wang <haonan.wang2@amd.com>
Reviewed-by: Eric Yang <eric.yang2@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Added HPO HW control shutdown support

[Why]
HPO is only used for DP2.0. HPO HW control should be
disable when not being used to save power.

[How]
Shutdown HPO HW control during init hw.
Shutdown HPO HW control during stream disable.
Enable HPO HW control during stream enable if DP2.0.

Acked-by: Aurabindo Pillai <aurabindo.pillai@amd.com>
Signed-off-by: Jake Wang <haonan.wang2@amd.com>
Reviewed-by: Nicholas Kazlauskas <Nicholas.Kazlauskas@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: fix register write sequence for LINK_SQUARE_PATTERN

[why&how]
write LINK_SQUARE_PATTERN_num + 1 for square pulse pattern.
Specs requirement to write this register prior to write LINK_QUAL_LANEX_SET.

Acked-by: Aurabindo Pillai <aurabindo.pillai@amd.com>
Signed-off-by: Wenjing Liu <wenjing.liu@amd.com>
Reviewed-by: George Shen <george.shen@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Clear encoder assignments when state cleared.

[Why]
State can be cleared without removing individual streams (by
calling dc_remove_stream_from_ctx()). This can leave the
encoder assignment module in an incoherent state and cause
future assignments to be incorrect.

[How]
Clear encoder assignments when committing 0 streams or
re-initializing hardware.

Acked-by: Aurabindo Pillai <aurabindo.pillai@amd.com>
Signed-off-by: Jimmy Kizito <Jimmy.Kizito@amd.com>
Reviewed-by: Jun Lei <Jun.Lei@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Force disable planes on any pipe split change

[Why]
In scenario when 1 display connected with pipe split (2 pipes in use)
and 3 new displays simultaneously hotplugged via MST hub (4 pipes in use),
mpcc may get reprogram to other vtg, remaining busy.
In this case waiting for mpcc idle timeouts with error like this:
[drm] REG_WAIT timeout 1us * 100000 tries - mpc2_assert_idle_mpcc
RIP: 0010:mpc2_assert_mpcc_idle_before_connect
Call Trace:
dcn20_update_mpcc
dcn20_program_front_end_for_ctx
dc_commit_state
amdgpu_dm_atomic_commit_tail
...

[How]
Add pipe split change condition to disable dangling plane.

Acked-by: Aurabindo Pillai <aurabindo.pillai@amd.com>
Signed-off-by: Roman Li <Roman.Li@amd.com>
Reviewed-by: Nicholas Kazlauskas <Nicholas.Kazlauskas@amd.com>
Reviewed-by: Dmytro Laktyushkin <Dmytro.Laktyushkin@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Fix bpc calculation for specific encodings

[Why]
1. YCbCr 4:2:2 8bpc/10bpc modes are blocked for HDMI by policy
2. A YCbCr 4:2:0 calculation error blocked some 4:2:0 timing modes

[How]
YCbCr 4:2:2 8bpc/10bpc modes are allowed for HDMI
Fix YCbCr 4:2:0 calculation error

Acked-by: Aurabindo Pillai <aurabindo.pillai@amd.com>
Signed-off-by: Bing Guo <Bing.Guo@amd.com>
Reviewed-by: Chris Park <chris.park@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: avoid link loss short pulse stuck the system

[Why]
MST monitor sends link loss short pulse continuous but sink is
occupy by HDMI input to lead link training fail.

[How]
disable link once retraining fail.

Acked-by: Aurabindo Pillai <aurabindo.pillai@amd.com>
Signed-off-by: Yu-ting Shen <yu-tshen@amd.com>
Reviewed-by: Wenjing Liu <Wenjing.Liu@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Fix dummy p-state hang on monitors with extreme timing

[WHY]
It was found that the system would hang on a dummy pstate when playing
4k60 videos on a 1080p 390Hz monitor.

[HOW]
Properly select the dummy_pstate_latency_ms when firmware assisted
memory clock switching is enabled instead of assuming that the highest
latency would work for every monitor timing.

Acked-by: Aurabindo Pillai <aurabindo.pillai@amd.com>
Signed-off-by: Felipe Clark <felclark@amd.com>
Reviewed-by: Jun Lei <Jun.Lei@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Fix dcn10_log_hubp_states printf format string

Fix spacing issue for the format string.

Addresses-Coverity-ID: 1446765: ("Invalid printf format string")

Acked-by: Aurabindo Pillai <aurabindo.pillai@amd.com>
Signed-off-by: Anson Jacob <Anson.Jacob@amd.com>
Reviewed-by: Rodrigo Siqueira <Rodrigo.Siqueira@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: dsc engine not disabled after unplug dsc mst hub

[WHY]
If timing and bpp of displays on mst hub are not changed,
pbn, slot_num for displays should not be changed. Linux
user mode may initiate atomic_check with different display
configuration after set mode finished. This will call to
amdgpu_dm to re-compute payload, slot_num of displays and
saved to dm_connect_state. stream->timing.flags.dsc, pbn,
slot_num are updated to values which may be different from
that were used for set mode. when dsc hub with 3 4k@60hz dp
connected, 3 dsc engines are enabled. timing.flags.dsc = 1.
timing.flags.dsc are changed to 0 due to atomic check. when
dsc hub is unplugged, amdgpu driver check timing.flags.dsc
for last mode set and find out flags.dsc = 0, then does not
disable dsc.

[HOW]
check status of  displays on dsc mst hubs. re-compute pbn,
slot_num, timing.flags.dsc only if there is mode, connect
or enable/disable change.

Acked-by: Aurabindo Pillai <aurabindo.pillai@amd.com>
Signed-off-by: Hersen Wu <hersenwu@amd.com>
Reviewed-by: Mikita Lipski <Mikita.Lipski@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: remove duplicated kfd_resume_iommu

Remove duplicated kfd_resume_iommu which already runs
in mdgpu_amdkfd_device_init.

Tested-By: Ken Moffat <zarniwhoop@ntlworld.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: James Zhu <James.Zhu@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: update RLC_PG_DELAY_3 Value to 200us for yellow carp

For yellow carp, the desired CGPG hysteresis value is 0x4E20.

Signed-off-by: Aaron Liu <aaron.liu@amd.com>
Acked-by: Huang Rui <ray.huang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Look at firmware version to determine using dmub on dcn21

commit 652de07addd2 ("drm/amd/display: Fully switch to dmub for all dcn21
asics") switched over to using dmub on Renoir to fix Gitlab 1735, but this
implied a new dependency on newer firmware which might not be met on older
kernel versions.

Since sw_init runs before hw_init, there is an opportunity to determine
whether or not the firmware version is new to adjust the behavior.

Cc: Roman.Li@amd.com
BugLink: https://gitlab.freedesktop.org/drm/amd/-/issues/1772
BugLink: https://gitlab.freedesktop.org/drm/amd/-/issues/1735
Fixes: 652de07addd2 ("drm/amd/display: Fully switch to dmub for all dcn21 asics")
Signed-off-by: Mario Limonciello <mario.limonciello@amd.com>
Acked-by: Alex Deucher <alexander.deucher@amd.com>
Reviewed-by: Roman Li <Roman.Li@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu/pm: Don't show pp_power_profile_mode for unsupported devices

For ASICs not supporting power profile mode, don't show the attribute.
Verify that the function has been implemented by the subsystem.

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

drm/amd/pm: Adjust returns when power_profile_mode is not supported

This better aligns that the caller can make a mistake with the buffer
and -EINVAL should be returned, but if the hardware doesn't support
the feature it should be -EOPNOTSUPP.

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

drm/amd/pm: Add missing mutex for pp_get_power_profile_mode

Prevent possible issues from set and get being called simultaneously.

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

drm/amdgpu/pm: drop pp_power_profile_mode support for yellow carp

This was added by commit bd8dcea93a7d ("drm/amd/pm: add callbacks to
read/write sysfs file pp_power_profile_mode") but the feature was
deprecated from PMFW.  Remove it from the driver.

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

drm/amdkfd: update gfx target version for Renoir

Previously Renoir compiler gfx target version was forced to Raven.
Update driver side for completeness.

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

drm/amdgpu: Convert SMU version to decimal in debugfs

This is more useful when talking to the SMU team to have the information
in this format, save one less step to manually do it.

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

drm/amdkfd: Handle incomplete migration to system memory

If some pages fail to migrate to system memory, don't update
prange->actual_loc = 0. This prevents endless CPU page faults after
partial migration failures due to contested page locks.

Migration to RAM must be complete during migrations from VRAM to VRAM and
during evictions. Implement retry and fail if the migration to RAM fails.

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

drm/amdkfd: Avoid thrashing of stack and heap

Stack and heap pages tend to be shared by many small allocations.
Concurrent access by CPU and GPU is therefore likely, which can lead to
thrashing. Avoid this by setting the preferred location to system memory.

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

drm/amdkfd: Fix SVM_ATTR_PREFERRED_LOC

The preferred location should be used as the migration destination
whenever it is accessible by the faulting GPU. System memory is always
accessible. Peer memory is accessible if it's in the same XGMI hive.

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

drm/amdgpu: use correct register mask to extract field

Aldebaran has different register mask definitions for
regiter MC_VM_XGMI_LFB_CNTL. Use the correct masks
to interpret fields of this register.

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

drm/amd/amdgpu: fix bad job hw_fence use after free in advance tdr

[Why]
In advance tdr mode, the real bad job will be resubmitted twice, while
in drm_sched_resubmit_jobs_ext, there's a dma_fence_put, so the bad job
is put one more time than other jobs.

[How]
Adding dma_fence_get before resbumit job in
amdgpu_device_recheck_guilty_jobs and put the fence for normal jobs

Signed-off-by: Jingwen Chen <Jingwen.Chen2@amd.com>
Reviewed-by: Andrey Grodzovsky <andrey.grodzovsky@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

Merge tag 'amd-drm-next-5.16-2021-10-29' of https://gitlab.freedesktop.org/agd5f/linux into drm-next

amd-drm-next-5.16-2021-10-29:

amdgpu:
- RAS fixes
- Fix a potential memory leak in device tear down
- Add a stutter mode quirk
- Misc display fixes
- Further display FP refactoring
- Display USB4 fixes
- Display DP2.0 fixes
- DCN 3.1 fixes
- Display 8 ch audio fix
- Fix DMA mask regression for SI parts
- Aldebaran fixes

amdkfd:
- userptr fix
- BO lifetime fix
- Misc code cleanup

UAPI:
- Minor header cleanup (no functional change)

Signed-off-by: Dave Airlie <airlied@redhat.com>
From: Alex Deucher <alexander.deucher@amd.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20211029184338.4863-1-alexander.deucher@amd.com

drm/amdgpu/gmc6: fix DMA mask from 44 to 40 bits

The DMA mask on SI parts is 40 bits not 44.  Copy
paste typo.

Fixes: 244511f386ccb9 ("drm/amdgpu: simplify and cleanup setting the dma mask")
Bug: https://gitlab.freedesktop.org/drm/amd/-/issues/1762
Acked-by: Christian König <christian.koenig@amd.com>
Tested-by: Paul Menzel <pmenzel@molgen.mpg.de>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: MST support for DPIA

[Why]
- DPIA MST slot registers are not programmed during payload
allocation and hence MST does not work with DPIA.
- HPD RX interrupts are not handled for DPIA.

[How]
- Added inbox command to program the MST slots whenever
  payload allocation happens for DPIA links.
- Added support for handling HPD RX interrupts

Signed-off-by: Meenakshikumar Somasundaram <meenakshikumar.somasundaram@amd.com>
Reviewed-by: Jun Lei <Jun.Lei@amd.com>
Acked-by: Nicholas Kazlauskas <nicholas.kazlauskas@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: Fix even more out of bound writes from debugfs

CVE-2021-42327 was fixed by:

commit f23750b5b3d98653b31d4469592935ef6364ad67
Author: Thelford Williams <tdwilliamsiv@gmail.com>
Date:   Wed Oct 13 16:04:13 2021 -0400

    drm/amdgpu: fix out of bounds write

but amdgpu_dm_debugfs.c contains more of the same issue so fix the
remaining ones.

v2:
* Add missing fix in dp_max_bpc_write (Harry Wentland)

Fixes: 918698d5c2b5 ("drm/amd/display: Return the number of bytes parsed than allocated")
Signed-off-by: Patrik Jakobsson <pjakobsson@suse.de>
Reviewed-by: Harry Wentland <harry.wentland@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu/discovery: add SDMA IP instance info for soc15 parts

Add secondary instance version info for soc15 parts.

Reviewed-by: Luben Tuikov <luben.tuikov@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu/discovery: add UVD/VCN IP instance info for soc15 parts

Add secondary instance version info for vega20, arcturure, and
aldebaran.

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

drm/amdgpu/UAPI: rearrange header to better align related items

Move the RAS query parameters to align with the INFO query where
they are used.  No functional change.

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

drm/amd/display: Enable dpia in dmub only for DCN31 B0

[Why]
DMUB binary is common for both A0 and B0. Hence, driver should
notify FW about the support for DPIA in B0.

[How]
Added dpia_supported bit in dmub_fw_boot_options and will be set
only for B0.

Assign dpia_supported to true before dm_dmub_hw_init
in B0 case.

v2: fix build without CONFIG_DRM_AMD_DC_DCN (Alex)

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

drm/amd/display: Fix USB4 hot plug crash issue

[Why]
Notify data from outbox corrupt, the notify type should be 2 (HPD) instead of 0
(No data). We copied the address instead of the value. The memory might be
freed in the end of outbox IRQ

[How]
We should allocate the memory of notify and copy the whole content from outbox to
hpd handle function

Fixes: 88f52b1fff891e ("drm/amd/display: Support for SET_CONFIG processing with DMUB")
Signed-off-by: Jude Shih <shenshih@amd.com>
Reviewed-by: Nicholas Kazlauskas <nicholas.kazlauskas@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Fix deadlock when falling back to v2 from v3

[Why]
A deadlock in the kernel occurs when we fallback from the V3 to V2
add_topology_to_display or remove_topology_to_display because they
both try to acquire the dtm_mutex but recursive locking isn't
supported on mutex_lock().

[How]
Make the mutex_lock/unlock more fine grained and move them up such that
they're only required for the psp invocation itself.

Fixes: bf62221e9d0e ("drm/amd/display: Add DCN3.1 HDCP support")

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

drm/amd/display: Fallback to clocks which meet requested voltage on DCN31

[WHY]
On certain configs, SMU clock table voltages don't match which cause parser
to behave incorrectly by leaving dcfclk and socclk table entries unpopulated.

[HOW]
Currently the function that finds the corresponding clock for a given voltage
only checks for exact voltage level matches. In the case that no match gets
found, parser now falls back to searching for the max clock which meets the
requested voltage (i.e. its corresponding voltage is below requested).

Signed-off-by: Michael Strauss <michael.strauss@amd.com>
Reviewed-by: Nicholas Kazlauskas <nicholas.kazlauskas@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: move FPU associated DCN301 code to DML folder

[Why & How]
As part of the FPU isolation work documented in
https://patchwork.freedesktop.org/series/93042/, isolate
code that uses FPU in DCN301 to DML, where all FPU code
should locate.

Cc: Christian König <christian.koenig@amd.com>
Cc: Harry Wentland <harry.wentland@amd.com>
Cc: Rodrigo Siqueira <Rodrigo.Siqueira@amd.com>
Tested-by: Zhan Liu <Zhan.Liu@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Qingqing Zhuo <Qingqing.Zhuo@amd.com>
Reviewed-by: Rodrigo Siqueira <Rodrigo.Siqueira@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: fix link training regression for 1 or 2 lane

[why]
We have a regression that cause maximize lane settings to use
uninitialized data from unused lanes.
This will cause link training to fail for 1 or 2 lanes because the lane
adjust is populated incorrectly sometimes.

v2: fix build without CONFIG_DRM_AMD_DC_DCN (Alex)

Reviewed-by: Eric Yang <eric.yang2@amd.com>
Acked-by: Rodrigo Siqueira <Rodrigo.Siqueira@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Wenjing Liu <wenjing.liu@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: add two lane settings training options

[why]
option 1: disallow different lanes to have different lane settings
option 2: dpcd lane settings will always use the same hw lane settings
even if it doesn't match requested lane adjust

Reviewed-by: Jun Lei <Jun.Lei@amd.com>
Acked-by: Rodrigo Siqueira <Rodrigo.Siqueira@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Wenjing Liu <wenjing.liu@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: decouple hw_lane_settings from dpcd_lane_settings

[why]
As DP features expands, we have encountered many situations where we
must configure a different DPCD lane setting from hw lane settings we
output.  The change is to decouple hw lane settings from dpcd lane
settings to provide flexibility to configure dpcd and hw individually.

Reviewed-by: Jun Lei <Jun.Lei@amd.com>
Acked-by: Rodrigo Siqueira <Rodrigo.Siqueira@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Wenjing Liu <wenjing.liu@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: implement decide lane settings

[why]
Decouple lane settings decision logic all to its own function. The
function takes in lane adjust array and link training settings and
decide what hw lane setting and dpcd lane setting should be used.

Reviewed-by: Jun Lei <Jun.Lei@amd.com>
Acked-by: Rodrigo Siqueira <Rodrigo.Siqueira@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Wenjing Liu <wenjing.liu@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: adopt DP2.0 LT SCR revision 8

[how]
revision 8 SCR requires DP Source to write TPS2 and FFE lane adjustment
in one 5 byte write aux transaction.
It specifies to read aux rd interval value as soon as we turn on TPS1
pattern.

Cc: Wayne Lin <wayne.lin@amd.com>
Acked-by: Rodrigo Siqueira <Rodrigo.Siqueira@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Wenjing Liu <wenjing.liu@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: FEC configuration for dpia links in MST mode

[Why]
To fix the check condition for fec enable for dpia links in MST mode.

[How]
dc_link_should_enable_fec() to be used to check whether fec should be
enabled in MST mode.

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

drm/amd/display: FEC configuration for dpia links

[Why]
To fix the check condition for fec enable for dpia links.

[How]
dc_link_should_enable_fec() to be used to check whether fec should be
enabled.

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

drm/amd/display: Add workaround flag for EDID read on certain docks

[Why]
Certain docks appear to NAK I2C writes to the segment pointer with the
MOT (middle of transaction) bit clear. This behaviour can cause EDID
reads from higher segments to fail.

[How]
Add workaround flag for links which connect to docks exhibiting this
issue.

Cc: Wayne Lin <wayne.lin@amd.com>
Reviewed-by: Jun Lei <Jun.Lei@amd.com>
Acked-by: Rodrigo Siqueira <Rodrigo.Siqueira@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Jimmy Kizito <Jimmy.Kizito@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Set phy_mux_sel bit in dmub scratch register

[Why]
B0 has pipe mux for DIGC and DIGD which can be connected to PHYF/PHYG or
PHYC/PHY D.

[How]
Based on chip internal hardware revision id determine it is B0 and set
DMUB scratch register so DMUBFW can connect the display pipe is
connected correctly to the dig.

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

drm/amd/display: Manually adjust strobe for DCN303

why:
DCN303's 4 channel SOC BB causes problems at strobe

how:
workaround to manually adjust strobe calculation using FCLK
restrict.

Reviewed-by: Jun Lei <Jun.Lei@amd.com>
Acked-by: Agustin Gutierrez <agustin.gutierrez@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Martin Leung <Martin.Leung@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: 3.2.159

This new DC version brings improvements in the following areas:
- Improvements for USB4;
- Isolate FPU code for DCN20, DCN301, and DSC;
- Fixes on Linking training;
- Refactoring some parts of the code, such as PSR;

Acked-by: Agustin Gutierrez <agustin.gutierrez@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Aric Cyr <aric.cyr@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: [FW Promotion] Release 0.0.90

Acked-by: Agustin Gutierrez <agustin.gutierrez@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Anthony Koo <Anthony.Koo@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Remove unused macros

fixed16_to_double & fixed16_to_double_to_cpu are not used.

Reviewed-by: Rodrigo Siqueira <Rodrigo.Siqueira@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Acked-by: Agustin Gutierrez <agustin.gutierrez@amd.com>
Signed-off-by: Anson Jacob <Anson.Jacob@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: allow windowed mpo + odm

This change adds a config flag to allow non fullscreen MPO during ODM.
Scaling calculation will still fail configurations where video is only
one one side of the screen.

Reviewed-by: Aric Cyr <Aric.Cyr@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Acked-by: Agustin Gutierrez <agustin.gutierrez@amd.com>
Signed-off-by: Dmytro Laktyushkin <Dmytro.Laktyushkin@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: set Layout properly for 8ch audio at timing validation

Why:
For audio packet type 0x02, there are 2 Layouts:
   Layout = 0 for 2 channels
and Layout = 1 for > 2 channels.
Layout will affect bandwidth check.

Currently, for HDMI FRL, Layout field isn't set and has a default value
of 0, so theoretically only 2-channel audio for audio packet type 0x02
is supported now.

How:
1. Set Layout properly according to maximum audio channel numbers for
audios with audio packet type 0x02.
2. 8ch LPCM audio is not supported for timing modes with v_active <= 576.

Reviewed-by: Chris Park <chris.park@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Acked-by: Agustin Gutierrez <agustin.gutierrez@amd.com>
Signed-off-by: Bing Guo <Bing.Guo@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Fix 3DLUT skipped programming

We found a compilation error that we thought was caused by the 3DLUT
patch; later on, we figured out the root cause of the problem, but we
already applied the revert in the wrong patch. This commit brings it
back  the 3DLUT fix.

Acked-by: Agustin Gutierrez <agustin.gutierrez@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Aric Cyr <aric.cyr@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: 3.2.158

Acked-by: Agustin Gutierrez <agustin.gutierrez@amd.com>
Signed-off-by: Aric Cyr <aric.cyr@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: [FW Promotion] Release 0.0.89

- Add flag to control root clock gating in init_hw
- Add flag to indicate a diags environment is being used

Acked-by: Agustin Gutierrez <agustin.gutierrez@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Anthony Koo <Anthony.Koo@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Handle I2C-over-AUX write channel status update

[Why]
When writing long AUX commands some sinks will respond will write status
update requiring source to read status.

[How]
When a write request is replied with data (AUX_ACK_M), retry a read of
write status to determine when the write is completed.

Reviewed-by: Martin Leung <Martin.Leung@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Acked-by: Agustin Gutierrez <agustin.gutierrez@amd.com>
Signed-off-by: Aric Cyr <aric.cyr@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Add comment for preferred_training_settings

[Why]
Currently the naming of preferred_training_settings is ambiguous and has
caused confusion regarding its purpose and usage.

[How]
Add comment to clarify the intention.

Reviewed-by: Wenjing Liu <Wenjing.Liu@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Acked-by: Agustin Gutierrez <agustin.gutierrez@amd.com>
Signed-off-by: George Shen <george.shen@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Implement fixed DP drive settings

[Why]
Currently there are use cases that require DP link to maintain fixed VS
and PE in HW regardless of what the sink requests. BIOS integrated info
table will specify whether we need to use the fixed drive settings, and
the drive settings to use.

[How]
Implement changes to parse the integrated info table and set the fixed
drive settings accordingly.

Reviewed-by: Wenjing Liu <Wenjing.Liu@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Acked-by: Agustin Gutierrez <agustin.gutierrez@amd.com>
Signed-off-by: George Shen <george.shen@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: restyle dcn31 resource header inline with other asics

Style change for better consistency across codebase

Reviewed-by: Nicholas Kazlauskas <Nicholas.Kazlauskas@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Acked-by: Agustin Gutierrez <agustin.gutierrez@amd.com>
Signed-off-by: Dmytro Laktyushkin <Dmytro.Laktyushkin@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: clean up dcn31 revision check

This is unnecessary in clk_mgr

Reviewed-by: Nicholas Kazlauskas <Nicholas.Kazlauskas@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Acked-by: Agustin Gutierrez <agustin.gutierrez@amd.com>
Signed-off-by: Dmytro Laktyushkin <Dmytro.Laktyushkin@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Defer GAMCOR and DSCL power down sequence to vupdate

[WHY]
Every other CM LUT power down sequence is deferred to next vupdate as
memory powerdown updates immediately while selecting LUTs is double
buffered.  Previous update to defer LUT power down missed GAMCOR and
DSCL, causing some visible flicker when entering/exiting fullscreen
video playback.

[HOW]
Update dpp deferred update loop to check for valid DPPs in res_pool
instead of referencing dcn_ip which turns out to not be populated during
runtime.  Move GAMCOR and DSCL powerdown to dpp deferred updates.

Reviewed-by: Haonan Wang <Haonan.Wang2@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Acked-by: Agustin Gutierrez <agustin.gutierrez@amd.com>
Signed-off-by: Michael Strauss <michael.strauss@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Set i2c memory to light sleep during hw init

[WHY]
i2c memory doesn't get set to light sleep on hw init as intended

[HOW]
Set i2c to light sleep after reg gets zeroed, ensuring memory power
control doesn't get disabled for any other DIO memory

Reviewed-by: Haonan Wang <Haonan.Wang2@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Acked-by: Agustin Gutierrez <agustin.gutierrez@amd.com>
Signed-off-by: Michael Strauss <michael.strauss@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: fix a crash on USB4 over C20 PHY

[Why]
Crash when USB4 is connected.

[How]
Added an ASIC specific code guard.

Reviewed-by: Nikola Cornij <Nikola.Cornij@amd.com>
Reviewed-by: Wenjing Liu <Wenjing.Liu@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Acked-by: Agustin Gutierrez <agustin.gutierrez@amd.com>
Signed-off-by: Ahmad Othman <Ahmad.Othman@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: move FPU associated DSC code to DML folder

[Why & How]
As part of the FPU isolation work documented in
https://patchwork.freedesktop.org/series/93042/, isolate code that uses
FPU in DSC to DML, where all FPU code should locate.

This change does not refactor any functions but move code around.

Cc: Christian König <christian.koenig@amd.com>
Cc: Hersen Wu <hersenxs.wu@amd.com>
Cc: Anson Jacob <Anson.Jacob@amd.com>
Cc: Harry Wentland <harry.wentland@amd.com>
Reviewed-by: Rodrigo Siqueira <Rodrigo.Siqueira@amd.com>
Acked-by: Agustin Gutierrez <agustin.gutierrez@amd.com>
Tested-by: Anson Jacob <Anson.Jacob@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Qingqing Zhuo <qingqing.zhuo@amd.com>
Acked-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Add support for USB4 on C20 PHY for DCN3.1

[Why]
Created new fields that matches new B0 structs On DCN31 the mapping of
DIO output to PHY differs from A0 to B0 boards with new PHY C20 & this
new mapping needed to be handled.

[How]
Mapped new structure based on new structs Added logic for mapping over
A0 and B0 boards Hooked all new structs together.

Reviewed-by: Wenjing Liu <Wenjing.Liu@amd.com>
Acked-by: Agustin Gutierrez <agustin.gutierrez@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Ahmad Othman <Ahmad.Othman@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: dc_link_set_psr_allow_active refactoring

[Why]
To expose new power optimization flags to PSR interface. It allows the
PSR related power features can be enabled separately base on different
use scenarios.

Reviewed-by: Anthony Koo <Anthony.Koo@amd.com>
Acked-by: Agustin Gutierrez <agustin.gutierrez@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Robin Chen <po-tchen@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Get ceiling for v_total calc

Updating certain variable blanking calculations to use ceiling function.

Reviewed-by: Chris Park <chris.park@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Acked-by: Agustin Gutierrez <agustin.gutierrez@amd.com>
Signed-off-by: Bing Guo <Bing.Guo@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: dcn20_resource_construct reduce scope of FPU enabled

Limit when FPU is enabled to only functions that does FPU operations for
dcn20_resource_construct, which gets called during driver
initialization.

Enabling FPU operation disables preemption.  Sleeping functions(mutex
(un)lock, memory allocation using GFP_KERNEL, etc.) should not be called
when preemption is disabled.

Fixes the following case caught by enabling
CONFIG_DEBUG_ATOMIC_SLEEP in kernel config
[    1.338434] BUG: sleeping function called from invalid context at kernel/locking/mutex.c:281
[    1.347395] in_atomic(): 1, irqs_disabled(): 0, non_block: 0, pid: 197, name: systemd-udevd
[    1.356356] CPU: 7 PID: 197 Comm: systemd-udevd Not tainted 5.13.0+ #3
[    1.356358] Hardware name: System manufacturer System Product Name/PRIME X570-PRO, BIOS 3405 02/01/2021
[    1.356360] Call Trace:
[    1.356361]  dump_stack+0x6b/0x86
[    1.356366]  ___might_sleep.cold+0x87/0x98
[    1.356370]  __might_sleep+0x4b/0x80
[    1.356372]  mutex_lock+0x21/0x50
[    1.356376]  smu_get_uclk_dpm_states+0x3f/0x80 [amdgpu]
[    1.356538]  pp_nv_get_uclk_dpm_states+0x35/0x50 [amdgpu]
[    1.356711]  init_soc_bounding_box+0xf9/0x210 [amdgpu]
[    1.356892]  ? create_object+0x20d/0x340
[    1.356897]  ? dcn20_resource_construct+0x46f/0xd30 [amdgpu]
[    1.357077]  dcn20_resource_construct+0x4b1/0xd30 [amdgpu]
...

Tested on: 5700XT (NAVI10 0x1002:0x731F 0x1DA2:0xE410 0xC1)

Cc: Christian König <christian.koenig@amd.com>
Cc: Hersen Wu <hersenxs.wu@amd.com>
Cc: Anson Jacob <Anson.Jacob@amd.com>
Cc: Harry Wentland <harry.wentland@amd.com>
Reviewed-by: Rodrigo Siqueira <Rodrigo.Siqueira@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Acked-by: Agustin Gutierrez <agustin.gutierrez@amd.com>
Signed-off-by: Anson Jacob <Anson.Jacob@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Align bw context with hw config when system resume

[Why]
When the vbios config and driver config are different, if we update
clock to lower before call program_timing and program_pixel_clk, garbage
appear.

[How]
Align bw context with hw config when system resume

Reviewed-by: Anthony Koo <Anthony.Koo@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Acked-by: Agustin Gutierrez <agustin.gutierrez@amd.com>
Signed-off-by: Lewis Huang <Lewis.Huang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu/pm: look up current_level for asics without pm callback

For asics without a callback, use the current level rather than 0xff.
This can avoid an unnecesary forced level set on older asics when
set by the user.

Reviewed-by: Kevin Wang <kevinyang.wang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu/display: add quirk handling for stutter mode

Stutter mode is a power saving feature on GPUs, however at
least one early raven system exhibits stability issues with
it.  Add a quirk to disable it for that system.

Bug: https://bugzilla.kernel.org/show_bug.cgi?id=214417
Fixes: 005440066f929b ("drm/amdgpu: enable gfxoff again on raven series (v2)")
Reviewed-by: Harry Wentland <harry.wentland@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: remove GPRs init for ALDEBARAN in gpu reset (v3)

Remove GPRs init for ALDEBARAN in gpu reset temporarily, will add the init once the
algorithm is stable.

v2: Only remove GPRs init in gpu reset.
v3: Suspend needs it, only skip it in gpu reset.

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

drm/amdkfd: Remove cu mask from struct queue_properties(v2)

Actually, cu_mask has been copied to mqd memory and
does't have to persist in queue_properties. Remove it
from queue_properties.

And use struct mqd_update_info to store such properties,
then pass it to update queue operation.

v2:
* Rename pqm_update_queue to pqm_update_queue_properties.
* Rename struct queue_update_info to struct mqd_update_info.
* Rename pqm_set_cu_mask to pqm_update_mqd.

Suggested-by: Felix Kuehling <Felix.Kuehling@amd.com>
Signed-off-by: Lang Yu <lang.yu@amd.com>
Reviewed-by: Felix Kuehling <Felix.Kuehling@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdkfd: Add an optional argument into update queue operation(v2)

Currently, queue is updated with data in queue_properties.
And all allocated resource in queue_properties will not
be freed until the queue is destroyed.

But some properties(e.g., cu mask) bring some memory
management headaches(e.g., memory leak) and make code
complex. Actually they have been copied to mqd and
don't have to persist in queue_properties.

Add an argument into update queue to pass such properties,
then we can remove them from queue_properties.

v2: Don't use void *.

Suggested-by: Felix Kuehling <Felix.Kuehling@amd.com>
Signed-off-by: Lang Yu <lang.yu@amd.com>
Reviewed-by: Felix Kuehling <Felix.Kuehling@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: skip GPRs init for some CU settings on ALDEBARAN

Skip GPRs init in specific condition since current GPRs init algorithm only works for some CU settings.

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

drm/amdgpu: Update TA version output in driver

TA version should only be displayed in firmware version column.

Signed-off-by: Candice Li <candice.li@amd.com>
Reviewed-by: John Clements <john.clements@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: fix a potential memory leak in amdgpu_device_fini_sw()

amdgpu_fence_driver_sw_fini() should be executed before
amdgpu_device_ip_fini(), otherwise fence driver resource
won't be properly freed as adev->rings have been tore down.

Fixes: 72c8c97b1522 ("drm/amdgpu: Split amdgpu_device_fini into early and late")

Signed-off-by: Lang Yu <lang.yu@amd.com>
Reviewed-by: Andrey Grodzovsky <andrey.grodzovsky@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdkfd: Separate pinned BOs destruction from general routine

Currently, all kfd BOs use same destruction routine. But pinned
BOs are not unpinned properly. Separate them from general routine.

v2 (Felix):
Add safeguard to prevent user space from freeing signal BO.
Kunmap signal BO in the event of setting event page error.
Just kunmap signal BO to avoid duplicating the code.

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

drm/amdkfd: restore userptr ignore bad address error

The userptr can be unmapped by application and still registered to
driver, restore userptr work return user pages will get -EFAULT bad
address error. Pretend this error as succeed. GPU access this userptr
will have VM fault later, it is better than application soft hangs with
stalled user mode queues.

v2: squash in warning fix (Alex)

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