drm/amdkfd: Update mapping if range attributes changed

Change SVM range mapping flags or access attributes don't trigger
migration, if range is already mapped on GPUs we should update GPU
mapping and pass flush_tlb flag true to amdgpu vm.

Change SVM range preferred_loc or migration granularity don't need
update GPU mapping, skip the validate_and_map.

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

drm/amdkfd: Add SVM range mapped_to_gpu flag

To avoid unnecessary unmap SVM range from GPUs if range is not mapped on
GPUs when migrating the range. This flag will also be used to flush TLB
when updating the existing mapping on GPUs.

It is protected by prange->migrate_mutex and mmap read lock in MMU
notifier callback.

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

drm/amd/display: 3.2.183

This version brings along following fixes:
- Keep tracking of DSC packed PPS for future use
- Maintain current link settings in link loss interrupt
- Remove DDC write and read size check
- Read PSR-SU cap DPCD for specific panel
- Don't pass HostVM by default on DCN3.1
- Reset cached PSR parameters after hibernate
- Add audio readback registers
- Update dcn315 clk table read

Acked-by: Tom Chung <chiahsuan.chung@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: Keep track of DSC packed PPS

[Why]
Store current packed PPS data in dc_stream_state for future use.

Reviewed-by: Aric Cyr <Aric.Cyr@amd.com>
Acked-by: Tom Chung <chiahsuan.chung@amd.com>
Signed-off-by: Ilya Bakoulin <Ilya.Bakoulin@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Remove unused integer

Integer no longer needed.

Reviewed-by: Martin Leung <Martin.Leung@amd.com>
Acked-by: Tom Chung <chiahsuan.chung@amd.com>
Signed-off-by: Dillon Varone <dillon.varone@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Maintain current link settings in link loss interrupt

[Why]
DP compliance test case 400.3.2.3 is failed because in link loss interrupt
the current link settings is not used in the DP link training.

[How]
In link loss interrupt, use the current link settings in the following DP
link training.

Reviewed-by: Wenjing Liu <Wenjing.Liu@amd.com>
Acked-by: Tom Chung <chiahsuan.chung@amd.com>
Signed-off-by: Gary Li <garyli12@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Remove ddc write and read size checking

[Why]
Customer found I2C over AUX using ADL_Display_DDCBlockAccess_Get
will fail when sending more than 256 bytes of data;

[How]
Remove the write and read size checking to allow sending data more
than 256 bytes;

Reviewed-by: Martin Leung <Martin.Leung@amd.com>
Acked-by: Tom Chung <chiahsuan.chung@amd.com>
Signed-off-by: Leo Ma <hanghong.ma@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: read PSR-SU cap DPCD for specific panel

[why & how]
For some specific eDP panel, we'd check the PSR-SU cap during boot
by reading the vendor specific DPCD, otherwise it will cause to
false report the eDP panel which supports PSR-SU as an non-PSR-SU
panel.

- add the vendor specific DPCD address in ddc_service_types header
- if specific eDP panel detected, check vendor specific DPCD for
  PSR-SU cap

Reviewed-by: Aurabindo Jayamohanan Pillai <Aurabindo.Pillai@amd.com>
Acked-by: Tom Chung <chiahsuan.chung@amd.com>
Signed-off-by: David Zhang <dingchen.zhang@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Don't pass HostVM by default on DCN3.1

[WHY]
Roll back previous change to stop passing this value by default, instead
add a debug flag to override to previous behaviour (or force HostVM calcs)

Reviewed-by: Nicholas Kazlauskas <Nicholas.Kazlauskas@amd.com>
Acked-by: Tom Chung <chiahsuan.chung@amd.com>
Signed-off-by: Michael Strauss <michael.strauss@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Reset cached PSR parameters after hibernate

[WHY]
After hibernate system might be using old invalid psr_power_opt and
psr_allow_active that never get reset

[HOW]
Reset cached Panel Self Refresh parameters when PSR is first configured
for eDP in dc_link_setup_psr.

Reviewed-by: Harry Vanzylldejong <harry.vanzylldejong@amd.com>
Acked-by: Tom Chung <chiahsuan.chung@amd.com>
Signed-off-by: Evgenii Krasnikov <Evgenii.Krasnikov@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Add Audio readback registers

[Why]
Can be useful for verifying the correctness of audio output.

Reviewed-by: Aric Cyr <Aric.Cyr@amd.com>
Acked-by: Tom Chung <chiahsuan.chung@amd.com>
Signed-off-by: Ilya Bakoulin <Ilya.Bakoulin@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: update dcn315 clk table read

Clean up the sequence by making sure clk_mgr always builds a
reasonable clock table regardless of what we read from smu
by moving all defaults from resource soc struct to clk_mgr.

Now the only thing resource soc update does is read
the clock table and apply any DC specific policy decisions
to how clocks are populated in dml soc.

Reviewed-by: Qingqing Zhuo <Qingqing.Zhuo@amd.com>
Acked-by: Tom Chung <chiahsuan.chung@amd.com>
Signed-off-by: Dmytro Laktyushkin <Dmytro.Laktyushkin@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: 3.2.182

This version brings along following improvements:
- Fix HDCP QUERY Error for eDP and Tiled
- Insert smu busy status before sending another request

Acked-by: Tom Chung <chiahsuan.chung@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: Fix HDCP QUERY Error for eDP and Tiled

[WHY]
For dio_output_encoder ID we are relying on SW concept which is
invisible to HW

[HOW]
Needed to create separate cases for when DPIA and non DPIA for
dio link encoder ID

Reviewed-by: Wenjing Liu <Wenjing.Liu@amd.com>
Reviewed-by: James Zhang <james.zhang@amd.com>
Acked-by: Tom Chung <chiahsuan.chung@amd.com>
Signed-off-by: Mustapha Ghaddar <mghaddar@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Insert smu busy status before sending another request

[why]
Need to check if result register is busy before sending another request

[how]
Call method to check if result register is busy

Reviewed-by: Charlene Liu <Charlene.Liu@amd.com>
Acked-by: Tom Chung <chiahsuan.chung@amd.com>
Signed-off-by: Oliver Logush <oliver.logush@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdkfd: Ignore bogus signals from MEC efficiently

MEC firmware sometimes sends signal interrupts without a valid context ID
on end of pipe events that don't intend to signal any HSA signals.
This triggers the slow path in kfd_signal_event_interrupt that scans the
entire event page for signaled events. Detect these signals in the top
half interrupt handler to stop processing them as early as possible.

Because we now always treat event ID 0 as invalid, reserve that ID during
process initialization.

v2: Update firmware version checks to support more GPUs

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: Remove useless kfree

After alloc fail, we do not need to kfree.

Signed-off-by: Haowen Bai <baihaowen@meizu.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: Ta fw needs to be loaded for SRIOV aldebaran

Load ta fw during psp_init_sriov_microcode to enable XGMI.
It is required to be loaded by both guest and host starting
from Arcturus. Cap fw needs to be loaded first.

Signed-off-by: David Yu <David.Yu@amd.com>
Reviewed-by: Shaoyun.liu <Shaoyun.liu@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/pm: fix the deadlock issue observed on SI

The adev->pm.mutx is already held at the beginning of
amdgpu_dpm_compute_clocks/amdgpu_dpm_enable_uvd/amdgpu_dpm_enable_vce.
But on their calling path, amdgpu_display_bandwidth_update will be
called and thus its sub functions amdgpu_dpm_get_sclk/mclk. They
will then try to acquire the same adev->pm.mutex and deadlock will
occur.

By placing amdgpu_display_bandwidth_update outside of adev->pm.mutex
protection(considering logically they do not need such protection) and
restructuring the call flow accordingly, we can eliminate the deadlock
issue. This comes with no real logics change.

Fixes: 3712e7a49459 ("drm/amd/pm: unified lock protections in amdgpu_dpm.c")
Reported-by: Paul Menzel <pmenzel@molgen.mpg.de>
Reported-by: Arthur Marsh <arthur.marsh@internode.on.net>
Link: https://lore.kernel.org/all/9e689fea-6c69-f4b0-8dee-32c4cf7d8f9c@molgen.mpg.de/
BugLink: https://gitlab.freedesktop.org/drm/amd/-/issues/1957
Signed-off-by: Evan Quan <evan.quan@amd.com>
Reviewed-by: Lijo Lazar <lijo.lazar@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: add RAS fatal error interrupt handler

The fatal error handler is independent from general ras interrupt
handler since there is no related IH ring.

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: add RAS poison consumption handler (v2)

Add support for general RAS poison consumption handler.

v2: remove callback function for poison consumption.

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: add RAS poison creation handler (v2)

Prepare for the implementation of poison consumption handler.

v2: separate umc handler from poison creation.

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: use kvcalloc() instead of kvmalloc() in kfd_migrate

simplify programming with existing functions.

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

drm/amd/amdgpu: Update PF2VF header

- In the latest version of the header, there is a variable name change.
  This should not cause any backward compatibility since the variable is
  at the same offset in the struct.

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

drm/amd/amdgpu: Properly indent PF2VF header

- Clean up the identation in the header file

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

drm/amd/amdgpu: Update MIT license in SRIOV msg header

- Update MIT license header

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

drm/amdgpu/display: make hubp31_program_extended_blank static

It's not used outside of dcn31_hubp.c.

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

drm/amd/display: Fix memory leak in dcn21_clock_source_create

When dcn20_clk_src_construct() fails, we need to release clk_src.

Fixes: 6f4e6361c3ff ("drm/amd/display: Add Renoir resource (v2)")
Signed-off-by: Miaoqian Lin <linmq006@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Remove useless code

aux_rep only memset but no use at all, so we drop it.

Signed-off-by: Haowen Bai <baihaowen@meizu.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: don't runtime suspend if there are displays attached (v3)

We normally runtime suspend when there are displays attached if they
are in the DPMS off state, however, if something wakes the GPU
we send a hotplug event on resume (in case any displays were connected
while the GPU was in suspend) which can cause userspace to light
up the displays again soon after they were turned off.

Prior to
commit 087451f372bf76 ("drm/amdgpu: use generic fb helpers instead of setting up AMD own's."),
the driver took a runtime pm reference when the fbdev emulation was
enabled because we didn't implement proper shadowing support for
vram access when the device was off so the device never runtime
suspended when there was a console bound.  Once that commit landed,
we now utilize the core fb helper implementation which properly
handles the emulation, so runtime pm now suspends in cases where it did
not before.  Ultimately, we need to sort out why runtime suspend in not
working in this case for some users, but this should restore similar
behavior to before.

v2: move check into runtime_suspend
v3: wake ups -> wakeups in comment, retain pm_runtime behavior in
    runtime_idle callback

Fixes: 087451f372bf76 ("drm/amdgpu: use generic fb helpers instead of setting up AMD own's.")
Link: https://lore.kernel.org/r/20220403132322.51c90903@darkstar.example.org/
Tested-by: Michele Ballabio <ballabio.m@gmail.com>
Reviewed-by: Evan Quan <evan.quan@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

Revert "drm/amdkfd: only allow heavy-weight TLB flush on some ASICs for SVM too"

This reverts commit 36bf93216ecbe399c40c5e0486f0f0e3a4afa69e.

It causes SVM regressions on Vega10 with XNACK-ON. Just revert it
at the moment.

./kfdtest --gtest_filter=KFDSVMRangeTest.MigratePolicyTest

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

drm/amdgpu: Add debugfs TA load/unload/invoke support

v1:
Add debugfs support to load/unload/invoke TA in runtime.

v2:
1. Update some variables to static.
2. Use PAGE_ALIGN to calculate shared buf size directly.
3. Remove fp check.
4. Update debugfs from read to write.

Signed-off-by: John Clements <john.clements@amd.com>
Signed-off-by: Candice Li <candice.li@amd.com>
Reviewed-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: Use indirect buffer and save response status for TA load/invoke

The upcoming TA debugfs interface needs to use indirect buffer
when performing TA invoke and check psp response status for TA
load and invoke.

Signed-off-by: John Clements <john.clements@amd.com>
Signed-off-by: Candice Li <candice.li@amd.com>
Reviewed-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: CRIU add support for GWS queues

Add support to checkpoint/restore GWS (Global Wave Sync) queues.

Signed-off-by: David Yat Sin <david.yatsin@amd.com>
Reviewed-by: Felix Kuehling <Felix.Kuehling@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdkfd: Fix GWS queue count

dqm->gws_queue_count and pdd->qpd.mapped_gws_queue need to be updated
each time the queue gets evicted.

Fixes: b8020b0304c8 ("drm/amdkfd: Enable over-subscription with >1 GWS queue")
Signed-off-by: David Yat Sin <david.yatsin@amd.com>
Reviewed-by: Felix Kuehling <Felix.Kuehling@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

MAINTAINERS: add docs entry to AMDGPU

To make sure maintainers of amdgpu drivers are aware of any changes
 in their documentation, add its entry to MAINTAINERS.

Acked-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Tales Lelo da Aparecida <tales.aparecida@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

Documentation/gpu: Add entries to amdgpu glossary

Add missing acronyms to the amdgppu glossary.

Closes: https://gitlab.freedesktop.org/drm/amd/-/issues/1939
Acked-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Tales Lelo da Aparecida <tales.aparecida@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/radeon/kms: change evergreen_default_state table from global to static

evergreen_default_state and evergreen_default_size are only
used in evergreen.c.  Single file symbols should be static.
So move their definitions to evergreen_blit_shaders.h
and change their storage-class-specifier to static.

Remove unneeded evergreen_blit_shader.c

evergreen_ps/vs definitions were removed with
commit 4f8629675800 ("drm/radeon/kms: remove r6xx+ blit copy routines")
So their declarations in evergreen_blit_shader.h
are not needed, so remove them.

Signed-off-by: Tom Rix <trix@redhat.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: add virtual_setup_stream_attribute decl to header

Smatch reports this issue
virtual_link_hwss.c:32:6: warning: symbol
  'virtual_setup_stream_attribute' was not declared.
  Should it be static?

virtual_setup_stream_attribute is only used in
virtual_link_hwss.c, but the other functions in the
file are declared in the header file and used elsewhere.
For consistency, add the virtual_setup_stream_attribute
decl to virtual_link_hwss.h.

Signed-off-by: Tom Rix <trix@redhat.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/pm: fix double free in si_parse_power_table()

In function si_parse_power_table(), array adev->pm.dpm.ps and its member
is allocated. If the allocation of each member fails, the array itself
is freed and returned with an error code. However, the array is later
freed again in si_dpm_fini() function which is called when the function
returns an error.

This leads to potential double free of the array adev->pm.dpm.ps, as
well as leak of its array members, since the members are not freed in
the allocation function and the array is not nulled when freed.
In addition adev->pm.dpm.num_ps, which keeps track of the allocated
array member, is not updated until the member allocation is
successfully finished, this could also lead to either use after free,
or uninitialized variable access in si_dpm_fini().

Fix this by postponing the free of the array until si_dpm_fini() and
increment adev->pm.dpm.num_ps everytime the array member is allocated.

Signed-off-by: Keita Suzuki <keitasuzuki.park@sslab.ics.keio.ac.jp>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: make hubp1_wait_pipe_read_start() static

It's a local function, let's make it static.

AGD: remove prototype in dcn10_hubp.h

Signed-off-by: Tales Lelo da Aparecida <tales.aparecida@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

amdgpu/pm: Clarify documentation of error handling in send_smc_mesg

Clarify the smu_cmn_send_smc_msg_with_param documentation to mention two
cases exist where messages are silently dropped with no error returned.
These cases occur in unusual situations where either:
 1. the message type is not allowed to a virtual GPU, or
 2. a PCI recovery is underway and the HW is not yet in sync with the SW

For more details see
 commit 4ea5081c82c4 ("drm/amd/powerplay: enable SMC message filter")
 commit bf36b52e781d ("drm/amdgpu: Avoid accessing HW when suspending SW state")

(v2)
  Reworked with suggestions from Luben & Paul

(v3)
  Updated wording as per Luben's feedback
  Corrected error stating all messages denied on virtual GPU
  (each GPU has mask of which messages are allowed)

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

drm/amdgpu/pm: fix the null pointer while the smu is disabled

It needs to check if the pp_funcs is initialized while release the
context, otherwise it will trigger null pointer panic while the software
smu is not enabled.

[ 1109.404555] BUG: kernel NULL pointer dereference, address: 0000000000000078
[ 1109.404609] #PF: supervisor read access in kernel mode
[ 1109.404638] #PF: error_code(0x0000) - not-present page
[ 1109.404657] PGD 0 P4D 0
[ 1109.404672] Oops: 0000 [#1] PREEMPT SMP NOPTI
[ 1109.404701] CPU: 7 PID: 9150 Comm: amdgpu_test Tainted: G           OEL    5.16.0-custom #1
[ 1109.404732] Hardware name: innotek GmbH VirtualBox/VirtualBox, BIOS VirtualBox 12/01/2006
[ 1109.404765] RIP: 0010:amdgpu_dpm_force_performance_level+0x1d/0x170 [amdgpu]
[ 1109.405109] Code: 5d c3 44 8b a3 f0 80 00 00 eb e5 66 90 0f 1f 44 00 00 55 48 89 e5 41 57 41 56 41 55 41 54 53 48 83 ec 08 4c 8b b7 f0 7d 00 00 <49> 83 7e 78 00 0f 84 f2 00 00 00 80 bf 87 80 00 00 00 48 89 fb 0f
[ 1109.405176] RSP: 0018:ffffaf3083ad7c20 EFLAGS: 00010282
[ 1109.405203] RAX: 0000000000000000 RBX: ffff9796b1c14600 RCX: 0000000002862007
[ 1109.405229] RDX: ffff97968591c8c0 RSI: 0000000000000001 RDI: ffff9796a3700000
[ 1109.405260] RBP: ffffaf3083ad7c50 R08: ffffffff9897de00 R09: ffff979688d9db60
[ 1109.405286] R10: 0000000000000000 R11: ffff979688d9db90 R12: 0000000000000001
[ 1109.405316] R13: ffff9796a3700000 R14: 0000000000000000 R15: ffff9796a3708fc0
[ 1109.405345] FS:  00007ff055cff180(0000) GS:ffff9796bfdc0000(0000) knlGS:0000000000000000
[ 1109.405378] CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
[ 1109.405400] CR2: 0000000000000078 CR3: 000000000a394000 CR4: 00000000000506e0
[ 1109.405434] Call Trace:
[ 1109.405445]  <TASK>
[ 1109.405456]  ? delete_object_full+0x1d/0x20
[ 1109.405480]  amdgpu_ctx_set_stable_pstate+0x7c/0xa0 [amdgpu]
[ 1109.405698]  amdgpu_ctx_fini.part.0+0xcb/0x100 [amdgpu]
[ 1109.405911]  amdgpu_ctx_do_release+0x71/0x80 [amdgpu]
[ 1109.406121]  amdgpu_ctx_ioctl+0x52d/0x550 [amdgpu]
[ 1109.406327]  ? _raw_spin_unlock+0x1a/0x30
[ 1109.406354]  ? drm_gem_handle_delete+0x81/0xb0 [drm]
[ 1109.406400]  ? amdgpu_ctx_get_entity+0x2c0/0x2c0 [amdgpu]
[ 1109.406609]  drm_ioctl_kernel+0xb6/0x140 [drm]

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

drm/amdkfd: only allow heavy-weight TLB flush on some ASICs for SVM too

The idea is from
commit a50fe7078035 ("drm/amdkfd: Only apply heavy-weight TLB flush on Aldebaran")
and
commit f61c40c0757a ("drm/amdkfd: enable heavy-weight TLB flush on Arcturus").

At the moment, heavy-weight TLB could cause problems on ASICs except
Aldebaran and Arcturus.

A simple hipMallocManaged/hipFree program could trigger this issue.

[   97.787657] amdgpu 0000:01:00.0: amdgpu: wait for kiq fence error: 0.
[  106.868758] amdgpu: qcm fence wait loop timeout expired
[  106.868966] amdgpu: The cp might be in an unrecoverable state due to an unsuccessful queues preemption
[  106.869203] amdgpu: Failed to evict process queues
[  106.869261] amdgpu: Failed to quiesce KFD

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: move kfd_flush_tlb_after_unmap into kfd_priv.h

To make kfd_flush_tlb_after_unmap visible in kfd_svm.c,
move it into kfd_priv.h. And change it to an inline function.

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/amd/amdgpu: Remove static from variable in RLCG Reg RW

[why]
These static variables save the RLC Scratch registers address.
When we install multiple GPUs (for example: XGMI setting) and
multiple GPUs call the function at same time. The RLC Scratch
registers address are changed each other. Then it caused
reading/writing from/to wrong GPU.

[how]
Removed the static from the variables. The variables are
on the stack.

Fixes: 5d447e29670148 ("drm/amdgpu: add helper for rlcg indirect reg access")
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Gavin Wan <Gavin.Wan@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdkfd: fix race condition in kfd_wait_on_events

Add the waiters to the wait queue during initialization, while holding the
event spinlock. Otherwise the waiter will not get activated if the event
signals before being added to the wait queue.

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>

Revert "drm/amd/display: Pass HostVM enable flag into DCN3.1 DML"

This reverts commit 863fa85e6a01c8b239009825dd9de1f64d7d020a.

While we were testing DCN3.1 with a hub, we noticed that only one of 2
connected displays lights up when using some specific display
resolution. In summary, this was the setup:

1. Displays:
 * Sharp LQ156M1JW26 (eDP): 1080@240
 * BENQ SW320 (DP): 4k@60
 * BENQ EX3203R (DP): 4k@60
2. Hub: Club3D CSV-7300
3. ASIC: DCN3.1

After bisecting this issue, we figured out the commit mentioned above
introduced this issue. We are investigating why this patch introduced
this regression, but we need to revert it for now.

Cc: Harry Wentland <harry.wentland@amd.com>
Cc: Mark Broadworth <Mark.Broadworth@amd.com>
Cc: Michael Strauss <michael.strauss@amd.com>
Acked-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Rodrigo Siqueira <Rodrigo.Siqueira@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: Fix one use-after-free of VM

VM might already be freed when amdgpu_vm_tlb_seq_cb() is called.
We see the calltrace below.

Fix it by keeping the last flush fence around and wait for it to signal

BUG kmalloc-4k (Not tainted): Poison overwritten

0xffff9c88630414e8-0xffff9c88630414e8 @offset=5352. First byte 0x6c
instead of 0x6b Allocated in amdgpu_driver_open_kms+0x9d/0x360 [amdgpu]
age=44 cpu=0 pid=2343
 __slab_alloc.isra.0+0x4f/0x90
 kmem_cache_alloc_trace+0x6b8/0x7a0
 amdgpu_driver_open_kms+0x9d/0x360 [amdgpu]
 drm_file_alloc+0x222/0x3e0 [drm]
 drm_open+0x11d/0x410 [drm]
Freed in amdgpu_driver_postclose_kms+0x3e9/0x550 [amdgpu] age=22 cpu=1
pid=2485
 kfree+0x4a2/0x580
 amdgpu_driver_postclose_kms+0x3e9/0x550 [amdgpu]
 drm_file_free+0x24e/0x3c0 [drm]
 drm_close_helper.isra.0+0x90/0xb0 [drm]
 drm_release+0x97/0x1a0 [drm]
 __fput+0xb6/0x280
 ____fput+0xe/0x10
 task_work_run+0x64/0xb0

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

drm/amdkfd: potential NULL dereference in kfd_set/reset_event()

If lookup_event_by_id() returns a NULL "ev" pointer then the
spin_lock(&ev->lock) will crash.  This was detected by Smatch:

    drivers/gpu/drm/amd/amdgpu/../amdkfd/kfd_events.c:644 kfd_set_event()
    error: we previously assumed 'ev' could be null (see line 639)

Fixes: 5273e82c5f47 ("drm/amdkfd: Improve concurrency of event handling")
Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Reviewed-by: Felix Kuehling <Felix.Kuehling@amd.com>
Signed-off-by: Felix Kuehling <Felix.Kuehling@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdkfd: Cleanup IO links during KFD device removal

Currently, the IO-links to the device being removed from topology,
are not cleared. As a result, there would be dangling links left in
the KFD topology. This patch aims to fix the following:
1. Cleanup all IO links to the device being removed.
2. Ensure that node numbering in sysfs and nodes proximity domain
   values are consistent after the device is removed:
   a. Adding a device and removing a GPU device are made mutually
      exclusive.
   b. The global proximity domain counter is no longer required to be
      an atomic counter. A normal 32-bit counter can be used instead.
3. Update generation_count to let user-mode know that topology has
   changed due to device removal.

CC: Shuotao Xu <shuotaoxu@microsoft.com>
Reviewed-by: Shuotao Xu <shuotaoxu@microsoft.com>
Reviewed-by: Felix Kuehling <Felix.Kuehling@amd.com>
Signed-off-by: Mukul Joshi <mukul.joshi@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/amdgpu: Not request init data for MS_HYPERV with vega10

MS_HYPERV with vega10 doesn't have the interface to process
request init data msg.

Check hypervisor type to not send the request for MS_HYPERV.

Signed-off-by: Yongqiang Sun <yongqiang.sun@amd.com>
Reviewed-by: Alice Wong <shiwei.wong@amd.com>
Acked-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdkfd: shrink bitmap size in struct svm_validate_context

A MAX_GPU_INSTANCE bits bitmap will suffice.

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/amd/display: extract update stream allocation to link_hwss

[why]
Extract update stream allocation table into link hwss as part of the
link hwss refactor work.

Reviewed-by: George Shen <George.Shen@amd.com>
Reviewed-by: Fangzhi Zuo <Jerry.Zuo@amd.com>
Acked-by: Pavle Kotarac <Pavle.Kotarac@amd.com>
Signed-off-by: Wenjing Liu <wenjing.liu@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: implement shared PSR-SU sink validation helper

[why]
creating a generic helper for AMD specific PSR-SU sink validation.
Moving the function to the power module to reference it across all
OS.

[how]
- drop PSRSU specific sink validation helper and move to power
  module by reading PSR version and other PSR caps
- call the new helper from linux DM (amdgpu_dm_psr)

Acked-by: Pavle Kotarac <Pavle.Kotarac@amd.com>
Acked-by: Tom Chung <chiahsuan.chung@amd.com>
Signed-off-by: David Zhang <dingchen.zhang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm: add PSR2 support and capability definition as per eDP 1.5

[why & how]
As per eDP 1.5 spec, add the below two DPCD bit fields for PSR-SU
support and capability:

1. DP_PSR2_WITH_Y_COORD_ET_SUPPORTED
2. DP_PSR2_SU_AUX_FRAME_SYNC_NOT_NEEDED

changes in v2
------------------
* fixed the typo
* explicitly list what DPCD bit fields are added

Signed-off-by: David Zhang <dingchen.zhang@amd.com>
Reviewed-by: Harry Wentland <harry.wentland@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: 3.2.181

Title: DC Patches Apri 6, 2022

This DC patchset brings improvements in multiple areas. In summary, we highlight:

*Disabling Z10 on DCN31
*Fix issue breaking 32bit Linux build
*Fix inconsistent timestamp type
*Add DCN30 support FEC init
*Fix crash on setting VRR with no display connected
*Disable FEC if DSC not supported for EDP
*Add odm seamless boot support
*Select correct DTO source
*Power down hardware if timer not trigger

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

drm/amd/display: Select correct DTO source

[WHY&HOW]
Change criteria for setting DTO source value, and always set it regardless of
the signal type.

Reviewed-by: Ariel Bernstein <Eric.Bernstein@amd.com>
Acked-by: Pavle Kotarac <Pavle.Kotarac@amd.com>
Signed-off-by: Dillon Varone <dillon.varone@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdkfd: Asynchronously free events

The synchronize_rcu call in destroy_events can take several ms, which
noticeably slows down applications destroying many events. Use kfree_rcu
to free the event structure asynchronously and eliminate the
synchronize_rcu call in the user thread.

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/amd/display: expose skip_detection_link_training to debugfs

[Why] within dc link detecion, dp link training will be
executed for external sst dp. for debug purpose, we may
need skip dp link training.

[How] expose dc debug option to skip_detection_link_training
to debugfs

Reviewed-by: Roman Li <Roman.Li@amd.com>
Acked-by: Pavle Kotarac <Pavle.Kotarac@amd.com>
Signed-off-by: hersen wu <hersenxs.wu@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Add dtb clock to dc_clocks

Reviewed-by: Ariel Bernstein <Eric.Bernstein@amd.com>
Acked-by: Pavle Kotarac <Pavle.Kotarac@amd.com>
Signed-off-by: Dillon Varone <dillon.varone@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Fix inconsistent timestamp type

[WHY]
An unsigned int timestamp variable is assigned with an unsigned
long long value. Also, the assignment directly converts the
tick value to us without using built-in get elapsed time function.

[HOW]
Cast the assigned value correctly and also use built-in function
to get the timestamp in the unit we want.

v2: squash in 64 bit division fix

Reviewed-by: Aric Cyr <Aric.Cyr@amd.com>
Acked-by: Pavle Kotarac <Pavle.Kotarac@amd.com>
Signed-off-by: Angus Wang <Angus.Wang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: add DCN30 support FEC init

[Why]
FEC init used on DCN30.

[How]
Check fec active when HW init.

Co-authored-by: Jingwen Zhu <Jingwen.Zhu@amd.com>
Reviewed-by: Wenjing Liu <Wenjing.Liu@amd.com>
Acked-by: Pavle Kotarac <Pavle.Kotarac@amd.com>
Signed-off-by: Jingwen Zhu <Jingwen.Zhu@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Add odm seamless boot support

[WHY]
Implement changes to transition from Pre-OS odm to
Post-OS odm support. Seamless boot case is also
considered.

[HOW]
Revised validation logic when marking for seamless
boot. Init resources accordingly when Pre-OS has
odm enabled. Reset odm and det size when transitioning
Pre-OS odm to Post-OS non-odm to avoid corruption.
Apply logic to set odm accordingly upon commit.

Reviewed-by: Nicholas Kazlauskas <Nicholas.Kazlauskas@amd.com>
Acked-by: Pavle Kotarac <Pavle.Kotarac@amd.com>
Signed-off-by: Duncan Ma <Duncan.Ma@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Always update DPPCLK_DTO when clock increase

[why]
Need to update the update_clock sequence to a fully tested sequence for
dcn30

[how]
Removed the check to see if clock is lowered

Reviewed-by: Charlene Liu <Charlene.Liu@amd.com>
Acked-by: Pavle Kotarac <Pavle.Kotarac@amd.com>
Signed-off-by: Oliver Logush <oliver.logush@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Insert pulling smu busy status before sending another request

[why]
Make sure smu is not busy before sending another request, this is to
prevent stress failures from MS.

[how]
Check to make sure the SMU fw busy signal is cleared before sending
another request

Reviewed-by: Charlene Liu <Charlene.Liu@amd.com>
Reviewed-by: Nicholas Kazlauskas <Nicholas.Kazlauskas@amd.com>
Acked-by: Pavle Kotarac <Pavle.Kotarac@amd.com>
Signed-off-by: Oliver Logush <oliver.logush@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Power down hardware if timer not trigger

[WHY]
In headless systems, if SetMode/Power down timer
is not called, hardware will not be powered down
causing HW/SW discrepancies. Powering down hardware
on SetPowerState to D3 will ensure SW/HW state is accurate.

[HOW]
1. If PowerDownThread timer is not trigger but OS call
SetPowerState to D3, power down hardware.
2. Update HDMI hang w/a to apply to all TMDS signals on
headless system

Reviewed-by: Martin Leung <Martin.Leung@amd.com>
Acked-by: Pavle Kotarac <Pavle.Kotarac@amd.com>
Signed-off-by: Paul Hsieh <paul.hsieh@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: remove dtbclk_ss compensation for dcn316

[why]
dcn316's dtbclk is from non_ss clock source.
no compensation required here.

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

drm/amd/display: ensure PSR force_static flag can always be set

[HOW&WHY]
Make sure psr_force_static() can always be called regardless of
psr_allow_active value.

Reviewed-by: Harry Vanzylldejong <harry.vanzylldejong@amd.com>
Reviewed-by: Evgenii Krasnikov <Evgenii.Krasnikov@amd.com>
Reviewed-by: Nicholas Choi <Nicholas.Choi@amd.com>
Acked-by: Pavle Kotarac <Pavle.Kotarac@amd.com>
Signed-off-by: Evgenii Krasnikov <Evgenii.Krasnikov@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Fix crash on setting VRR with no display connected

[HOW&WHY]
VRR was getting set at the same time
the timing generator would be null when there was no display
connected. Added null check to the timing generator variable
so it does not get referenced if it is null.

Reviewed-by: Harry Vanzylldejong <harry.vanzylldejong@amd.com>
Reviewed-by: Evgenii Krasnikov <Evgenii.Krasnikov@amd.com>
Reviewed-by: Nicholas Choi <Nicholas.Choi@amd.com>
Acked-by: Pavle Kotarac <Pavle.Kotarac@amd.com>
Signed-off-by: Harry VanZyllDeJong <hvanzyll@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Disable FEC if DSC not supported for EDP

[WHY]
Screen was seen corrupted for a few ms
when switching both ways. There was also
not enough bandwidth for HDR to be
enabled in HG disabled mode.
This was due to FEC being
enabled although DSC was not supported
or disabled for the EDP.

[HOW]
Check for EDP DSC support in DC caps
or if DSC should be disabled for EDP
before enabling FEC for EDP.

Reviewed-by: Harry Vanzylldejong <harry.vanzylldejong@amd.com>
Reviewed-by: Evgenii Krasnikov <Evgenii.Krasnikov@amd.com>
Reviewed-by: Nicholas Choi <Nicholas.Choi@amd.com>
Acked-by: Pavle Kotarac <Pavle.Kotarac@amd.com>
Signed-off-by: Iswara Nagulendran <inagulen@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/radeon: remove r600_blit_shaders.[c|h]

The only use of the global variables in r600_blit_shaders.c
were in the old drivers/gpu/drm/radeon/r600_blit.c
This file was removed in
commit 8333f607a631 ("drm/radeon: remove UMS support")

So remove the r600_blit_shaders.[c|h] files

Signed-off-by: Tom Rix <trix@redhat.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Disallow entering PSR when panel is disconnected

[WHY]
The dGPU cannot enter PSR when it is not connected to a panel.

[HOW]
Added a check to dc_link_set_psr_allow_active
which returns early if panel is disconnected.

Reviewed-by: Harry Vanzylldejong <harry.vanzylldejong@amd.com>
Reviewed-by: Evgenii Krasnikov <Evgenii.Krasnikov@amd.com>
Reviewed-by: Nicholas Choi <Nicholas.Choi@amd.com>
Acked-by: Pavle Kotarac <Pavle.Kotarac@amd.com>
Signed-off-by: Max Erenberg <merenber@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Disabling Z10 on DCN31

[WHY]
Z10 is should not be enabled by default on DCN31.

[HOW]
Using DC debug flags to disable Z10 by default on DCN31.

Reviewed-by: Eric Yang <Eric.Yang2@amd.com>
Acked-by: Pavle Kotarac <Pavle.Kotarac@amd.com>
Signed-off-by: Saaem Rizvi <syerizvi@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: do not wait for mpc idle if tg is disabled

[Why]
When booting, the driver waits for the MPC idle bit to be set as part of
pipe initialization. However, on some systems this occurs before OTG is
enabled, and since the MPC idle bit won't be set until the vupdate
signal occurs (which requires OTG to be enabled), this never happens and
the wait times out. This can add hundreds of milliseconds to the boot
time.

[How]
Do not wait for mpc idle if tg is disabled

Reviewed-by: Jun Lei <Jun.Lei@amd.com>
Acked-by: Pavle Kotarac <Pavle.Kotarac@amd.com>
Signed-off-by: Josip Pavic <Josip.Pavic@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: undo clearing of z10 related function pointers

[Why]
Z10 and S0i3 have some shared path. Previous code clean up ,
incorrectly removed these pointers, which breaks s0i3 restore

[How]
Do not clear the function pointers based on Z10 disable.

Reviewed-by: Nicholas Kazlauskas <Nicholas.Kazlauskas@amd.com>
Acked-by: Pavle Kotarac <Pavle.Kotarac@amd.com>
Signed-off-by: Eric Yang <Eric.Yang2@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: add umc query error status function

In order to debug ras error, driver will print IPID/SYND/MISC0
register value if detect correctable or uncorrectable error.
Provide umc_query_error_status_helper function to reduce code
redundancy.

Signed-off-by: Stanley.Yang <Stanley.Yang@amd.com>
Reviewed-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/display: fix 64 bit divide in freesync code

Use div_u64() rather than a a 64 bit divide.

Fixes: 3fe5739db48843 ("drm/amd/display: Add flip interval workaround")
Reviewed-by: Nathan Chancellor <nathan@kernel.org>
Reported-by: kernel test robot <lkp@intel.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Cc: Angus Wang <Angus.Wang@amd.com>
Cc: Anthony Koo <Anthony.Koo@amd.com>
Cc: Aric Cyr <Aric.Cyr@amd.com>
Cc: Nathan Chancellor <nathan@kernel.org>

drm/amdgpu: Fix incorrect enum type

Instead of the 'amdgpu_ring_priority_level' type,
the 'amdgpu_gfx_pipe_priority' type was used,
which is an error when setting ring priority.
This is a minor error, but may cause problems in the future.

Instead of AMDGPU_RING_PRIO_2 = 2, we can use AMDGPU_RING_PRIO_MAX = 3,
but AMDGPU_RING_PRIO_2 = 2 is used for compatibility with
AMDGPU_GFX_PIPE_PRIO_HIGH = 2, and not change the behavior of the
code.

Signed-off-by: Grigory Vasilyev <h0tc0d3@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/radeon: change cayman_default_state table from global to static

cayman_default_state and cayman_default_size are only
used in ni.c.  Single file symbols should be static.
So move their definitions to cayman_blit_shaders.h
and change their storage-class-specifier to static.

Remove unneeded cayman_blit_shader.c

cayman_ps/vs definitions were removed with
commit 4f8629675800 ("drm/radeon/kms: remove r6xx+ blit copy routines")
So their declarations in cayman_blit_shader.h
are not needed, so remove them.

Signed-off-by: Tom Rix <trix@redhat.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/amdgpu: Update debugfs GCA data

The data revision was not changed to 5 from 4 when the CG flags
were extended to 64-bits.  Since this was missed I took
the opportunity to add future upper 64-bits of PG flags
as well so we don't need to bump it again when that comes.

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

drm/amd/amdgpu: Fix asm/hypervisor.h build error.

Add CONFIG_X86 check to fix the build error.

Fixes: 49aa98ca30cd18 ("drm/amd/amdgpu: Only reserve vram for firmware with vega9 MS_HYPERV host.")
Reported-by: kernel test robot <lkp@intel.com>
Signed-off-by: Yongqiang Sun <yongqiang.sun@amd.com>
Acked-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdkfd: Handle drain retry fault race with XNACK mode change

Application could change XNACK enabled to disabled while KFD is draining
stale retry fault, therefore the check for whether to drain retry faults
must be before the check for whether xnack_enabled, to avoid report
incorrect vm fault after application changes XNACK mode.

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

drm/amdgpu: Use flexible array member

Use flexible array member in ip discovery struct as recommended[1].

[1] https://www.kernel.org/doc/html/latest/process/deprecated.html#zero-length-and-one-element-arrays

v2: squash in struct_size fixes

Signed-off-by: Lijo Lazar <lijo.lazar@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: expand cg_flags from u32 to u64

With this, we can support more CG flags.

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

drm/amd/amdgpu: Only reserve vram for firmware with vega9 MS_HYPERV host.

driver loading failed on VEGA10 SRIOV VF with linux host due to a wide
range of stolen reserved vram.
Since VEGA10 SRIOV VF need to reserve vram for firmware with windows
Hyper_V host specifically, check hypervisor type to only reserve
memory for it, and the range of the reserved vram can be limited
to between 5M-7M area.

Fixes: faad5ccac1eaae ("drm/amdgpu: Add stolen reserved memory for MI25 SRIOV.")
Signed-off-by: Yongqiang Sun <yongqiang.sun@amd.com>
Acked-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdkfd: Fix NULL pointer dereference

Check that adev->gfx.ras is valid before using it.

Fixes: 6475ae2b742876 ("drm/amdgpu: add UTCL2 RAS poison query for Aldebaran (v2)")
CC: Tao Zhou <tao.zhou1@amd.com>
Signed-off-by: Felix Kuehling <Felix.Kuehling@amd.com>
Reviewed-by: Mukul Joshi <mukul.joshi@amd.com>
Reviewed-by: Hawking Zhang <Hawking.Zhang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: cleanup extern usage in function definition

Smatch reports this issue
hdcp1_execution.c:500:29: warning: function
  'mod_hdcp_hdcp1_dp_execution' with external linkage
  has definition

The storage-class-specifier extern is not needed in a
definition, so remove it.

Reviewed-by: Harry Wentland <harry.wentland@amd.com>
Signed-off-by: Tom Rix <trix@redhat.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: Enable gfxoff quirk on MacBook Pro

Enabling gfxoff quirk results in perfectly usable graphical user
interface on MacBook Pro (15-inch, 2019) with Radeon Pro Vega 20 4 GB.

Without the quirk, X server is completely unusable as every few seconds
there is gpu reset due to ring gfx timeout.

Signed-off-by: Tomasz Moń <desowin@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: Ensure HDA function is suspended before ASIC reset

DP/HDMI audio on AMD PRO VII stops working after S3:
[  149.450391] amdgpu 0000:63:00.0: amdgpu: MODE1 reset
[  149.450395] amdgpu 0000:63:00.0: amdgpu: GPU mode1 reset
[  149.450494] amdgpu 0000:63:00.0: amdgpu: GPU psp mode1 reset
[  149.983693] snd_hda_intel 0000:63:00.1: refused to change power state from D0 to D3hot
[  150.003439] amdgpu 0000:63:00.0: refused to change power state from D0 to D3hot
...
[  155.432975] snd_hda_intel 0000:63:00.1: CORB reset timeout#2, CORBRP = 65535

The offending commit is daf8de0874ab5b ("drm/amdgpu: always reset the asic in
suspend (v2)"). Commit 34452ac3038a7 ("drm/amdgpu: don't use BACO for
reset in S3 ") doesn't help, so the issue is something different.

Assuming that to make HDA resume to D0 fully realized, it needs to be
successfully put to D3 first. And this guesswork proves working, by
moving amdgpu_asic_reset() to noirq callback, so it's called after HDA
function is in D3.

Fixes: daf8de0874ab5b ("drm/amdgpu: always reset the asic in suspend (v2)")
Signed-off-by: Kai-Heng Feng <kai.heng.feng@canonical.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: fix VCN 3.1.2 firmware name

Drop the trailing vcn.

Fixes: afc2f276057ea1 ("drm/amdgpu/vcn: add vcn support for vcn 3.1.2")
Reviewed-by: James Zhu <James.Zhu@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: don't ignore alpha property on pre-multiplied mode

"Pre-multiplied" is the default pixel blend mode for KMS/DRM, as
documented in supported_modes of drm_plane_create_blend_mode_property():
https://cgit.freedesktop.org/drm/drm-misc/tree/drivers/gpu/drm/drm_blend.c

In this mode, both 'pixel alpha' and 'plane alpha' participate in the
calculation, as described by the pixel blend mode formula in KMS/DRM
documentation:

out.rgb = plane_alpha * fg.rgb +
          (1 - (plane_alpha * fg.alpha)) * bg.rgb

Considering the blend config mechanisms we have in the driver so far,
the alpha mode that better fits this blend mode is the
_PER_PIXEL_ALPHA_COMBINED_GLOBAL_GAIN, where the value for global_gain
is the plane alpha (global_alpha).

With this change, alpha property stops to be ignored. It also addresses
Bug: https://gitlab.freedesktop.org/drm/amd/-/issues/1734

v2:
 * keep the 8-bit value for global_alpha_value (Nicholas)
 * correct the logical ordering for combined global gain (Nicholas)
 * apply to dcn10 too (Nicholas)

Signed-off-by: Melissa Wen <mwen@igalia.com>
Tested-by: Rodrigo Siqueira <Rodrigo.Siqueira@amd.com>
Reviewed-by: Harry Wentland <harry.wentland@amd.com>
Tested-by: Simon Ser <contact@emersion.fr>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdkfd: Improve concurrency of event handling

Use rcu_read_lock to read p->event_idr concurrently with other readers
and writers. Use p->event_mutex only for creating and destroying events
and in kfd_wait_on_events.

Protect the contents of the kfd_event structure with a per-event
spinlock that can be taken inside the rcu_read_lock critical section.

This eliminates contention of p->event_mutex in set_event, which tends
to be on the critical path for dispatch latency even when busy waiting
is used. It also eliminates lock contention in event interrupt handlers.
Since the p->event_mutex is now used much less, the impact of requiring
it in kfd_wait_on_events should also be much smaller.

This should improve event handling latency for processes using multiple
GPUs concurrently.

v2: Reschedule the worker periodically to avoid soft lockup warnings

Signed-off-by: Felix Kuehling <Felix.Kuehling@amd.com>
Reviewed-by: Sean Keely <Sean.Keely@amd.com> # v1
Tested-by: Sanjay Tripathi <sanjay.tripathi@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/dc: remove duplicate include

'dm_services.h' included in 'freesync,c' is duplicated, so remove one.

Reported-by: Zeal Robot <zealci@zte.com.cn>
Signed-off-by: Lv Ruyi <lv.ruyi@zte.com.cn>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu/smu10: fix SoC/fclk units in auto mode

SMU takes clock limits in Mhz units.  socclk and fclk were
using 10 khz units in some cases.  Switch to Mhz units.
Fixes higher than required SoC clocks.

Fixes: 97cf32996c46d9 ("drm/amd/pm: Removed fixed clock in auto mode DPM")
Reviewed-by: Paul Menzel <pmenzel@molgen.mpg.de>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: update dcn315 clock table read

[Why & How]
Make dcn315 base its clock table off dcfclk rather than fclk.

This change also adds some sanity checking to make sure an
empty pmfw table does not result in invalid dal clocks.

Reviewed-by: Charlene Liu <Charlene.Liu@amd.com>
Acked-by: Qingqing Zhuo <qingqing.zhuo@amd.com>
Signed-off-by: Dmytro Laktyushkin <Dmytro.Laktyushkin@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: Remove leftover igp_lane_info

Variable igp_lane_info always is 0. 0 & any value = 0 and false.
In this way, all сonditional statements will false.

The code was leftover from when the code was ported from radeon
where igp_lane_info was derived from the vbios on supported
platforms.

[update commit message - Alex]

Signed-off-by: Grigory Vasilyev <h0tc0d3@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/radeon: change si_default_state table from global to static

Smatch reports these issues
si_blit_shaders.c:31:11: warning: symbol 'si_default_state'
  was not declared. Should it be static?
si_blit_shaders.c:253:11: warning: symbol 'si_default_size'
  was not declared. Should it be static?

Both symbols are only used in si.c.  Single file symbols
should be static.  So move the definition of
si_default_state and si_default_size to si_blit_shader.h
and change their storage-class-specifier to static.

Remove unneeded si_blit_shader.c

Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Tom Rix <trix@redhat.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu/display: change pipe policy for DCN 2.1

Fixes crash on MST Hub disconnect.

Bug: https://gitlab.freedesktop.org/drm/amd/-/issues/1849
Fixes: ee2698cf79cc ("drm/amd/display: Changed pipe split policy to allow for multi-display pipe split")
Signed-off-by: Benjamin Marty <info@benjaminmarty.ch>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/radeon: simplify if-if to if-else

Replace `if (!ret)` with `else` for simplification.

Signed-off-by: Guo Zhengkui <guozhengkui@vivo.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>