drm/amdgpu: Fix GTT size reporting in amdgpu_ioctl

The commit below changed the TTM manager size unit from pages to
bytes, but failed to adjust the corresponding calculations in
amdgpu_ioctl.

Fixes: dfa714b88eb0 ("drm/amdgpu: remove GTT accounting v2")
Bug: https://gitlab.freedesktop.org/drm/amd/-/issues/1930
Bug: https://gitlab.freedesktop.org/mesa/mesa/-/issues/6642
Tested-by: Martin Roukala <martin.roukala@mupuf.org>
Tested-by: Mike Lothian <mike@fireburn.co.uk>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Michel Dänzer <mdaenzer@redhat.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu/pm: remove the repeated EnableGfxImu message sending

The EnableGfxImu message will be issued in the set_gfx_power_up_by_imu.

Signed-off-by: Tim Huang <tim.huang@amd.com>
Reviewed-by: Yifan Zhang <yifan1.zhang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu/pm: correct the firmware flag address for SMU IP v13.0.4

For SMU IP v13.0.4, the smnMP1_FIRMWARE_FLAGS address is different,
we need this to correct the reading address.

Signed-off-by: Tim Huang <tim.huang@amd.com>
Reviewed-by: Yifan Zhang <yifan1.zhang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: ignore modifiers when checking for format support

[Why&How]
There are cases where swizzle modes are set but modifiers arent. For
such a userspace, we need not check modifiers while checking
compatibilty in the drm hook for checking plane format.

Ignore checking modifiers but check the DCN generation for the
supported swizzle mode.

v2: squash in unused variable removal (Alex)

Signed-off-by: Aurabindo Pillai <aurabindo.pillai@amd.com>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd: disable GPA mode in backdoor load

GPA mode should be disabled in direct load.

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

drm/amdgpu/discovery: enable vcn/jpeg v4_0_2

Enable vcn/jpeg 4_0_2.

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

drm/amd/pm: fix driver reload SMC firmware fail issue for smu13

issue calltrace:
[  402.773695] [drm] failed to load ucode SMC(0x2C)
[  402.773754] [drm] psp gfx command LOAD_IP_FW(0x6) failed and response status is (0x0)
[  402.773762] [drm:psp_load_smu_fw [amdgpu]] *ERROR* PSP load smu failed!
[  402.966758] [drm:psp_v13_0_ring_destroy [amdgpu]] *ERROR* Fail to stop psp ring
[  402.966949] [drm:psp_hw_init [amdgpu]] *ERROR* PSP firmware loading failed
[  402.967116] [drm:amdgpu_device_fw_loading [amdgpu]] *ERROR* hw_init of IP block <psp> failed -22
[  402.967252] amdgpu 0000:03:00.0: amdgpu: amdgpu_device_ip_init failed
[  402.967255] amdgpu 0000:03:00.0: amdgpu: Fatal error during GPU init

if not reset mp1 state during kernel driver unload, it will cause psp
load pmfw fail at the second time.

add PPSMC_MSG_PrepareMp1ForUnload support for smu_v13_0_0/smu_v13_0_7

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

drm/amd/pm: support BAMACO reset on smu_v13_0_7

support BAMACO reset on smu_v13_0_7, take BAMACO as a subset of BACO
for the low latency, and it only happens on specific platforms.

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

drm/amd/pm: enable gfxoff on smu_v13_0_7

enable gfxoff on smu_v13_0_7

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

drm/amd/pm: update the driver if header for smu_v13_0_7

update the driver if header for smu_v13_0_7

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

drm/amdgpu: Stop any pending reset if another in progress.

We skip rest requests if another one is already in progress.

Signed-off-by: Andrey Grodzovsky <andrey.grodzovsky@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: Rename amdgpu_device_gpu_recover_imp back to amdgpu_device_gpu_recover

We removed the wrapper that was queueing the recover function
into reset domain queue who was using this name.

Signed-off-by: Andrey Grodzovsky <andrey.grodzovsky@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: Add work_struct for GPU reset from kfd.

We need to have a work_struct to cancel this reset if another
already in progress.

Signed-off-by: Andrey Grodzovsky <andrey.grodzovsky@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: Add work_struct for GPU reset from debugfs

We need to have a work_struct to cancel this reset if another
already in progress.

Signed-off-by: Andrey Grodzovsky <andrey.grodzovsky@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/admgpu: Serialize RAS recovery work directly into reset domain queue.

Save the extra usless work schedule.

Signed-off-by: Andrey Grodzovsky <andrey.grodzovsky@amd.com>
Acked-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: Cache result of last reset at reset domain level.

Will be read by executors of async reset like debugfs.

Signed-off-by: Andrey Grodzovsky <andrey.grodzovsky@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

Revert "workqueue: remove unused cancel_work()"

This reverts commit 6417250d3f894e66a68ba1cd93676143f2376a6f.

amdpgu need this function in order to prematurly stop pending
reset works when another reset work already in progress.

Acked-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Andrey Grodzovsky <andrey.grodzovsky@amd.com>
Reviewed-by: Lai Jiangshan<jiangshanlai@gmail.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdkfd: Remove field io_link_count from struct kfd_topology_device

The field is redundant and does not serve any functional role

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

drm/amdgpu/display: Fix compilation issues

[Why]
Getting below build errors:
drivers/gpu/drm/amd/amdgpu/../display/dc/bios/bios_parser2.c:1419:3: error: unannotated fall-through between switch labels [-Werror,-Wimplicit-fallthrough]
                default:
                ^
drivers/gpu/drm/amd/amdgpu/../display/dc/bios/bios_parser2.c:1535:2: error: unannotated fall-through between switch labels [-Werror,-Wimplicit-fallthrough]
        default:
        ^

drivers/gpu/drm/amd/amdgpu/../display/dc/bios/bios_parser2.c:3276:20: error: variable 'v1_5' is uninitialized when used here [-Werror,-Wuninitialized]
                        for (i = 0; i < v1_5->number_of_path; ++i)
                                        ^~~~

[How]
Fix compilation issues

Signed-off-by: Chandan Vurdigere Nataraj <chandan.vurdigerenataraj@amd.com>
Reviewed-by: Leo Li <sunpeng.li@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu/display: Remove unnecessary typecasts and fix build issues

[Why]
Getting below errors:
drivers/gpu/drm/amd/amdgpu/../display/dc/dml/dcn32/display_mode_vba_32.c:1414:5: error: implicit conversion from enumeration type 'enum scan_direction_class' to different enumeration type 'enum dm_rotation_angle' [-Werror,-Wenum-conversion]
                                mode_lib->vba.SourceScan[k],
                                ^~~~~~~~~~~~~~~~~~~~~~~~~~~
drivers/gpu/drm/amd/amdgpu/../display/dc/dml/dcn32/display_mode_vba_32.c:1744:22: error: implicit conversion from enumeration type 'enum scan_direction_class' to different enumeration type 'enum dm_rotation_angle' [-Werror,-Wenum-conversion]
                        && (!(!IsVertical(mode_lib->vba.SourceScan[k])) || mode_lib->vba.DCCEnable[k] == true)) {
                               ~~~~~~~~~~ ^~~~~~~~~~~~~~~~~~~~~~~~~~~
2 errors generated.

drivers/gpu/drm/amd/amdgpu/../display/dc/dml/dcn32/display_mode_vba_util_32.c:5484:18: error: implicit conversion from enumeration type 'RequestType' to different enumeration type 'enum RequestType' [-Werror,-Wenum-conversion]
                        RequestLuma = REQ_256Bytes;
                                    ~ ^~~~~~~~~~~~
18 errors of similar kind

[How]
1. Add typecast at relevant places
2. Move the enum RequestType definition ahead of declarations

Signed-off-by: Chandan Vurdigere Nataraj <chandan.vurdigerenataraj@amd.com>
Reviewed-by: Leo Li <sunpeng.li@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: always flush the TLB on gfx8

The TLB on GFX8 stores each block of 8 PTEs where any of the valid bits
are set.

Fixes: 5255e146c99a ("drm/amdgpu: rework TLB flushing")
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Tested-by: Michal Kubecek <mkubecek@suse.cz>
Signed-off-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: fix limiting AV1 to the first instance on VCN3

The job is not yet initialized here.

Bug: https://gitlab.freedesktop.org/drm/amd/-/issues/2037
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Tested-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Signed-off-by: Christian König <christian.koenig@amd.com>
Fixes: cdc7893fc93f ("drm/amdgpu: use job and ib structures directly in CS parsers")
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/pm: enable BACO on smu_v13_0_7

enable BACO on smu_v13_0_7

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

drm/amd/pm: add interface to deallocate power_context for smu_v13_0_7

add interface to deallocate power_context for smu_v13_0_7

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

drm/amdgpu/jpeg2: Add jpeg vmid update under IB submit

Add jpeg vmid update under IB submit

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

drm/amdgpu/mes: only invalid/prime icache when finish loading both pipe MES FWs.

invalid/prime icahce operation takes effect both pipes cuconrrently,
therefore CP_MES_IC_BASE_LO/HI and CP_MES_MDBASE_LO/HI both have to be
set before prime icache. Otherwise MES hardware gets garbage data in
above regsters and causes page fault

[  470.873200] amdgpu 0000:33:00.0: amdgpu: [gfxhub] page fault (src_id:0 ring:217 vmid:0 pasid:0, for process  pid 0 thread  pid 0)
[  470.873222] amdgpu 0000:33:00.0: amdgpu:   in page starting at address 0x000092cb89b00000 from client 10
[  470.873234] amdgpu 0000:33:00.0: amdgpu: GCVM_L2_PROTECTION_FAULT_STATUS:0x00000BB3
[  470.873242] amdgpu 0000:33:00.0: amdgpu:      Faulty UTCL2 client ID: CPC (0x5)
[  470.873247] amdgpu 0000:33:00.0: amdgpu:      MORE_FAULTS: 0x1
[  470.873251] amdgpu 0000:33:00.0: amdgpu:      WALKER_ERROR: 0x1
[  470.873256] amdgpu 0000:33:00.0: amdgpu:      PERMISSION_FAULTS: 0xb
[  470.873260] amdgpu 0000:33:00.0: amdgpu:      MAPPING_ERROR: 0x1
[  470.873264] amdgpu 0000:33:00.0: amdgpu:      RW: 0x0

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

drm/amd/pm: drop unneeded dpm features disablement for SMU 13.0.0/7

PMFW will handle that properly. Driver involvement may cause some
unexpected issues.

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

drm/amd/pm: drop redundant declarations

Drop those redundant declarations in smu_v13_0.h.

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

drm/amd/pm: enable mode1 reset support for SMU 13.0.0

Fulfill the interfaces for mode1 reset related.

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

drm/amdgpu: avoid to perform undesired clockgating operation

Make sure the clockgating feature is supported before action.
Otherwise, the feature may be disabled unexpectedly on enablement
request.

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

drm/amdgpu: enable ASPM support for PCIE 7.4.0/7.6.0

Enable ASPM support for PCIE 7.4.0 and 7.6.0.

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/amd/pm: optimize the interface for dpm feature status query

Drop extra CMN2ASIC_MAPPING_FEATURE transform. Also some cosmetic
fixes for better readability.

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

drm/amd/pm: drop unneeded thermal_controller_type check

As there is actually no direct dependence between them.

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

drm/amd/pm: enable FW CTF feature for SMU 13.0.0

The new 78.40.0 PMFW has this feature supported.

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

drm/amdgpu: fix scratch register access method in SRIOV

The scratch register should be accessed through MMIO instead of RLCG
in SRIOV, since it being used in RLCG register access function.

Fixes: d54762cc3e6a ("drm/amdgpu: nuke dynamic gfx scratch reg allocation")
Signed-off-by: ZhenGuo Yin <zhenguo.yin@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/radeon: Initialize fences array entries in radeon_sa_bo_next_hole

Similar to the handling of amdgpu_sa_bo_next_hole in commit 6a15f3ff19a8
("drm/amdgpu: Initialize fences array entries in amdgpu_sa_bo_next_hole"),
we thought a patch might be needed here as well.

The entries were only initialized once in radeon_sa_bo_new. If a fence
wasn't signalled yet in the first radeon_sa_bo_next_hole call, but then
got signalled before a later radeon_sa_bo_next_hole call, it could
destroy the fence but leave its pointer in the array, resulting in
use-after-free in radeon_sa_bo_new.

Signed-off-by: Xiaohui Zhang <xiaohuizhang@ruc.edu.cn>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdkfd: Document and fix GTT BO kmap API

Removed an unused parameter from two functions and added kernel-doc
comments.

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: Reduce frame size in the bouding box for DCN21

GCC throw warnings for the function dcn21_update_bw_bounding_box and
dcn316_update_bw_bounding_box due to its frame size that looks like
this:

 error: the frame size of 1936 bytes is larger than 1024 bytes [-Werror=frame-larger-than=]

For fixing this issue I dropped an intermadiate variable.

Cc: Stephen Rothwell <sfr@canb.auug.org.au>
Cc: Hamza Mahfooz <hamza.mahfooz@amd.com>
Cc: Aurabindo Pillai <aurabindo.pillai@amd.com>
Reviewed-by: Harry Wentland <harry.wentland@amd.com>
Reviewed-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/amdkfd: Extend KFD device topology to surface peer-to-peer links

Extend KFD device topology to surface peer-to-peer links among
GPU devices connected over PCIe or xGMI. Enabling HSA_AMD_P2P is
REQUIRED to surface peer-to-peer links.

Prior to this KFD did not expose to user mode any P2P links or
indirect links that go over two or more direct hops. Old versions
of the Thunk used to make up their own P2P and indirect links without
the information about peer-accessibility and chipset support available
to the kernel mode driver. In this patch we expose P2P links in a new
sysfs directory to provide more reliable P2P link information to user
mode.

Old versions of the Thunk will continue to work as before and ignore
the new directory. This avoids conflicts between P2P links exposed by
KFD and P2P links created by the Thunk itself. New versions of the Thunk
will use only the P2P links provided in the new p2p_links directory, if
it exists, or fall back to the old code path on older KFDs that don't
expose p2p_links.

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

drm/radeon: integer overflow in radeon_mode_dumb_create()

Similar to the handling of amdgpu_mode_dumb_create in commit 54ef0b5461c0
("drm/amdgpu: integer overflow in amdgpu_mode_dumb_create()"),
we thought a patch might be needed here as well.

args->size is a u64.  arg->pitch and args->height are u32.  The
multiplication will overflow instead of using the high 32 bits as
intended.

Signed-off-by: Xiaohui Zhang <xiaohuizhang@ruc.edu.cn>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: simplify amdgpu_ucode_get_load_type()

This is the same as the default case, so drop the extra
logic.

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

drm/amd/display: Reduce frame size in the bouding box for DCN31/316

GCC throw warnings for the function dcn31_update_bw_bounding_box and
dcn316_update_bw_bounding_box due to its frame size that looks like
this:

 error: the frame size of 1936 bytes is larger than 1024 bytes [-Werror=frame-larger-than=]

For fixing this issue I dropped an intermadiate variable.

Cc: Stephen Rothwell <sfr@canb.auug.org.au>
Cc: Hamza Mahfooz <hamza.mahfooz@amd.com>
Cc: Aurabindo Pillai <aurabindo.pillai@amd.com>
Reviewed-by: Harry Wentland <harry.wentland@amd.com>
Reviewed-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/amd/display: Reduce frame size in the bouding box for DCN301

GCC throw warnings for the function dcn301_fpu_update_bw_bounding_box
due to its frame size that looks like this:

 error: the frame size of 1936 bytes is larger than 1024 bytes [-Werror=frame-larger-than=]

For fixing this issue I dropped an intermadiate variable.

Cc: Stephen Rothwell <sfr@canb.auug.org.au>
Cc: Hamza Mahfooz <hamza.mahfooz@amd.com>
Cc: Aurabindo Pillai <aurabindo.pillai@amd.com>
Reviewed-by: Harry Wentland <harry.wentland@amd.com>
Reviewed-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/amd/display: fix null pointer deref error

[Why]
0 was passed in place of a pointer which triggered null pointer
dereference.

Causes a backtrace like:

[   41.159466] RIP: 0010:dccg31_set_audio_dtbclk_dto+0x10/0x120 [amdgpu]
[   41.159928] Code: c0 00 00 00 6a 01 8b 92 84 01 00 00 52 0f b6 40 61 e9 30 ff ff ff 0f 1f 40 00 0f 1f 44 00 00 55 48 89 e5 41 56 41 55 41 54 53 <44> 8b 66 18 48 89 fb 48 8b 47 30 48 8b 3f 45 85 e4 74 09 4c 63 6e
[   41.159932] RSP: 0018:ffffaee54055afd0 EFLAGS: 00010246
[   41.159936] RAX: 0000000000000000 RBX: 0000000000000000 RCX: ffff917445901800
[   41.159939] RDX: ffffffffc15e6ca0 RSI: 0000000000000000 RDI: ffff91744dbd8c60
[   41.159941] RBP: ffffaee54055aff0 R08: 0000000000000100 R09: ffffaee54055afe8
[   41.159944] R10: 0000000000000001 R11: 0000000000000009 R12: ffff91747ca201f0
[   41.159946] R13: ffff91747ca20000 R14: ffff917448720438 R15: ffff91747ca20000
[   41.159948] FS:  00007f5e13e5f740(0000) GS:ffff91775ca40000(0000) knlGS:0000000000000000
[   41.159951] CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
[   41.159954] CR2: 0000000000000018 CR3: 000000010fea0000 CR4: 00000000003506e0
[   41.159956] Call Trace:
[   41.159959]  <TASK>
[   41.159964]  dce110_setup_audio_dto.isra.0+0xd8/0x1f0 [amdgpu]
[   41.160411]  dce110_apply_ctx_to_hw+0x1aa/0x780 [amdgpu]
[   41.160842]  ? __free_pages+0x88/0xb0
[   41.160850]  ? kfree+0x360/0x3e0
[   41.160857]  dc_commit_state+0x337/0xac0 [amdgpu]
[   41.161135]  amdgpu_dm_atomic_commit_tail+0x5e3/0x2680 [amdgpu]

[How]
Pass in a pointer that contains nullified parameters instead of null
pointer.

Fixes: 405bb9eea36a ("drm/amd/display: Implement DTBCLK ref switching on dcn32")
Signed-off-by: Aurabindo Pillai <aurabindo.pillai@amd.com>
Reviewed-by: Rodrigo Siqueira <Rodrigo.Siqueira@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: Add peer-to-peer support among PCIe connected AMD GPUs

Add support for peer-to-peer communication among AMD GPUs over PCIe
bus. Support REQUIRES enablement of config HSA_AMD_P2P.

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

drm/amdkfd: Define config HSA_AMD_P2P to support peer-to-peer

Extend current kernel config requirements of amdgpu by adding config
HSA_AMD_P2P. Enabling HSA_AMD_P2P is REQUIRED to support peer-to-peer
communication between AMD GPU devices over PCIe bus

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

drm/amd/display: Reduce frame size in the bouding box for DCN20

GCC throw warnings for the function dcn20_update_bounding_box due to its
frame size that looks like this:

 error: the frame size of 1936 bytes is larger than 1024 bytes [-Werror=frame-larger-than=]

This commit fixes this issue by eliminating an intermediary variable
that creates a large array.

Cc: Stephen Rothwell <sfr@canb.auug.org.au>
Cc: Hamza Mahfooz <hamza.mahfooz@amd.com>
Cc: Aurabindo Pillai <aurabindo.pillai@amd.com>
Reviewed-by: Harry Wentland <harry.wentland@amd.com>
Reviewed-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/amd/display: Remove duplicated macro

Reviewed-by: Harry Wentland <harry.wentland@amd.com>
Reviewed-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: Add MODE register to wave debug info in gfx11

All other chips, from gfx6-gfx10, now include the MODE register at the
end of the wave debug state. This appears to have been missed in gfx11,
so this patch adds in MODE to the debug state for gfx11.

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

drm/amd/amdgpu: Fix alignment issue

Fix alignment problems reported by zuul for the
commit b07d1d73b09e ("drm/amd/amdgpu: Enable high priority gfx queue")

Fixes: b07d1d73b09e ("drm/amd/amdgpu: Enable high priority gfx queue")
Signed-off-by: Arunpravin Paneer Selvam <Arunpravin.PaneerSelvam@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdkfd:Fix fw version for 10.3.6

fix fw error when loading fw for 10.3.6

Signed-off-by: Jesse Zhang <Jesse.Zhang@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Reviewed-by: Mario Limonciello <mario.limonciello@amd.com>
Reviewed-by: Felix Kuehling <Felix.Kuehling@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: refactor dirty rect dmub command decision

[Why]
To wrap the decision logic of sending dirty rect dmub command
for both frame update and cursor update path.

Signed-off-by: Robin Chen <po-tchen@amd.com>
Acked-by: Leo Li <sunpeng.li@amd.com>
Reviewed-by: Harry Wentland <harry.wentland@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: pass panel instance in DMUB dirty rect command

[Why]

In cases where there are multiple eDP instances, DMUB needs to know
which instance the command is for. Today, the field for specifying the
panel_inst exists in both dmub_cmd_update_dirty_rect_data and
dmub_cmd_update_cursor_info_data.

For cursor updates, we already specify the panel_inst, but that's not
the case for dirty_rect updates. Today, a value of '0' is used (due
to initial memsetting of the cmd struct to 0)

[how]

In dc_dmub_update_dirty_rect(), Call dc_get_edp_link_panel_inst() to get
the panel_inst, and fill it in the DMUB cmd struct.

v2: Update commit message for clarity.

Signed-off-by: Mikita Lipski <mikita.lipski@amd.com>
Signed-off-by: David Zhang <dingchen.zhang@amd.com>
Acked-by: Leo Li <sunpeng.li@amd.com>
Reviewed-by: Harry Wentland <harry.wentland@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Add PSR-SU-RC support in DC

[Why]

PSR-SU Rate Control - or PSR-SU-RC - enables PSR-SU panels to work with
variable refresh rate to allow for more power savings. Lowering the
refresh rate can increase PSR residency by expanding the eDP main link
shut down duration. It can also lower panel power consumption.

There is a complication with PSR, since the eDP main link can be shut
down. Therefore, the timing controller (TCON) on the eDP sink nees to be
able to scan out its remote buffer independent of the main link. To
allow the eDP source to specify the sink's refresh rate while the link
is off, vendor-specific DPCD registers are used. This allows the eDP
source to then "Rate Control" the panel during PSR active.

[How]

Add DC support to communicate with PSR-SU-RC supported eDP sinks. The
sink will need to know the desired VTotal during PSR active.

This change only adds support to DC, support in amdgpu_dm is still
pending to enable this fully.

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

drm/amd/display: expose AMD specific DPCD for PSR-SU-RC support

[why & how]

Expose vendor specific DPCD registers for rate controlling the eDP sink
TCON's refresh rate during PSR active. When used in combination with
PSR-SU and Freesync, it is called PSR-SU Rate Contorol, or PSR-SU-RC for
short.

v2: Add all DPCD registers required

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

drm/amd/display: 3.2.189

This version brings along the following:
- DPP DTO fix
- Transient encoder fix
- Restrict the reading of LTTPR capabilities in LTTPR mode
- Increase maximum stages for BB
- Distinguish HDMI DTO from DP DTO

Acked-by: Hamza Mahfooz <hamza.mahfooz@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: Add HDMI member to DTO

[Why]
For Pixel Rate control, when on HDMI, HDMI DTO
should be selected instead of DP DTO.

[How]
Add HDMI member to dtbclk_dto_params, so it can be used tell apart HDMI
and DP DTO in the future.

Reviewed-by: Alvin Lee <Alvin.Lee2@amd.com>
Acked-by: Hamza Mahfooz <hamza.mahfooz@amd.com>
Signed-off-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: Extend soc BB capabilitiy

[why]
Some parts are consuming dangerously close to maximum number of states
supported when updating the BB (i.e. 8).

[how]
Change maximum stages from 9 to 20.

Reviewed-by: Alvin Lee <Alvin.Lee2@amd.com>
Acked-by: Hamza Mahfooz <hamza.mahfooz@amd.com>
Signed-off-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 entry into transient encoder assignment mode.

[Why]
In some scenarios it is possible for the encoder assignment module to be
set to "transient" mode even though there are no new encoder
assignments.

This can lead to incorrect results when querying encoder assignment,
which in turn can cause incorrect displays to be manipulated.

[How]
Only allow encoder assignment to be in transient mode of operation when
there are valid new encoder assignments.

Reviewed-by: Meenakshikumar Somasundaram <Meenakshikumar.Somasundaram@amd.com>
Reviewed-by: Jun Lei <Jun.Lei@amd.com>
Acked-by: Hamza Mahfooz <hamza.mahfooz@amd.com>
Signed-off-by: Jimmy Kizito <Jimmy.Kizito@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Fix dpp dto for disabled pipes

[Why]
When switching from 1 pipe to 4to1 mpc combine,
DppDtoClk aren't enabled for the disabled pipes
pior to programming the pipes. Upon optimizing
bandwidth, DppDto are enabled causing intermittent
underflow.

[How]
Update dppclk dto whenever pipe are flagged to
enable.

Reviewed-by: Dmytro Laktyushkin <Dmytro.Laktyushkin@amd.com>
Reviewed-by: Nicholas Kazlauskas <Nicholas.Kazlauskas@amd.com>
Reviewed-by: Hansen Dsouza <Hansen.Dsouza@amd.com>
Acked-by: Hamza Mahfooz <hamza.mahfooz@amd.com>
Signed-off-by: Duncan Ma <duncan.ma@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: 3.2.188

This version brings along the following:
- Optimized blank calculations
- More robust DP MST hotplug support
- eDP bug fix relating to ODM
- Revert a patch that caused a regression with DP
- min comp buffer size fix
- Make DP easier to debug
- Calculate the maximum OLED brightness correctly
- 3 plane MPO.

Acked-by: Hamza Mahfooz <hamza.mahfooz@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 0.0.119.0

Reviewed-by: Aric Cyr <Aric.Cyr@amd.com>
Acked-by: Hamza Mahfooz <hamza.mahfooz@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: Enable 3 plane MPO

[why and how]
3 plane MPO is a new feature missing in a few resource files
Enable 3 plane MPO by setting slave planes to 2

Reviewed-by: Krunoslav Kovac <Krunoslav.Kovac@amd.com>
Reviewed-by: Aric Cyr <Aric.Cyr@amd.com>
Acked-by: Hamza Mahfooz <hamza.mahfooz@amd.com>
Signed-off-by: Hansen Dsouza <Hansen.Dsouza@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Cap OLED brightness per max frame-average luminance

[Why]
For OLED eDP the Display Manager uses max_cll value as a limit
for brightness control.
max_cll defines the content light luminance for individual pixel.
Whereas max_fall defines frame-average level luminance.
The user may not observe the difference in brightness in between
max_fall and max_cll.
That negatively impacts the user experience.

[How]
Use max_fall value instead of max_cll as a limit for brightness control.

Reviewed-by: Rodrigo Siqueira <Rodrigo.Siqueira@amd.com>
Acked-by: Hamza Mahfooz <hamza.mahfooz@amd.com>
Signed-off-by: Roman Li <roman.li@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: add DP sanity checks during enable stream

[why]
1. When HPD deassertion is pulled in the middle of
enabe stream link training, we will abort current training
and turn off PHY. This causes current link settings
to be zeroed this causes later stream enablement
sequence to fail as we prefer to carry on enablement
process despite of link training failure for SST.
2. When HPD is toggled after detection before before
the enable stream sequence as a result. There could be
a race condition where we could end up enable stream based
on the previous link even though the link is updated
after the HPD toggle. This causes an issue where our link
bandwidth is no longer enough to accommodate the timing
therefore causes us to oversubscribe MST payload time
slots. As discussed we decided to add basic sanity check
to make sure that our code can handle the oversubscription
failure silently without system hang.

[how]
1. Keep PHY powered on when HPD is deasserted during
enable stream and wait for the detection sequence to power
it off later.
2. Do not allocate payload if the required timeslot for
current timing is greater than 64 timeslots.

Reviewed-by: Aric Cyr <Aric.Cyr@amd.com>
Reviewed-by: George Shen <George.Shen@amd.com>
Acked-by: Hamza Mahfooz <hamza.mahfooz@amd.com>
Signed-off-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: Correct min comp buffer size

[Why]
In 3-way mpo pipes, there is a case that we
overbook the CRB buffer size. At rare instances,
overbooking the crb will cause underflow. This only
happens when det_size changes dynamically
based on pipe_cnt.

[How]
Set min compbuff size to 1 segment when preparing BW.

Reviewed-by: Dmytro Laktyushkin <Dmytro.Laktyushkin@amd.com>
Acked-by: Hamza Mahfooz <hamza.mahfooz@amd.com>
Signed-off-by: Duncan Ma <duncan.ma@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Blank eDP on enable drv if odm enabled

[Why]
For panels with pixel clock > 1200MHz that require ODM
in pre-OS, when driver is disabled in OS, odm is enabled.
Upon driver enablement, corruption is seen if
odm was originally enabled. DP_PIXEL_COMBINE and
pixelclk must be programmed prior to programming the
optc-odm registers. However, eDP displays aren't blanked
prior to initializing odm in this case.

[How]
Upon driver enablement, check whether odm is enabled,
if so, blank eDP prior to programming optc-odm
registers.

Reviewed-by: Nicholas Kazlauskas <Nicholas.Kazlauskas@amd.com>
Acked-by: Hamza Mahfooz <hamza.mahfooz@amd.com>
Signed-off-by: Duncan Ma <duncan.ma@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Avoid reading LTTPR caps in non-LTTPR mode.

[Why]
LTTPR capable devices on the DisplayPort path may assume that
extended LTTPR AUX timeouts will be used after LTTPR capabilities
are read.

When DPTX operates in non-LTTPR mode, AUX timeouts are not
extended and this can result in AUX transactions timing out.

[How]
Use shared helper function to determine LTTPR mode and do not
read LTTPR capabilities in non-LTTPR mode.

Reviewed-by: Mustapha Ghaddar <Mustapha.Ghaddar@amd.com>
Reviewed-by: Meenakshikumar Somasundaram <Meenakshikumar.Somasundaram@amd.com>
Reviewed-by: Jun Lei <Jun.Lei@amd.com>
Acked-by: Hamza Mahfooz <hamza.mahfooz@amd.com>
Signed-off-by: Jimmy Kizito <Jimmy.Kizito@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

Revert "drm/amd/display: Pass the new context into disable OTG WA"

This reverts commit 8440f57532496d398a461887e56ca6f45089fbcf.

Causes a hang when hotplugging DP, shutting down system, or
enabling dual eDP.

Reviewed-by: Dmytro Laktyushkin <Dmytro.Laktyushkin@amd.com>
Acked-by: Hamza Mahfooz <hamza.mahfooz@amd.com>
Signed-off-by: Nicholas Kazlauskas <nicholas.kazlauskas@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

Revert "drm/amdgpu: Ensure the DMA engine is deactivated during set ups"

This reverts commit b992a19085885c096b19625a85c674cb89829ca1.

This causes regression in GPU reset related test.

Cc: Alexander Deucher <Alexander.Deucher@amd.com>
Cc: ricetons@gmail.com
Signed-off-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/soc21: add mode2 asic reset for SMU IP v13.0.4

Set the default reset method to mode2 for SMU IP v13.0.4

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

drm/radeon: fix potential buffer overflow in ni_set_mc_special_registers()

The last case label can write two buffers 'mc_reg_address[j]' and
'mc_data[j]' with 'j' offset equal to SMC_NISLANDS_MC_REGISTER_ARRAY_SIZE
since there are no checks for this value in both case labels after the
last 'j++'.

Instead of changing '>' to '>=' there, add the bounds check at the start
of the second 'case' (the first one already has it).

Also, remove redundant last checks for 'j' index bigger than array size.
The expression is always false. Moreover, before or after the patch
'table->last' can be equal to SMC_NISLANDS_MC_REGISTER_ARRAY_SIZE and it
seems it can be a valid value.

Detected using the static analysis tool - Svace.
Fixes: 69e0b57a91ad ("drm/radeon/kms: add dpm support for cayman (v5)")
Signed-off-by: Alexey Kodanev <aleksei.kodanev@bell-sw.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu/display: fix DCN3.2 Makefiles for non-x86

Add proper handling for PPC64.

Reviewed-by: Harry Wentland <harry.wentland@amd.com>
Reported-by: kernel test robot <lkp@intel.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu/display: make some functions static

Fixes "no previous prototype" warnings.

Reviewed-by: Harry Wentland <harry.wentland@amd.com>
Reported-by: kernel test robot <lkp@intel.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu/discovery: add comments about VCN instance handling

Add comments to clarify code that is safe, but triggers and
smatch warning.

Link: https://lists.freedesktop.org/archives/amd-gfx/2022-June/079905.html
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Cc: Dan Carpenter <dan.carpenter@oracle.com>

drm/amd/display: Detect dpcd_rev when hotplug mst monitor

[Why]
Once mst topology is constructed, later on new connected monitors
are reported to source by CSN message. Within CSN, there is no
carried info of DPCD_REV comparing to LINK_ADDRESS reply. As the
result, we might leave some ports connected to DP but without DPCD
revision number which will affect us determining the capability of
the DP Rx.

[How]
Send out remote DPCD read when the port's dpcd_rev is 0x0 in
detect_ctx(). Firstly, read out the value from DPCD 0x2200. If the
return value is 0x0, it's likely the DP1.2 DP Rx then we reques
revision from DPCD 0x0 again.

Reviewed-by: Hersen Wu <hersenwu@amd.com>
Acked-by: Hamza Mahfooz <hamza.mahfooz@amd.com>
Signed-off-by: Wayne Lin <Wayne.Lin@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Update optimized blank calc and programming

[Why]
The existing calculations in DCN3.1 were placeholder and need to be
replaced with HW team approved calculations.

[How]
The new calculations add new parameters to the bounding box and pipe
params - VblankNom and the bounding box default.

The placeholder calculations are dropped from DCN3.1 in the meantime
while we work out hardware approved replacements.
Also fix a bug where we wipe out other register contents with a REG_SET
instead of a REG_UPDATE for the register we were programming the
min_dst_y_next_start_optimized.

Reviewed-by: Nicholas Kazlauskas <Nicholas.Kazlauskas@amd.com>
Acked-by: Hamza Mahfooz <hamza.mahfooz@amd.com>
Signed-off-by: Gabe Teeger <gabe.teeger@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Implement MPO PSR SU

[WHY]

For additional power savings, PSR SU (also referred to as PSR2) can be
enabled on eDP panels with PSR SU support.

PSR2 saves more power compared to PSR1 by allowing more opportunities
for the display hardware to be shut down. In comparison to PSR1, Shut
down can now occur in-between frames, as well as in display regions
where there is no visible update. In otherwords, it allows for some
display hw components to be enabled only for a **selectively updated**
region of the visible display. Hence PSR SU.

[HOW]

To define the SU region, support from the OS is required. OS needs to
inform driver of damaged regions that need to be flushed to the eDP
panel. Today, such support is lacking in most compositors.

Therefore, an in-between solution is to implement PSR SU for MPO and
cursor scenarios. The plane bounds can be used to define the damaged
region to be flushed to panel. This is achieved by:

* Leveraging dm_crtc_state->mpo_requested flag to identify when MPO is
  enabled.
* If MPO is enabled, only add updated plane bounds to dirty region.
  Determine plane update by either:
    * Existence of drm damaged clips attached to the plane (added by a
      damage-aware compositor)
    * Change in fb id (flip)
    * Change in plane bounds (position and dimensions)
* If cursor is enabled, the old_pos and new_pos of cursor plus cursor
  size is used as damaged regions(*).

(*) Cursor updates follow a different code path through DC. PSR SU for
cursor is already implemented in DC, and the only thing required to
enable is to set DC_PSR_VERSION_SU_1 on the eDP link. See
dcn10_dmub_update_cursor_data().

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

drm/amd/display: update cursor position to DMUB FW

[why]
To involve the cursor position into dirty rectangle calculation.

[how]
- separate plane and cursor update by different DMUB command
- send the cursor information while cursor updating, when updating
  cursor position/attribute, store cursor pos/attr to hubp, and
  notify dmub FW to exit psr before program cursor registers

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

drm/amd/display: calculate psr config settings in runtime in DM

[why]
Currently the psr configuration parameters are hardcoded before
feeding into the DC helper before passing to DMUB FW. We'd rework
to call a shared helper to calculate/update generic psr config
fields which are relying on the stream timing and eDP sink PSR
caps to avoid hard-coding.

[how]
- drop part of hard-coded psr config fields by replacing w/ the
  call of helper from DM before feeding into DC link setup psr
  helper
- For those DM specific psr config fields, e.g. allow smu opt, is
  not to be set/updated from the shared helper but to rely on the
  DC feature mask
- for the psr version field in psr_config structure, since only
  the field psr_version of DC link psr_settings matters for that
  fed to DMUB FW, thus no need to set/update the psr_version field
  of psr_config structure.

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

drm/amd/display: add shared helpers to update psr config fields to power module

[why]
Currently the amdgpu DM psr configuration parameters are hardcoded
before feeding into the DC helper to setup PSR. We would define a
helper which is to calculate parts of the psr config fields to
avoid hard-coding.

[how]
To make helper shareable, declare and define the helper in the
module_helper, to set/update below fields:
- psr remote buffer setup time
- sdp tx line number deadline
- line time in us
- su_y_granularity
- su_granularity_required
- psr_frame_capture_indication_req
- psr_exit_link_training_required

add another helper to check given the stream context, if there is
only one stream and the output is eDP panel connected.

changes in v2:
------------------
- add detailed comment for how psr setup time is calculated as per
  eDP 1.5 spec

Cc: Chandan Vurdigerenataraj <chandan.vurdigerenataraj@amd.com>
Signed-off-by: David Zhang <dingchen.zhang@amd.com>
Acked-by: Leo Li <sunpeng.li@amd.com>
Reviewed-by: Harry Wentland <harry.wentland@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: PSRSU+DSC WA for specific TCON

[why]
Some specific TCON chip has HW limitation to support PSRSU+DSC.

[how]
Force ffu mode when DSC enabled if we detect it is the specific
model from sink OUI DPCD. And disable ABM update for this case.

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

drm/amd/display: use HW lock mgr for PSR-SU

[why]
Feature requires synchronization of dig, pipe, and cursor locking
between driver and DMUB fw for PSR-SU

[how]
return True if PSR-SU in the checker should_use_dmub_lock()

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

drm/amd/display: Set PSR level to enable ALPM by default

[Why & How]
While support ALPM, do ALPM state transition while PSR entry/exit.
ALPM is needed for PSR-SU feature, and since the function is ready,
we'd enable it by default.

- Add psr level definition to enable/disable ALPM and set ALPM
  powerdone mode.
- Enable ALPM by default

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

drm/amd/display: fix system hang when PSR exits

[why]
When DC driver send PSR exit dmub command to DMUB FW, it might not
wait until PSR exit. Then it may hit the following deadlock situation.
1. DC driver send HW LOCK command to DMUB FW due to frame update
2. DMUB FW Set the HW lock
3. DMUB execute PSR exit sequence and stuck at polling DPG Pending
register due to the HW Lock is set
4. DC driver ask DMUB FW to unlock HW lock, but DMUB FW is polling
DPG pending register

[how]
The reason why DC driver doesn't wait until PSR exit is because some of
the PSR state machine state is not update the dc driver. So when DC
driver read back the PSR state, it take the state for PSR inactive.

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

drm/amd/display: add vline time in micro sec to PSR context

[why]
The current PSR SU programming margin is fixed base on FHD 60HZ
panel. If the resolution and refresh rate become higher, the time
of current margin might not cover the programming SU time.

[how]
Notice that the programming SU time is the same among different
panels.

Instead of fixing the margin with target line number, change the
margin unit to micro second which indicate the time needed for
programming SU. Then FW set the margin line number base on the
line time and margin time.

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

drm/amd/display: Set default value of line_capture_indication

[Why & how]
We only support line capture indication as 0 for PSRSU

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

drm/amd/display: Passing Y-granularity to dmub fw

[Why]
The Y-granularity panel parameter indicate the grid
pattern granularity in the Y direction for PSRSU.

[How]
Send the Y-granularity data by PSR_COPY_SETTINGS dmub command.

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

drm/amd/amdgpu: add pipe1 hardware support

Enable pipe1 support starting from SIENNA CICHLID asic

Signed-off-by: Arunpravin Paneer Selvam <Arunpravin.PaneerSelvam@amd.com>
Acked-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/amdgpu: Enable high priority gfx queue

Starting from SIENNA CICHLID asic supports two gfx pipes, enabling
two graphics queues, 1 on each pipe, pipe0 queue0 would be the normal
piority queue and pipe1 queue0 would be the high priority queue

Only one queue per pipe is visble to SPI, SPI looks at the priority
value assigned to CP_GFX_HQD_QUEUE_PRIORITY from each of the queue's
HQD/MQD.

Create contexts applying AMDGPU_CTX_PRIORITY_HIGH which submits job
to the high priority queue on GFX pipe1. There would be starvation
of LP workload if HP workload is always available.

v2:
  - remove unnecessary check(Nirmoy)
  - make pipe1 hardware support a separate patch(Nirmoy)
  - remove duplicate code(Shashank)
  - add CSA support for second gfx pipe(Alex)

v3(Christian):
  - fix incorrect indentation
  - merge COMPUTE and GFX switch cases as both calls the same function.

v4:
  - rebase w/ latest code base

Signed-off-by: Arunpravin Paneer Selvam <Arunpravin.PaneerSelvam@amd.com>
Acked-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: adding device coredump support

Added device coredump information:
- Kernel version
- Module
- Time
- VRAM status
- Guilty process name and PID
- GPU register dumps
v1 -> v2: Variable name change
v1 -> v2: NULL check
v1 -> v2: Code alignment
v1 -> v2: Adding dummy amdgpu_devcoredump_free
v1 -> v2: memset reset_task_info to zero
v2 -> v3: add CONFIG_DEV_COREDUMP for variables
v2 -> v3: remove NULL check on amdgpu_devcoredump_read

Signed-off-by: Somalapuram Amaranath <Amaranath.Somalapuram@amd.com>
Reviewed-by: Shashank Sharma <Shashank.sharma@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: save the reset dump register value for devcoredump

Allocate memory for register value and use the same values for devcoredump.
v1 -> v2: Change krealloc_array() to kmalloc_array()
v2 -> v3: Fix alignment

Signed-off-by: Somalapuram Amaranath <Amaranath.Somalapuram@amd.com>
Reviewed-by: Shashank Sharma <Shashank.sharma@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: program PSR2 DPCD Configuration

[Why]
To support PSR2 Source DPCD configuration

[How]
Update the PSR2 Source DPCD settings while the PSR2 enabled

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

drm/amd/display: revise Start/End SDP data

[why & how]
We need to implement the VSC packet rev4 that is required by PSRSU.

Follow the eDP 1.5 spec pg. 257

changes in v2:
-------------------
- set vsc packet rev2 for PSR1

Cc: Chandan Vurdigerenataraj <chandan.vurdigerenataraj@amd.com>
Signed-off-by: David Zhang <dingchen.zhang@amd.com>
Acked-by: Leo Li <sunpeng.li@amd.com>
Reviewed-by: Harry Wentland <harry.wentland@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: update GSP1 generic info packet for PSRSU

[why & how]
Based on PSRSU specification, every selective update frame need to use
two SDP to indicate the frame active range. So we occupy another GSP1
for PSRSU execution.

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

drm/amd/display: combine dirty rectangles in DMUB FW

[why]
In PSR-SU design, the DMUB FW handles the combination of multiple
dirty rectangles.

[how]
- create DC dmub update dirty rectangle helper which sends the
  dirty rectangles per pipe from DC to DMUB, and DMUB FW will
  handle to combine the dirty RECTs
- call the helper from DC commit plane update function.

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

drm/amd/display: feed PSR-SU as psr version to dmub FW

[why & how]
set psr version as PSR-SU in kernel-FW interface function to ensure
the correct dmub command parameter is fed into FW.

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

drm/amd/display: align dmub cmd header to latest dmub FW to support PSR-SU

[why]
PSR-SU is implemented in upstreamed dmub FW but not enabled on
DM and DC. We'd add necessary and missing definitions in dmub
cmd header to align w/ the up-to-date DMUB FW for PSR-SU support.

[how]
Add definitions and items below into dmub cmd header:
- DMUB psr version enumeration for PSR-SU
- dirty rectangle structure
- psr debug flag of forcing full frame update
- dmub command of updating dirty rectangle and cursor infor
- dmub psr command type of setting sink vtotal in PSR active
- dmub psr su debug flags structure
- dmub cmd structure for
  - updating dirty rectangle
  - cursor infor
  - setting sink vtotal
- dmub ringbuffer command items

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

drm/amdgpu/display/dc: Add ACP_DATA register

Define ixAZALIA_F0_CODEC_PIN_CONTROL_ACP_DATA
Define AZALIA_F0_CODEC_PIN_CONTROL_ACP_DATA__SUPPORTS_AI_MASK/SHIFT

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

drm/amdgpu/display: Protect some functions with CONFIG_DRM_AMD_DC_DCN

Protect remove_hpo_dp_link_enc_from_ctx() and release_hpo_dp_link_enc()
with CONFIG_DRM_AMD_DC_DCN as the functions are only called from code
that is protected by CONFIG_DRM_AMD_DC_DCN.  Fixes build fail with
-Werror=unused-function.

Fixes: 9b0e0d433f74 ("drm/amd/display: Add dependant changes for DCN32/321")
Reported-by: Stephen Rothwell <sfr@canb.auug.org.au>
Reviewed-by: Harry Wentland <harry.wentland@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Cc: Aurabindo Pillai <aurabindo.pillai@amd.com>