drm/amdgpu: Use explicit VCN instance 0 in SR-IOV init

vcn_v2_0_start_sriov() declares a local variable "i" initialized to zero
and uses it only as the instance index in SOC15_REG_OFFSET(UVD, i, ...).
The value is never changed and all other fields are taken from
adev->vcn.inst[0], so this path only ever programs VCN instance 0.

This triggered a Smatch:
warn: iterator 'i' not incremented

Replace the dummy iterator with an explicit instance index of 0 in
SOC15_REG_OFFSET() calls.

Fixes: dd26858a9cd8 ("drm/amdgpu: implement initialization part on VCN2.0 for SRIOV")
Reported by: Dan Carpenter <dan.carpenter@linaro.org>
Cc: darlington Opara <darlington.opara@amd.com>
Cc: Jinage Zhao <jiange.zhao@amd.com>
Cc: Monk Liu <Monk.Liu@amd.com>
Cc: Emily Deng <Emily.Deng@amd.com>
Cc: Christian König <christian.koenig@amd.com>
Cc: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Srinivasan Shanmugam <srinivasan.shanmugam@amd.com>
Reviewed-by: Emily Deng <Emily.Deng@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: enable CP interrupt for gfx v12_1 in frontdoor loading case

Enable cp interrupt for event detection since GFX CGCG and LS
has been enabled by firmware.

v2: enable CP INT by merely checking fw_load_type

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

drm/amdkfd: Apply VGPR bank state fixup on gfx12.1 trap exit

- Identify co-issue of S_SET_VGPR_MSB and VALU with banked VGPR
- Restore previous bank setting when exiting the trap

v2:
- Refine VOP3PX2 detection
- Improve load pipelining
- Fix a comment typo

Signed-off-by: Jay Cornwall <jay.cornwall@amd.com>
Reviewed-by: Lancelot Six <lancelot.six@amd.com>
Cc: Joseph Greathouse <joseph.greathouse@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdkfd: Fix VGPR bank state save in gfx12.1 trap handler

S_SETREG_IMM32_B32 does not apply a mask to the MODE bank bits.
SRC2 is consequently unconditonally cleared during context save.

Use S_SETREG_B32 instead to preserve SRC2.

Signed-off-by: Jay Cornwall <jay.cornwall@amd.com>
Reviewed-by: Lancelot Six <lancelot.six@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: Add sysfs up clean for gfx_v12_1

Add sysfs clean up for gfx_v12_1 during gfx fini sequence. This will
prevent following crash while reloading driver

2645.490824] R13: 000055d0cb186330 R14: 000055d0cb185ed0 R15: 000055d0cb188f40
[ 2645.490825]  </TASK>
[ 2645.490836] amdgpu 0000:02:00.0: amdgpu: failed to create xcp sysfs files
[ 2645.490937] amdgpu 0000:02:00.0: amdgpu: sw_init of IP block <gfx_v12_1> failed -17
[ 2645.491018] amdgpu 0000:02:00.0: amdgpu: amdgpu_device_ip_init failed
[ 2645.491098] amdgpu 0000:02:00.0: amdgpu: Fatal error during GPU init
[ 2645.491547] amdgpu 0000:02:00.0: amdgpu: amdgpu: finishing device.
[ 2648.549939] ------------[ cut here ]------------
[ 2648.549942] WARNING: CPU: 0 PID: 2459 at /tmp/amd.aIpOeG3c/amd/amdgpu/amdgpu_irq.c:

Signed-off-by: Asad Kamal <asad.kamal@amd.com>
Reviewed-by: Lijo Lazar <lijo.lazar@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdkfd: Add metadata ring buffer for compute

Add support for separate ring-buffer for metadata packets when using
compute queues. Userspace application allocate the metadata ring-buffer
and the queue ring-buffer with a single allocation. The metadata
ring-buffer starts after the queue ring-buffer.

Signed-off-by: David Yat Sin <David.YatSin@amd.com>
Reviewed-by: Philip Yang <Philip.Yang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/amdgpu : Use the MES INV_TLBS API for tlb invalidation on gfx12_1

Signed-off-by: Shaoyun Liu <shaoyun.liu@amd.com>
Reviewed-by: Prike Liang <Prike.Liang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: Update TCP Control register on GFX 12.1

Update TCP CNTL register to disable some features not supported
on GFX 12.1.

Signed-off-by: Mukul Joshi <mukul.joshi@amd.com>
Reviewed-by: Alex Sierra <alex.sierra@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdkfd: Add back CWSR trap handler for GFX 12.1

CWSR Trap handler for GFX 12.1 was missed when merging changes
from 6.14 NPI branch to 6.16 NPI branch. This change adds back
the CWSR trap handler for GFX 12.1.

Signed-off-by: Mukul Joshi <mukul.joshi@amd.com>
Reviewed-by: Alex Sierra <alex.sierra@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: Cleanup gmc_v12_1 after 6.16 merge

After the 6.16 merge, some changes not applicable to GFX 12.1 were
added in the gmc_v12_1_get_vm_pte function. Additionally, add the
case for MTYPE RW for GFX 12.1.

Signed-off-by: Mukul Joshi <mukul.joshi@amd.com>
Reviewed-by: Alex Sierra <alex.sierra@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: Disable TCP Early Write Ack for GFX 12.1

Disable the TCP Early Write Ack feature on GFX 12.1.

Signed-off-by: Mukul Joshi <mukul.joshi@amd.com>
Reviewed-by: Alex Sierra <alex.sierra@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdkfd: enable precise memory operations for gfx1250

Enable precise memory for GFX 1250.

Signed-off-by: Jonathan Kim <jonathan.kim@amd.com>
Reviewed-by: Mukul Joshi <mukul.joshi@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdkfd: fix partitioned gfx12 address watch enablement

GFX 12 devices that support spatial partitioning should use the WREG32
per XCC macro when updating address watch settings, similar to GFX 9
devices that support spatial partitioning.

Signed-off-by: Jonathan Kim <jonathan.kim@amd.com>
Reviewed-by: Mukul Joshi <mukul.joshi@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdkfd: Implement CU Masking for GFX 12.1

Add CU masking implementation for GFX 12.1. Add a local
implementation for GFX 12.1 instead of using the generic
function defined in kfd_mqd_manager.c because of some
quirks in the way CU mask is handled on GFX 12.1.

Signed-off-by: Mukul Joshi <mukul.joshi@amd.com>
Reviewed-by: Alex Sierra <alex.sierra@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: skip gfxhub tlb flush if gfx is power off

Skip for gfxhub tlb flush for gc v12_1 if gfx is not poweron.

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

drm/amdkfd: Add gfx_v12_1_kfd2kgd interface for GFX12_1

Create new kfd2kgd interface for gfx v12_1, based on gfx v12.
Support register program accoding to xcc id.
V2: Fix SDMA register address for muti-xcc.

Signed-off-by: Likun Gao <Likun.Gao@amd.com>
Reviewed-by: Mukul Joshi <mukul.joshi@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: update mcm_addr_lut data for imu v12_1

Support for partition mode to program MCM_ADDR_LUT.

v2: clean up (Alex)

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

drm/amdgpu: Init mcm_addr look up table

Encode mcm address look up table in SPX mode
as a temp solution.

v2: fill in when interface is ready (Alex)

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

drm/amdgpu: Always set PTE.B for device memory on GFX 12.1

On GFX 12.1, we need to set the atomics bit (PTE.B) always for
device memory.

Signed-off-by: Mukul Joshi <mukul.joshi@amd.com>
Reviewed-by: Alex Sierra <alex.sierra@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu/gfx12.1: Don't fetch default register values from hardware in mqd init

1. We can't assure the fetched values are always default register values.
   Observing non-zero cp_hqd_pq_rptr in mes_v12_1_self_test->init_mqd()
   where no GRBM_GFX_CNTL is specified.

2. See commit fc3c139cf043 ("drm/amdgpu/gfx12: don't read registers in mqd init").

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

drm/amd: Convert DRM_*() to drm_*()

The drm_*() macros include the device which is helpful for debugging
issues in multi-GPU systems.

Signed-off-by: Mario Limonciello (AMD) <superm1@kernel.org>
Reviewed-by: Aurabindo Pillai <aurabindo.pillai@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd: Drop amdgpu prefix from message prints

Hardcoding the prefix isn't necessary when using drm_* or dev_*
message prints.

Signed-off-by: Mario Limonciello (AMD) <superm1@kernel.org>
Reviewed-by: Aurabindo Pillai <aurabindo.pillai@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd: Convert amdgpu_display from DRM_* to drm_ macros

drm_* macros show the device they were called with which is helpful
in multi-GPU systems.

Signed-off-by: Mario Limonciello (AMD) <superm1@kernel.org>
Reviewed-by: Aurabindo Pillai <aurabindo.pillai@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Fix DPMS log printing

[Why]
Spaces before newline are not necessary. Inserting newlines in
multi-line strings are harder to follow when tracing messages.

[How]
Drop extra new lines and split multi-line messages into one print
per line.

Signed-off-by: Mario Limonciello (AMD) <superm1@kernel.org>
Reviewed-by: Aurabindo Pillai <aurabindo.pillai@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd: Drop dev_fmt prefix

The `amdgpu:` prefix in dev_fmt() isn't needed because the core
already includes the driver in the print.

Signed-off-by: Mario Limonciello (AMD) <superm1@kernel.org>
Reviewed-by: Aurabindo Pillai <aurabindo.pillai@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd: Pass `adev` to amdgpu_gfx_parse_disable_cu()

In order for messages to be attribute to the correct device
amdgpu_gfx_parse_disable_cu() needs to know what device is being
operated on.  Pass the argument in.

Signed-off-by: Mario Limonciello (AMD) <superm1@kernel.org>
Reviewed-by: Aurabindo Pillai <aurabindo.pillai@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd: Add correct prefix for VBIOS message

It's not obvious which GPU the ATOM BIOS message goes with. Use
drm_info() to show the correct one.

Signed-off-by: Mario Limonciello (AMD) <superm1@kernel.org>
Reviewed-by: Aurabindo Pillai <aurabindo.pillai@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdkfd: Correct the topology message for APUs

At bootup on a Strix machine the following message comes up:
```
amdgpu: Topology: Add dGPU node [0x150e:0x1002]
```

This is an APU though. Clarify the messaging by only offer a
"CPU node" or "GPU node" message. Also set the message as
VID:DID instead which is how other messages work.

Signed-off-by: Mario Limonciello (AMD) <superm1@kernel.org>
Reviewed-by: Aurabindo Pillai <aurabindo.pillai@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdkfd: Fix signal_eviction_fence() bool return value

signal_eviction_fence() is declared to return bool, but returns -EINVAL
when no eviction fence is present.  This makes the "no fence" or "the
NULL-fence" path evaluate to true and triggers a Smatch warning.

v2: Return true instead to explicitly indicate that there is no eviction
fence to signal and that eviction is already complete. This matches the
existing caller logic where a NULL fence means "nothing to do" and
allows restore handling to proceed normally. (Christian)

Fixes the below:
drivers/gpu/drm/amd/amdgpu/../amdkfd/kfd_process.c:2099 signal_eviction_fence()
warn: '(-22)' is not bool

drivers/gpu/drm/amd/amdgpu/../amdkfd/kfd_process.c
    2090 static bool signal_eviction_fence(struct kfd_process *p)
                ^^^^

    2091 {
    2092         struct dma_fence *ef;
    2093         bool ret;
    2094
    2095         rcu_read_lock();
    2096         ef = dma_fence_get_rcu_safe(&p->ef);
    2097         rcu_read_unlock();
    2098         if (!ef)
--> 2099                 return -EINVAL;

This should be either true or false.
Probably true because presumably
it has been tested?

    2100
    2101         ret = dma_fence_check_and_signal(ef);
    2102         dma_fence_put(ef);
    2103
    2104         return ret;
    2105 }

Fixes: 37865e02e6cc ("drm/amdkfd: Fix eviction fence handling")
Reported by: Dan Carpenter <dan.carpenter@linaro.org>
Cc: Philip Yang <Philip.Yang@amd.com>
Cc: Gang BA <Gang.Ba@amd.com>
Cc: Felix Kuehling <felix.kuehling@amd.com>
Signed-off-by: Srinivasan Shanmugam <srinivasan.shanmugam@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/pm: fix wrong pcie parameter on navi1x

fix wrong pcie dpm parameter on navi1x

Fixes: 1a18607c07bb ("drm/amd/pm: override pcie dpm parameters only if it is necessary")
Closes: https://gitlab.freedesktop.org/drm/amd/-/issues/4671
Signed-off-by: Yang Wang <kevinyang.wang@amd.com>
Co-developed-by: Kenneth Feng <kenneth.feng@amd.com>
Signed-off-by: Kenneth Feng <kenneth.feng@amd.com>
Acked-by: Alex Deucher <alexander.deucher@amd.com>
Reviewed-by: Lijo Lazar <lijo.lazar@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd: Drop "amdgpu kernel modesetting enabled" message

The behavior for amdgpu was changed with commit e00e5c223878
("drm/amdgpu: adjust drm_firmware_drivers_only() handling") to
potentially allow loading even if nomodeset was set, so the
message is no longer accurate.

Just drop it to avoid confusion.

Fixes: e00e5c223878 ("drm/amdgpu: adjust drm_firmware_drivers_only() handling")
Signed-off-by: Mario Limonciello (AMD) <superm1@kernel.org>
Reviewed-by: Aurabindo Pillai <aurabindo.pillai@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: Add address checking for uniras

Add address checking for uniras

Signed-off-by: Jinzhou Su <jinzhou.su@amd.com>
Reviewed-by: YiPeng Chai <YiPeng.Chai@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/radeon: Remove __counted_by from ClockInfoArray.clockInfo[]

clockInfo[] is a generic uchar pointer to variable sized structures
which vary from ASIC to ASIC.

Closes: https://gitlab.freedesktop.org/drm/amd/-/issues/4374
Reviewed-by: Lijo Lazar <lijo.lazar@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: add support for MMHUB IP version 3.4.0

This initializes MMHUB IP version 3.4.0.

v2: squash in clients table update (Alex)

Reviewed-by: Mario Limonciello <mario.limonciello@amd.com>
Signed-off-by: Tim Huang <tim.huang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: add support for HDP IP version 6.1.1

This initializes HDP IP version 6.1.1.

Reviewed-by: Mario Limonciello <mario.limonciello@amd.com>
Signed-off-by: Tim Huang <tim.huang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: add support for IH IP version 6.1.1

This initializes IH IP version 6.1.1.

Reviewed-by: Mario Limonciello <mario.limonciello@amd.com>
Signed-off-by: Tim Huang <tim.huang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: add support for NBIO IP version 7.11.4

This initializes NBIO IP version 7.11.4.

Reviewed-by: Mario Limonciello <mario.limonciello@amd.com>
Signed-off-by: Tim Huang <tim.huang@amd.com>
Signed-off-by: Pratik Vishwakarma <Pratik.Vishwakarma@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: add support for SDMA IP version 6.1.4

This initializes SDMA IP version 6.1.4.

Reviewed-by: Mario Limonciello <mario.limonciello@amd.com>
Signed-off-by: Tim Huang <tim.huang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: add support for GC IP version 11.5.4

This initializes GC IP version 11.5.4.

v2: squash in RLC offset fix

Reviewed-by: Mario Limonciello <mario.limonciello@amd.com>
Signed-off-by: Tim Huang <tim.huang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: Fix xcc_id input for soc_v1_0_grbm_select

Ensure the GRBM_GFX_CNTL is programmed correctly

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

drm/amdgpu: Do not initialize imu callback for vf

Not needed in guest environment

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

drm/amdgpu: make normalize reg addr to common func for soc v1

Normalize registers address to local xcc address for sdma v7_1.
Merge normalize register address function to an common function
for soc v1.

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

drm/amdgpu: Setup MTYPE on SOC models for GFX 12.1

Fix it to apply for all models.

Signed-off-by: Mukul Joshi <mukul.joshi@amd.com>
Reviewed-by: Alex Sierra <alex.sierra@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: Report correct compute partition mode on GFX 12.1

PSP programs the NBIO partition status register. In the absence of PSP,
read the current compute partition from the GFX IMU register instead of
NBIO.

Signed-off-by: Mukul Joshi <mukul.joshi@amd.com>
Reviewed-by: Alex Sierra <alex.sierra@amd.com>
Reviewed-by: Michael Chen <michael.chen@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdkfd: Send MES packets on correct XCC on GFX 12.1

Send the Set_Shader_Debugger packet on the correct MES pipe when
partition mode is set to non-SPX mode.

Signed-off-by: Mukul Joshi <mukul.joshi@amd.com>
Reviewed-by: Alex Sierra <alex.sierra@amd.com>
Reviewed-by: Michael Chen <michael.chen@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdkfd: Add/remove queues on the correct XCC on GFX 12.1

On GFX 12.1, pass the xcc id of the master XCC to choose the correct
MES Pipe to send the add_queue/remove_queue requests to MES.

Signed-off-by: Mukul Joshi <mukul.joshi@amd.com>
Reviewed-by: Alex Sierra <alex.sierra@amd.com>
Reviewed-by: Michael Chen <michael.chen@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdkfd: Don't partition VMID space on GFX 12.1

There is no need to partition VMID space on GFX 12.1 when
operating in CPX mode as SDMA is not sharing MMHUB on GFX 12.1.

Signed-off-by: Mukul Joshi <mukul.joshi@amd.com>
Reviewed-by: Alex Sierra <alex.sierra@amd.com>
Reviewed-by: Michael Chen <michael.chen@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: Rework MES initialization on GFX 12.1

Currently, only SPX mode works on GFX 12.1. This patch reworks
the MES initialization to get other non-SPX modes working. For example,
for CPX mode, coop_enable bit needs to be set to 0. The shared command
buffer initialization is also not needed in CPX mode.
The shared command buffer initialization needs further improvements which
will be handled in later patches.

Signed-off-by: Mukul Joshi <mukul.joshi@amd.com>
Reviewed-by: Alex Sierra <alex.sierra@amd.com>
Reviewed-by: Michael Chen <michael.chen@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: Use correct MES pipe in non-SPX mode on GFX 12.1

On GFX 12.1, use the correct MES pipe instance for readiness before
sending MES commands on that pipe. Additionally, send the TLB requests
on the correct MES pipe in non-SPX modes.

Signed-off-by: Mukul Joshi <mukul.joshi@amd.com>
Reviewed-by: Alex Sierra <alex.sierra@amd.com>
Reviewed-by: Michael Chen <michael.chen@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: adjust xcc_id program logic for sdma v7_1

Adjust program logic for sdam v7_1, only use physical xcc_id
when program register to support compute partition.

Signed-off-by: Likun Gao <Likun.Gao@amd.com>
Reviewed-by: Lijo Lazar <lijo.lazar@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: adjust xcc logic for gfxhub v12_1

Adjust xcc_id logic to only use physical xcc_id when program
register, (use logic xcc_id by default), to fit for compute
partition.

Signed-off-by: Likun Gao <Likun.Gao@amd.com>
Reviewed-by: Lijo Lazar <lijo.lazar@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: adjust xcc_cp_resume function for gfx_v12_1

Adjust gfx_v12_1_xcc_cp_resume function to program
cp resume per xcc_id (logic xcc number) to fix for
xcp_resume.
V2: Allocate compute microcode bo when sw init

Signed-off-by: Likun Gao <Likun.Gao@amd.com>
Reviewed-by: Lijo Lazar <lijo.lazar@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdkfd: Add SDMA queue quantum support for GFX12.1

    program SDMAx_QUEUEx_SCHEDULE_CNTL for context switch due to
    quantum in KFD for GFX12.1

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

drm/amdkfd: Set SDMA_QUEUEx_IB_CNTL/SWITCH_INSIDE_IB

    When submitting MQD to CP, set SDMA_QUEUEx_IB_CNTL/SWITCH_INSIDE_IB bit
    so it'll allow SDMA preemption if there is a massive command buffer of
    long-running SDMA commands.

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

drm/amdgpu: disable burst for gfx v12_1

Disable burst in GL1A and GLARBA for gfx v12_1.

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

drm/amdgpu: Setup Retry based thrashing prevention on GFX 12.1

Enable the new UTCL0 retry-based thrashing prevention on GFX 12.1.

Signed-off-by: Mukul Joshi <mukul.joshi@amd.com>
Reviewed-by: Alex Sierra <alex.sierra@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: Program IH_VMID_LUT_INDEX register on GFX 12.1

For querying VMID <-> PASID mapping on GFX 12.1, we need to first
program the IH_VMID_LUT_INDEX before fetching the LUT mapping. Without
this TLB flush may not work.

Signed-off-by: Mukul Joshi <mukul.joshi@amd.com>
Reviewed-by: Michael Chen <michael.chen@amd.com>
Reviewed-by: Alex Sierra <alex.sierra@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/ras: Support physical address convert

Support physical address convert to current NPS
pages in uniras.

Signed-off-by: Jinzhou Su <jinzhou.su@amd.com>
Reviewed-by: YiPeng Chai <YiPeng.Chai@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu/gfx_v12_1: add mqd_stride_size input parameter

mqd_stride_size is used to calculate the next mqd offset
for cooperative dispatch.

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

drm/amdkfd: Fix a couple of spelling mistakes

There are a couple of spelling mistakes, one in a pr_warn message
and one in a seq_printf message. Fix these.

Signed-off-by: Colin Ian King <colin.i.king@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: Describe @AMD_IP_BLOCK_TYPE_RAS in amd_ip_block_type enum

Sphinx reports kernel-doc warning:

WARNING: ./drivers/gpu/drm/amd/include/amd_shared.h:113 Enum value 'AMD_IP_BLOCK_TYPE_RAS' not described in enum 'amd_ip_block_type'

Describe the value to fix it.

Fixes: 7169e706c82d ("drm/amdgpu: Add ras module ip block to amdgpu discovery")
Signed-off-by: Bagas Sanjaya <bagasdotme@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Don't use kernel-doc comment in dc_register_software_state struct

Sphinx reports kernel-doc warning:

WARNING: ./drivers/gpu/drm/amd/display/dc/dc.h:2796 This comment starts with '/**', but isn't a kernel-doc comment. Refer to Documentation/doc-guide/kernel-doc.rst
 * Software state variables used to program register fields across the display pipeline

Don't use kernel-doc comment syntax to fix it.

Fixes: b0ff344fe70c ("drm/amd/display: Add interface to capture expected HW state from SW state")
Signed-off-by: Bagas Sanjaya <bagasdotme@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Reduce number of arguments of dcn30's CalculateWatermarksAndDRAMSpeedChangeSupport()

CalculateWatermarksAndDRAMSpeedChangeSupport() has a large number of
parameters, which must be passed on the stack. Most of the parameters
between the two callsites are the same, so they can be accessed through
the existing mode_lib pointer, instead of being passed as explicit
arguments. Doing this reduces the stack size of
dml30_ModeSupportAndSystemConfigurationFull() from 1912 bytes to 1840
bytes building for x86_64 with clang-22, helping stay under the 2048
byte limit for display_mode_vba_30.c.

Additionally, now that there is a pointer to mode_lib->vba available,
use 'v' consistently throughout the entire function.

Signed-off-by: Nathan Chancellor <nathan@kernel.org>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Reduce number of arguments of dcn30's CalculatePrefetchSchedule()

After an innocuous optimization change in clang-22,
dml30_ModeSupportAndSystemConfigurationFull() is over the 2048 byte
stack limit for display_mode_vba_30.c.

  drivers/gpu/drm/amd/amdgpu/../display/dc/dml/dcn30/display_mode_vba_30.c:3529:6: warning: stack frame size (2096) exceeds limit (2048) in 'dml30_ModeSupportAndSystemConfigurationFull' [-Wframe-larger-than]
   3529 | void dml30_ModeSupportAndSystemConfigurationFull(struct display_mode_lib *mode_lib)
        |      ^

With clang-21, this function was already close to the limit:

  drivers/gpu/drm/amd/amdgpu/../display/dc/dml/dcn30/display_mode_vba_30.c:3529:6: warning: stack frame size (1912) exceeds limit (1586) in 'dml30_ModeSupportAndSystemConfigurationFull' [-Wframe-larger-than]
   3529 | void dml30_ModeSupportAndSystemConfigurationFull(struct display_mode_lib *mode_lib)
        |      ^

CalculatePrefetchSchedule() has a large number of parameters, which must
be passed on the stack. Most of the parameters between the two callsites
are the same, so they can be accessed through the existing mode_lib
pointer, instead of being passed as explicit arguments. Doing this
reduces the stack size of dml30_ModeSupportAndSystemConfigurationFull()
from 2096 bytes to 1912 bytes with clang-22.

Closes: https://github.com/ClangBuiltLinux/linux/issues/2117
Signed-off-by: Nathan Chancellor <nathan@kernel.org>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Apply e4479aecf658 to dml

After an innocuous optimization change in clang-22, allmodconfig (which
enables CONFIG_KASAN and CONFIG_WERROR) breaks with:

  drivers/gpu/drm/amd/amdgpu/../display/dc/dml/dcn32/display_mode_vba_32.c:1724:6: error: stack frame size (3144) exceeds limit (3072) in 'dml32_ModeSupportAndSystemConfigurationFull' [-Werror,-Wframe-larger-than]
   1724 | void dml32_ModeSupportAndSystemConfigurationFull(struct display_mode_lib *mode_lib)
        |      ^

With clang-21, this function was already pretty close to the existing
limit of 3072 bytes.

  drivers/gpu/drm/amd/amdgpu/../display/dc/dml/dcn32/display_mode_vba_32.c:1724:6: error: stack frame size (2904) exceeds limit (2048) in 'dml32_ModeSupportAndSystemConfigurationFull' [-Werror,-Wframe-larger-than]
   1724 | void dml32_ModeSupportAndSystemConfigurationFull(struct display_mode_lib *mode_lib)
        |      ^

A similar situation occurred in dml2, which was resolved by
commit e4479aecf658 ("drm/amd/display: Increase sanitizer frame larger
than limit when compile testing with clang") by increasing the limit for
clang when compile testing with certain sanitizer enabled, so that
allmodconfig (an easy testing target) continues to work.

Apply that same change to the dml folder to clear up the warning for
allmodconfig, unbreaking the build.

Closes: https://github.com/ClangBuiltLinux/linux/issues/2135
Signed-off-by: Nathan Chancellor <nathan@kernel.org>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/radeon : Use devm_i2c_add_adapter instead of i2c_add_adapter

Replace i2c_add_adapter() with devm_i2c_add_adapter() and remove all
associated cleanup, as devm_i2c_add_adapter() handles adapter teardown
automatically.

Signed-off-by: Erick Karanja <karanja99erick@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: Update AMDGPU_INFO_UQ_FW_AREAS query for sdma

Add a query for sdma queues.  Userspace can use this to
query the size of the CSA buffers for sdma user queues.

Proposed userspace:
https://gitlab.freedesktop.org/yogeshmohan/mesa/-/commits/userq_query

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

drm/amdgpu: Update AMDGPU_INFO_UQ_FW_AREAS query for compute

Add a query for compute queues.  Userspace can use this to
query the size of the EOP buffers for compute user queues.

Proposed userspace:
https://gitlab.freedesktop.org/yogeshmohan/mesa/-/commits/userq_query

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

drm/radeon: Convert legacy DRM logging in evergreen.c to drm_* helpers

Replace DRM_DEBUG(), DRM_ERROR(), and DRM_INFO() calls with the
corresponding drm_dbg(), drm_err(), and drm_info() helpers in the
radeon driver.

The drm_*() logging helpers take a struct drm_device * argument,
allowing the DRM core to prefix log messages with the correct device
name and instance. This is required to correctly distinguish log
messages on systems with multiple GPUs.

This change aligns radeon with the DRM TODO item:
"Convert logging to drm_* functions with drm_device parameter".

Signed-off-by: Abhishek Rajput <abhiraj21put@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: Add gfx v12_1 interrupt source header

To acommandate specific interrupt source for gfx v12_1

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

drm/amdkfd: Override KFD SVM mappings for GFX 12.1

Override the local MTYPE mappings in KFD SVM code with mtype_local
modprobe param for GFX 12.1.

Signed-off-by: Mukul Joshi <mukul.joshi@amd.com>
Reviewed-by: Alex Sierra <alex.sierra@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: correct rlc autoload for xcc harvest

If the number instances of firmware is RLC_NUM_INS_CODE0(Only 1 inst),
need to copy it directly for rlcautolad.
For the firmware which instances number bigger than 1, only copy for
enabled XCC to save copy time.

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

drm/amdgpu: add gfx sysfs support for gfx_v12_1

Add gfx sysfs support for gfx_v12_1.

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

drm/amdgpu/mes_v12_1: fix mes access xcd register

Fix to use local register offset inside die for mes fw accessing
local/remote xcd register.

Signed-off-by: Jack Xiao <Jack.Xiao@amd.com>
Reviewed-by: Likun Gao <Likun.Gao@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: normalize reg addr as local xcc for gfx v12_1

Normalize registers address to local xcc address for gfx v12_1.

Signed-off-by: Likun Gao <Likun.Gao@amd.com>
Reviewed-by: Lijo Lazar <lijo.lazar@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: support xcc harvest for ih translate

Support xcc harvest for ih translate to logic xcc.
V2: Only check available instances

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

drm/amdgpu: Correct inst_id input from physical to logic

Correct inst_id input from physical to logic for sdma v7_1.
V2: Show real instance number on logic xcc.

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

drm/amdgpu: use physical xcc id to get rrmt

Use physical xcc_id to get rrmt on misc_op for mes v12_1.

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

drm/radeon: Convert logging in radeon_display.c to drm_* helpers

Replace DRM_ERROR() and DRM_INFO() calls in
drivers/gpu/drm/radeon/radeon_display.c with the corresponding
drm_err() and drm_info() helpers.

The drm_*() logging functions take a struct drm_device * argument,
allowing the DRM core to prefix log messages with the correct device
name and instance. This is required to correctly distinguish log
messages on systems with multiple GPUs.

This change aligns radeon with the DRM TODO item:
"Convert logging to drm_* functions with drm_device parameter".

Acked-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Mukesh Ogare <mukeshogare871@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdkfd: Fix improper NULL termination of queue restore SMI event string

Pass character "0" rather than NULL terminator to properly format
queue restoration SMI events. Currently, the NULL terminator precedes
the newline character that is intended to delineate separate events
in the SMI event buffer, which can break userspace parsers.

Signed-off-by: Brian Kocoloski <brian.kocoloski@amd.com>
Reviewed-by: Philip Yang <Philip.Yang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: Correct xcc_id input to GET_INST from physical to logic

Correct xcc_id input to GET_INST from physical to logic for
gfx_v12_1.

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

drm/amdgpu: Fix CP_MEC_MDBASE in multi-xcc for gfx v12_1

Need to allocate memory for MEC FW data and program
registers CP_MEC_MDBASE for each XCC respectively.

Signed-off-by: Michael Chen <michael.chen@amd.com>
Acked-by: Harish Kasiviswanathan <Harish.Kasiviswanathan@amd.com>
Reviewed-by: Shaoyun.liu <Shaoyun.liu@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: Support 57bit fault address for GFX 12.1.0

The gmc fault virtual address is up to 57bit for 5 level page table,
this also works with 48bit virtual address for 4 level page table.

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

drm/amdgpu: Add pde3 table invalidation request for GFX 12.1.0

Set pde3 invalidation request bit during tlb flush for up to 5 level
page table.

Signed-off-by: Philip Yang <Philip.Yang@amd.com>
Acked-by: Christian König <christian.koenig@amd.com>
Acked-by: Felix Kuehling <felix.kuehling@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdkfd: Update LDS, Scratch base for 57bit address

For 5-level page tables, update compute vmid sh_mem_base LDS aperture
and Scratch aperture base address to above 57-bit, use the same setting
from gfx vmid, we can remove the duplicate macro.

Update queue pdd lds_base and scratch_base to the same value as
sh_mem_base setting. Then application get process apertures return the
correct value to access LDS and Scratch memory for 57bit address 5-level
page tables. This may pass to MES in future when mapping queue.

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

drm/amdgpu: Enable 5-level page table for GFX 12.1.0

GFX 12.1.0 support 57bit virtual, 52bit physical address, set PDE
max_level to 4, min_vm_size to 128PB to enable GPU vm 5-level page
tables to support 57bit virtual address.

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

drm/amdgpu: init RS64_MEC_P2/P3_STACK for gfx12.1

Add GFX12.1 MEC P2/P3 STACK firmware init.

Signed-off-by: Feifei Xu <Feifei.Xu@amd.com>
Reviewed-by: Lijo Lazar <lijo.lazar@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: Fix CU info calculations for GFX 12.1

This patch fixes the CU info calculations for gfx 12.1.

Signed-off-by: Mukul Joshi <mukul.joshi@amd.com>
Reviewed-by: Lijo Lazar <lijo.lazar@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdkfd: Update CWSR area calculations for GFX 12.1

Update the SGPR, VGPR, HWREG size and number of waves supported
for GFX 12.1 CWSR memory limits. The CU calculation changed in
topology, as a result, the values need to be updated.

Signed-off-by: Mukul Joshi <mukul.joshi@amd.com>
Reviewed-by: Feifei Xu <Feifei.Xu@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: Add soc v1_0 ih client id table

To acommandate the specific ih client for soc v1_0

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

drm/amdgpu: Flush TLB on all XCCs on GFX 12.1

Currently, the driver code is flushing TLB on XCC 0 only.
Fix it by flushing on all XCCs within the partition.

Signed-off-by: Mukul Joshi <mukul.joshi@amd.com>
Reviewed-by: Felix Kuehling <felix.kuehling@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/pm: restore SCLK settings after S0ix resume

User-configured SCLK(GPU core clock)frequencies were not persisting
across S0ix suspend/resume cycles on smu v14 hardware.
The issue occurred because of the code resetting clock frequency
to zero during resume.

This patch addresses the problem by:
- Preserving user-configured values in driver and sets the
  clock frequency across resume
- Preserved settings are sent to the hardware during resume

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

drm/amdgpu: do not use amdgpu_bo_gpu_offset_no_check individually

This should not be used indiviually, use amdgpu_bo_gpu_offset
with bo reserved.

v3 - unpin bo in queue destroy (Christian)
v2 - pin bo so that offset returned won't change after unlock (Christian)

Signed-off-by: Saleemkhan Jamadar <saleemkhan083@gmail.com>
Suggested-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: Change set ip clock/power gating param

It's not required to use generic void *, change to struct amdgpu_device *.

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

drm/amdgpu: Use helper to get ip block

Replace individual searches with the utility function get_ip_block

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

drm/amdgpu: Move ip block related functions

Move ip block related functions to amdgpu_ip.c. No functional change
intended.

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

drm/amdgpu: fix a job->pasid access race in gpu recovery

Avoid a possible UAF in GPU recovery due to a race between
the sched timeout callback and the tdr work queue.

The gpu recovery function calls drm_sched_stop() and
later drm_sched_start().  drm_sched_start() restarts
the tdr queue which will eventually free the job.  If
the tdr queue frees the job before time out callback
completes, the job will be freed and we'll get a UAF
when accessing the pasid.  Cache it early to avoid the
UAF.

Example KASAN trace:
[  493.058141] BUG: KASAN: slab-use-after-free in amdgpu_device_gpu_recover+0x968/0x990 [amdgpu]
[  493.067530] Read of size 4 at addr ffff88b0ce3f794c by task kworker/u128:1/323
[  493.074892]
[  493.076485] CPU: 9 UID: 0 PID: 323 Comm: kworker/u128:1 Tainted: G            E       6.16.0-1289896.2.zuul.bf4f11df81c1410bbe901c4373305a31 #1 PREEMPT(voluntary)
[  493.076493] Tainted: [E]=UNSIGNED_MODULE
[  493.076495] Hardware name: TYAN B8021G88V2HR-2T/S8021GM2NR-2T, BIOS V1.03.B10 04/01/2019
[  493.076500] Workqueue: amdgpu-reset-dev drm_sched_job_timedout [gpu_sched]
[  493.076512] Call Trace:
[  493.076515]  <TASK>
[  493.076518]  dump_stack_lvl+0x64/0x80
[  493.076529]  print_report+0xce/0x630
[  493.076536]  ? _raw_spin_lock_irqsave+0x86/0xd0
[  493.076541]  ? __pfx__raw_spin_lock_irqsave+0x10/0x10
[  493.076545]  ? amdgpu_device_gpu_recover+0x968/0x990 [amdgpu]
[  493.077253]  kasan_report+0xb8/0xf0
[  493.077258]  ? amdgpu_device_gpu_recover+0x968/0x990 [amdgpu]
[  493.077965]  amdgpu_device_gpu_recover+0x968/0x990 [amdgpu]
[  493.078672]  ? __pfx_amdgpu_device_gpu_recover+0x10/0x10 [amdgpu]
[  493.079378]  ? amdgpu_coredump+0x1fd/0x4c0 [amdgpu]
[  493.080111]  amdgpu_job_timedout+0x642/0x1400 [amdgpu]
[  493.080903]  ? pick_task_fair+0x24e/0x330
[  493.080910]  ? __pfx_amdgpu_job_timedout+0x10/0x10 [amdgpu]
[  493.081702]  ? _raw_spin_lock+0x75/0xc0
[  493.081708]  ? __pfx__raw_spin_lock+0x10/0x10
[  493.081712]  drm_sched_job_timedout+0x1b0/0x4b0 [gpu_sched]
[  493.081721]  ? __pfx__raw_spin_lock_irq+0x10/0x10
[  493.081725]  process_one_work+0x679/0xff0
[  493.081732]  worker_thread+0x6ce/0xfd0
[  493.081736]  ? __pfx_worker_thread+0x10/0x10
[  493.081739]  kthread+0x376/0x730
[  493.081744]  ? __pfx_kthread+0x10/0x10
[  493.081748]  ? __pfx__raw_spin_lock_irq+0x10/0x10
[  493.081751]  ? __pfx_kthread+0x10/0x10
[  493.081755]  ret_from_fork+0x247/0x330
[  493.081761]  ? __pfx_kthread+0x10/0x10
[  493.081764]  ret_from_fork_asm+0x1a/0x30
[  493.081771]  </TASK>

Fixes: a72002cb181f ("drm/amdgpu: Make use of drm_wedge_task_info")
Link: https://github.com/HansKristian-Work/vkd3d-proton/pull/2670
Cc: SRINIVASAN.SHANMUGAM@amd.com
Cc: vitaly.prosyak@amd.com
Cc: christian.koenig@amd.com
Suggested-by: Matthew Brost <matthew.brost@intel.com>
Reviewed-by: Srinivasan Shanmugam <srinivasan.shanmugam@amd.com>
Reviewed-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/amd/display: Promote DC to 3.2.363

This version brings along the following updates:

- Replay Video Conferencing V2
- Fix scratch registers offsets for DCN35 and DCN351
- Fix DP no audio issue
- Add use_max_lsw parameter
- Fix presentation of Z8 efficiency
- Add USB-C DP Alt Mode lane limitation in DCN32
- Support DRR granularity
- Don't disable DPCD mst_en if sink connected
- Set enable_legacy_fast_update to false for DCN35/351
- Split update_planes_and_stream_v3 into parts (V2)

Signed-off-by: Taimur Hassan <Syed.Hassan@amd.com>
Signed-off-by: Chenyu Chen <chen-yu.chen@amd.com>
Acked-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: [FW Promotion] Release 0.1.40.0

Summary for changes in firmware:
* Update DCHVM restore sequence for dcn35
* Add 2 new debug polling methods for dchvm "busy" during IPS entry for DCN35

Signed-off-by: Taimur Hassan <Syed.Hassan@amd.com>
Signed-off-by: Chenyu Chen <chen-yu.chen@amd.com>
Acked-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: Split update_planes_and_stream_v3 into parts (V2)

[Why]
Currently all of the preparation and execution of plane update is done
under a DC lock, blocking other code from accessing DC for longer than
strictly necessary.

[How]
Break the v3 update flow into 3 parts:
    * prepare - locked, calculate update flow and modify DC state
    * execute - unlocked, program hardware
    * cleanup - locked, finalize DC state and free temp resources
Legacy v2 flow too compilicated to break down for now, link new API
with old by executing everything in slightly misnamed prepare stage.

V2:
Keep the new code structure, but point all users back at the old code,
until fully tested.

Reviewed-by: Nicholas Kazlauskas <nicholas.kazlauskas@amd.com>
Signed-off-by: Dominik Kaszewski <dominik.kaszewski@amd.com>
Signed-off-by: Roman Li <roman.li@amd.com>
Signed-off-by: Chenyu Chen <chen-yu.chen@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>