drm/amdgpu/userq: declutter the code with goto

Clean up the amdgpu_userq_create function clean up in
failure condition using goto method. This avoid replication
of cleanup for every failure condition.

Signed-off-by: Sunil Khatri <sunil.khatri@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu/userq: defer queue publication until create completes

The userq create path publishes queues to global xarrays such as
userq_doorbell_xa and userq_xa before creation was fully complete.
Later on if create queue fails, teardown could free an already
visible queue, opening a UAF race with concurrent queue walkers.
Also calling amdgpu_userq_put in such cases complicates the cleanup.

Solution is to defer queue publication until create succeeds and no
partially initialized queue is exposed.

Signed-off-by: Sunil Khatri <sunil.khatri@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: Check for multiplication overflow in checkpoint stack size

get_checkpoint_info() in kfd_mqd_manager_v9.c finds 32-bit value
ctl_stack_size by multiplying two 32-bit values. This can overflow to a
lower value, which could result in copying outside the bounds of
a buffer in checkpoint_mqd() in the same file.

Put in a check for the overflow, and fail with -EINVAL if detected.

v2: use check_mul_overflow()

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

drm/amdkfd: fix the warning for potential insecure string

Below is the warning thrown by the clang compiler:
linux/drivers/gpu/drm/amd/amdgpu/../amdkfd/kfd_process.c:588:9: warning: format string is not a string literal (potentially insecure) [-Wformat-security]
                                           stats_dir_filename);
                                           ^~~~~~~~~~~~~~~~~~
linux/drivers/gpu/drm/amd/amdgpu/../amdkfd/kfd_process.c:588:9: note: treat the string as an argument to avoid this
                                           stats_dir_filename);
                                           ^
                                           "%s",
linux/drivers/gpu/drm/amd/amdgpu/../amdkfd/kfd_process.c:635:18: warning: format string is not a string literal (potentially insecure) [-Wformat-security]
                                           p->kobj, counters_dir_filename);
                                                    ^~~~~~~~~~~~~~~~~~~~~
linux/drivers/gpu/drm/amd/amdgpu/../amdkfd/kfd_process.c:635:18: note: treat the string as an argument to avoid this
                                           p->kobj, counters_dir_filename);
                                                    ^
                                                    "%s",

Signed-off-by: Sunil Khatri <sunil.khatri@amd.com>
CC: Philip Yang <philip.yang@amd.com>
CC: Felix Kuehling <felix.kuehling@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: fix warning for potentially insecure string

linux/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c:2358:24: warning: format string is not a string literal (potentially insecure) [-Wformat-security]
                         sprintf(ring->name, amdgpu_sw_ring_name(i));
                                             ^~~~~~~~~~~~~~~~~~~~~~
linux/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c:2358:24: note: treat the string as an argument to avoid this
                         sprintf(ring->name, amdgpu_sw_ring_name(i));

Signed-off-by: Sunil Khatri <sunil.khatri@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/pm: Enable aid/xcd/hbm temperature reporting

Enable aid/xcd/hbm temperature reporting for smu_v13_0_12 via gpu
metrics

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/amd/pm: Add aid/xcd/hbm temperatures

Add aid, xcd & hbm temperatures to gpu metrics for smu_v13_0_12

v2: Use correct umc control per stack (Lijo)

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/amdgpu: include ip discovery data in devcoredump

This is the best way to describe the GPU to a tool loading
the devcoredump.

Signed-off-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: fix gpu idle power consumption issue for gfx v12

Older versions of the MES firmware may cause abnormal GPU power consumption.
When performing inference tasks on the GPU (e.g., with Ollama using ROCm),
the GPU may show abnormal power consumption in idle state and incorrect GPU load information.
This issue has been fixed in firmware version 0x8b and newer.

Closes: https://github.com/ROCm/ROCm/issues/5706
Signed-off-by: Yang Wang <kevinyang.wang@amd.com>
Acked-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: Fix kernel-doc comments for some LUT properties

The following members of struct amdgpu_mode_info do not have valid
references in the related kernel-doc sections:

 - plane_shaper_lut_property
 - plane_shaper_lut_size_property,
 - plane_lut3d_size_property

Correct all affected comment blocks.

Fixes: f545d82479b4 ("drm/amd/display: add plane shaper LUT and TF driver-specific properties")
Fixes: 671994e3bf33 ("drm/amd/display: add plane 3D LUT driver-specific properties")
Reviewed-by: Melissa Wen <mwen@igalia.com>
Signed-off-by: Cristian Ciocaltea <cristian.ciocaltea@collabora.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu/amdgpu_connectors: remove amdgpu_connector_free_edid

Now that we are using struct drm_edid, we can just call drm_edid_free
directly. Remove the function and update calls to drm_edid_free.

Signed-off-by: Joshua Peisach <jpeisach@ubuntu.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu/amdgpu_connectors: use struct drm_edid instead of struct edid

Some amdgpu code is still using deprecated edid functions. Switch to
the newer functions and update the amdgpu_connector struct's edid type
to the drm_edid type.

At the same time, use the raw EDID when we need to for speaker
allocations and for determining if the input is digital.

Signed-off-by: Joshua Peisach <jpeisach@ubuntu.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu/gfx12.1: add support for disable_kq

Plumb in support for disabling kernel queues and make it
the default.  For testing, kernel queues can be re-enabled
by setting amdgpu.user_queue=0

v2: integrate feedback from Lijo

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

drm/amd: Fix a few more NULL pointer dereference in device cleanup

I found a few more paths that cleanup fails due to a NULL version pointer
on unsupported hardware.

Add NULL checks as applicable.

Fixes: 39fc2bc4da00 ("drm/amdgpu: Protect GPU register accesses in powergated state in some paths")
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Mario Limonciello <mario.limonciello@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: gfx 12.1 cleanups

Remove some remnants from when the code was forked
from gfx 12.0.

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

drm/radeon: Test for fbdev GEM object with generic helper

Replace radeon's test for the fbdev GEM object with a call to the
generic helper.

Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Thomas Zimmermann <tzimmermann@suse.de>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: Move test for fbdev GEM object into generic helper

Provide a generic helper that tests if fbdev emulation is backed by
a specific GEM object. Not all drivers use client buffers (yet), hence
also test against the first GEM object in the fbdev framebuffer.

Convert amdgpu. The helper will also be useful for radeon.

Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Thomas Zimmermann <tzimmermann@suse.de>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/pm: add missing od setting PP_OD_FEATURE_ZERO_FAN_BIT for smu v14

add missing od setting PP_OD_FEATURE_ZERO_FAN_BIT for smu v14.0.2/14.0.3

Fixes: 9710b84e2a6a ("drm/amd/pm: add overdrive support on smu v14.0.2/3")
Closes: https://gitlab.freedesktop.org/drm/amd/-/issues/5018
Signed-off-by: Yang Wang <kevinyang.wang@amd.com>
Acked-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu/userq: remove queue from doorbell xa during clean up

If function amdgpu_userq_map_helper fails we do need to clean
up and remove the queue from the userq_doorbell_xa.

Signed-off-by: Sunil Khatri <sunil.khatri@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu/userq: remove queue from doorbell xarray

In case of failure in xa_alloc, remove the queue during
clean up from the userq_doorbell_xa.

Signed-off-by: Sunil Khatri <sunil.khatri@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd: Fix NULL pointer dereference in device cleanup

When GPU initialization fails due to an unsupported HW block
IP blocks may have a NULL version pointer. During cleanup in
amdgpu_device_fini_hw, the code calls amdgpu_device_set_pg_state and
amdgpu_device_set_cg_state which iterate over all IP blocks and access
adev->ip_blocks[i].version without NULL checks, leading to a kernel
NULL pointer dereference.

Add NULL checks for adev->ip_blocks[i].version in both
amdgpu_device_set_cg_state and amdgpu_device_set_pg_state to prevent
dereferencing NULL pointers during GPU teardown when initialization has
failed.

Fixes: 39fc2bc4da00 ("drm/amdgpu: Protect GPU register accesses in powergated state in some paths")
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Mario Limonciello <mario.limonciello@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/pm: add missing od setting PP_OD_FEATURE_ZERO_FAN_BIT for smu v13

add missing od setting PP_OD_FEATURE_ZERO_FAN_BIT for smu v13.0.0/13.0.7

Fixes: cfffd980bf21 ("drm/amd/pm: add zero RPM OD setting support for SMU13")
Closes: https://gitlab.freedesktop.org/drm/amd/-/issues/5018
Signed-off-by: Yang Wang <kevinyang.wang@amd.com>
Acked-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: Fix mutex handling in amdgpu_benchmark_do_move() v3

amdgpu_benchmark_do_move() can exit the loop early if
amdgpu_copy_buffer() or dma_fence_wait() fails.

In the error path, the function jumps to the exit label
without releasing adev->mman.default_entity.lock, which
leaves the mutex held and results in a lock imbalance.

This can block subsequent users of default_entity and
potentially cause deadlocks.

Move the mutex_unlock() to the common exit path so the
lock is released on both success and error returns.

This fixes:
drivers/gpu/drm/amd/amdgpu/amdgpu_benchmark.c:57 amdgpu_benchmark_do_move()
warn: inconsistent returns '&adev->mman.default_entity.lock'.

v2:
- Drop unrelated initialization of 'r'
- Keep the change focused on the mutex imbalance fix (Pierre).

v3:
- Removed empty line

Fixes: 30f2daedf4d8 ("drm/amdgpu: add missing lock in amdgpu_benchmark_do_move")
Reported-by: Dan Carpenter <dan.carpenter@linaro.org>
Cc: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@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: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/pm: Avoid overflow when sorting pp_feature list

pp_features sorting uses int8_t sort_feature[] to store driver
feature enum indices. On newer ASICs the enum index can exceed 127,
causing signed overflow and silently dropping entries from the output.
Switch the array to int16_t so all enum indices are preserved.

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/amd/pm: Add pm firmware info to dmesg log

Add PMFW info to dmesg log for SMUv13 SOCs. It's helpful as diagnostic
data for some driver load issues.

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

drm/amdgpu/psp: Use Indirect access address for GFX to PSP mailbox

The reason the RAP is not granting access to 0x58200 is that
a dedicated RSMU slot would have to be spent for this address range,
and MPASP is close to running out of RSMU slots.

This will help to fix PSP TOC load failure during secureboot.
GFX Driver Need to use indirect access for SMN address regs.

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

drm/amdgpu: Remove redundant missing hw ip handling

Now that it is guaranteed there can be no entity if there is no hw ip
block we can remove the open coded protection during CS parsing.

Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@igalia.com>
References: 55414ad5c983 ("drm/amdgpu: error out on entity with no run queue")
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: Reject impossible entities early

Currently there are two different behaviour modes when userspace tries to
operate on not present HW IP blocks. On a machine without UVD, VCE and VPE
blocks, this can be observed for example like this:

$ sudo ./amd_fuzzing --r cs-wait-fuzzing
...
amd_cs_wait_fuzzing DRM_IOCTL_AMDGPU_CTX r 0
amd_cs_wait_fuzzing AMDGPU_WAIT_CS AMD_IP_GFX r 0
amd_cs_wait_fuzzing AMDGPU_WAIT_CS AMD_IP_COMPUTE r 0
amd_cs_wait_fuzzing AMDGPU_WAIT_CS AMD_IP_DMA r 0
amd_cs_wait_fuzzing AMDGPU_WAIT_CS AMD_IP_UVD r -1
amd_cs_wait_fuzzing AMDGPU_WAIT_CS AMD_IP_VCE r 0
amd_cs_wait_fuzzing AMDGPU_WAIT_CS AMD_IP_UVD_ENC r -1
amd_cs_wait_fuzzing AMDGPU_WAIT_CS AMD_IP_VCN_DEC r 0
amd_cs_wait_fuzzing AMDGPU_WAIT_CS AMD_IP_VCN_ENC r 0
amd_cs_wait_fuzzing AMDGPU_WAIT_CS AMD_IP_VCN_JPEG r 0
amd_cs_wait_fuzzing AMDGPU_WAIT_CS AMD_IP_VPE r 0

We can see that UVD returns an errno (-EINVAL) from the CS_WAIT ioctl,
while VCE and VPE return unexpected successes.

The difference stems from the fact the UVD is a load balancing engine
which retains the context, so with a workaround implemented in
amdgpu_ctx_init_entity(), but which does not account for the fact hardware
block may not be present.

This causes a single NULL scheduler to be passed to
drm_sched_entity_init(), which immediately rejects this with -EINVAL.

The not present VCE and VPE cases on the other hand pass zero schedulers
to drm_sched_entity_init(), which is explicitly allowed and results in
unusable entities.

As the UVD case however shows, call paths can handle the errors, so we can
consolidate this into a single path which will always return -EINVAL if
the HW IP block is not present.

We do this by rejecting it early and not calling drm_sched_entity_init()
when there is no backing hardware.

This also removes the need for the drm_sched_entity_init() to handle the
zero schedulers and NULL scheduler cases, which means that we can follow
up by removing the special casing from the DRM scheduler.

Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@igalia.com>
References: f34e8bb7d6c6 ("drm/sched: fix null-ptr-deref in init entity")
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu/userq: refcount userqueues to avoid any race conditions

To avoid race condition and avoid UAF cases, implement kref
based queues and protect the below operations using xa lock
a. Getting a queue from xarray
b. Increment/Decrement it's refcount

Every time some one want to access a queue, always get via
amdgpu_userq_get to make sure we have locks in place and get
the object if active.

A userqueue is destroyed on the last refcount is dropped which
typically would be via IOCTL or during fini.

v2: Add the missing drop in one the condition in the signal ioclt [Alex]

v3: remove the queue from the xarray first in the free queue ioctl path
    [Christian]

- Pass queue to the amdgpu_userq_put directly.
- make amdgpu_userq_put xa_lock free since we are doing put for each get
  only and final put is done via destroy and we remove the queue from xa
  with lock.
- use userq_put in fini too so cleanup is done fully.

v4: Use xa_erase directly rather than doing load and erase in free
    ioctl. Also remove some of the error logs which could be exploited
    by the user to flood the logs [Christian]

Signed-off-by: Sunil Khatri <sunil.khatri@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: Fix use-after-free race in VM acquire

Replace non-atomic vm->process_info assignment with cmpxchg()
to prevent race when parent/child processes sharing a drm_file
both try to acquire the same VM after fork().

Reviewed-by: Harish Kasiviswanathan <Harish.Kasiviswanathan@amd.com>
Signed-off-by: Alysa Liu <Alysa.Liu@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/pm: remove invalid gpu_metrics.energy_accumulator on smu v13.0.x

v1:
The metrics->EnergyAccumulator field has been deprecated on newer pmfw.

v2:
add smu 13.0.0/13.0.7/13.0.10 support.

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

drm/amd/ras: adapt sync info func for pmfw eeprom

adapt sync info func for pmfw eeprom

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

drm/amd/ras: add check func for pmfw eeprom

add check func for pmfw eeprom

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

drm/amdgpu: GFX12.1 scratch memory limit up to 57-bit

The scratch aperture or gmc private aperture in flat memory contains
57 bits of data on gfx v12.1.0 compared to the 32 bits from previous.

Add new helper kfd_init_apertures_v12 for gfx version >= v12.1.0 which
supports 57-bit VA space.

v2:
  - update pdd->scratch_limit (Yu, Lang)
  - update fixes tag (Felix Kuehling)
  - add helper kfd_init_apertures_v12

Fixes: db1882b3ff0c ("drm/amdkfd: Update LDS, Scratch base for 57bit address")
Signed-off-by: Philip Yang <Philip.Yang@amd.com>
Reviewed-by: Lang Yu <lang.yu@amd.com>
Acked-by: Felix Kuehling <felix.kuehling@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/ras: add initialization func for pmfw eeprom

add initialization func for pmfw eeprom

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

drm/amd/ras: adapt page retirement process for pmfw eeprom

read bad page data from pmfw eeprom when retirement
is triggered, use timestamp read from eeprom

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

drm/amd/ras: add read func for pmfw eeprom

add read func for pmfw eeprom, and adapt address converting
for bad pages loaded from pmfw eeprom

v2: change label 'Out' to 'out'

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

drm/amd/ras: make MCA IPID parse global

add a new IPID parse interface for umc, so we can
implement it for each ASIC, and so we can call it
in other blocks

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

drm/amd/ras: add append func for pmfw eeprom

add append func for pmfw eeprom

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

drm/amd/ras: add check safety watermark func for pmfw eeprom

add check safety watermark func for pmfw eeprom

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

drm/amd/ras: Add table reset func for pmfw eeprom

add table reset func for pmfw eeprom, add smu eeprom control
structure

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

drm/amd/display: remove extra ; from statement, remove extra tabs

There is a statement that has a ;; at the end, remove the extraneous ;
and remove extra tabs in the code block.

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

drm/amdgpu: Use get_smn_base in aqua_vanjaram

Use get_smn_base interface to get IP die instance's base offset in
aqua_vanjaram. encode_ext_smn_addressing is not used.

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

drm/amdgpu: Add smn callbacks to register block

Add smn block to register access and callback interface definition to
get smn base.

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

drm/amdgpu: Move pcie lock to register block

Move pcie register access lock to register access block.

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

drm/amdgpu: Add pcie64 extended to register block

Add extended pcie 64-bit access method to register access block.

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

drm/amdgpu: Add pcie64 indirect to register block

Move 64-bit pcie indirect read/writes to register access block.

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

drm/amdgpu: Add pcie ext access to register block

Move pcie extended access (64-bit address) to register access block.

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

drm/amdgpu: Add pcie indirect to register block

Move pcie indirect access to register access block.

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

drm/amdgpu: Add pciep method to register block

Move pcie port method to register access block.

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

drm/amdgpu: Add audio method to register block

Move audio endpoint callbacks to register access block.

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

drm/amdgpu: Add se cac method to register block

Move se cac access callbacks to register access block.

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

drm/amdgpu: Add gc cac method to register block

Move gc cac access callbacks to register access block.

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

drm/amdgpu: Add didt method to register block

Move didt callbacks to register access block.

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

drm/amdgpu: Drop redundant syncobj handle limit checks in userq ioctls

Clang warns that comparing a __u16 value against 65536 is always false.

num_syncobj_handles is defined as __u16 in both the userq signal and
wait ioctl argument structs, so it can never exceed 65535. The checks
against AMDGPU_USERQ_MAX_HANDLES are therefore redundant and trigger
-Wtautological-constant-out-of-range-compare.

Fixes: Clang -Wtautological-constant-out-of-range-compare in userq
signal/wait ioctls

Fixes: d8e760b7996d ("drm/amdgpu: update type for num_syncobj_handles in drm_amdgpu_userq_signal")
Fixes: c561d2320492 ("drm/amdgpu: update type for num_syncobj_handles in drm_amdgpu_userq_wait")
Cc: Alex Deucher <alexander.deucher@amd.com>
Cc: Christian König <christian.koenig@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/ras: add wrapper funcs for pmfw eeprom

add wrapper funcs for pmfw eeprom interface to make them
easier to be called

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

drm/amd/ras: add uniras smu feature flag init func

add flag to indicate if pmfw eeprom is supported or
not, and initialize it

v2: change copyright from 2025 to 2026

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

drm/amd/ras: add pmfw eeprom smu interfaces

add smu interfaces and its data structures for
pmfw eeprom in uniras

v2: add 'const' to smu messages array, and specify
    index for each member when initializing.

Signed-off-by: Gangliang Xie <ganglxie@amd.com>
Reviewed-by: Tao Zhou <tao.zhou1@amd.com>
Reviewed-by: Yang Wang <kevinyang.wang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/pm: add feature query interface for uniras

add amdgpu_smu_ras_feature_is_enabled to query one feature
is supported or not

v2: change default return value from -EOPNOTSUPP to 0

Signed-off-by: Gangliang Xie <ganglxie@amd.com>
Reviewed-by: Tao Zhou <tao.zhou1@amd.com>
Reviewed-by: Yang Wang <kevinyang.wang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/pm: add pmfw eeprom messages into uniras interface

add pmfw eeprom related messages into smu_v13_0_6_ras_send_msg

v2: add sriov check before sending smu commands

Signed-off-by: Gangliang Xie <ganglxie@amd.com>
Reviewed-by: Tao Zhou <tao.zhou1@amd.com>
Reviewed-by: Yang Wang <kevinyang.wang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: Add uvd indirect to register block

Add uvd indirect method to register access block and replace the
existing calls from adev.

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

drm/amdgpu: Add smc method to register block

Define register access block which consolidates different register access
methods. Add smc method to register access block.

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

drm/amdgpu: clear related counter after RAS eeprom reset

Make eeprom data and its counter consistent.

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

drm/amdgpu: compatible with specific RAS old eeprom format

Handle RAS eeprom record when UMC_CHANNEL_IDX_V2 is set.

v2: get UMC_CHANNEL_IDX_V2 flag before the clear of it.

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

drm/amdgpu: update type for num_syncobj_handles in drm_amdgpu_userq_wait

update the type for num_syncobj_handles from __u32 to _u16 with
required padding.

This breaks the UAPI for big-endian platforms but this is deliberate
and harmless since userqueues is still a beta feature. It is enabled
via module parameter and need the right fw support to work.

Signed-off-by: Sunil Khatri <sunil.khatri@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: update type for num_syncobj_handles in drm_amdgpu_userq_signal

update the type for num_syncobj_handles from __u64 to _u16 with
required padding.

This breaks the UAPI for big-endian platforms but this is deliberate
and harmless since userqueues is still a beta feature. It is enabled
via module parameter and need the right fw support to work.

Signed-off-by: Sunil Khatri <sunil.khatri@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu/userq: change queue id type to u32 from int

queue id always remain a positive value and should
be of type unsigned.

With this we also dont need to typecast the id to other
types specially in xarray functions.

Signed-off-by: Sunil Khatri <sunil.khatri@amd.com>
Acked-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Promote DC to 3.2.372

This version brings along the following updates:

- Prevent integer overflow when mhz to khz
- Remove always-false branches
- Remove redundant initializers
- Silence unused variable warning
- Initialize replay_state to PR_STATE_INVALID
- Fallback to boot snapshot for dispclk
- Skip cursor cache reset if hubp powergating is disabled

Reviewed-by: Leo Li <sunpeng.li@amd.com>
Signed-off-by: Taimur Hassan <Syed.Hassan@amd.com>
Signed-off-by: Alex Hung <alex.hung@amd.com>
Tested-by: Dan Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: Fix static assertion failure issue

Since the PAGE_SIZE is 8KB on sparc64, the size of
structure amdsriov_ras_telemetry will exceed 64KB,
so use absolute value to fix the buffer size.

Fixes the issue:
 drivers/gpu/drm/amd/amdgpu/amdgv_sriovmsg.h:522:2: error: static
 assertion failed due to requirement 'sizeof(struct
 amdsriov_ras_telemetry) <= 64 << 10': amdsriov_ras_telemetry must be 64 KB
 |  sizeof(struct amdsriov_ras_telemetry) <=
AMD_SRIOV_MSG_RAS_TELEMETRY_SIZE_KB_V1 << 10,
 drivers/gpu/drm/amd/amdgpu/amdgv_sriovmsg.h:522:40: note:
expression evaluates to '115616 <= 65536'
 |   sizeof(struct amdsriov_ras_telemetry) <=
AMD_SRIOV_MSG_RAS_TELEMETRY_SIZE_KB_V1 << 10,

Fixes: cb48a6b2b61d ("drm/amd/ras: use dedicated memory as vf ras command buffer")
Reported-by: kernel test robot <lkp@intel.com>
Closes: https://lore.kernel.org/oe-kbuild-all/202602261700.rVOLIw4l-lkp@intel.com/
Signed-off-by: YiPeng Chai <YiPeng.Chai@amd.com>
Reviewed-by: Tao Zhou <tao.zhou1@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Prevent integer overflow when mhz to khz

[WHAT]
Cast to long long before multiplication to prevent overflow
when converting mhz to khz by multiplying by 1000.

This is reported as INTEGER_OVERFLOW errors by Coverity.

Reviewed-by: Roman Li <roman.li@amd.com>
Signed-off-by: Alex Hung <alex.hung@amd.com>
Tested-by: Dan Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Remove always-false branches

[WHAT]
program_prealpha_dealpha and hpo_frl_stream_enc_acquired are always
false and all branches depending on them will never be taken.

This is reported as DEADCODE errors by Coverity.

Reviewed-by: Roman Li <roman.li@amd.com>
Signed-off-by: Alex Hung <alex.hung@amd.com>
Tested-by: Dan Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Remove redundant initializers

[WHAT]
Remove unnecessary default value assignments for variables that
are unconditionally assigned before use.

Linux kernel code style prefers no assignments during initialization
when variables are assigned unconditionally as they can obscures
the actual data flow. In addition, compilers will be able to catch them
if variables are used without being updated later in all conditions.

This is reported as UNUSED_VALUE errors by Coverity.

Reviewed-by: Roman Li <roman.li@amd.com>
Signed-off-by: Alex Hung <alex.hung@amd.com>
Tested-by: Dan Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Silence unused variable warning

[WHY & HOW]
Remove unused dpp_pipe_count variable.

Reviewed-by: Austin Zheng <austin.zheng@amd.com>
Signed-off-by: Clay King <clayking@amd.com>
Signed-off-by: Alex Hung <alex.hung@amd.com>
Tested-by: Dan Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Initialize replay_state to PR_STATE_INVALID

[WHY & HOW]
Initialize the replay_state variable to PR_STATE_INVALID instead of
PR_STATE_0 before retrieving the actual replay state.

Reviewed-by: Wenjing Liu <wenjing.liu@amd.com>
Signed-off-by: Ivan Lipski <ivan.lipski@amd.com>
Signed-off-by: Alex Hung <alex.hung@amd.com>
Tested-by: Dan Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Fallback to boot snapshot for dispclk

[WHY & HOW]
If the dentist is unavailable, fallback to reading CLKIP via the boot
snapshot to get the current dispclk.

Reviewed-by: Nicholas Kazlauskas <nicholas.kazlauskas@amd.com>
Signed-off-by: Dillon Varone <Dillon.Varone@amd.com>
Signed-off-by: Alex Hung <alex.hung@amd.com>
Cc: Mario Limonciello <mario.limonciello@amd.com>
Cc: Alex Deucher <alexander.deucher@amd.com>
Tested-by: Dan Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/pm: fix kernel-doc warning for smu_msg_v1_send_msg()

Warning: drivers/gpu/drm/amd/amdgpu/../pm/swsmu/smu_cmn.c:415 expecting prototype for smu_msg_proto_v1_send_msg(). Prototype was for smu_msg_v1_send_msg() instead

Fixes: 4f379370a49c ("drm/amd/pm: Add smu message control block")
Signed-off-by: Yujie Liu <yujie.liu@intel.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/ras: fix kernel-doc warning for ras_eeprom_append()

Warning: drivers/gpu/drm/amd/amdgpu/../ras/rascore/ras_eeprom.c:845 function parameter 'ras_core' not described in 'ras_eeprom_append'
Warning: drivers/gpu/drm/amd/amdgpu/../ras/rascore/ras_eeprom.c:845 expecting prototype for ras_core_eeprom_append(). Prototype was for ras_eeprom_append() instead

Fixes: 5c3be5defc92 ("drm/amd/ras: Add eeprom ras functions")
Signed-off-by: Yujie Liu <yujie.liu@intel.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: fix kernel-doc warning for amdgpu_ttm_alloc_mmio_remap_bo()

Warning: drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c:1923 expecting prototype for amdgpu_ttm_mmio_remap_bo_init(). Prototype was for amdgpu_ttm_alloc_mmio_remap_bo() instead

Fixes: 96e97a562d06 ("drm/amdgpu: Drop MMIO_REMAP domain bit and keep it Internal")
Signed-off-by: Yujie Liu <yujie.liu@intel.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/ras: Fix type size of remainder argument

Forcing an int to be dereferenced at uint64_t for div64_u64_rem() runs
the risk of endian confusion and stack overflowing writes. Seen while
preparing to enable -Warray-bounds globally:

In file included from ../arch/x86/include/asm/processor.h:35,
                 from ../include/linux/sched.h:13,
                 from ../include/linux/ratelimit.h:6,
                 from ../include/linux/dev_printk.h:16,
                 from ../drivers/gpu/drm/amd/amdgpu/../ras/ras_mgr/ras_sys.h:29,
                 from ../drivers/gpu/drm/amd/amdgpu/../ras/rascore/ras.h:27,
                 from ../drivers/gpu/drm/amd/amdgpu/../ras/rascore/ras_core.c:24:
In function 'div64_u64_rem',
    inlined from 'ras_core_convert_timestamp_to_time' at ../drivers/gpu/drm/amd/amdgpu/../ras/rascore/ras_core.c:72:9:
../include/linux/math64.h:56:20: error: array subscript 'u64 {aka long long unsigned int}[0]' is partly outside array bounds of 'int[1]' [-Werror=array-bounds=]
   56 |         *remainder = dividend % divisor;
      |         ~~~~~~~~~~~^~~~~~~~~~~~~~~~~~~~
../drivers/gpu/drm/amd/amdgpu/../ras/rascore/ras_core.c: In function 'ras_core_convert_timestamp_to_time':
../drivers/gpu/drm/amd/amdgpu/../ras/rascore/ras_core.c:70:19: note: object 'remaining_seconds' of size 4
   70 |         int days, remaining_seconds;
      |                   ^~~~~~~~~~~~~~~~~

Use a 64-bit type for the remainder calculation, but leave
remaining_seconds as 32-bit to avoid 64-bit division later. The value of
remainder will always be less than seconds_per_day, so there's no
truncation risk.

Fixes: ace232eff50e ("drm/amdgpu: Add ras module files into amdgpu")
Signed-off-by: Kees Cook <kees@kernel.org>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: Move register access functions

Move register access methods from amdgpu_device.c to a dedicated file.

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

drm/amdgpu: Enable DPG support for VCN5

This will set DPG flags for enabling power gating on GFX11_5_4

Signed-off-by: sguttula <suresh.guttula@amd.com>
Reviewed-by: Pratik Vishwakarma <Pratik.Vishwakarma@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Enable DEGAMMA and reject COLOR_PIPELINE+DEGAMMA_LUT

[WHAT]
Create DEGAMMA properties even if color pipeline is enabled, and enforce
the mutual exclusion in atomic check by rejecting any commit that
attempts to enable both COLOR_PIPELINE on the plane and DEGAMMA_LUT on
the CRTC simultaneously.

Fixes: 18a4127e9315 ("drm/amd/display: Disable CRTC degamma when color pipeline is enabled")
Closes: https://gitlab.freedesktop.org/drm/amd/-/issues/4963
Reviewed-by: Melissa Wen <mwen@igalia.com>
Reviewed-by: Harry Wentland <harry.wentland@amd.com>
Signed-off-by: Alex Hung <alex.hung@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Use mpc.preblend flag to indicate 3D LUT

[WHAT]
New ASIC's 3D LUT is indicated by mpc.preblend.

Fixes: 0de2b1afea8d ("drm/amd/display: add 3D LUT colorop")
Reviewed-by: Melissa Wen <mwen@igalia.com>
Reviewed-by: Harry Wentland <harry.wentland@amd.com>
Signed-off-by: Alex Hung <alex.hung@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdkfd: fix CWSR trap handler

Fix up what looks like a bad merge.

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

drm/amd: Disable MES LR compute W/A

A workaround was introduced in commit 1fb710793ce2 ("drm/amdgpu: Enable
MES lr_compute_wa by default") to help with some hangs observed in gfx1151.

This WA didn't fully fix the issue.  It was actually fixed by adjusting
the VGPR size to the correct value that matched the hardware in commit
b42f3bf9536c ("drm/amdkfd: bump minimum vgpr size for gfx1151").

There are reports of instability on other products with newer GC microcode
versions, and I believe they're caused by this workaround. As we don't
need the workaround any more, remove it.

Fixes: b42f3bf9536c ("drm/amdkfd: bump minimum vgpr size for gfx1151")
Acked-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Mario Limonciello <mario.limonciello@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: Fix error handling in slot reset

If the device has not recovered after slot reset is called, it goes to
out label for error handling. There it could make decision based on
uninitialized hive pointer and could result in accessing an uninitialized
list.

Initialize the list and hive properly so that it handles the error
situation and also releases the reset domain lock which is acquired
during error_detected callback.

Fixes: 732c6cefc1ec ("drm/amdgpu: Replace tmp_adev with hive in amdgpu_pci_slot_reset")
Signed-off-by: Lijo Lazar <lijo.lazar@amd.com>
Reviewed-by: Ce Sun <cesun102@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/ras: Handle check address validity in SR-IOV

Handle check address validity command in SR-IOV
guest.

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/vcn5: Add SMU dpm interface type

This will set AMDGPU_VCN_SMU_DPM_INTERFACE_* smu_type
based on soc type and fixing ring timeout issue seen
for DPM enabled case.

Signed-off-by: sguttula <suresh.guttula@amd.com>
Reviewed-by: Pratik Vishwakarma <Pratik.Vishwakarma@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/ras: Add function to convert retired address

Add function to convert retired address in SR-IOV
guest.

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/amd/ras: Handle address check in SR-IOV guest

Handle address check validity command in SR-IOV guest

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/amd/ras: Add convert retired address structure

Add convert retired address command and structure
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/amd/ras: Add address check structure

Add address check command and data structure
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/amdgpu: Fix locking bugs in error paths

Do not unlock psp->ras_context.mutex if it has not been locked. This has
been detected by the Clang thread-safety analyzer.

Cc: Alex Deucher <alexander.deucher@amd.com>
Cc: Christian König <christian.koenig@amd.com>
Cc: YiPeng Chai <YiPeng.Chai@amd.com>
Cc: Hawking Zhang <Hawking.Zhang@amd.com>
Cc: amd-gfx@lists.freedesktop.org
Fixes: b3fb79cda568 ("drm/amdgpu: add mutex to protect ras shared memory")
Acked-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Bart Van Assche <bvanassche@acm.org>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: Unlock a mutex before destroying it

Mutexes must be unlocked before these are destroyed. This has been detected
by the Clang thread-safety analyzer.

Cc: Alex Deucher <alexander.deucher@amd.com>
Cc: Christian König <christian.koenig@amd.com>
Cc: Yang Wang <kevinyang.wang@amd.com>
Cc: Hawking Zhang <Hawking.Zhang@amd.com>
Cc: amd-gfx@lists.freedesktop.org
Fixes: f5e4cc8461c4 ("drm/amdgpu: implement RAS ACA driver framework")
Reviewed-by: Yang Wang <kevinyang.wang@amd.com>
Acked-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Bart Van Assche <bvanassche@acm.org>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Use GFP_ATOMIC in dc_create_stream_for_sink

This can be called while preemption is disabled, for example by
dcn32_internal_validate_bw which is called with the FPU active.

Fixes "BUG: scheduling while atomic" messages I encounter on my Navi31
machine.

Signed-off-by: Natalie Vock <natalie.vock@gmx.de>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: add upper bound check on user inputs in wait ioctl

Huge input values in amdgpu_userq_wait_ioctl can lead to a OOM and
could be exploited.

So check these input value against AMDGPU_USERQ_MAX_HANDLES
which is big enough value for genuine use cases and could
potentially avoid OOM.

v2: squash in Srini's fix

Signed-off-by: Sunil Khatri <sunil.khatri@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: add upper bound check on user inputs in signal ioctl

Huge input values in amdgpu_userq_signal_ioctl can lead to a OOM and
could be exploited.

So check these input value against AMDGPU_USERQ_MAX_HANDLES
which is big enough value for genuine use cases and could
potentially avoid OOM.

Signed-off-by: Sunil Khatri <sunil.khatri@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu/userq: Use drm_gem_objects_lookup in amdgpu_userq_wait_ioctl

Use the existing helper instead of open coding it

Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@igalia.com>
Reviewed-by: Sunil Khatri <sunil.khatrti@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu/userq: Use drm_gem_objects_lookup in amdgpu_userq_signal_ioctl

Use the existing helper instead of open coding it.

Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@igalia.com>
Reviewed-by: Sunil Khatri <sunil.khatrti@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/ras: use dedicated memory as vf ras command buffer

Use dedicated memory as vf ras command buffer.

V2:
  Add lock to ensure serialization of sending vf ras commands.

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