drm/amd/display: DML2.1 Post-Si Cleanup

[Why]
There are a few cleanup and refactoring tasks that need to be done
with the DML2.1 wrapper and DC interface to remove dependencies on
legacy structures and N-1 prototypes.

[How]
Implemented pipe_ctx->global_sync.
Implemented new functions to use pipe_ctx->hubp_regs and
pipe_ctx->global_sync:
- hubp_setup2
- hubp_setup_interdependent2
- Several other new functions for DCN 4.01 to support newer structures

Removed dml21_update_pipe_ctx_dchub_regs
Removed dml21_extract_legacy_watermark_set
Removed dml21_populate_pipe_ctx_dlg_param
Removed outdated dcn references in DML2.1 wrapper.

Reviewed-by: Austin Zheng <austin.zheng@amd.com>
Reviewed-by: Dillon Varone <dillon.varone@amd.com>
Signed-off-by: Rafal Ostrowski <rostrows@amd.com>
Signed-off-by: Tom Chung <chiahsuan.chung@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Update chip_cap defines and usage

[WHY]
The defines have also been updated with prefix AMD_ and atomfirmware.h
has been temporarily updated with both sets of defines to allow the
transition.
This update is being made to standardize workaround chip_cap flags,
in order to support more workaround flags in the future.

[HOW]
Updated EXT_DISPLAY_PATH_CAPS__DP_FIXED_VS_EN define, the flag is now
an enum masked by EXT_DISPLAY_PATH_CAPS__EXT_CHIP_MASK. All checks for
DP_FIXED_VS_EN are now performed by masking with EXT_CHIP_MASK and
checking for an exact match rather than the previous bitwise AND check.

Reviewed-by: Wenjing Liu <wenjing.liu@amd.com>
Signed-off-by: Michael Strauss <michael.strauss@amd.com>
Signed-off-by: Tom Chung <chiahsuan.chung@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 0.0.248.0

Refactoring some flags for replay

Acked-by: Wayne Lin <wayne.lin@amd.com>
Signed-off-by: Taimur Hassan <Syed.Hassan@amd.com>
Signed-off-by: Tom Chung <chiahsuan.chung@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Add replay desync error count tracking and reset functionality

[Why & How]
Build-up get/reset desync error count interface and implement the functions.

Reviewed-by: ChunTao Tso <chuntao.tso@amd.com>
Reviewed-by: Robin Chen <robin.chen@amd.com>
Signed-off-by: Jack Chang <jack.chang@amd.com>
Signed-off-by: Tom Chung <chiahsuan.chung@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Log Hard Min Clocks and Phantom Pipe Status

[WHY]
On entering/exiting idle power, certain parameters would be
very useful to know for power profiling purposes.

[HOW]
This commit adds certain hard min clocks and pipe types
to log output on idle optimization enter/exit.

Reviewed-by: Alvin Lee <alvin.lee2@amd.com>
Signed-off-by: Sung Lee <Sung.Lee@amd.com>
Signed-off-by: Tom Chung <chiahsuan.chung@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Limit Scaling Ratio on DCN3.01

[why]
Underflow and flickering was occuring due to high scaling ratios
when resizing videos.

[how]
Limit the scaling ratios by increasing the max scaling factor

Reviewed-by: Nicholas Kazlauskas <nicholas.kazlauskas@amd.com>
Signed-off-by: Gabe Teeger <Gabe.Teeger@amd.com>
Signed-off-by: Tom Chung <chiahsuan.chung@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu/smu13: update powersave optimizations

Only apply when compute profile is selected.  This is
the only supported configuration.  Selecting other
profiles can lead to performane degradations.

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

drm/amd/display: add CEC notifier to amdgpu driver

This patch adds the cec_notifier feature to amdgpu driver.
The changes will allow amdgpu driver code to notify EDID
and HPD changes to an eventual CEC adapter.

Signed-off-by: Kun Liu <Kun.Liu2@amd.com>
Reviewed-by: Mario Limonciello <mario.limonciello@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: fill the ucode bo during psp resume for SRIOV

refill the ucode bo during psp resume for SRIOV, otherwise ucode load
will fail after VM hibernation and fb clean.

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

drm/amdgpu/gfx10: Enable cleaner shader for GFX10.3.2/10.3.4/10.3.5 GPUs

Enable the cleaner shader for GFX10.3.2/10.3.4/10.3.5 GPUs to provide
data isolation between GPU workloads. The cleaner shader is responsible
for clearing the Local Data Store (LDS), Vector General Purpose
Registers (VGPRs), and Scalar General Purpose Registers (SGPRs), which
helps prevent data leakage and ensures accurate computation results.

This update extends cleaner shader support to GFX10.3.2/10.3.4/10.3.5
GPUs, previously available for GFX10.3.0. It enhances security by
clearing GPU memory between processes and maintains a consistent GPU
state across KGD and KFD workloads.

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

drm/amdgpu: fix gpu recovery disable with per queue reset

Per queue reset should be bypassed when gpu recovery is disabled
with module parameter.

Fixes: ee0a469cf917 ("drm/amdkfd: support per-queue reset on gfx9")
Signed-off-by: Jonathan Kim <jonathan.kim@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: wrong array index to get ip block for PSP

The adev->ip_blocks array is not indexed by AMD_IP_BLOCK_TYPE_xxx,
instead we should call amdgpu_device_ip_get_ip_block() to get the
corresponding IP block oject.

Fix some checkpatch issues (Alex)

Signed-off-by: Jiang Liu <gerry@linux.alibaba.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: tear down ttm range manager for doorbell in amdgpu_ttm_fini()

Tear down ttm range manager for doorbell in function amdgpu_ttm_fini(),
to avoid memory leakage.

Fixes: 792b84fb9038 ("drm/amdgpu: initialize ttm for doorbells")
Signed-off-by: Jiang Liu <gerry@linux.alibaba.com>
Signed-off-by: Kent Russell <kent.russell@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: fix incorrect number of active RBs for gfx12

The RB bitmap should be global active RB bitmap &
active RB bitmap based on active SA.

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

drm/amdgpu: fix incorrect active RB bitmap in setup RBs

The RB bitmap width per SA may be 0x1 for some ASICs.
Use the actual bitmap of SA instead of 0x3 to determine
the active RB bitmap.

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

drm/amdkfd: Move gfx12 trap handler to separate file

gfx12 derivatives will have substantially different trap handler
implementations from gfx10/gfx11. Add a separate source file for
gfx12+ and remove unneeded conditional code.

No functional change.

v2: Revert copyright date to 2018, minor comment fixes

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

drm/amdgpu: Fix shift type in amdgpu_debugfs_sdma_sched_mask_set()

The "mask" and "val" variables are type u64.  The problem is that the
BIT() macros are type unsigned long which is just 32 bits on 32bit
systems.

It's unlikely that people will be using this driver on 32bit kernels
and even if they did we only use the lower AMDGPU_MAX_SDMA_INSTANCES (16)
bits.  So this bug does not affect anything in real life.

Still, for correctness sake, u64 bit masks should use BIT_ULL().

Fixes: d2e3961ae371 ("drm/amdgpu: add amdgpu_sdma_sched_mask debugfs")
Signed-off-by: Dan Carpenter <dan.carpenter@linaro.org>
Reviewed-by: Mario Limonciello <mario.limonciello@amd.com>
Link: https://lore.kernel.org/r/d39a9325-87a4-4543-b6ec-1c61fca3a6fc@stanley.mountain
Signed-off-by: Mario Limonciello <mario.limonciello@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: enable gfx12 queue reset flag

Enable the kgq and kcq queue reset flag

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

drm/amd/display: Increase sanitizer frame larger than limit when compile testing with clang

Commit 24909d9ec7c3 ("drm/amd/display: Overwriting dualDPP UBF values
before usage") added a new warning in dml2/display_mode_core.c when
building allmodconfig with clang:

  drivers/gpu/drm/amd/amdgpu/../display/dc/dml2/display_mode_core.c:6268:13: error: stack frame size (3128) exceeds limit (3072) in 'dml_prefetch_check' [-Werror,-Wframe-larger-than]
   6268 | static void dml_prefetch_check(struct display_mode_lib_st *mode_lib)
        |             ^

Commit be4e3509314a ("drm/amd/display: DML21 Reintegration For Various
Fixes") introduced one in dml2_core/dml2_core_dcn4_calcs.c with the same
configuration:

  drivers/gpu/drm/amd/amdgpu/../display/dc/dml2/dml21/src/dml2_core/dml2_core_dcn4_calcs.c:7236:13: error: stack frame size (3256) exceeds limit (3072) in 'dml_core_mode_support' [-Werror,-Wframe-larger-than]
   7236 | static bool dml_core_mode_support(struct dml2_core_calcs_mode_support_ex *in_out_params)
        |             ^

In the case of the first warning, the stack usage was already at the
limit at the parent change, so the offending change was rather
innocuous. In the case of the second warning, there was a rather
dramatic increase in stack usage compared to the parent:

  drivers/gpu/drm/amd/amdgpu/../display/dc/dml2/dml21/src/dml2_core/dml2_core_dcn4_calcs.c:7032:13: error: stack frame size (2696) exceeds limit (2048) in 'dml_core_mode_support' [-Werror,-Wframe-larger-than]
   7032 | static bool dml_core_mode_support(struct dml2_core_calcs_mode_support_ex *in_out_params)
        |             ^

This is an unfortunate interaction between an issue with stack slot
reuse in LLVM that gets exacerbated by sanitization (which gets enabled
with all{mod,yes}config) and function calls using a much higher number
of parameters than is typical in the kernel, necessitating passing most
of these values on the stack.

While it is possible that there should be source code changes to address
these warnings, this code is difficult to modify for various reasons, as
has been noted in other changes that have occurred for similar reasons,
such as commit 6740ec97bcdb ("drm/amd/display: Increase frame warning
limit with KASAN or KCSAN in dml2").

Increase the frame larger than limit when compile testing with clang and
the sanitizers enabled to avoid this breakage in all{mod,yes}config, as
they are commonly used and valuable testing targets. While it is not the
best to hide this issue, it is not really relevant when compile testing,
as the sanitizers are commonly stressful on optimizations and they are
only truly useful at runtime, which COMPILE_TEST states will not occur
with the current build.

Reported-by: kernel test robot <lkp@intel.com>
Closes: https://lore.kernel.org/oe-kbuild-all/202412121748.chuX4sap-lkp@intel.com/
Signed-off-by: Nathan Chancellor <nathan@kernel.org>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu/pm: Implement SDMA queue reset for different asic

Implement sdma queue reset by SMU_MSG_ResetSDMA2

Signed-off-by: Jesse Zhang <jesse.zhang@amd.com>
Suggested-by: Tim Huang <tim.huang@amd.com>
Reviewed-by: Tim Huang <tim.huang@amd.com>
Acked-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu/pm: add definition PPSMC_MSG_ResetSDMA2

add the PPSMC_MSG_ResetSDMA2 definition for smu 13.0.6

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

drm/amdgpu/sdma4.4.2: add apu support in sdma queue reset

Remove apu check in sdma queue reset.

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

drm/amdgpu: Remove unnecessary NULL check

container_of cannot return NULL, so it is unnecessary to check for
NULL after gem_to_amdgpu_bo, which is just a container_of call

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

drm/amd/pm: Fill max mem bw & total app clk counter

Fill max memory bandwidth and total app clock counter to metrics v1_7

v2: Remove unnecessary check

v3: Add app clock counter support for apu

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: Update SMUv13.0.6 PMFW headers

Update pmfw headers for smuv13.0.6 to pmfw version 85.121

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: Add a lock when accessing the buddy trim function

When running YouTube videos and Steam games simultaneously,
the tester found a system hang / race condition issue with
the multi-display configuration setting. Adding a lock to
the buddy allocator's trim function would be the solution.

<log snip>
[ 7197.250436] general protection fault, probably for non-canonical address 0xdead000000000108
[ 7197.250447] RIP: 0010:__alloc_range+0x8b/0x340 [amddrm_buddy]
[ 7197.250470] Call Trace:
[ 7197.250472]  <TASK>
[ 7197.250475]  ? show_regs+0x6d/0x80
[ 7197.250481]  ? die_addr+0x37/0xa0
[ 7197.250483]  ? exc_general_protection+0x1db/0x480
[ 7197.250488]  ? drm_suballoc_new+0x13c/0x93d [drm_suballoc_helper]
[ 7197.250493]  ? asm_exc_general_protection+0x27/0x30
[ 7197.250498]  ? __alloc_range+0x8b/0x340 [amddrm_buddy]
[ 7197.250501]  ? __alloc_range+0x109/0x340 [amddrm_buddy]
[ 7197.250506]  amddrm_buddy_block_trim+0x1b5/0x260 [amddrm_buddy]
[ 7197.250511]  amdgpu_vram_mgr_new+0x4f5/0x590 [amdgpu]
[ 7197.250682]  amdttm_resource_alloc+0x46/0xb0 [amdttm]
[ 7197.250689]  ttm_bo_alloc_resource+0xe4/0x370 [amdttm]
[ 7197.250696]  amdttm_bo_validate+0x9d/0x180 [amdttm]
[ 7197.250701]  amdgpu_bo_pin+0x15a/0x2f0 [amdgpu]
[ 7197.250831]  amdgpu_dm_plane_helper_prepare_fb+0xb2/0x360 [amdgpu]
[ 7197.251025]  ? try_wait_for_completion+0x59/0x70
[ 7197.251030]  drm_atomic_helper_prepare_planes.part.0+0x2f/0x1e0
[ 7197.251035]  drm_atomic_helper_prepare_planes+0x5d/0x70
[ 7197.251037]  drm_atomic_helper_commit+0x84/0x160
[ 7197.251040]  drm_atomic_nonblocking_commit+0x59/0x70
[ 7197.251043]  drm_mode_atomic_ioctl+0x720/0x850
[ 7197.251047]  ? __pfx_drm_mode_atomic_ioctl+0x10/0x10
[ 7197.251049]  drm_ioctl_kernel+0xb9/0x120
[ 7197.251053]  ? srso_alias_return_thunk+0x5/0xfbef5
[ 7197.251056]  drm_ioctl+0x2d4/0x550
[ 7197.251058]  ? __pfx_drm_mode_atomic_ioctl+0x10/0x10
[ 7197.251063]  amdgpu_drm_ioctl+0x4e/0x90 [amdgpu]
[ 7197.251186]  __x64_sys_ioctl+0xa0/0xf0
[ 7197.251190]  x64_sys_call+0x143b/0x25c0
[ 7197.251193]  do_syscall_64+0x7f/0x180
[ 7197.251197]  ? srso_alias_return_thunk+0x5/0xfbef5
[ 7197.251199]  ? amdgpu_display_user_framebuffer_create+0x215/0x320 [amdgpu]
[ 7197.251329]  ? drm_internal_framebuffer_create+0xb7/0x1a0
[ 7197.251332]  ? srso_alias_return_thunk+0x5/0xfbef5

Signed-off-by: Arunpravin Paneer Selvam <Arunpravin.PaneerSelvam@amd.com>
Fixes: 4a5ad08f5377 ("drm/amdgpu: Add address alignment support to DCC buffers")
Acked-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdkfd: always include uapi header in priv.h

The header usr/linux/kfd_ioctl.h is a duplicate of uapi/linux/kfd_ioctl.h.
And it is actually not a file in the source code tree.
Ideally, the usr version should be updated whenever the source code is recompiled.

However, I have noticed a discrepancy between the two headers
even after rebuilding the kernel.

This commit modifies kfd_priv.h to always include the header from uapi to ensure
the latest changes are reflected. We should always include the source
code header other than the duplication.

Signed-off-by: Zhu Lingshan <lingshan.zhu@amd.com>
Reviewed-by: Felix Kuehling <felix.kuehling@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: reduce RLC safe mode request for gfx clock gating

The driver can only request one time for the power safe mode instead of
polling and disabling the power feature each time prior to program the
GFX clock gating control registers. This update will reduce the latency
on the GFX clock gating entry.

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

Revert "drm/amd/display: Optimize cursor position updates"

This reverts commit 88c7c56d07c108ed4de319c8dba44aa4b8a38dd1.

SW and HW state are not always matching in some cases causing cursor to
be disabled.

Signed-off-by: Aurabindo Pillai <aurabindo.pillai@amd.com>
Reviewed-by: Leo Li <sunpeng.li@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu/gfx10: Add cleaner shader for GFX10.3.0

This commit adds the cleaner shader microcode for GFX10.3.0 GPUs. The
cleaner shader is a piece of GPU code that is used to clear or
initialize certain GPU resources, such as Local Data Share (LDS), Vector
General Purpose Registers (VGPRs), and Scalar General Purpose Registers
(SGPRs).

Clearing these resources is important for ensuring data isolation
between different workloads running on the GPU. Without the cleaner
shader, residual data from a previous workload could potentially be
accessed by a subsequent workload, leading to data leaks and incorrect
computation results.

The cleaner shader microcode is represented as an array of 32-bit words
(`gfx_10_3_0_cleaner_shader_hex`). This array is the binary
representation of the cleaner shader code, which is written in a
low-level GPU instruction set.

When the cleaner shader feature is enabled, the AMDGPU driver loads this
array into a specific location in the GPU memory. The GPU then reads
this memory location to fetch and execute the cleaner shader
instructions.

The cleaner shader is executed automatically by the GPU at the end of
each workload, before the next workload starts. This ensures that all
GPU resources are in a clean state before the start of each workload.

This addition is part of the cleaner shader feature implementation. The
cleaner shader feature helps resource utilization by cleaning up GPU
resources after they are used. It also enhances security and reliability
by preventing data leaks between workloads.

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

drm/amd/pm:  fix BUG: scheduling while atomic

atomic scheduling will be triggered in interrupt handler for
AC/DC mode switch as following backtrace.
Call Trace:
 <IRQ>
 dump_stack_lvl
 __schedule_bug
 __schedule
 schedule
 schedule_preempt_disabled
 __mutex_lock
 smu_cmn_send_smc_msg_with_param
 smu_v13_0_irq_process
 amdgpu_irq_dispatch
 amdgpu_ih_process
 amdgpu_irq_handler
 __handle_irq_event_percpu
 handle_irq_event
 handle_edge_irq
 __common_interrupt
 common_interrupt
 </IRQ>
 <TASK>
 asm_common_interrupt

Reviewed-by: Lijo Lazar <lijo.lazar@amd.com>
Reviewed-by: Kenneth Feng <kenneth.feng@amd.com>
Signed-off-by: Kun Liu <Kun.Liu2@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: Fix error handling in amdgpu_ras_add_bad_pages

It ensures that appropriate error codes are returned when an error
condition is detected

Fixes the below;
drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c:2849 amdgpu_ras_add_bad_pages() warn: missing error code here? 'amdgpu_umc_pages_in_a_row()' failed.
drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c:2884 amdgpu_ras_add_bad_pages() warn: missing error code here? 'amdgpu_ras_mca2pa()' failed.

v2: s/-EIO/-EINVAL, retained the use of -EINVAL from
    amdgpu_umc_pages_in_a_row & and amdgpu_ras_mca2pa_by_idx, when the
    RAS context is not initialized or the convert_ras_err_addr function is
    unavailable. (Thomas)

V3: Returning 0 as the absence of eh_data is acceptable. (Tao)

Fixes: a8d133e625ce ("drm/amdgpu: parse legacy RAS bad page mixed with new data in various NPS modes")
Reported-by: Dan Carpenter <dan.carpenter@linaro.org>
Cc: YiPeng Chai <yipeng.chai@amd.com>
Cc: Tao Zhou <tao.zhou1@amd.com>
Cc: Hawking Zhang <Hawking.Zhang@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: Tao Zhou <tao.zhou1@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdkfd: wq_release signals dma_fence only when available

kfd_process_wq_release() signals eviction fence by
dma_fence_signal() which wanrs if dma_fence
is NULL.

kfd_process->ef is initialized by kfd_process_device_init_vm()
through ioctl. That means the fence is NULL for a new
created kfd_process, and close a kfd_process right
after open it will trigger the warning.

This commit conditionally signals the eviction fence
in kfd_process_wq_release() only when it is available.

[  503.660882] WARNING: CPU: 0 PID: 9 at drivers/dma-buf/dma-fence.c:467 dma_fence_signal+0x74/0xa0
[  503.782940] Workqueue: kfd_process_wq kfd_process_wq_release [amdgpu]
[  503.789640] RIP: 0010:dma_fence_signal+0x74/0xa0
[  503.877620] Call Trace:
[  503.880066]  <TASK>
[  503.882168]  ? __warn+0xcd/0x260
[  503.885407]  ? dma_fence_signal+0x74/0xa0
[  503.889416]  ? report_bug+0x288/0x2d0
[  503.893089]  ? handle_bug+0x53/0xa0
[  503.896587]  ? exc_invalid_op+0x14/0x50
[  503.900424]  ? asm_exc_invalid_op+0x16/0x20
[  503.904616]  ? dma_fence_signal+0x74/0xa0
[  503.908626]  kfd_process_wq_release+0x6b/0x370 [amdgpu]
[  503.914081]  process_one_work+0x654/0x10a0
[  503.918186]  worker_thread+0x6c3/0xe70
[  503.921943]  ? srso_alias_return_thunk+0x5/0xfbef5
[  503.926735]  ? srso_alias_return_thunk+0x5/0xfbef5
[  503.931527]  ? __kthread_parkme+0x82/0x140
[  503.935631]  ? __pfx_worker_thread+0x10/0x10
[  503.939904]  kthread+0x2a8/0x380
[  503.943132]  ? __pfx_kthread+0x10/0x10
[  503.946882]  ret_from_fork+0x2d/0x70
[  503.950458]  ? __pfx_kthread+0x10/0x10
[  503.954210]  ret_from_fork_asm+0x1a/0x30
[  503.958142]  </TASK>
[  503.960328] ---[ end trace 0000000000000000 ]---

Fixes: 967d226eaae8 ("dma-buf: add WARN_ON() illegal dma-fence signaling")
Signed-off-by: Zhu Lingshan <lingshan.zhu@amd.com>
Reviewed-by: Felix Kuehling <felix.kuehling@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: Fix for MEC SJT FW Load Fail on VF

Users might switch to ROCM build does not include MEC SJT FW and driver
needs to consider this case.w

Signed-off-by: yfeng1 <yfeng1@amd.com>
Reviewed-by: Hawking Zhang <Hawking.Zhang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Promote DC to 3.2.315

This version brings along the following:
- Add Interface to Dump DSC Caps from dm
- Add DP required HBlank size calc to link interface
- Add 6bpc RGB case for dcn32 output bpp calculations
- Add VC for VESA Aux Backlight Control
- Add support for setting multiple CRC windows in dc
- Clean up SPL code and outdated interfaces in dcn401_clk_mgr
- Disable replay and psr while VRR is enabled
- Fix PSR-SU not support but still call the  amdgpu_dm_psr_enable
- Implement Replay Low Hz Visual Confirm
- Extend dc_stream_get_crc to support 2nd crc engine
- Update power gating logic for DCN35 hw

Reviewed-by: Roman Li <roman.li@amd.com>
Signed-off-by: Martin Leung <Martin.Leung@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: Extend capability to get multiple ROI CRCs

[Why & How]
We already extend our dm, dc and dmub to support setting of multiple CRC
instances, now extend the capability to return back the ROI/CRC pair result
from psp by specifying activated ROI instances.

Reviewed-by: HaoPing Liu <haoping.liu@amd.com>
Signed-off-by: Wayne Lin <Wayne.Lin@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 VC for VESA Aux Backlight Control

[WHY]
There is no way to distinguish
the static backlight control type
being used and the VABC support
without the use of a debugger or
reading DPCD registers.

[HOW]
Add Visual Confirm support
for VESA Aux-based Backlight Control.

Reviewed-by: Harry Vanzylldejong <harry.vanzylldejong@amd.com>
Signed-off-by: Iswara Nagulendran <Iswara.Nagulendran@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: Clean up SPL code

[Why & How]
Use helper functions for checking formats
Apply cositing offset in rotation case

Reviewed-by: Navid Assadian <navid.assadian@amd.com>
Signed-off-by: Samson Tam <Samson.Tam@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 check for granularity in dml ceil/floor helpers

[Why]
Wrapper functions for dcn_bw_ceil2() and dcn_bw_floor2()
should check for granularity is non zero to avoid assert and
divide-by-zero error in dcn_bw_ functions.

[How]
Add check for granularity 0.

Cc: Mario Limonciello <mario.limonciello@amd.com>
Reviewed-by: Alvin Lee <alvin.lee2@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 6bpc RGB case for dcn32 output bpp calculations

[Why]
Current DCN32 calculation doesn't consider RGB 6bpc for the DP case.
This results in an invalid output bpp being calculated when DSC is not
enabled in the configuration, failing the mode validation.

[How]
Add special case to handle 6bpc RGB in the output bpp calculation.

Reviewed-by: Alvin Lee <alvin.lee2@amd.com>
Signed-off-by: George Shen <george.shen@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: Disable replay and psr while VRR is enabled

[Why]
Replay and PSR will cause some video corruption while VRR is enabled.

[How]
1. Disable the Replay and PSR while VRR is enabled.
2. Change the amdgpu_dm_crtc_vrr_active() parameter to const.
   Because the function will only read data from dm_crtc_state.

Reviewed-by: Sun peng Li <sunpeng.li@amd.com>
Signed-off-by: Tom Chung <chiahsuan.chung@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: Fix PSR-SU not support but still call the amdgpu_dm_psr_enable

[Why]
The enum DC_PSR_VERSION_SU_1 of psr_version is 1 and
DC_PSR_VERSION_UNSUPPORTED is 0xFFFFFFFF.

The original code may has chance trigger the amdgpu_dm_psr_enable()
while psr version is set to DC_PSR_VERSION_UNSUPPORTED.

[How]
Modify the condition to psr->psr_version == DC_PSR_VERSION_SU_1

Reviewed-by: Sun peng Li <sunpeng.li@amd.com>
Signed-off-by: Tom Chung <chiahsuan.chung@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 HBlank reduction DPCD write to DPMS sequence

[Why]
Certain small HBlank timings may not have a large enough HBlank to
support audio when low bpp DSC is enabled. HBlank expansion by the
source can solve this problem, but requires the branch/sink to support
HBlank reduction.

[How]
Update DPMS sequence to call DM to perform DPCD write to enable HBlank
reduction on the branch/sink. Add stub in dm_helpers to be implemented
later.

Reviewed-by: Michael Strauss <michael.strauss@amd.com>
Reviewed-by: Wenjing Liu <wenjing.liu@amd.com>
Signed-off-by: George Shen <george.shen@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: Optimize cursor position updates

[why]
Updating the cursor enablement register can be a slow operation and accumulates
when high polling rate cursors cause frequent updates asynchronously to the
cursor position.

[how]
Since the cursor enable bit is cached there is no need to update the
enablement register if there is no change to it.  This removes the
read-modify-write from the cursor position programming path in HUBP and
DPP, leaving only the register writes.

Reviewed-by: Josip Pavic <josip.pavic@amd.com>
Signed-off-by: Aric Cyr <Aric.Cyr@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: fix init_adj offset for cositing in SPL

[Why & How]
init_adj offset is applied when cosited not interstitial
Adjust cositing offset in SPL

Reviewed-by: Jun Lei <jun.lei@amd.com>
Signed-off-by: Samson Tam <Samson.Tam@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: Ensure correct GFX tiling info passed to DML

[Why]
To ensure DML validation receives the correct tiling information,
such as swizzle mode or array mode, based on the active GFX format

[How]
- For new GFX format passed swizzle_mode to DML.
- For legacy GFX format passed array_mode to DML.
- Dynamically determined the appropriate tiling info based on the
  active GFX format.

[Description]
This commit ensures that the correct GFX tiling information is passed
to DML. Depending on the active GFX format, the appropriate tiling info
is passed to DML. This change accommodates the different GFX formats
supported by latest platforms, ensuring compatibility and proper
DML validation.

Reviewed-by: Alvin Lee <alvin.lee2@amd.com>
Signed-off-by: Karthi Kandasamy <karthi.kandasamy@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: Update dc_tiling_info union to structure

[WHY]
The `dc_tiling_info` union previously did not have a field to
specify the active GFX format, assuming only one format would
be used per DCN version. from DCN4+, support for switching
between different GFX formats is introduced, requiring a way
to track which format is currently in use.

[HOW]
Updated the `dc_tiling_info` union to include a new field that
explicitly indicates the currently used GFX format.
This allows the system to determine the active GFX format
and take the correct programming path accordingly.

[Description]
The union `dc_tiling_info` has been updated to support multiple
GFX formats by adding a new field for identifying the active format.
This update ensures that the correct programming path is followed
based on the selected format. All references to `dc_tiling_info`
in the codebase have been updated to reflect the new structure.

Reviewed-by: Alvin Lee <alvin.lee2@amd.com>
Signed-off-by: Karthi Kandasamy <karthi.kandasamy@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: modify init dc_power_state

[why]
initialize the power state for dc use,
but dc_set_power_state it not called at D3.
It would cause can't recognize last power state

[how]
remove initialize the power state for dc use, it is not necessary.

Reviewed-by: Nicholas Kazlauskas <nicholas.kazlauskas@amd.com>
Signed-off-by: Brandon Syu <Brandon.Syu@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: have pretrain for dpia

[WHY]
We like to have pretrain for dpia link so that dp and dp tunneling
have aligned behavior. The Main difficult for dpia pretrain is that
encoder can not get corresponded dpia port when link detection
in current implementation.

[HOW]
1. create enable/disable dpia output functions for dcn35 encoder
and have dpia_id and other necessary info as inputs.
2. dcn35 dpia use the new functions to enable/disable output.
3. have a option to enable/disable the change.

Reviewed-by: Wenjing Liu <wenjing.liu@amd.com>
Reviewed-by: Meenakshikumar Somasundaram <meenakshikumar.somasundaram@amd.com>
Signed-off-by: Peichen Huang <PeiChen.Huang@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: Implement Replay Low Hz Visual Confirm

[why]
Add new Visual confirm color for Replay Low Hz.

Reviewed-by: Robin Chen <robin.chen@amd.com>
Signed-off-by: Dennis.Chan <Dennis.Chan@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 Interface to Dump DSC Caps from dm

No common dsc params found between encoder and decoder is one
of the reason that could prevent dsc from properly enabled.

Dump the params to a specific timing to help locate possible
invalid dsc params in either encoder or decoder side.

Reviewed-by: Aurabindo Pillai <aurabindo.pillai@amd.com>
Signed-off-by: Fangzhi Zuo <Jerry.Zuo@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 guards around MAX/MIN

MAX/MIN macros maybe defined already, hence add a guard around them to
prevent errors that complain about redefinition like:

drivers/gpu/drm/amd/amdgpu/../dal-dev/modules/hdcp/hdcp_ddc.c:31: error: "MIN" redefined [-Werror]
   31 | #define MIN(a, b) ((a) < (b) ? (a) : (b))
      |
In file included from ./include/linux/kernel.h:28,
                 from ./include/linux/cpumask.h:11,
                 from ./include/linux/smp.h:13,
                 from ./include/linux/lockdep.h:14,
                 from ./include/linux/spinlock.h:63,
                 from ./include/linux/mmzone.h:8,
                 from ./include/linux/gfp.h:7,
                 from ./include/linux/slab.h:16,
                 from drivers/gpu/drm/amd/amdgpu/../display/dc/os_types.h:37,
                 from drivers/gpu/drm/amd/amdgpu/../display/modules/inc/mod_hdcp.h:29,
                 from drivers/gpu/drm/amd/amdgpu/../display/modules/hdcp/hdcp.h:29,
                 from drivers/gpu/drm/amd/amdgpu/../display/modules/hdcp/hdcp_ddc.c:29:
./include/linux/minmax.h:329: note: this is the location of the previous definition
  329 | #define MIN(a,b) __cmp(min,a,b)
      |
cc1: all warnings being treated as errors

Reviewed-by: Sun peng Li <sunpeng.li@amd.com>
Signed-off-by: Aurabindo Pillai <aurabindo.pillai@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: correct type mismatches in comparisons in DML2

[Why]
Comparisons were made between unsigned char and unsigned int.

[How]
Corrected by changing variable types.

Reviewed-by: Nicholas Kazlauskas <nicholas.kazlauskas@amd.com>
Signed-off-by: Natanel Roizenman <Natanel.Roizenman@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: Adjust dm to use supported interfaces for setting multiple crc windows

[Why & How]
We actually have the capability to calculate independent CRC for 2 crc window
at the same time. Extend dm with the capability by having array to
configure/maintain multiple crc windows. Add the flexibility but use 1st CRC
instance only for now. Can change to use the 2nd CRC instance if needed.

Reviewed-by: HaoPing Liu <haoping.liu@amd.com>
Signed-off-by: Wayne Lin <Wayne.Lin@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: Extend dc_stream_get_crc to support 2nd crc engine

[Why & How]
Since now we can set multiple crc windows for secure display, add a new input
parameter for dc_stream_get_crc to indicate to fetch crc from which crc
engine.

Reviewed-by: HaoPing Liu <haoping.liu@amd.com>
Signed-off-by: Wayne Lin <Wayne.Lin@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 support for setting multiple CRC windows in dc

[Why & How]
Have to support multiple CRC windows setting to dmub. Add new dmub forward
functions for supporting/forwarding multiple crc windows setting to dmub.

Reviewed-by: HaoPing Liu <haoping.liu@amd.com>
Signed-off-by: Wayne Lin <Wayne.Lin@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 expanded HBlank field to dc_crtc_timing

[Why]
For DP HBlank expansion/reduction, the HBlank parameters of the original
EDID timing needs to be notified to the sink in order for the timing to
be reduced back to the original HBlank size.

[How]
Add parameter in dc_crtc_timing to track the increased HBlank.

Reviewed-by: Michael Strauss <michael.strauss@amd.com>
Signed-off-by: George Shen <george.shen@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 required HBlank size calc to link interface

[Why]
Some features, such as HBlank expansion/reduction, needs to know how
much HBlank is required to support basic audio.

[How]
Add interface to link to calculate required HBlank size for a given
link + timing combination to support basic audio (i.e. 2-channel 48KHz).

Reviewed-by: Wenjing Liu <wenjing.liu@amd.com>
Signed-off-by: George Shen <george.shen@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: Parse RECEIVE_PORT0_CAP capabilities from DPCD

[Why]
DPCD register RECEIVE_PORT0_CAP contains HBlank expansion/reduction
capabilities of a DP device. These capabilities are required to enable
HBlank expansion/reduction logic.

[How]
Read raw RECEIVE_PORT0_CAP register values and store parsed fields.

Reviewed-by: Wenjing Liu <wenjing.liu@amd.com>
Signed-off-by: George Shen <george.shen@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: Cleanup outdated interfaces in dcn401_clk_mgr

[WHY&HOW]
- Remove legacy update clocks sequence
- FCLK P-State allow message is not required

Reviewed-by: Alvin Lee <alvin.lee2@amd.com>
Signed-off-by: Dillon Varone <dillon.varone@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: power up all gating blocks when releasing hw DCN35

[WHY & HOW]
Driver disable will deallocate framebuffer to reset IPS state, this will cause IPS start with
INIT state to blindly power gate ONO region to break power sequence. All the gating blocks
should be powered up when releasing hw to ensure all the power optimizations are identical to pre-OS.

Reviewed-by: Nicholas Kazlauskas <nicholas.kazlauskas@amd.com>
Reviewed-by: Duncan Ma <duncan.ma@amd.com>
Signed-off-by: Yihan Zhu <Yihan.Zhu@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: update sequential pg logic DCN35

[WHY & HOW]
No check for HUBP/DPP power gating when DSC instance is still running. Avoid HUBP/DPP to
power gate when corresponding DSC block is still running in the power gating calculation.

Reviewed-by: Nicholas Kazlauskas <nicholas.kazlauskas@amd.com>
Reviewed-by: Duncan Ma <duncan.ma@amd.com>
Signed-off-by: Yihan Zhu <Yihan.Zhu@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/amdkfd: fixed page fault when enable MES shader debugger

Initialize the process context address before setting the shader debugger.

[  260.781212] amdgpu 0000:03:00.0: amdgpu: [gfxhub] page fault (src_id:0 ring:32 vmid:0 pasid:0)
[  260.781236] amdgpu 0000:03:00.0: amdgpu:   in page starting at address 0x0000000000000000 from client 10
[  260.781255] amdgpu 0000:03:00.0: amdgpu: GCVM_L2_PROTECTION_FAULT_STATUS:0x00040A40
[  260.781270] amdgpu 0000:03:00.0: amdgpu:      Faulty UTCL2 client ID: CPC (0x5)
[  260.781284] amdgpu 0000:03:00.0: amdgpu:      MORE_FAULTS: 0x0
[  260.781296] amdgpu 0000:03:00.0: amdgpu:      WALKER_ERROR: 0x0
[  260.781308] amdgpu 0000:03:00.0: amdgpu:      PERMISSION_FAULTS: 0x4
[  260.781320] amdgpu 0000:03:00.0: amdgpu:      MAPPING_ERROR: 0x0
[  260.781332] amdgpu 0000:03:00.0: amdgpu:      RW: 0x1
[  260.782017] amdgpu 0000:03:00.0: amdgpu: [gfxhub] page fault (src_id:0 ring:32 vmid:0 pasid:0)
[  260.782039] amdgpu 0000:03:00.0: amdgpu:   in page starting at address 0x0000000000000000 from client 10
[  260.782058] amdgpu 0000:03:00.0: amdgpu: GCVM_L2_PROTECTION_FAULT_STATUS:0x00040A41
[  260.782073] amdgpu 0000:03:00.0: amdgpu:      Faulty UTCL2 client ID: CPC (0x5)
[  260.782087] amdgpu 0000:03:00.0: amdgpu:      MORE_FAULTS: 0x1
[  260.782098] amdgpu 0000:03:00.0: amdgpu:      WALKER_ERROR: 0x0
[  260.782110] amdgpu 0000:03:00.0: amdgpu:      PERMISSION_FAULTS: 0x4
[  260.782122] amdgpu 0000:03:00.0: amdgpu:      MAPPING_ERROR: 0x0
[  260.782137] amdgpu 0000:03:00.0: amdgpu:      RW: 0x1
[  260.782155] amdgpu 0000:03:00.0: amdgpu: [gfxhub] page fault (src_id:0 ring:32 vmid:0 pasid:0)
[  260.782166] amdgpu 0000:03:00.0: amdgpu:   in page starting at address 0x0000000000000000 from client 10

Fixes: 438b39ac74e2 ("drm/amdkfd: pause autosuspend when creating pdd")
Closes: https://gitlab.freedesktop.org/drm/amd/-/issues/3849
Signed-off-by: Jesse Zhang <jesse.zhang@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Reapply fdedd77b0eb3

Commit 2563391e57b5 ("drm/amd/display: DML2.1 resynchronization") blew
away the compiler warning fix from commit 2fde4fdddc1f
("drm/amd/display: Avoid -Wenum-float-conversion in
add_margin_and_round_to_dfs_grainularity()"), causing the warning to
reappear.

  drivers/gpu/drm/amd/amdgpu/../display/dc/dml2/dml21/src/dml2_dpmm/dml2_dpmm_dcn4.c:183:58: error: arithmetic between enumeration type 'enum dentist_divider_range' and floating-point type 'double' [-Werror,-Wenum-float-conversion]
    183 |         divider = (unsigned int)(DFS_DIVIDER_RANGE_SCALE_FACTOR * (vco_freq_khz / clock_khz));
        |                                  ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ ^ ~~~~~~~~~~~~~~~~~~~~~~~~~~
  1 error generated.

Apply the fix again to resolve the warning.

Re-apply again after commit be4e3509314a ("drm/amd/display: DML21 Reintegration For Various Fixes")
This should be making its way back to the original DML trees this time. (Alex)

Fixes: be4e3509314a ("drm/amd/display: DML21 Reintegration For Various Fixes")
Closes: https://gitlab.freedesktop.org/drm/amd/-/issues/3841
Signed-off-by: Nathan Chancellor <nathan@kernel.org>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: fix divide error in DM plane scale calcs

dm_get_plane_scale doesn't take into account plane scaled size equal to
zero, leading to a kernel oops due to division by zero. Fix by setting
out-scale size as zero when the dst size is zero, similar to what is
done by drm_calc_scale(). This issue started with the introduction of
cursor ovelay mode that uses this function to assess cursor mode changes
via dm_crtc_get_cursor_mode() before checking plane state.

[Dec17 17:14] Oops: divide error: 0000 [#1] PREEMPT SMP NOPTI
[  +0.000018] CPU: 5 PID: 1660 Comm: surface-DP-1 Not tainted 6.10.0+ #231
[  +0.000007] Hardware name: Valve Jupiter/Jupiter, BIOS F7A0131 01/30/2024
[  +0.000004] RIP: 0010:dm_get_plane_scale+0x3f/0x60 [amdgpu]
[  +0.000553] Code: 44 0f b7 41 3a 44 0f b7 49 3e 83 e0 0f 48 0f a3 c2 73 21 69 41 28 e8 03 00 00 31 d2 41 f7 f1 31 d2 89 06 69 41 2c e8 03 00 00 <41> f7 f0 89 07 e9 d7 d8 7e e9 44 89 c8 45 89 c1 41 89 c0 eb d4 66
[  +0.000005] RSP: 0018:ffffa8df0de6b8a0 EFLAGS: 00010246
[  +0.000006] RAX: 00000000000003e8 RBX: ffff9ac65c1f6e00 RCX: ffff9ac65d055500
[  +0.000003] RDX: 0000000000000000 RSI: ffffa8df0de6b8b0 RDI: ffffa8df0de6b8b4
[  +0.000004] RBP: ffff9ac64e7a5800 R08: 0000000000000000 R09: 0000000000000a00
[  +0.000003] R10: 00000000000000ff R11: 0000000000000054 R12: ffff9ac6d0700010
[  +0.000003] R13: ffff9ac65d054f00 R14: ffff9ac65d055500 R15: ffff9ac64e7a60a0
[  +0.000004] FS:  00007f869ea00640(0000) GS:ffff9ac970080000(0000) knlGS:0000000000000000
[  +0.000004] CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
[  +0.000003] CR2: 000055ca701becd0 CR3: 000000010e7f2000 CR4: 0000000000350ef0
[  +0.000004] Call Trace:
[  +0.000007]  <TASK>
[  +0.000006]  ? __die_body.cold+0x19/0x27
[  +0.000009]  ? die+0x2e/0x50
[  +0.000007]  ? do_trap+0xca/0x110
[  +0.000007]  ? do_error_trap+0x6a/0x90
[  +0.000006]  ? dm_get_plane_scale+0x3f/0x60 [amdgpu]
[  +0.000504]  ? exc_divide_error+0x38/0x50
[  +0.000005]  ? dm_get_plane_scale+0x3f/0x60 [amdgpu]
[  +0.000488]  ? asm_exc_divide_error+0x1a/0x20
[  +0.000011]  ? dm_get_plane_scale+0x3f/0x60 [amdgpu]
[  +0.000593]  dm_crtc_get_cursor_mode+0x33f/0x430 [amdgpu]
[  +0.000562]  amdgpu_dm_atomic_check+0x2ef/0x1770 [amdgpu]
[  +0.000501]  drm_atomic_check_only+0x5e1/0xa30 [drm]
[  +0.000047]  drm_mode_atomic_ioctl+0x832/0xcb0 [drm]
[  +0.000050]  ? __pfx_drm_mode_atomic_ioctl+0x10/0x10 [drm]
[  +0.000047]  drm_ioctl_kernel+0xb3/0x100 [drm]
[  +0.000062]  drm_ioctl+0x27a/0x4f0 [drm]
[  +0.000049]  ? __pfx_drm_mode_atomic_ioctl+0x10/0x10 [drm]
[  +0.000055]  amdgpu_drm_ioctl+0x4e/0x90 [amdgpu]
[  +0.000360]  __x64_sys_ioctl+0x97/0xd0
[  +0.000010]  do_syscall_64+0x82/0x190
[  +0.000008]  ? __pfx_drm_mode_createblob_ioctl+0x10/0x10 [drm]
[  +0.000044]  ? srso_return_thunk+0x5/0x5f
[  +0.000006]  ? drm_ioctl_kernel+0xb3/0x100 [drm]
[  +0.000040]  ? srso_return_thunk+0x5/0x5f
[  +0.000005]  ? __check_object_size+0x50/0x220
[  +0.000007]  ? srso_return_thunk+0x5/0x5f
[  +0.000005]  ? srso_return_thunk+0x5/0x5f
[  +0.000005]  ? drm_ioctl+0x2a4/0x4f0 [drm]
[  +0.000039]  ? __pfx_drm_mode_createblob_ioctl+0x10/0x10 [drm]
[  +0.000043]  ? srso_return_thunk+0x5/0x5f
[  +0.000005]  ? srso_return_thunk+0x5/0x5f
[  +0.000005]  ? __pm_runtime_suspend+0x69/0xc0
[  +0.000006]  ? srso_return_thunk+0x5/0x5f
[  +0.000005]  ? amdgpu_drm_ioctl+0x71/0x90 [amdgpu]
[  +0.000366]  ? srso_return_thunk+0x5/0x5f
[  +0.000006]  ? syscall_exit_to_user_mode+0x77/0x210
[  +0.000007]  ? srso_return_thunk+0x5/0x5f
[  +0.000005]  ? do_syscall_64+0x8e/0x190
[  +0.000006]  ? srso_return_thunk+0x5/0x5f
[  +0.000006]  ? do_syscall_64+0x8e/0x190
[  +0.000006]  ? srso_return_thunk+0x5/0x5f
[  +0.000007]  entry_SYSCALL_64_after_hwframe+0x76/0x7e
[  +0.000008] RIP: 0033:0x55bb7cd962bc
[  +0.000007] Code: 4c 89 6c 24 18 4c 89 64 24 20 4c 89 74 24 28 0f 57 c0 0f 11 44 24 30 89 c7 48 8d 54 24 08 b8 10 00 00 00 be bc 64 38 c0 0f 05 <49> 89 c7 48 83 3b 00 74 09 4c 89 c7 ff 15 62 64 99 00 48 83 7b 18
[  +0.000005] RSP: 002b:00007f869e9f4da0 EFLAGS: 00000217 ORIG_RAX: 0000000000000010
[  +0.000007] RAX: ffffffffffffffda RBX: 00007f869e9f4fb8 RCX: 000055bb7cd962bc
[  +0.000004] RDX: 00007f869e9f4da8 RSI: 00000000c03864bc RDI: 000000000000003b
[  +0.000003] RBP: 000055bb9ddcbcc0 R08: 00007f86541b9920 R09: 0000000000000009
[  +0.000004] R10: 0000000000000004 R11: 0000000000000217 R12: 00007f865406c6b0
[  +0.000003] R13: 00007f86541b5290 R14: 00007f865410b700 R15: 000055bb9ddcbc18
[  +0.000009]  </TASK>

Fixes: 1b04dcca4fb1 ("drm/amd/display: Introduce overlay cursor mode")
Link: https://gitlab.freedesktop.org/drm/amd/-/issues/3729
Reported-by: Fabio Scaccabarozzi <fsvm88@gmail.com>
Co-developed-by: Fabio Scaccabarozzi <fsvm88@gmail.com>
Signed-off-by: Fabio Scaccabarozzi <fsvm88@gmail.com>
Signed-off-by: Melissa Wen <mwen@igalia.com>
Reviewed-by: Rodrigo Siqueira <Rodrigo.Siqueira@amd.com>
Signed-off-by: Rodrigo Siqueira <Rodrigo.Siqueira@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: increase MAX_SURFACES to the value supported by hw

As the hw supports up to 4 surfaces, increase the maximum number of
surfaces to prevent the DC error when trying to use more than three
planes.

[drm:dc_state_add_plane [amdgpu]] *ERROR* Surface: can not attach plane_state 000000003e2cb82c! Maximum is: 3

Link: https://gitlab.freedesktop.org/drm/amd/-/issues/3693
Signed-off-by: Melissa Wen <mwen@igalia.com>
Reviewed-by: Rodrigo Siqueira <Rodrigo.Siqueira@amd.com>
Signed-off-by: Rodrigo Siqueira <Rodrigo.Siqueira@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: fix page fault due to max surface definition mismatch

DC driver is using two different values to define the maximum number of
surfaces: MAX_SURFACES and MAX_SURFACE_NUM. Consolidate MAX_SURFACES as
the unique definition for surface updates across DC.

It fixes page fault faced by Cosmic users on AMD display versions that
support two overlay planes, since the introduction of cursor overlay
mode.

[Nov26 21:33] BUG: unable to handle page fault for address: 0000000051d0f08b
[  +0.000015] #PF: supervisor read access in kernel mode
[  +0.000006] #PF: error_code(0x0000) - not-present page
[  +0.000005] PGD 0 P4D 0
[  +0.000007] Oops: Oops: 0000 [#1] PREEMPT SMP NOPTI
[  +0.000006] CPU: 4 PID: 71 Comm: kworker/u32:6 Not tainted 6.10.0+ #300
[  +0.000006] Hardware name: Valve Jupiter/Jupiter, BIOS F7A0131 01/30/2024
[  +0.000007] Workqueue: events_unbound commit_work [drm_kms_helper]
[  +0.000040] RIP: 0010:copy_stream_update_to_stream.isra.0+0x30d/0x750 [amdgpu]
[  +0.000847] Code: 8b 10 49 89 94 24 f8 00 00 00 48 8b 50 08 49 89 94 24 00 01 00 00 8b 40 10 41 89 84 24 08 01 00 00 49 8b 45 78 48 85 c0 74 0b <0f> b6 00 41 88 84 24 90 64 00 00 49 8b 45 60 48 85 c0 74 3b 48 8b
[  +0.000010] RSP: 0018:ffffc203802f79a0 EFLAGS: 00010206
[  +0.000009] RAX: 0000000051d0f08b RBX: 0000000000000004 RCX: ffff9f964f0a8070
[  +0.000004] RDX: ffff9f9710f90e40 RSI: ffff9f96600c8000 RDI: ffff9f964f000000
[  +0.000004] RBP: ffffc203802f79f8 R08: 0000000000000000 R09: 0000000000000000
[  +0.000005] R10: 0000000000000000 R11: 0000000000000000 R12: ffff9f96600c8000
[  +0.000004] R13: ffff9f9710f90e40 R14: ffff9f964f000000 R15: ffff9f96600c8000
[  +0.000004] FS:  0000000000000000(0000) GS:ffff9f9970000000(0000) knlGS:0000000000000000
[  +0.000005] CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
[  +0.000005] CR2: 0000000051d0f08b CR3: 00000002e6a20000 CR4: 0000000000350ef0
[  +0.000005] Call Trace:
[  +0.000011]  <TASK>
[  +0.000010]  ? __die_body.cold+0x19/0x27
[  +0.000012]  ? page_fault_oops+0x15a/0x2d0
[  +0.000014]  ? exc_page_fault+0x7e/0x180
[  +0.000009]  ? asm_exc_page_fault+0x26/0x30
[  +0.000013]  ? copy_stream_update_to_stream.isra.0+0x30d/0x750 [amdgpu]
[  +0.000739]  ? dc_commit_state_no_check+0xd6c/0xe70 [amdgpu]
[  +0.000470]  update_planes_and_stream_state+0x49b/0x4f0 [amdgpu]
[  +0.000450]  ? srso_return_thunk+0x5/0x5f
[  +0.000009]  ? commit_minimal_transition_state+0x239/0x3d0 [amdgpu]
[  +0.000446]  update_planes_and_stream_v2+0x24a/0x590 [amdgpu]
[  +0.000464]  ? srso_return_thunk+0x5/0x5f
[  +0.000009]  ? sort+0x31/0x50
[  +0.000007]  ? amdgpu_dm_atomic_commit_tail+0x159f/0x3a30 [amdgpu]
[  +0.000508]  ? srso_return_thunk+0x5/0x5f
[  +0.000009]  ? amdgpu_crtc_get_scanout_position+0x28/0x40 [amdgpu]
[  +0.000377]  ? srso_return_thunk+0x5/0x5f
[  +0.000009]  ? drm_crtc_vblank_helper_get_vblank_timestamp_internal+0x160/0x390 [drm]
[  +0.000058]  ? srso_return_thunk+0x5/0x5f
[  +0.000005]  ? dma_fence_default_wait+0x8c/0x260
[  +0.000010]  ? srso_return_thunk+0x5/0x5f
[  +0.000005]  ? wait_for_completion_timeout+0x13b/0x170
[  +0.000006]  ? srso_return_thunk+0x5/0x5f
[  +0.000005]  ? dma_fence_wait_timeout+0x108/0x140
[  +0.000010]  ? commit_tail+0x94/0x130 [drm_kms_helper]
[  +0.000024]  ? process_one_work+0x177/0x330
[  +0.000008]  ? worker_thread+0x266/0x3a0
[  +0.000006]  ? __pfx_worker_thread+0x10/0x10
[  +0.000004]  ? kthread+0xd2/0x100
[  +0.000006]  ? __pfx_kthread+0x10/0x10
[  +0.000006]  ? ret_from_fork+0x34/0x50
[  +0.000004]  ? __pfx_kthread+0x10/0x10
[  +0.000005]  ? ret_from_fork_asm+0x1a/0x30
[  +0.000011]  </TASK>

Fixes: 1b04dcca4fb1 ("drm/amd/display: Introduce overlay cursor mode")
Suggested-by: Leo Li <sunpeng.li@amd.com>
Link: https://gitlab.freedesktop.org/drm/amd/-/issues/3693
Signed-off-by: Melissa Wen <mwen@igalia.com>
Reviewed-by: Rodrigo Siqueira <Rodrigo.Siqueira@amd.com>
Signed-off-by: Rodrigo Siqueira <Rodrigo.Siqueira@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Remove unnecessary amdgpu_irq_get/put

[WHY & HOW]
commit 7fb363c57522 ("drm/amd/display: Let drm_crtc_vblank_on/off manage interrupts")
lets drm_crtc_vblank_* to manage interrupts in amdgpu_dm_crtc_set_vblank,
and amdgpu_irq_get/put do not need to be called here.  Part of that
patch got lost somehow, so fix it up.

Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Reviewed-by: Leo Li <sunpeng.li@amd.com>
Signed-off-by: Alex Hung <alex.hung@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: Handle NULL bo->tbo.resource (again) in amdgpu_vm_bo_update

Third time's the charm, I hope?

Fixes: d3116756a710 ("drm/ttm: rename bo->mem and make it a pointer")
Closes: https://gitlab.freedesktop.org/drm/amd/-/issues/3837
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/admgpu: replace kmalloc() and memcpy() with kmemdup()

The static analyser tool gave the following advice:

./drivers/gpu/drm/amd/amdgpu/amdgpu_virt.c:1266:7-14: WARNING opportunity for kmemdup

 → 1266         tmp = kmalloc(used_size, GFP_KERNEL);
   1267         if (!tmp)
   1268                 return -ENOMEM;
   1269
 → 1270         memcpy(tmp, &host_telemetry->body.error_count, used_size);

Replacing kmalloc() + memcpy() with kmemdump() doesn't change semantics.
Original code works without fault, so this is not a bug fix but proposed improvement.

Link: https://lwn.net/Articles/198928/
Fixes: 84a2947ecc85 ("drm/amdgpu: Implement virt req_ras_err_count")
Cc: Alex Deucher <alexander.deucher@amd.com>
Cc: "Christian König" <christian.koenig@amd.com>
Cc: Xinhui Pan <Xinhui.Pan@amd.com>
Cc: David Airlie <airlied@gmail.com>
Cc: Simona Vetter <simona@ffwll.ch>
Cc: Zhigang Luo <Zhigang.Luo@amd.com>
Cc: Victor Skvortsov <victor.skvortsov@amd.com>
Cc: Hawking Zhang <Hawking.Zhang@amd.com>
Cc: Lijo Lazar <lijo.lazar@amd.com>
Cc: Yunxiang Li <Yunxiang.Li@amd.com>
Cc: Jack Xiao <Jack.Xiao@amd.com>
Cc: Vignesh Chander <Vignesh.Chander@amd.com>
Cc: Danijel Slivka <danijel.slivka@amd.com>
Cc: amd-gfx@lists.freedesktop.org
Cc: dri-devel@lists.freedesktop.org
Cc: linux-kernel@vger.kernel.org
Reviewed-by: Lijo Lazar <lijo.lazar@amd.com>
Signed-off-by: Mirsad Todorovac <mtodorovac69@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Fix NULL pointer dereference in dmub_tracebuffer_show

It corrects the issue by checking if 'adev->dm.dmub_srv' is NULL before
accessing its 'meta_info' member. This ensures that we do not
dereference a NULL pointer.

Fixes the below:
drivers/gpu/drm/amd/amdgpu/../display/amdgpu_dm/amdgpu_dm_debugfs.c:917 dmub_tracebuffer_show()
warn: address of 'adev->dm.dmub_srv->meta_info' is non-NULL

drivers/gpu/drm/amd/amdgpu/../display/amdgpu_dm/amdgpu_dm_debugfs.c
    901 static int dmub_tracebuffer_show(struct seq_file *m, void *data)
    902 {
    903         struct amdgpu_device *adev = m->private;
    904         struct dmub_srv_fb_info *fb_info = adev->dm.dmub_fb_info;
    905         struct dmub_fw_meta_info *fw_meta_info = &adev->dm.dmub_srv->meta_info;
                                                         ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ Even if adev->dm.dmub_srv is NULL, the address of ->meta_info can't be NULL

    906         struct dmub_debugfs_trace_entry *entries;
    907         uint8_t *tbuf_base;
    908         uint32_t tbuf_size, max_entries, num_entries, first_entry, i;
    909
    910         if (!fb_info)
    911                 return 0;
    912
    913         tbuf_base = (uint8_t *)fb_info->fb[DMUB_WINDOW_5_TRACEBUFF].cpu_addr;
    914         if (!tbuf_base)
    915                 return 0;
    916
--> 917         tbuf_size = fw_meta_info ? fw_meta_info->trace_buffer_size :
                            ^^^^^^^^^^^^ Always non-NULL

    918                                    DMUB_TRACE_BUFFER_SIZE;
    919         max_entries = (tbuf_size - sizeof(struct dmub_debugfs_trace_header)) /
    920                       sizeof(struct dmub_debugfs_trace_entry);
    921
    922         num_entries =

v2: Initialize struct dmub_fw_meta_info *fw_meta_info to NULL (Dan Carpenter)

Fixes: 5a498172c8d0 ("drm/amd/display: Make DMCUB tracebuffer debugfs chronological")
Cc: Leo Li <sunpeng.li@amd.com>
Cc: Tom Chung <chiahsuan.chung@amd.com>
Cc: Rodrigo Siqueira <Rodrigo.Siqueira@amd.com>
Cc: Roman Li <roman.li@amd.com>
Cc: Alex Hung <alex.hung@amd.com>
Cc: Aurabindo Pillai <aurabindo.pillai@amd.com>
Cc: Harry Wentland <harry.wentland@amd.com>
Cc: Hamza Mahfooz <hamza.mahfooz@amd.com>
Reported-by: Dan Carpenter <dan.carpenter@linaro.org>
Signed-off-by: Srinivasan Shanmugam <srinivasan.shanmugam@amd.com>
Reviewed-by: Roman Li <roman.li@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: Show warning message if IH ring overflow

If IH primary ring and KFD ih fifo overflows, we may miss CP, SDMA
interrupts and cause application soft hang. Show warning message with
ring name if overflow happens.

Add function to get ih ring name to avoid duplicating it. To keep
warning message consistent between GPU generations, change all
*_ih.c except ASICs older than Vega which has only one ih ring.

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

drm/amdkfd: Improve signal event slow path

If event slot is not signaled, kfd_signal_event_interrupt goes to slow
path to scan all event slots to find the signaled event, this is needed
for old ASICs that don't have the event ID or the event IDs are
incorrect in the IH payload.

There is case that GPU signal the same event twice, then driver process
the first event interrupt, set_event and event slot is auto-reset, then
for the second event interrupt, KFD goes to slow path as event is not
signaled, just drop the second event interrupt because the application
only need wakeup once.

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

drm/amdkfd: Queue interrupt work to different CPU

For CPX mode, each KFD node has interrupt worker to process ih_fifo to
send events to user space. Currently all interrupt workers of same adev
queue to same CPU, all workers execution are actually serialized and
this cause KFD ih_fifo overflow when CPU usage is high.

Use per-GPU unbounded highpri queue with number of workers equals to
number of partitions, let queue_work select the next CPU round robin
among the local CPUs of same NUMA.

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

drm/amdgpu: Optimize gfx v9 GPU page fault handling

After GPU page fault, there are lots of page fault interrupts generated
at short period even with CAM filter enabled because the fault address
is different. Each page fault copy to KFD ih fifo to send event to user
space by KFD interrupt worker, this could cause KFD ih fifo overflow
while other processes generate events at same time.

KFD process is aborted after GPU page fault, we only need one GPU page
fault interrupt sent to KFD ih fifo to send memory exception event to
user space.

Incease KFD ih fifo size to 2 times of IH primary ring size, to handle
the burst events case.

This patch handle the gfx v9 path, cover retry on/off and CAM filter
on/off cases.

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

drm/amdkfd: KFD interrupt access ih_fifo data in-place

To handle 40000 to 80000 interrupts per second running CPX mode with 4
streams/queues per KFD node, KFD interrupt handler becomes the
performance bottleneck.

Remove the kfifo_out memcpy overhead by accessing ih_fifo data in-place
and updating rptr with kfifo_skip_count.

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

drm/amdgpu: partially revert "reduce reset time"

This partially reverts commit 194eb174cbe4fe2b3376ac30acca2dc8c8beca00.

This commit introduced a new state variable into adev without even
remotely worrying about CPU barriers.

Since we already have the amdgpu_in_reset() function exactly for this
use case partially revert that.

Signed-off-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: set the VM pointer to NULL in amdgpu_job_prepare

As soon as the prepare phase is completed the VM might be released,
better set it to NULL.

Signed-off-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 amdgpu_coredump

The VM pointer might already be outdated when that function is called.
Use the PASID instead to gather the information instead.

Signed-off-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: Enable psp v14_0_3 RAS support for non-SRIOV configurations.

Enable psp v14_0_3 RAS support for non-SRIOV configurations.

Signed-off-by: Candice Li <candice.li@amd.com>
Reviewed-by: Hawking Zhang <Hawking.Zhang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: Don't enable sdma 4.4.5 CTXEMPTY interrupt

The sdma context empty interrupt is dropped in amdgpu_irq_dispatch
as unregistered interrupt src_id 243, this interrupt accounts to 1/3 of
total interrupts and causes IH primary ring overflow when running
stressful benchmark application. Disable this interrupt has no side
effect found.

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

drm/amdgpu: Fix potential integer overflow in scheduler mask calculations

The use of 1 << i in scheduler mask calculations can result in an
unintentional integer overflow due to the expression being
evaluated as a 32-bit signed integer.

This patch replaces 1 << i with 1ULL << i to ensure the operation
is performed as a 64-bit unsigned integer, preventing overflow

Discovered in coverity scan, CID 1636393, 1636175, 1636007, 1635853

Fixes: c5c63d9cb5d3 ("drm/amdgpu: add amdgpu_gfx_sched_mask and amdgpu_compute_sched_mask debugfs")
Signed-off-by: Karol Przybylski <karprzy7@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu/smu14.0.2: fix IP version check

Use the helper function rather than reading it directly.

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

drm/amdgpu/gfx12: fix IP version check

Use the helper function rather than reading it directly.

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

drm/amdgpu/mmhub4.1: fix IP version check

Use the helper function rather than reading it directly.

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

drm/amdgpu/nbio7.11: fix IP version check

Use the helper function rather than reading it directly.

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

drm/amdgpu/nbio7.0: fix IP version check

Use the helper function rather than reading it directly.

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

drm/amdgpu/nbio7.7: fix IP version check

Use the helper function rather than reading it directly.

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

drm/amd/display: 3.2.314

DC 3.2.314 contains some improvements as summarized below:

* Update DML21 code.
* Fixes for FAMS2 interface.
* HDMI fixes.
* Compilation warning fixes.

Signed-off-by: Aric Cyr <aric.cyr@amd.com>
Acked-by: Rodrigo Siqueira <rodrigo.siqueira@amd.com>
Signed-off-by: Rodrigo Siqueira <rodrigo.siqueira@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Disable MPC rate control on ODM pipe update

[Why]
Seamless boot skips MPC init for the active pipe, resulting in stale MPC
rate control state being retained. This will cause issues since other
logic assumes it is disabled (as DCN30 and newer does not need it).

[How]
Disable MPC rate control on ODM pipe update to cover the seamless boot
case.

Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Reviewed-by: Alvin Lee <alvin.lee2@amd.com>
Signed-off-by: George Shen <george.shen@amd.com>
Signed-off-by: Rodrigo Siqueira <rodrigo.siqueira@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Block Invalid TMDS operation

[Why]
When sink type is TMDS, PHY programming does not block against pixel
clock greater than 600MHz.

[How]
Based on sink type, block greater than 600MHz phy programming.

Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Reviewed-by: Aric Cyr <aric.cyr@amd.com>
Signed-off-by: Chris Park <chris.park@amd.com>
Signed-off-by: Dillon Varone <dillon.varone@amd.com>
Signed-off-by: Rodrigo Siqueira <rodrigo.siqueira@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Re-validate streams on commit_streams

To prevent invalid HW programming, streams should be revalidated first
before committing to HW.

Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Reviewed-by: Aric Cyr <aric.cyr@amd.com>
Signed-off-by: Dillon Varone <dillon.varone@amd.com>
Signed-off-by: Rodrigo Siqueira <rodrigo.siqueira@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

Revert "drm/amd/display: Fix green screen issue after suspend"

This reverts commit 87b7ebc2e16c14d32a912f18206a4d6cc9abc3e8.

A long time ago, we had an issue with the Raven system when it was
connected to two displays: one with DP and another with HDMI. After the
system woke up from suspension, we saw a solid green screen caused by an
underflow generated by bad DCC metadata. To workaround this issue, the
'commit 87b7ebc2e16c ("drm/amd/display: Fix green screen issue after
suspend")' was introduced to disable the DCC for a few frames after in
the resume phase. However, in hindsight, this solution was probably a
workaround at the kernel level for some issues from another part
(probably other driver components or user space). After applying this
patch and trying to reproduce the green issue in a similar hardware
system but using the latest kernel and userspace, we cannot see the
issue, which makes this workaround obsolete and creates extra
unnecessary complexity to the code; for all of this reason, this commit
reverts the original change.

Cc: Mario Limonciello <mario.limonciello@amd.com>
Cc: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Reviewed-by: Aurabindo Pillai <aurabindo.pillai@amd.com>
Signed-off-by: Rodrigo Siqueira <Rodrigo.Siqueira@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Fix uninitialized variables in amdgpu_dm_debugfs

[WHAT]
Some fields in struct dc_link_settings and link_training_settings are
not initialized and using them can cause unexpected results.

[HOW]
Initialize struct dc_link_settings and link_training_settings to zero.

Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Reviewed-by: Rodrigo Siqueira <rodrigo.siqueira@amd.com>
Signed-off-by: Alex Hung <alex.hung@amd.com>
Signed-off-by: Rodrigo Siqueira <rodrigo.siqueira@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Apply (some) policy for DML2 formulation on DCN35/DCN351

[Why]
Dropping the entirety of dml2_policy_build_synthetic_soc_states exposes
an issue for states that cannot be filled via bbox_overrides and rely on
the default parameters that may or may not be present depending on the
DM.

For amdgpu_dm this results in missing parameters for most of the struct
in higher states:

- sr_exit_time_us
- sr_enter_plus_exit_time_us
- sr_exit_z8_time_us
- sr_enter_plus_exit_z8_time_us
- urgent_latency_pixel_data_only_us
- urgent_latency_pixel_mixed_with_vm_data_us
- urgent_latency_vm_data_only_us
- dram_clock_change_latency_us
- fclk_change_latency_us
- usr_retraining_latency_us
- writeback_latency_us
- urgent_latency_adjustment_fabric_clock_component_us
- urgent_latency_adjustment_fabric_clock_reference_mhz
- dscclk_mhz
- phyclk_mhz
- phyclk_d18_mhz
- phyclk_d32_mhz
- use_ideal_dram_bw_strobe

[How]
Copy from the first state, applying a minimal policy to set max clocks
for SOC independent values.

Then copy the SOC dependent ones from the states modified by
bbox_overrides.

Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Reviewed-by: Aurabindo Pillai <aurabindo.pillai@amd.com>
Signed-off-by: Nicholas Kazlauskas <nicholas.kazlauskas@amd.com>
Signed-off-by: Rodrigo Siqueira <rodrigo.siqueira@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: delete legacy code

Delete unused code.

Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Reviewed-by: Jun Lei <jun.lei@amd.com>
Signed-off-by: Shunlu Zhang <Shunlu.Zhang@amd.com>
Signed-off-by: Rodrigo Siqueira <rodrigo.siqueira@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Add new message for DF throttling optimization on dcn401

[WHY]
When effective bandwidth from the SoC is enough to perform SubVP
prefetchs, then DF throttling is not required.

[HOW]
Provide SMU the required clocks for which DF throttling is not required.

Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Reviewed-by: Alvin Lee <alvin.lee2@amd.com>
Signed-off-by: Dillon Varone <dillon.varone@amd.com>
Signed-off-by: Rodrigo Siqueira <rodrigo.siqueira@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: DML21 Reintegration For Various Fixes

Reintegrate latest DML21 code.

Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Reviewed-by: Dillon Varone <dillon.varone@amd.com>
Signed-off-by: Austin Zheng <Austin.Zheng@amd.com>
Signed-off-by: Rodrigo Siqueira <rodrigo.siqueira@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Fix brightness adjustment on MiniLED

[Why]
Older Asics were changed to target new DCN while still needing older
support causing brightness adjustments to fail.

[How]
Reverted the DCN targets on required DCNs

Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Reviewed-by: Iswara Nagulendran <iswara.nagulendran@amd.com>
Signed-off-by: Harry VanZyllDeJong <hvanzyll@amd.com>
Signed-off-by: Rodrigo Siqueira <rodrigo.siqueira@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Fix Mode Cutoff in DSC Passthrough to DP2.1 Monitor

Source --> DP2.1 MST hub --> DP1.4/2.1 monitor

When change from DP1.4 to DP2.1 from monitor manual, modes higher than
4k120 are all cutoff by mode validation. Switch back to DP1.4 gets all
the modes up to 4k240 available to be enabled by dsc passthrough.

[why]
Compared to DP1.4 link from hub to monitor, DP2.1 link has larger
full_pbn value that causes overflow in the process of doing conversion
from pbn to kbps.

[how]
Change the data type accordingly to fit into the data limit during
conversion calculation.

Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Reviewed-by: Wayne Lin <wayne.lin@amd.com>
Signed-off-by: Fangzhi Zuo <Jerry.Zuo@amd.com>
Signed-off-by: Rodrigo Siqueira <rodrigo.siqueira@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>