drm/amdgpu: Add more types for boot time error reporting

Data abort exception and unknown errors are supported.

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

drm/amdgpu: force to use legacy inv in mmhub

MMHUB v4.1.0 only support fixed cache mode, so
only use legacy invalidation accordingly.

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

drm/amdgpu: optimize the padding for gfx12

Adding NOP packets one by one in the ring
does not use the CP efficiently.

Solution:
Use CP optimization while adding NOP packet's so PFP
can discard NOP packets based on information of count
from the Header instead of fetching all NOP packets
one by one.

Reviewed-by: Christian König <christian.koenig@amd.com>
Cc: Christian König <christian.koenig@amd.com>
Cc: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Cc: Tvrtko Ursulin <tursulin@igalia.com>
Cc: Marek Olšák <marek.olsak@amd.com>
Signed-off-by: Sunil Khatri <sunil.khatri@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Add NULL check for function pointer in dcn32_set_output_transfer_func

This commit adds a null check for the set_output_gamma function pointer
in the dcn32_set_output_transfer_func function. Previously,
set_output_gamma was being checked for null, but then it was being
dereferenced without any null check. This could lead to a null pointer
dereference if set_output_gamma is null.

To fix this, we now ensure that set_output_gamma is not null before
dereferencing it. We do this by adding a null check for set_output_gamma
before the call to set_output_gamma.

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>
Signed-off-by: Srinivasan Shanmugam <srinivasan.shanmugam@amd.com>
Reviewed-by: Tom Chung <chiahsuan.chung@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Add NULL check for function pointer in dcn401_set_output_transfer_func

This commit adds a null check for the set_output_gamma function pointer
in the dcn401_set_output_transfer_func function. Previously,
set_output_gamma was being checked for null, but then it was being
dereferenced without any null check. This could lead to a null pointer
dereference if set_output_gamma is null.

To fix this, we now ensure that set_output_gamma is not null before
dereferencing it. We do this by adding a null check for set_output_gamma
before the call to set_output_gamma.

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>
Signed-off-by: Srinivasan Shanmugam <srinivasan.shanmugam@amd.com>
Reviewed-by: Tom Chung <chiahsuan.chung@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Add NULL check for function pointer in dcn20_set_output_transfer_func

This commit adds a null check for the set_output_gamma function pointer
in the dcn20_set_output_transfer_func function. Previously,
set_output_gamma was being checked for null at line 1030, but then it
was being dereferenced without any null check at line 1048. This could
potentially lead to a null pointer dereference error if set_output_gamma
is null.

To fix this, we now ensure that set_output_gamma is not null before
dereferencing it. We do this by adding a null check for set_output_gamma
before the call to set_output_gamma at line 1048.

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>
Signed-off-by: Srinivasan Shanmugam <srinivasan.shanmugam@amd.com>
Reviewed-by: Tom Chung <chiahsuan.chung@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: use CPU for page table update if SDMA is unavailable

avoid using SDMA if it is unavailable.

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

drm/amdgpu: optimize the padding for gfx11

Adding NOP packets one by one in the ring
does not use the CP efficiently.

Solution:
Use CP optimization while adding NOP packet's so PFP
can discard NOP packets based on information of count
from the Header instead of fetching all NOP packets
one by one.

Reviewed-by: Christian König <christian.koenig@amd.com>
Cc: Christian König <christian.koenig@amd.com>
Cc: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Cc: Tvrtko Ursulin <tursulin@igalia.com>
Cc: Marek Olšák <marek.olsak@amd.com>
Signed-off-by: Sunil Khatri <sunil.khatri@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdkfd: fix debug watchpoints for logical devices

The number of watchpoints should be set and constrained per logical
partition device, not by the socket device.

Signed-off-by: Jonathan Kim <jonathan.kim@amd.com>
Reviewed-by: Harish Kasiviswanathan <harish.kasiviswanathan@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: do not call insert_nop fn for zero count

Do not make a function call for zero size NOP as it
does not add anything in the ring and is unnecessary
function call.

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

drm/amdkfd: support per-queue reset on gfx9

Support per-queue reset for GFX9.  The recommendation is for the driver
to target reset the HW queue via a SPI MMIO register write.

Since this requires pipe and HW queue info and MEC FW is limited to
doorbell reports of hung queues after an unmap failure, scan the HW
queue slots defined by SET_RESOURCES first to identify the user queue
candidates to reset.

Only signal reset events to processes that have had a queue reset.

If queue reset fails, fall back to GPU reset.

Signed-off-by: Jonathan Kim <jonathan.kim@amd.com>
Reviewed-by: Felix Kuehling <felix.kuehling@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: optimize the padding for gfx10

Adding NOP packets one by one in the ring
does not use the CP efficiently.

Solution:
Use CP optimization while adding NOP packet's so PFP
can discard NOP packets based on information of count
from the Header instead of fetching all NOP packets
one by one.

Cc: Christian König <christian.koenig@amd.com>
Cc: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Cc: Tvrtko Ursulin <tursulin@igalia.com>
Cc: Marek Olšák <marek.olsak@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Sunil Khatri <sunil.khatri@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Align 'dpp401_dscl_program_isharp' with actual function parameters

This commit corrects the function comment for
'dpp401_dscl_program_isharp' in 'dcn401_dpp_dscl.c'. The comment
previously included a description for a  non-existent parameter
'bs_coeffs_updated'. This parameter description has been removed to
reflect the function's actual parameters.

Fixes the below with gcc W=1:
drivers/gpu/drm/amd/amdgpu/../display/dc/dpp/dcn401/dcn401_dpp_dscl.c:981: warning: Excess function parameter 'bs_coeffs_updated' description in 'dpp401_dscl_program_isharp'

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>
Signed-off-by: Srinivasan Shanmugam <srinivasan.shanmugam@amd.com>
Reviewed-by: Tom Chung <chiahsuan.chung@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: Clean up the register dump via debugfs list

debugfs register list for dump is cleaned as it have
some issues related to proper power state of the IP
before register read.

Since the above mentioned is removed we no longer want
this to be dumped part of the devcoredump and hence
removed.

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

drm/amdgpu: Remove debugfs amdgpu_reset_dump_register_list

There are some problem with existing amdgpu_reset_dump_register_list
debugfs node. It is supposed to read a list of registers but there
could be cases when the IP is not in correct power state. Register
read in such cases could lead to more problems.

We are taking care of all such power states in devcoredump and
dumping the registers of need for debugging. So cleaning this code
and we dont need this functionality via debugfs anymore.

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

drm/amd/pm: add overdrive support on smu v14.0.2/3

add overdrive support on smu v14.0.2/3

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

drm/amd/display: Handle null 'stream_status' in 'planes_changed_for_existing_stream'

This commit adds a null check for 'stream_status' in the function
'planes_changed_for_existing_stream'. Previously, the code assumed
'stream_status' could be null, but did not handle the case where it was
actually null. This could lead to a null pointer dereference.

Reported by smatch:
drivers/gpu/drm/amd/amdgpu/../display/dc/core/dc_resource.c:3784 planes_changed_for_existing_stream() error: we previously assumed 'stream_status' could be null (see line 3774)

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>
Signed-off-by: Srinivasan Shanmugam <srinivasan.shanmugam@amd.com>
Reviewed-by: Tom Chung <chiahsuan.chung@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/pm: update powerplay structure on smu v14.0.2/3

update powerplay structure on smu v14.0.2/3

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

drm/amdkfd: Fix compile error if HMM support not enabled

Fixes the below if kernel config not enable HMM support

>> drivers/gpu/drm/amd/amdgpu/../amdkfd/kfd_queue.c:107:26: error:
implicit declaration of function 'svm_range_from_addr'

>> drivers/gpu/drm/amd/amdgpu/../amdkfd/kfd_queue.c:107:24: error:
assignment to 'struct svm_range *' from 'int' makes pointer from integer
without a cast [-Wint-conversion]

>> drivers/gpu/drm/amd/amdgpu/../amdkfd/kfd_queue.c:111:28: error:
invalid use of undefined type 'struct svm_range'

Fixes: b049504e211e ("drm/amdkfd: Validate user queue svm memory residency")
Reported-by: kernel test robot <lkp@intel.com>
Closes: https://lore.kernel.org/oe-kbuild-all/202407252127.zvnxaKRA-lkp@intel.com/
Signed-off-by: Philip Yang <Philip.Yang@amd.com>
Reviewed-by: Felix Kuehling <felix.kuehling@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: 3.2.294

This version brings along the following:

- SPL improvements.
- Address coverity issues.
- DML2 fixes.
- Code cleanup.
- DIO and DCCG refactor.
- Improve the PSR state.

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

drm/amd/display: Remove unused code

Remove function pointers that were never used.

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

drm/amd/display: Add missing registers for dcn32

Add missing debug registers for DCN32.

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

drm/amd/display: Add dcc propagation value

Initialize the field dcc_meta_propagation_delay_us with 10 ms.

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

drm/amd/display: Add missing mcache registers

Add missing register programming for mcache in DCN401.

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

drm/amd/display: Remove duplicated code

DCN_MINIMUM_DISPCLK_Khz and DCN_MINIMUM_DPPCLK_Khz is declared twice.
This commit removes that duplication.

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

drm/amd/display: Add missing DET segments programming

The commit 5034b935f62a ("drm/amd/display: Modify DHCUB waterwark
structures and functions") introduced a code refactor for DCHUB, but
during the merge process into amd-staging-drm-next, the program det
segments were removed. This commit adds the DET segment programming for
DCN35.

Fixes: 5034b935f62a ("drm/amd/display: Modify DHCUB waterwark structures and functions")
Reviewed-by: Aurabindo Pillai <aurabindo.pillai@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Rodrigo Siqueira <Rodrigo.Siqueira@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: For FAMS2 don't program P-State force from driver

P-State force programming is handled entirely by FW in FAMS2.  Remove
any programming from driver side to prevent incorrect programming from
driver side (which may override FW programming)

Signed-off-by: Alvin Lee <Alvin.Lee2@amd.com>
Reviewed-by: Rodrigo Siqueira <Rodrigo.Siqueira@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Rodrigo Siqueira <Rodrigo.Siqueira@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Fix Cursor Offset in Scaled Scenarios

[WHY]
Cursor position code had improper offsets in scaled modes.

[HOW]
Adjust cursor scaling to account for cursor offsets properly.

Reviewed-by: Rodrigo Siqueira <Rodrigo.Siqueira@amd.com>
Signed-off-by: Sung Lee <sunglee@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Rodrigo Siqueira <Rodrigo.Siqueira@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Replace dm_execute_dmub_cmd with dc_wake_and_execute_dmub_cmd

In the commit c2cec7a872b6 ("drm/amd/display: Wake DMCUB before sending
a command for replay feature"), replaced dm_execute_dmub_cmd with
dc_wake_and_execute_dmub_cmd in multiple areas, but due to merge issues
the replacement of this function in the dmub_replay_copy_settings was
missed. This commit replaces the old dm_execute_dmub_cmd with
dc_wake_and_execute_dmub_cmd.

Fixes: 3601a35a2e9d ("drm/amd/display: Wake DMCUB before sending a command for replay feature")
Reviewed-by: Aurabindo Pillai <aurabindo.pillai@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Rodrigo Siqueira <Rodrigo.Siqueira@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Setup two pixel per container

SPL has a control field for controlling the two pixels per container
that is not in use yet. This commit adds a proper initialization for
this feature.

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

drm/amd/display: Re-order enum in a header file

Move the lb_memory_config close to the pixel format enums to improve the
code readability.

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

drm/amd/display: Get link index for AUX reply notification

The link index wasn't updated for the AUX reply notification. Get link
index based on DPIA instance for AUX reply notification.

Signed-off-by: Cruise <Cruise.Hung@amd.com>
Reviewed-by: Meenakshikumar Somasundaram <meenakshikumar.somasundaram@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Rodrigo Siqueira <rodrigo.siqueira@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Check UnboundedRequestEnabled's value

CalculateSwathAndDETConfiguration_params_st's UnboundedRequestEnabled is
a pointer (i.e. dml_bool_t *UnboundedRequestEnabled), and thus
p->UnboundedRequestEnabled checks its address, not bool value.

To check value, *p->UnboundedRequestEnabled is used instead.

This fixes 1 REVERSE_INULL issue reported by Coverity.

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

drm/amd/display: Underflow Seen on DCN401 eGPU

[WHY]
In dcn401 we read clock values before FW is loaded. These incorrect
values cause the driver to believe that we are running higher clocks
than what we actually have. This then causes corruption/underflow for
the eGPU.

[HOW]
When new values are read from HW, update internal structures to
propagate the new/correct value. Fixes issue

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

drm/amd/display: Enable aux transfer path via dmub for dp tunneling

[Why]
Aux transfer retries path does not support dp tunneling.

[How]
Based on ddc pin check, aux will be issued in legacy path or DMUB.

Signed-off-by: Meenakshikumar Somasundaram <meenakshikumar.somasundaram@amd.com>
Reviewed-by: Eric Yang <eric.yang@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@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 enable and disable functions for DCN35

Add new enable and disable functions based on DCCG spec.

Signed-off-by: Hansen Dsouza <Hansen.Dsouza@amd.com>
Reviewed-by: Muhammad Ahmed <ahmed.ahmed@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Rodrigo Siqueira <rodrigo.siqueira@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Use gpuvm_min_page_size_kbytes for DML2 surfaces

[Why]
It's currently hard coded to 256 when it should be using the SOC
provided values. This can result in corruption with linear surfaces
where we prefetch more PTE than the buffer can hold.

[How]
Update the min page size correctly for the plane.

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

drm/amd/display: Initialize get_bytes_per_element's default to 1

Variables, used as denominators and maybe not assigned to other values,
should not be 0. bytes_per_element_y & bytes_per_element_c are
initialized by get_bytes_per_element() which should never return 0.

This fixes 10 DIVIDE_BY_ZERO issues reported by Coverity.

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

drm/amd/display: Check null values from functions

Functions get_per_method_common_meta and get_expanded_strategy_list can
return null and thus it is necessary to check their returned values
before dereferencing.

This fixes 3 NULL_RETURNS issues reported by Coverity.

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

drm/amd/display: Add logs for debugging outbox

The DP tunnel AUX reply is received through Outbox1. Print the Outbox1
status if an issue occurs.

Signed-off-by: Cruise <Cruise.Hung@amd.com>
Reviewed-by: Nicholas Kazlauskas <nicholas.kazlauskas@amd.com>
Reviewed-by: Wayne Lin <wayne.lin@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@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 enable and disable functions

Add new enable and disable functions based on DCCG spec.

Signed-off-by: Hansen Dsouza <Hansen.Dsouza@amd.com>
Reviewed-by: Muhammad Ahmed <ahmed.ahmed@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Rodrigo Siqueira <rodrigo.siqueira@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/radeon: convert bios_hardcoded_edid to drm_edid

Instead of manually passing around 'struct edid *' and its size,
use 'struct drm_edid', which encapsulates a validated combination of
both.

As the drm_edid_ can handle NULL gracefully, the explicit checks can be
dropped.

Also save a few characters by transforming '&array[0]' to the equivalent
'array' and using 'max_t(int, ...)' instead of manual casts.

Signed-off-by: Thomas Weißschuh <linux@weissschuh.net>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: convert bios_hardcoded_edid to drm_edid

Instead of manually passing around 'struct edid *' and its size,
use 'struct drm_edid', which encapsulates a validated combination of
both.

As the drm_edid_ can handle NULL gracefully, the explicit checks can be
dropped.

Also save a few characters by transforming '&array[0]' to the equivalent
'array' and using 'max_t(int, ...)' instead of manual casts.

Signed-off-by: Thomas Weißschuh <linux@weissschuh.net>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: Fix APU handling in amdgpu_pm_load_smu_firmware()

We only need to skip this on modern APUs.  It's required
on older APUs as it's where start_smu gets called from.

Closes: https://gitlab.freedesktop.org/drm/amd/-/issues/3502
Fixes: 064d92436b69 ("drm/amd/pm: avoid to load smu firmware for APUs")
Reviewed-by: Tim Huang <tim.huang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Cc: Tim Huang <Tim.Huang@amd.com>

drm/amdgpu: trigger ip dump before suspend of IP's

Problem:
IP dump right now is done post suspend of all
IP's which for some IP's could change power
state and software state too which we do not want
to reflect in the dump as it might not be same at
the time of hang.

Solution:
IP should be dumped as close to the HW state when
the GPU was in hung state without trying to reinitialize
any resource.

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

drm/amd/display: Avoid overflow assignment in link_dp_cts

sampling_rate is an uint8_t but is assigned an unsigned int, and thus it
can overflow. As a result, sampling_rate is changed to uint32_t.

Similarly, LINK_QUAL_PATTERN_SET has a size of 2 bits, and it should
only be assigned to a value less or equal than 4.

This fixes 2 INTEGER_OVERFLOW issues reported by Coverity.

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

drm/amd/display: Use correct cm_helper function

Need to use cm3_helper function with DCN401 to avoid cases where high
RGB component values can get set to zero if using the TF curve generated
by cm_helper.

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

drm/amd/display: Add seamless boot support for more DIG operation modes

[Why]
When pre-OS firmware enables display support for displays that operate
the DIG in 2 pixels per cycle processing modes the inferred pixel rate
from get_pixel_clk_frequency_100hz does not account for the true pixel
rate since we're outputting 2 per cycle past the stream encoder.

This causes seamless boot validation to abort early.

[How]
Add a new stream encoder function for getting pixels per cycle from the
stream encoder. If the pixels per cycle is greater than 1 and the driver
policy is to enable 2 pixels per cycle for post-OS then allow seamless
boot to continue.

Signed-off-by: Nicholas Kazlauskas <nicholas.kazlauskas@amd.com>
Reviewed-by: Duncan Ma <duncan.ma@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Rodrigo Siqueira <rodrigo.siqueira@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Reset VRR config during resume

[Why]
After resume the system, the new_crtc_state->vrr_infopacket does not
synchronize with the current state.  It will affect the
update_freesync_state_on_stream() does not update the state correctly.

The previous patch causes a PSR SU regression that cannot let panel go
into self-refresh mode.

[How]
Reset the VRR config during resume to force update the VRR config later.

Fixes: eb6dfbb7a9c6 ("drm/amd/display: Reset freesync config before update new state")
Signed-off-by: Tom Chung <chiahsuan.chung@amd.com>
Reviewed-by: Sun peng Li <sunpeng.li@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Rodrigo Siqueira <rodrigo.siqueira@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Add a missing PSR state

[Why & How]
Add a missing PSR state to make the dmub_psr_get_state() return a
correct PSR state.

Signed-off-by: Tom Chung <chiahsuan.chung@amd.com>
Reviewed-by: Sun peng Li <sunpeng.li@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Rodrigo Siqueira <rodrigo.siqueira@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: sync dmub output event type.

[why]
dmubfw added a new event type, update amdgpu to avoid "notify type 6
invalid"

Signed-off-by: Charlene Liu <Charlene.Liu@amd.com>
Reviewed-by: Chris Park <chris.park@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Rodrigo Siqueira <rodrigo.siqueira@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: restore immediate_disable_crtc for w/a

[why]
immediate_disable_crtc does not reset ODM.  if switching to disable_crtc
which will disable ODM as well.  i.e. need to restore ODM mem cfg at
reenable it at end of w/a.

Signed-off-by: Charlene Liu <Charlene.Liu@amd.com>
Reviewed-by: Xi (Alex) Liu <xi.liu@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Rodrigo Siqueira <rodrigo.siqueira@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: increase mes log buffer size for gfx12

MES firmware requires larger log buffer for gfx12. Allocate
proper buffer respectively for gfx11 and gfx12.

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

drm/amd/display: Check stream_status before it is used

[WHAT & HOW]
dc_state_get_stream_status can return null, and therefore null must be
checked before stream_status is used.

This fixes 1 NULL_RETURNS issue reported by Coverity.

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

drm/amd/display: Check null pointers before using them

[WHAT & HOW]
dc_link is null checked previously in the same function, indicating it
might be null as reported by Coverity.

This fixes 1 FORWARD_NULL issue reported by Coverity.

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

drm/amd/display: Fix possible overflow in integer multiplication

[WHAT & HOW]
Integer multiplies integer may overflow in context that expects an
expression of unsigned long long (64 bits). This can be fixed by casting
integer to unsigned long long to force 64 bits results.

This fixes 2 OVERFLOW_BEFORE_WIDEN issues reported by Coverity.

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

drm/amd/display: Add option to disable unbounded req in DML21

Use debug option for disabling unbounded req in DML21

Signed-off-by: Alvin Lee <Alvin.Lee2@amd.com>
Reviewed-by: Austin Zheng <Austin.Zheng@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Rodrigo Siqueira <rodrigo.siqueira@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Refactor for dio

Moved files to respective folders to improve DIO code.

Signed-off-by: Bhuvanachandra Pinninti <bpinnint@amd.com>
Reviewed-by: Martin Leung <martin.leung@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Rodrigo Siqueira <rodrigo.siqueira@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Request 0MHz dispclk for zero display case

[Why]
If we aren't entering RCG/IPS2 or CLKSTOP is not supported by PMFW then
we should be requesting a dispclk value of 0MHz to PMFW.

Currenly we run at max clock since there's an assumption in APU clock
table formulation where we can run at any DISPCLK at any state so the
real clock value ends up as 1200Mhz - the maximum.

[How]
Set to 0 instead of the minimum value in the state array.

Signed-off-by: Nicholas Kazlauskas <nicholas.kazlauskas@amd.com>
Reviewed-by: Duncan Ma <duncan.ma@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Rodrigo Siqueira <rodrigo.siqueira@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: print VCN instance dump for valid instance

VCN dump is dependent on power state of the ip. Dump is
valid if VCN was powered up at the time of ip dump.

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

drm/amd/display: Add two dmmuy I2C entry for GPIO port mapping issue

[Why]
When only 4 I2C is declared, two dummies are required to correctly map
GPIO port.

[How]
Add one more I2C dummy entry to match GPIO port.

Signed-off-by: Chris Park <chris.park@amd.com>
Reviewed-by: Alvin Lee <alvin.lee2@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Rodrigo Siqueira <rodrigo.siqueira@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Run idle optimizations at end of vblank handler

[Why & How]
1. After allowing idle optimizations, hw programming is disallowed.
2. Before hw programming, we need to disallow idle optimizations.

Otherwise, in scenario 1, we will immediately kick hw out of idle
optimizations with register access.

Scenario 2 is less of a concern, since any register access will kick
hw out of idle optimizations. But we'll do it early for correctness.

Signed-off-by: Leo Li <sunpeng.li@amd.com>
Reviewed-by: Aurabindo Pillai <aurabindo.pillai@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Rodrigo Siqueira <rodrigo.siqueira@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Let drm_crtc_vblank_on/off manage interrupts

[Why]
We manage interrupts for CRTCs in two places:
1. In manage_dm_interrupts(), when CRTC get enabled or disabled
2. When drm_vblank_get/put() starts or kills the vblank counter, calling
   into amdgpu_dm_crtc_set_vblank()

The interrupts managed by these twp places should be identical.

[How]
Since manage_dm_interrupts() already use drm_crtc_vblank_on/off(), just
move all CRTC interrupt management into amdgpu_dm_crtc_set_vblank().

This has the added benefit of disabling all CRTC and HUBP interrupts
when there are no vblank requestors.

Note that there is a TODO item - unchanged from when it was first
introduced - to properly identify the HUBP instance from the OTG
instance, rather than just assume direct mapping.

Signed-off-by: Leo Li <sunpeng.li@amd.com>
Reviewed-by: Aurabindo Pillai <aurabindo.pillai@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Rodrigo Siqueira <rodrigo.siqueira@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: roll back quality EASF and ISHARP and dc dependency changes

[Why]
Seeing several regressions related to quality EASF and ISHARP changes
and removing dc dependency changes.

[How]
Roll back SPL changes

Signed-off-by: Samson Tam <Samson.Tam@amd.com>
Reviewed-by: Martin Leung <martin.leung@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Rodrigo Siqueira <rodrigo.siqueira@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdkfd: Fix missing error code in kfd_queue_acquire_buffers

The fix involves setting 'err' to '-EINVAL' before each 'goto
out_err_unreserve'.

Fixes the below:
drivers/gpu/drm/amd/amdgpu/../amdkfd/kfd_queue.c:265 kfd_queue_acquire_buffers()
warn: missing error code 'err'

drivers/gpu/drm/amd/amdgpu/../amdkfd/kfd_queue.c
    226 int kfd_queue_acquire_buffers(struct kfd_process_device *pdd, struct queue_properties *properties)
    227 {
    228         struct kfd_topology_device *topo_dev;
    229         struct amdgpu_vm *vm;
    230         u32 total_cwsr_size;
    231         int err;
    232
    233         topo_dev = kfd_topology_device_by_id(pdd->dev->id);
    234         if (!topo_dev)
    235                 return -EINVAL;
    236
    237         vm = drm_priv_to_vm(pdd->drm_priv);
    238         err = amdgpu_bo_reserve(vm->root.bo, false);
    239         if (err)
    240                 return err;
    241
    242         err = kfd_queue_buffer_get(vm, properties->write_ptr, &properties->wptr_bo, PAGE_SIZE);
    243         if (err)
    244                 goto out_err_unreserve;
    245
    246         err = kfd_queue_buffer_get(vm, properties->read_ptr, &properties->rptr_bo, PAGE_SIZE);
    247         if (err)
    248                 goto out_err_unreserve;
    249
    250         err = kfd_queue_buffer_get(vm, (void *)properties->queue_address,
    251                                    &properties->ring_bo, properties->queue_size);
    252         if (err)
    253                 goto out_err_unreserve;
    254
    255         /* only compute queue requires EOP buffer and CWSR area */
    256         if (properties->type != KFD_QUEUE_TYPE_COMPUTE)
    257                 goto out_unreserve;

This is clearly a success path.

    258
    259         /* EOP buffer is not required for all ASICs */
    260         if (properties->eop_ring_buffer_address) {
    261                 if (properties->eop_ring_buffer_size != topo_dev->node_props.eop_buffer_size) {
    262                         pr_debug("queue eop bo size 0x%lx not equal to node eop buf size 0x%x\n",
    263                                 properties->eop_buf_bo->tbo.base.size,
    264                                 topo_dev->node_props.eop_buffer_size);
--> 265                         goto out_err_unreserve;

This has err in the label name.  err = -EINVAL?

    266                 }
    267                 err = kfd_queue_buffer_get(vm, (void *)properties->eop_ring_buffer_address,
    268                                            &properties->eop_buf_bo,
    269                                            properties->eop_ring_buffer_size);
    270                 if (err)
    271                         goto out_err_unreserve;
    272         }
    273
    274         if (properties->ctl_stack_size != topo_dev->node_props.ctl_stack_size) {
    275                 pr_debug("queue ctl stack size 0x%x not equal to node ctl stack size 0x%x\n",
    276                         properties->ctl_stack_size,
    277                         topo_dev->node_props.ctl_stack_size);
    278                 goto out_err_unreserve;

err?

    279         }
    280
    281         if (properties->ctx_save_restore_area_size != topo_dev->node_props.cwsr_size) {
    282                 pr_debug("queue cwsr size 0x%x not equal to node cwsr size 0x%x\n",
    283                         properties->ctx_save_restore_area_size,
    284                         topo_dev->node_props.cwsr_size);
    285                 goto out_err_unreserve;

err?  Not sure.

    286         }
    287
    288         total_cwsr_size = (topo_dev->node_props.cwsr_size + topo_dev->node_props.debug_memory_size)
    289                           * NUM_XCC(pdd->dev->xcc_mask);
    290         total_cwsr_size = ALIGN(total_cwsr_size, PAGE_SIZE);
    291
    292         err = kfd_queue_buffer_get(vm, (void *)properties->ctx_save_restore_area_address,
    293                                    &properties->cwsr_bo, total_cwsr_size);
    294         if (!err)
    295                 goto out_unreserve;
    296
    297         amdgpu_bo_unreserve(vm->root.bo);
    298
    299         err = kfd_queue_buffer_svm_get(pdd, properties->ctx_save_restore_area_address,
    300                                        total_cwsr_size);
    301         if (err)
    302                 goto out_err_release;
    303
    304         return 0;
    305
    306 out_unreserve:
    307         amdgpu_bo_unreserve(vm->root.bo);
    308         return 0;
    309
    310 out_err_unreserve:
    311         amdgpu_bo_unreserve(vm->root.bo);
    312 out_err_release:
    313         kfd_queue_release_buffers(pdd, properties);
    314         return err;
    315 }

Fixes: 629568d25fea ("drm/amdkfd: Validate queue cwsr area and eop buffer size")
Reported-by: Dan Carpenter <dan.carpenter@linaro.org>
Cc: Philip Yang <Philip.Yang@amd.com>
Cc: Felix Kuehling <Felix.Kuehling@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: Philip Yang <Philip.Yang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Add null check for top_pipe_to_program in commit_planes_for_stream

This commit addresses a null pointer dereference issue in the
`commit_planes_for_stream` function at line 4140. The issue could occur
when `top_pipe_to_program` is null.

The fix adds a check to ensure `top_pipe_to_program` is not null before
accessing its stream_res. This prevents a null pointer dereference.

Reported by smatch:
drivers/gpu/drm/amd/amdgpu/../display/dc/core/dc.c:4140 commit_planes_for_stream() error: we previously assumed 'top_pipe_to_program' could be null (see line 3906)

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>
Signed-off-by: Srinivasan Shanmugam <srinivasan.shanmugam@amd.com>
Reviewed-by: Tom Chung <chiahsuan.chung@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Add null check for pipe_ctx->plane_state in dcn20_program_pipe

This commit addresses a null pointer dereference issue in the
`dcn20_program_pipe` function. The issue could occur when
`pipe_ctx->plane_state` is null.

The fix adds a check to ensure `pipe_ctx->plane_state` is not null
before accessing. This prevents a null pointer dereference.

Reported by smatch:
drivers/gpu/drm/amd/amdgpu/../display/dc/hwss/dcn20/dcn20_hwseq.c:1925 dcn20_program_pipe() error: we previously assumed 'pipe_ctx->plane_state' could be null (see line 1877)

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>
Signed-off-by: Srinivasan Shanmugam <srinivasan.shanmugam@amd.com>
Reviewed-by: Tom Chung <chiahsuan.chung@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: Add MFD support for ISP I2C bus

ISP I2C bus device can't be enumerated via ACPI mechanism
since it shares the memory map with the AMDGPU.
So use the MFD mechanism for registering the ISP I2C device
and add the required resources.

Signed-off-by: Venkata Narendra Kumar Gutta <vengutta@amd.com>
Acked-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: fix contiguous handling for IB parsing v2

Otherwise we won't get correct access to the IB.

v2: keep setting AMDGPU_GEM_CREATE_VRAM_CONTIGUOUS to avoid problems in
    the VRAM backend.

Signed-off-by: Christian König <christian.koenig@amd.com>
Closes: https://gitlab.freedesktop.org/drm/amd/-/issues/3501
Fixes: e362b7c8f8c7 ("drm/amdgpu: Modify the contiguous flags behaviour")
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Cc: stable@vger.kernel.org
Tested-by: Dave Airlie <airlied@redhat.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: add print support for vcn_v3_0 ip dump

Add support for logging the registers in devcoredump
buffer for vcn_v3_0.

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

drm/amdgpu: add vcn_v3_0 ip dump support

Add support of vcn ip dump in the devcoredump
for vcn_v3_0.

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

drm/amdgpu: add macro to calculate offset with instance

Add macro definition which calculate offset of the
register with index override.

This is useful in case when there is an array of
registers which is common for all instances.
To read registers in that case it is easy to define
registers once and the index value is manually passed
to calculate proper offset of register for each instance.

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

drm/amdgpu: add vcn ip dump ptr in vcn global struct

Add pointer to the vcn ip dump in the vcn global structure
to be accessible for all vcn version via global adev.

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

drm/amd/display: remove unneeded semicolon

No functional modification involved.

./drivers/gpu/drm/amd/display/dc/dml2/dml21/src/dml2_core/dml2_core_dcn4_calcs.c:481:2-3: Unneeded semicolon.
./drivers/gpu/drm/amd/display/dc/dml2/dml21/src/dml2_core/dml2_core_dcn4_calcs.c:3783:168-169: Unneeded semicolon.
./drivers/gpu/drm/amd/display/dc/dml2/dml21/src/dml2_core/dml2_core_dcn4_calcs.c:3782:166-167: Unneeded semicolon.

Reported-by: Abaci Robot <abaci@linux.alibaba.com>
Closes: https://bugzilla.openanolis.cn/show_bug.cgi?id=9575
Signed-off-by: Jiapeng Chong <jiapeng.chong@linux.alibaba.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdkfd: allow users to target recommended SDMA engines

Certain GPUs have better copy performance over xGMI on specific
SDMA engines depending on the source and destination GPU.
Allow users to create SDMA queues on these recommended engines.
Close to 2x overall performance has been observed with this
optimization.

Signed-off-by: Jonathan Kim <jonathan.kim@amd.com>
Reviewed-by: Felix Kuehling <felix.kuehling@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu/pm: support gpu_metrics sysfs interface for smu v14.0.2/3

support gpu_metrics sysfs interface for smu v14.0.2/3

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

drm/amd/display: use swap() in sort()

Use existing swap() function rather than duplicating its implementation.

./drivers/gpu/drm/amd/display/dc/dml2/dml21/src/dml2_pmo/dml2_pmo_dcn3.c:17:29-30: WARNING opportunity for swap().

Reported-by: Abaci Robot <abaci@linux.alibaba.com>
Closes: https://bugzilla.openanolis.cn/show_bug.cgi?id=9573
Signed-off-by: Jiapeng Chong <jiapeng.chong@linux.alibaba.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Reapply 2fde4fdddc1f

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.

Fixes: 2563391e57b5 ("drm/amd/display: DML2.1 resynchronization")
Signed-off-by: Nathan Chancellor <nathan@kernel.org>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Fix spelling mistake "tolarance" -> "tolerance"

There is a spelling mistake in a dml2_printf message. Fix it.

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

drm/radeon: properly handle vbios fake edid sizing

The comment in the vbios structure says:
// = 128 means EDID length is 128 bytes, otherwise the EDID length = ucFakeEDIDLength*128

This fake edid struct has not been used in a long time, so I'm
not sure if there were actually any boards out there with a non-128 byte
EDID, but align the code with the comment.

Reviewed-by: Thomas Weißschuh <linux@weissschuh.net>
Reported-by: Thomas Weißschuh <linux@weissschuh.net>
Link: https://lists.freedesktop.org/archives/amd-gfx/2024-June/109964.html
Fixes: c324acd5032f ("drm/radeon/kms: parse the extended LCD info block")
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: properly handle vbios fake edid sizing

The comment in the vbios structure says:
// = 128 means EDID length is 128 bytes, otherwise the EDID length = ucFakeEDIDLength*128

This fake edid struct has not been used in a long time, so I'm
not sure if there were actually any boards out there with a non-128 byte
EDID, but align the code with the comment.

Reviewed-by: Thomas Weißschuh <linux@weissschuh.net>
Reported-by: Thomas Weißschuh <linux@weissschuh.net>
Link: https://lists.freedesktop.org/archives/amd-gfx/2024-June/109964.html
Fixes: d38ceaf99ed0 ("drm/amdgpu: add core driver (v4)")
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdkfd: Validate queue cwsr area and eop buffer size

When creating KFD user compute queue, check if queue eop buffer size,
cwsr area size, ctl stack size equal to the size of KFD node
properities.

Check the entire cwsr area which may split into multiple svm ranges
aligned to granularity boundary.

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

drm/amdkfd: Store queue cwsr area size to node properties

Use the queue eop buffer size, cwsr area size, ctl stack size
calculation from Thunk, store the value to KFD node properties.

Those will be used to validate queue eop buffer size, cwsr area size,
ctl stack size when creating KFD user compute queue.

Those will be exposed to user space via sysfs KFD node properties, to
remove the duplicate calculation code from Thunk.

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

drm/amdgpu: reset vm state machine after gpu reset(vram lost)

[Why]
Page table of compute VM in the VRAM will lost after gpu reset.
VRAM won't be restored since compute VM has no shadows.

[How]
Use higher 32-bit of vm->generation to record a vram_lost_counter.
Reset the VM state machine when vm->genertaion is not equal to
the new generation token.

v2: Check vm->generation instead of calling drm_sched_entity_error
in amdgpu_vm_validate.
v3: Use new generation token instead of vram_lost_counter for check.

Signed-off-by: ZhenGuo Yin <zhenguo.yin@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Add null check for set_output_gamma in dcn30_set_output_transfer_func

This commit adds a null check for the set_output_gamma function pointer
in the  dcn30_set_output_transfer_func function. Previously,
set_output_gamma was being checked for nullity at line 386, but then it
was being dereferenced without any nullity check at line 401. This
could potentially lead to a null pointer dereference error if
set_output_gamma is indeed null.

To fix this, we now ensure that set_output_gamma is not null before
dereferencing it. We do this by adding a nullity check for
set_output_gamma before the call to set_output_gamma at line 401. If
set_output_gamma is null, we log an error message and do not call the
function.

This fix prevents a potential null pointer dereference error.

drivers/gpu/drm/amd/amdgpu/../display/dc/hwss/dcn30/dcn30_hwseq.c:401 dcn30_set_output_transfer_func()
error: we previously assumed 'mpc->funcs->set_output_gamma' could be null (see line 386)

drivers/gpu/drm/amd/amdgpu/../display/dc/hwss/dcn30/dcn30_hwseq.c
    373 bool dcn30_set_output_transfer_func(struct dc *dc,
    374                                 struct pipe_ctx *pipe_ctx,
    375                                 const struct dc_stream_state *stream)
    376 {
    377         int mpcc_id = pipe_ctx->plane_res.hubp->inst;
    378         struct mpc *mpc = pipe_ctx->stream_res.opp->ctx->dc->res_pool->mpc;
    379         const struct pwl_params *params = NULL;
    380         bool ret = false;
    381
    382         /* program OGAM or 3DLUT only for the top pipe*/
    383         if (pipe_ctx->top_pipe == NULL) {
    384                 /*program rmu shaper and 3dlut in MPC*/
    385                 ret = dcn30_set_mpc_shaper_3dlut(pipe_ctx, stream);
    386                 if (ret == false && mpc->funcs->set_output_gamma) {
                                            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^ If this is NULL

    387                         if (stream->out_transfer_func.type == TF_TYPE_HWPWL)
    388                                 params = &stream->out_transfer_func.pwl;
    389                         else if (pipe_ctx->stream->out_transfer_func.type ==
    390                                         TF_TYPE_DISTRIBUTED_POINTS &&
    391                                         cm3_helper_translate_curve_to_hw_format(
    392                                         &stream->out_transfer_func,
    393                                         &mpc->blender_params, false))
    394                                 params = &mpc->blender_params;
    395                          /* there are no ROM LUTs in OUTGAM */
    396                         if (stream->out_transfer_func.type == TF_TYPE_PREDEFINED)
    397                                 BREAK_TO_DEBUGGER();
    398                 }
    399         }
    400
--> 401         mpc->funcs->set_output_gamma(mpc, mpcc_id, params);
                ^^^^^^^^^^^^^^^^^^^^^^^^^^^^ Then it will crash

    402         return ret;
    403 }

Fixes: d99f13878d6f ("drm/amd/display: Add DCN3 HWSEQ")
Reported-by: Dan Carpenter <dan.carpenter@linaro.org>
Cc: Tom Chung <chiahsuan.chung@amd.com>
Cc: Rodrigo Siqueira <Rodrigo.Siqueira@amd.com>
Cc: Roman Li <roman.li@amd.com>
Cc: Hersen Wu <hersenxs.wu@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>
Signed-off-by: Srinivasan Shanmugam <srinivasan.shanmugam@amd.com>
Reviewed-by: Tom Chung <chiahsuan.chung@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: add missed harvest check for VCN IP v4/v5

To prevent below probe failure, add a check for models with VCN
IP v4.0.6 where VCN1 may be harvested.

v2:
Apply the same check to VCN IP v4.0 and v5.0.

[   54.070117] RIP: 0010:vcn_v4_0_5_start_dpg_mode+0x9be/0x36b0 [amdgpu]
[   54.071055] Code: 80 fb ff 8d 82 00 80 fe ff 81 fe 00 06 00 00 0f 43
c2 49 69 d5 38 0d 00 00 48 8d 71 04 c1 e8 02 4c 01 f2 48 89 b2 50 f6 02
00 <89> 01 48 8b 82 50 f6 02 00 48 8d 48 04 48 89 8a 50 f6 02 00 c7 00
[   54.072408] RSP: 0018:ffffb17985f736f8 EFLAGS: 00010286
[   54.072793] RAX: 00000000000000d6 RBX: ffff99a82f680000 RCX:
0000000000000000
[   54.073315] RDX: ffff99a82f680000 RSI: 0000000000000004 RDI:
ffff99a82f680000
[   54.073835] RBP: ffffb17985f73730 R08: 0000000000000001 R09:
0000000000000000
[   54.074353] R10: 0000000000000008 R11: ffffb17983c05000 R12:
0000000000000000
[   54.074879] R13: 0000000000000000 R14: ffff99a82f680000 R15:
0000000000000001
[   54.075400] FS:  00007f8d9c79a000(0000) GS:ffff99ab2f140000(0000)
knlGS:0000000000000000
[   54.075988] CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
[   54.076408] CR2: 0000000000000000 CR3: 0000000140c3a000 CR4:
0000000000750ef0
[   54.076927] PKRU: 55555554
[   54.077132] Call Trace:
[   54.077319]  <TASK>
[   54.077484]  ? show_regs+0x69/0x80
[   54.077747]  ? __die+0x28/0x70
[   54.077979]  ? page_fault_oops+0x180/0x4b0
[   54.078286]  ? do_user_addr_fault+0x2d2/0x680
[   54.078610]  ? exc_page_fault+0x84/0x190
[   54.078910]  ? asm_exc_page_fault+0x2b/0x30
[   54.079224]  ? vcn_v4_0_5_start_dpg_mode+0x9be/0x36b0 [amdgpu]
[   54.079941]  ? vcn_v4_0_5_start_dpg_mode+0xe6/0x36b0 [amdgpu]
[   54.080617]  vcn_v4_0_5_set_powergating_state+0x82/0x19b0 [amdgpu]
[   54.081316]  amdgpu_device_ip_set_powergating_state+0x64/0xc0
[amdgpu]
[   54.082057]  amdgpu_vcn_ring_begin_use+0x6f/0x1d0 [amdgpu]
[   54.082727]  amdgpu_ring_alloc+0x44/0x70 [amdgpu]
[   54.083351]  amdgpu_vcn_dec_sw_ring_test_ring+0x40/0x110 [amdgpu]
[   54.084054]  amdgpu_ring_test_helper+0x22/0x90 [amdgpu]
[   54.084698]  vcn_v4_0_5_hw_init+0x87/0xc0 [amdgpu]
[   54.085307]  amdgpu_device_init+0x1f96/0x2780 [amdgpu]
[   54.085951]  amdgpu_driver_load_kms+0x1e/0xc0 [amdgpu]
[   54.086591]  amdgpu_pci_probe+0x19f/0x550 [amdgpu]
[   54.087215]  local_pci_probe+0x48/0xa0
[   54.087509]  pci_device_probe+0xc9/0x250
[   54.087812]  really_probe+0x1a4/0x3f0
[   54.088101]  __driver_probe_device+0x7d/0x170
[   54.088443]  driver_probe_device+0x24/0xa0
[   54.088765]  __driver_attach+0xdd/0x1d0
[   54.089068]  ? __pfx___driver_attach+0x10/0x10
[   54.089417]  bus_for_each_dev+0x8e/0xe0
[   54.089718]  driver_attach+0x22/0x30
[   54.090000]  bus_add_driver+0x120/0x220
[   54.090303]  driver_register+0x62/0x120
[   54.090606]  ? __pfx_amdgpu_init+0x10/0x10 [amdgpu]
[   54.091255]  __pci_register_driver+0x62/0x70
[   54.091593]  amdgpu_init+0x67/0xff0 [amdgpu]
[   54.092190]  do_one_initcall+0x5f/0x330
[   54.092495]  do_init_module+0x68/0x240
[   54.092794]  load_module+0x201c/0x2110
[   54.093093]  init_module_from_file+0x97/0xd0
[   54.093428]  ? init_module_from_file+0x97/0xd0
[   54.093777]  idempotent_init_module+0x11c/0x2a0
[   54.094134]  __x64_sys_finit_module+0x64/0xc0
[   54.094476]  do_syscall_64+0x58/0x120
[   54.094767]  entry_SYSCALL_64_after_hwframe+0x6e/0x76

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

drm/amdgpu: skip kfd init if GFX is not ready.

avoid kfd init crash in that case.

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

drm/amdkfd: Validate user queue update

Ensure update queue new ring buffer is mapped on GPU with correct size.

Decrease queue old ring_bo queue_refcount and increase new ring_bo
queue_refcount.

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

drm/amdkfd: Validate user queue svm memory residency

Queue CWSR area maybe registered to GPU as svm memory, create queue to
ensure svm mapped to GPU with KFD_IOCTL_SVM_FLAG_GPU_ALWAYS_MAPPED flag.

Add queue_refcount to struct svm_range, to track queue CWSR area usage.

Because unmap mmu notifier callback return value is ignored, if
application unmap the CWSR area while queue is active, pr_warn message
in dmesg log. To be safe, evict user queue.

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

drm/amdgpu/gfx9.4.3: Enable bad opcode interrupt

For the bad opcode case, it will cause CP/ME hang.
The firmware will prevent the ME side from hanging by raising a bad opcode interrupt.
And the driver needs to perform a vmid reset when receiving the interrupt.

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

drm/amdgpu/gfx9: Enable bad opcode interrupt

For the bad opcode case, it will cause CP/ME hang.
The firmware will prevent the ME side from hanging by raising a bad opcode interrupt.
And the driver needs to perform a vmid reset when receiving the interrupt.

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

drm/amdgpu/gfx12: Enable bad opcode interrupt

For the bad opcode case, it will cause CP/ME hang.
The firmware will prevent the ME side from hanging by raising a bad opcode interrupt.
And the driver needs to perform a vmid reset when receiving the interrupt.

v2: update irq naming (drop priv) (Alex)

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

drm/amdgpu/gfx10: Enable bad opcode interrupt

For the bad opcode case, it will cause CP/ME hang.
The firmware will prevent the ME side from hanging by raising a bad opcode interrupt.
And the driver needs to perform a vmid reset when receiving the interrupt.

v2: update irq naming (drop priv) (Alex)

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

drm/amdgpu/gfx11: Enable bad opcode interrupt

For the bad opcode case, it will cause CP/ME hang.
The firmware will prevent the ME side from hanging by raising a bad opcode interrupt.
And the driver needs to perform a vmid reset when receiving the interrupt.

v2: update irq naming (drop priv) (Alex)

Acked-by: Felix Kuehling <felix.kuehling@amd.com>
Signed-off-by: Jesse Zhang <Jesse.Zhang@amd.com>
Reviewed-by: Prike Liang <Prike.Liang@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu/gfx: add bad opcode interrupt

Add the irq source for bad opcodes.

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

drm/amdgpu/gfx9: properly handle error ints on all pipes

Need to handle the interrupt enables for all pipes.

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

drm/amdgpu/gfx12: properly handle error ints on all pipes

Need to handle the interrupt enables for all pipes.

v2: fix indexing (Jessie)

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

drm/amdgpu/gfx11: properly handle error ints on all pipes

Need to handle the interrupt enables for all pipes.

v2: fix indexing (Jessie)

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

drm/amdgpu/gfx10: properly handle error ints on all pipes

Need to handle the interrupt enables for all pipes.

v2: fix indexing (Jessie)

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

drm/amdgpu/gfx12: enable wave kill for compute queues

It should work the same for compute as well as gfx.

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