Philip Yang [Fri, 16 Feb 2024 16:41:16 +0000 (11:41 -0500)]
drm/amdkfd: Output migrate end event if migrate failed
If page migration failed, also output migrate end event to match with
migrate start event, with failure error_code added to the end of the
migrate message macro. This will not break uAPI because application uses
old message macro sscanf drop and ignore the error_code.
Output GPU page fault restore end event if migration failed.
drm/amdgpu/gfx11: Implement cleaner shader support for GFX11 hardware
The patch modifies the gfx_v11_0_kiq_set_resources function to write
the cleaner shader's memory controller address to the ring buffer. It
also adds a new function, gfx_v11_0_ring_emit_cleaner_shader, which
emits the PACKET3_RUN_CLEANER_SHADER packet to the ring buffer.
This patch adds support for the PACKET3_RUN_CLEANER_SHADER packet in the
gfx_v11_0 module. This packet is used to emit the cleaner shader, which
is used to clear GPU memory before it's reused, helping to prevent data
leakage between different processes.
Finally, the patch updates the ring function structures to include the
new gfx_v11_0_ring_emit_cleaner_shader function. This allows the
cleaner shader to be emitted as part of the ring's operations.
drm/amd/pm: use pm_runtime_get_if_active for debugfs getters
Don't wake up the GPU for reading pm values. Instead, take a runtime
powermanagement ref when trying to read it if and only if the GPU is
already awake.
This avoids spurious wake ups (eg: from applets).
We use pm_runtime_get_if_in_active because we care about "is the GPU awake?"
not about "is the GPU awake and something else prevents suspend?".
Sunil Khatri [Wed, 2 Oct 2024 13:54:25 +0000 (19:24 +0530)]
drm/amdgpu: change the comment from handle to ip_block
htmldoc generation depend upon the input arguments etc
to generate the document. After update of handle to
ip_block then update needs in comments too to fix the
warnings.
drm/amdgpu/gfx10: Implement cleaner shader support for GFX10 hardware
The patch modifies the gfx_v10_0_kiq_set_resources function to write
the cleaner shader's memory controller address to the ring buffer. It
also adds a new function, gfx_v10_0_ring_emit_cleaner_shader, which
emits the PACKET3_RUN_CLEANER_SHADER packet to the ring buffer.
This patch adds support for the PACKET3_RUN_CLEANER_SHADER packet in the
gfx_v10_0 module. This packet is used to emit the cleaner shader, which
is used to clear GPU memory before it's reused, helping to prevent data
leakage between different processes.
Finally, the patch updates the ring function structures to include the
new gfx_v10_0_ring_emit_cleaner_shader function. This allows the
cleaner shader to be emitted as part of the ring's operations.
drm/amd/pm: don't update runpm last_usage on debugfs getter
Reading pm values from the GPU shouldn't prevent it to be suspended
by resetting the last active timestamp (eg: if an background app
monitors GPU sensors every second, it would prevent the autosuspend
sequence to trigger).
drm/amdgpu/gfx9: Add Cleaner Shader Deinitialization in gfx_v9_0 Module
This commit addresses an omission in the previous patch related to the
cleaner shader support for GFX9 hardware. Specifically, it adds the
necessary deinitialization code for the cleaner shader in the
gfx_v9_0_sw_fini function.
The added line amdgpu_gfx_cleaner_shader_sw_fini(adev); ensures that any
allocated resources for the cleaner shader are freed correctly, avoiding
potential memory leaks and ensuring that the GPU state is clean for the
next initialization sequence.
drm/amd/powerplay: Delete unused function and maths library
We start with the function 'atomctrl_calculate_voltage_evv_on_sclk'
which has been unused since 2016's commit e805ed83ba1c ("drm/amd/powerplay: delete useless files.")
Remove it.
It was also the last user of the entire fixed point maths library in
ppevvmath.h.
Some distributions have been patching amdgpu to enable overdrive by
default which may compromise stability. Furthermore when bug reports
are brought upstream it's not obvious that the system has been tampered
with.
When overdrive is enabled taint the kernel and leave a critical message
in the logs for users so that it's obvious in a bug report it's been
tampered with.
drm/amdkfd: Fix kdoc entry for 'get_wave_count()' function parameters
Update kdoc entries to reflect the function's parameters. The descriptor
for the 'queue_cnt' parameter has been added, and the incorrect mentions
of 'wave_cnt' and 'vmid', which are not parameters but local variables,
have been removed.
Fixes the below with gcc W=1:
drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v9.c:954: warning: Function parameter or struct member 'queue_cnt' not described in 'get_wave_count'
drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v9.c:954: warning: Excess function parameter 'wave_cnt' description in 'get_wave_count'
drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v9.c:954: warning: Excess function parameter 'vmid' description in 'get_wave_count'
drm/amdgpu: Flush tlb by VM_INVALIDATION packet in sdma_v5_2
In order for SDMA not to be switched between VM_INVALIDATION
request and ack, use an single VM_INVALIDATION packet in function
sdma_v5_2_ring_emit_vm_flush.
DC 3.2.303 contains some improvements as summarized below:
* Improve brightness control
* Add support for UHBR10 eDP
* OPTC required only for DTBCLK_P for dcn401
* Fix TBT monitor resume issue
* Code cleanup
drm/amd/display: Configure DTBCLK_P with OPTC only for dcn401
[WHY]
DTBCLK_P is used to generate virtual pixel clock, and to drive the HPO
stream encoder clock. Programming the required clock when
enabling/disabling both components can cause issues.
For example, if HPO is being disabled and clock source is changed to
REFCLK, virtual pixel rate will then be wrong, causing issues in CRTC.
[HOW]
Only program the DTBCLK_P when programming CRTC, as its expected it will
be enabled prior to HPO, and disabled after HPO in all valid cases.
drm/amd/display: Add IPS residency capture helpers to dc_dmub_srv
This enables starting and stopping IPS residency measurements
and querying the IPS residency information consisting of residency
percent, entry counter, total time active & inactive, and histograms
for the specified IPS mode.
drm/amd/display: Fix out-of-bounds access in 'dcn21_link_encoder_create'
An issue was identified in the dcn21_link_encoder_create function where
an out-of-bounds access could occur when the hpd_source index was used
to reference the link_enc_hpd_regs array. This array has a fixed size
and the index was not being checked against the array's bounds before
accessing it.
This fix adds a conditional check to ensure that the hpd_source index is
within the valid range of the link_enc_hpd_regs array. If the index is
out of bounds, the function now returns NULL to prevent undefined
behavior.
Alex Hung [Wed, 18 Sep 2024 21:33:27 +0000 (15:33 -0600)]
drm/amd/display: Remove always-false branches
[WHAT & HOW]
MacroTileSizeBytes is set to either 256 or 65535 and it is never
4096. Its branch is not taken, and should be removed. Similarly,
mode_422 is always 0 and thus ppe will always be 1. The ternary
operator should be removed.
This fixes 2 DEADCODE issues reported by Coverity.
Muyuan Yang [Wed, 5 Jun 2024 14:24:59 +0000 (10:24 -0400)]
drm/amd/display: Change Brightness Control Priority
Prioritize Aux-based over PWM-based brightness control
for more types of panels and introduce a new structure
to store and manage the type of brightness control used.
Tom Chung [Fri, 13 Sep 2024 07:44:40 +0000 (15:44 +0800)]
drm/amd/display: Fix system hang while resume with TBT monitor
[Why]
Connected with a Thunderbolt monitor and do the suspend and the system
may hang while resume.
The TBT monitor HPD will be triggered during the resume procedure
and call the drm_client_modeset_probe() while
struct drm_connector connector->dev->master is NULL.
It will mess up the pipe topology after resume.
[How]
Skip the TBT monitor HPD during the resume procedure because we
currently will probe the connectors after resume by default.
Ryan Seto [Fri, 13 Sep 2024 20:01:47 +0000 (16:01 -0400)]
drm/amd/display: Adjust PHY FSM transition to TX_EN-to-PLL_ON for TMDS
[Why]
If two monitors with TMDS signals were timing synced and one was
disconnected, the stream would go out of sync too early due to
the PLL turning off and the system could hang
[How]
On link disable output, change PHY FSM transition from TX_EN-to-PHY_OFF
to TX_EN-to-PLL_ON for TMDS
drm/amd/display: Wait For DET Update Should Use Current State
[Why]
Current state should be used when waiting for DET update
instead of new context.
For any streams decreasing in DET, pipes used in the current state
should be checked since those pipes need to free their DET before
DET can be reallocated.
[How]
Pass in current_state instead of context.
Use pipe from current_state instead of context. This assumes that
pipe in the current_state is an OTG_MASTER pipe if the pipe in the context is an OTG_MASTER pipe.
Alex Hung [Fri, 6 Sep 2024 17:39:18 +0000 (11:39 -0600)]
drm/amd/display: Add HDR workaround for specific eDP
[WHY & HOW]
Some eDP panels suffer from flicking when HDR is enabled in KDE. This
quirk works around it by skipping VSC that is incompatible with eDP
panels.
drm/amd/display: Clip rect size changes should be full updates
[Why]
In cases where an MPO plane is being dragged around partially off-screen,
it is possible to get a flip where the only scaling parameters to change
are the clip rect size and position. Currently, clip rect size changes
are considered medium updates, which can result in the clip rect being used
for HW programming being larger than the clip rect that was used for the
last DML validation. This can lead to mismatches in different parts of the
pipe and can result in a p-state hang.
[How]
- consider clip rect size changes scaling changes, therefore full updates
- refactor get_scaling_info_update_type for clarity
- remove clip_size_change update flag
Clip rect size changes were previously demoted from full updates as an
optimization when the MPO + ODM policy changed to always pre-allocate MPO
pipes, but it created the issue described above. Personally testing this
use case, the performance feels fine with full update spam, and we expect
this is a fairly infrequent use case. If the performance needs to be
optimized in the future, consider reworking the entire update type logic
to run a DML pass and determine the update type based on what DML says
will actually change.
[Why/How]
Certain transitions from HW to SW cursor can be missed, which will
result in black visual confirm instead of expected blue/red. Need to add
handling for cursor visual confirm to dc_update_visual_confirm_color.
drm/amd/display: Update Interface to Check UCLK DPM
[Why]
Videos using YUV420 format may result in high power being used.
Disabling MPO may result in lower power usage.
Update interface that can be used to check power profile of a dc_state.
[How]
Allow pstate switching in VBlank as last entry in strategy candidates.
Add helper functions that can be used to determine power level:
-get power profile after a dc_state has undergone full validation
[why]
The "dpp35_dppclk_control" routine is incorrectly disabling the register clock gater
when the DPP is enabled.
The "DISPCLK_R_GATE_DISABLE" should never be set to 1 in the normal operating mode.
This will disable the clock gater and the DPPCLK register clock branch will always be running.
As a consequence, the dynamic power will be higher than expected.
Michael Strauss [Tue, 3 Sep 2024 20:23:11 +0000 (16:23 -0400)]
drm/amd/display: Block UHBR Based On USB-C PD Cable ID
[WHY]
Currently the absence of UHBR cable ID caps from USB-C PD does not block UHBR
rates. In situations where DPCD reports valid UHBR capability but USB-C PD
does not, such as using a USB-C to DP1.4 dongle connected to a native DP2.1
cable, link loss and lightup failures can be seen as a result.
Additionally, in edge cases where a platform supports cable ID but DMUB
doesn't correctly return cable ID caps, driver currently also allows UHBR.
[HOW]
Block UHBR rates over DP alt mode if cable id indicates no UHBR support.
Additionally, block UHBR rates if a cable ID supported platform receives no
reply from a DMUB cable id query.
drm/amd/display: Set Pipe Unlock Order Outside of HWSEQ
[Why]
Current pipe unlock order set within dcn401_interdependent_update_lock.
Separate the logic from the functionality to allow for unit testing
and make it easier to debug.
[How]
Add a flag to indicate if a pipe unlock order should be set.
Create function to determine unlock order.
Indicate which pipes should be unlocked first using
array stored in dc scratch memory.
Pipes indicated in array can be unlocked in any order.
A stream without dsc_aux should not be eliminated from
the dsc determination. Whether it needs a dsc recompute depends on
whether its mode has changed or not. Eliminating such a no-dsc stream
from the dsc determination policy will end up with inconsistencies
in the new dc_state when compared to the current dc_state,
triggering a dsc recompute that should not have happened.
drm/amd/display: Monitor patch to call blank_stream() before otg off
[Why]
Turning off OTG before DIG is on and backlight is on, might cause:
DIG FIFO underflow.
EDP output unexpected video data. That might violate EDP spec.
EDP spec requires black light should be off before validate
video data is turned off.
Currently garbage issue only occurs in one type of EDP to MIPI converter.
[How]
Add monitor patch to call blank_stream() before disable_crtc().
That will be no impact for current sequence.
If there are more types of EDP panels meeting this issue later,
we might remove this monitor patch and fix the current sequence.
Alvin Lee [Mon, 9 Sep 2024 20:24:05 +0000 (16:24 -0400)]
drm/amd/display: Wait for all pending cleared before full update
[Description]
Before every full update we must wait for all pending updates to be
cleared - this is particularly important for minimal transitions
because if we don't wait for pending cleared, it will be as if
there was no minimal transition at all. In OTG we must read 3 different
status registers for pending cleared, one specifically for OTG updates,
one specifically for OPTC updates, and the last for surface related
updates.
drm/amd/display: Restore Optimized pbn Value if Failed to Disable DSC
Existing last step of dsc policy is to restore pbn value under minimum compression
when try to greedily disable dsc for a stream failed to fit in MST bw.
Optimized dsc params result from optimization step is not necessarily the minimum compression,
therefore it is not correct to restore the pbn under minimum compression rate.
Restore the pbn under minimum compression instead of the value from optimized pbn could result
in the dsc params not correct at the modeset where atomic_check failed due to not
enough bw. One or more monitors connected could not light up in such case.
Restore the optimized pbn value, instead of using the pbn value under minimum
compression.
[WHY & HOW]
Mismatch in DCN35 DML2 cause bw validation failed to acquire unexpected DPP pipe to cause
grey screen and system hang. Remove EnhancedPrefetchScheduleAccelerationFinal value override
to match HW spec.
Add partition config nodes in sysfs to get resource instance details for
a particular partition mode. A resource could be anything like an xcc,
vcn decoder, system dma units etc.
Details of various resource instances are available under
/sys/bus/pci/devices/.../compute_partition_config/
Select a partition configuration:
/sys/bus/pci/devices/.../compute_partition_config/xcp_config
Number of instances of a resource:
/sys/bus/pci/devices/.../compute_partition_config/<rsrc_name>/num_inst
Total partitions sharing the resource:
/sys/bus/pci/devices/.../compute_partition_config/<rsrc_name>/num_shared
Hamza Mahfooz [Fri, 9 Aug 2024 20:42:53 +0000 (16:42 -0400)]
drm/amd/display: change the panel power savings level without a modeset
We don't actually need to request that the compositor does a full
modeset to modify the panel power savings level, we can instead
just make a request to DMUB, to set the new level dynamically.
A reset on initialization will be needed if a new PSP TOS needs to be
loaded than the one currently active on the system. This is possible
only on SOCs which support a full device reset which results in unload
of active PSP TOS.
Add interface to check if a different TOS needs to be loaded than the
one which is which is already active on the SOC. Presently the interface
is restricted to specific variants of PSPv13.0.
Lijo Lazar [Mon, 26 Aug 2024 13:22:14 +0000 (18:52 +0530)]
drm/amdgpu: Refactor XGMI reset on init handling
Use XGMI hive information to rely on resetting XGMI devices on
initialization rather than using mgpu structure. mgpu structure may have
other devices as well.
Lijo Lazar [Fri, 30 Aug 2024 05:51:43 +0000 (11:21 +0530)]
drm/amdgpu: Add helper to initialize badpage info
Add a separate function to read badpage data during initialization.
Reading bad pages will need hardware access and cannot be done during
reset. Hence in cases where device needs a full reset during
init itself, attempting to read will cause a deadlock.
amdgpu_gfx_bit_to_me_queue has been unused since it was added in
commit 7470bfcf2014 ("drm/amdgpu: add helper function for gfx queue/bitmap
transition")
amdgpu_atpx_dgpu_req_power_for_displays has been unused since
commit bdb1ccb080da ("drm/amdgpu: remove ATPX_DGPU_REQ_POWER_FOR_DISPLAYS
check when hotplug-in")
amdgpu_atpx_get_dhandle has been unused since commit f9b7f3703ff9 ("drm/amdgpu/acpi: make ATPX/ATCS structures global (v2)")
amdgpu_device_ip_is_idle is unused.
It was renamed from 'amdgpu_is_idle' which was originally added in
commit 5dbbb60ba61e ("drm/amdgpu: add IP helpers for wait_for_idle and is_idle")
Add a callback interface to get the resource information of a partition
mode. Presently the information has number of resources and number of
entities sharing the resource.
Alex Deucher [Mon, 16 Sep 2024 17:16:53 +0000 (13:16 -0400)]
drm/amdgpu/gfx9: set additional bits on CP halt
Need to set the pipe reset and cache invalidation bits
on halt otherwise we can get stale state if the CP firmware
changes (e.g., on module unload and reload).
projects / linux.git / log