drm/amdgpu/userq: optimize enforce isolation and s/r

If user queues are disabled for all IPs in the case
of suspend and resume and for gfx/compute in the case
of enforce isolation, we can return early.

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

drm/amd/display: Remove unused *vbios_smu_set_dprefclk

rn_vbios_smu_set_dprefclk() was added in 2019 by
commit 4edb6fc91878 ("drm/amd/display: Add Renoir clock manager")
rv1_vbios_smu_set_dprefclk() was also added in 2019 by
commit dc88b4a684d2 ("drm/amd/display: make clk mgr soc specific")

neither have been used.

Remove them.

Signed-off-by: Dr. David Alan Gilbert <linux@treblig.org>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/radeon: Remove unused radeon_fence_wait_any

radeon_fence_wait_any() last use was removed in 2023's
commit 254986e324ad ("drm/radeon: Use the drm suballocation manager
implementation.")

Remove it.

Signed-off-by: Dr. David Alan Gilbert <linux@treblig.org>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/radeon/radeon_audio: Remove unused r600_hdmi_audio_workaround

The last use of r600_hdmi_audio_workaround() was removed by 2014's
commit 6e72376dcc66 ("radeon/audio: consolidate audio_mode_set()
functions")

Remove it.

Signed-off-by: Dr. David Alan Gilbert <linux@treblig.org>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: Print kernel message when error logged by scrub

Print a kernel message when the scrub bit of status register is set to
indicate that errors are being logged by the scrub.

Signed-off-by: Xiang Liu <xiang.liu@amd.com>
Reviewed-by: Hawking Zhang <Hawking.Zhang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: do not copy invalid CRTC timing info

Since b255ce4388e0, it is possible that the CRTC timing
information for the preferred mode has not yet been
calculated while amdgpu_dm_connector_mode_valid() is running.

In this case use the CRTC timing information of the actual mode.

Fixes: b255ce4388e0 ("drm/amdgpu: don't change mode in amdgpu_dm_connector_mode_valid()")
Closes: https://lore.kernel.org/all/ed09edb167e74167a694f4854102a3de6d2f1433.camel@irl.hu/
Closes: https://gitlab.freedesktop.org/drm/amd/-/issues/4085
Signed-off-by: Gergo Koteles <soyer@irl.hu>
Reviewed-by: Zaeem Mohamed <zaeem.mohamed@amd.com>
Tested-by: Mark Broadworth <mark.broadworth@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Correct prefetch calculation

[Why]
The minimum value of the dst_y_prefetch_equ was not correct
in prefetch calculation whice causes OPTC underflow.

[How]
Add the min operation of dst_y_prefetch_equ in prefetch calculation
for legacy DML.

Reviewed-by: Nicholas Kazlauskas <nicholas.kazlauskas@amd.com>
Signed-off-by: TungYu Lu <tungyu.lu@amd.com>
Signed-off-by: Zaeem Mohamed <zaeem.mohamed@amd.com>
Tested-by: Mark Broadworth <mark.broadworth@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Refactor SubVP cursor limiting logic

[WHY]
There are several gaps that can result in SubVP being enabled with
incompatible HW cursor sizes, and unjust restrictions to cursor size due
to wrong predictions on future usage of SubVP

[HOW]
- remove "prediction" logic in favor of tagging based on previous SubVP
  usage
- block SubVP if current HW cursor settings are incompatible
- provide interface for DM to determine if HW cursor should be disabled
  due to an attempt to enable SubVP

Reviewed-by: Alvin Lee <alvin.lee2@amd.com>
Signed-off-by: Dillon Varone <dillon.varone@amd.com>
Signed-off-by: Zaeem Mohamed <zaeem.mohamed@amd.com>
Tested-by: Mark Broadworth <mark.broadworth@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu/userq: add a helper to check which IPs are enabled

Add a helper to get a mask of IPs which support user queues.
Use this in the INFO IOCTL to get the IP mask to replace
the current code.

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

drm/amdgpu: Add queue id support to the user queue wait IOCTL

Add queue id support to the user queue wait IOCTL
drm_amdgpu_userq_wait structure.

This is required to retrieve the wait user queue and maintain
the fence driver references in it so that the user queue in
the same context releases their reference to the fence drivers
at some point before queue destruction.

Otherwise, we would gather those references until we
don't have any more space left and crash.

v2: Modify the UAPI comment as per the mesa and libdrm UAPI comment.

Libdrm MR: https://gitlab.freedesktop.org/mesa/drm/-/merge_requests/408
Mesa MR: https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/34493

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

drm/amdgpu/userq: enable support for secure queues

Enable users to create secure GFX/compute queues.

Reviewed-by: Sunil Khatri <sunil.khatri@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/amdgpu/userq/mes: pass the secure flag to mqd init

So that we initialize the MQD as a secure queue.

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

drm/amd/display: Fix pixel rate divider policy for 1 pixel per cycle config

[Why]
Pixel rate dividor was not programmed correctly for 1 pixel per cycle
configuration for empty tu case.

[How]
Included check for empty tu when pixel rate dividor values were selected.

Reviewed-by: Michael Strauss <michael.strauss@amd.com>
Signed-off-by: Meenakshikumar Somasundaram <meenakshikumar.somasundaram@amd.com>
Signed-off-by: Zaeem Mohamed <zaeem.mohamed@amd.com>
Tested-by: Mark Broadworth <mark.broadworth@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Default IPS to RCG_IN_ACTIVE_IPS2_IN_OFF

[Why]

Recent findings show negligible power savings between IPS2 and RCG
during static desktop. In fact, DCN related clocks are higher
when IPS2 is enabled vs RCG.

RCG_IN_ACTIVE is also the default policy for another OS supported by
DC, and it has faster entry/exit.

[How]

Remove previous logic that checked for IPS2 support, and just default
to `DMUB_IPS_RCG_IN_ACTIVE_IPS2_IN_OFF`.

Fixes: 199888aa25b3 ("drm/amd/display: Update IPS default mode for DCN35/DCN351")
Reviewed-by: Aurabindo Pillai <aurabindo.pillai@amd.com>
Signed-off-by: Leo Li <sunpeng.li@amd.com>
Signed-off-by: Zaeem Mohamed <zaeem.mohamed@amd.com>
Tested-by: Mark Broadworth <mark.broadworth@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Revert "not disable dtb as dto src at dpms off"

[why]
not all the asic using the same code path.
need to revisit and limit the impact.

This reverts commit 32be4e39f459f3ac9c191569ae8e3731cb82f7ab.

Reviewed-by: Gabe Teeger <gabe.teeger@amd.com>
Signed-off-by: Charlene Liu <Charlene.Liu@amd.com>
Signed-off-by: Zaeem Mohamed <zaeem.mohamed@amd.com>
Tested-by: Mark Broadworth <mark.broadworth@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Use 16ms AUX read interval for LTTPR with old sinks

[Why/How]
LTTPR are required to program DPCD 0000Eh to 0x4 (16ms) upon AUX read
reply to this register. Since old Sinks witih DPCD rev 1.1 and earlier
may not support this register, assume the mandatory value is programmed
by the LTTPR to avoid AUX timeout issues.

Reviewed-by: Wenjing Liu <wenjing.liu@amd.com>
Signed-off-by: George Shen <george.shen@amd.com>
Signed-off-by: Zaeem Mohamed <zaeem.mohamed@amd.com>
Tested-by: Mark Broadworth <mark.broadworth@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Fix ACPI edid parsing on some Lenovo systems

[Why]
The ACPI EDID in the BIOS of a Lenovo laptop includes 3 blocks, but
dm_helpers_probe_acpi_edid() has a start that is 'char'.  The 3rd
block index starts after 255, so it can't be indexed properly.
This leads to problems with the display when the EDID is parsed.

[How]
Change the variable type to 'short' so that larger values can be indexed.

Cc: Renjith Pananchikkal <renjith.pananchikkal@amd.com>
Reported-by: Mark Pearson <mpearson@lenovo.com>
Suggested-by: David Ober <dober@lenovo.com>
Fixes: c6a837088bed ("drm/amd/display: Fetch the EDID from _DDC if available for eDP")
Reviewed-by: Alex Hung <alex.hung@amd.com>
Signed-off-by: Mario Limonciello <mario.limonciello@amd.com>
Signed-off-by: Zaeem Mohamed <zaeem.mohamed@amd.com>
Tested-by: Mark Broadworth <mark.broadworth@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Promote DC to 3.2.329

Summary:

* Implement HDMI Read request
* RMCM and MCM 3DLUT support
* Enable urgent latency adjustment on DCN35
* Enable phy-ssc reduction by default

Signed-off-by: Taimur Hassan <Syed.Hassan@amd.com>
Signed-off-by: Zaeem Mohamed <zaeem.mohamed@amd.com>
Reviewed-by: Leo Li <sunpeng.li@amd.com>
Tested-by: Mark Broadworth <mark.broadworth@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: Allow P2P access through XGMI

If peer memory is accessible through XGMI, allow leaving it in VRAM
rather than forcing its migration to GTT on DMABuf attachment.

Signed-off-by: Felix Kuehling <felix.kuehling@amd.com>
Tested-by: Hao (Claire) Zhou <hao.zhou@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: enable phy-ssc reduction by default

[Why]
Reduction of phy-ssc is needed to support DP2 high pixel clock on dcn35x/36.
There's a special flag to enable it in dmub hw params.

[How]
Set hbr3_phy_ssc to true for dcn35, dcn351 and dcn36.

Reviewed-by: Charlene Liu <charlene.liu@amd.com>
Signed-off-by: Roman Li <Roman.Li@amd.com>
Signed-off-by: Zaeem Mohamed <zaeem.mohamed@amd.com>
Tested-by: Mark Broadworth <mark.broadworth@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Enable urgent latency adjustment on DCN35

[Why]

Urgent latency adjustment was disabled on DCN35 due to issues with P0
enablement on some platforms. Without urgent latency, underflows occur
when doing certain high timing configurations. After testing, we found
that reenabling urgent latency didn't reintroduce p0 support on multiple
platforms.

[How]

renable urgent latency on DCN35 and setting it to 3000 Mhz.

This reverts commit 3412860cc4c0c484f53f91b371483e6e4440c3e5.

Reviewed-by: Charlene Liu <charlene.liu@amd.com>
Signed-off-by: Nicholas Susanto <nsusanto@amd.com>
Signed-off-by: Zaeem Mohamed <zaeem.mohamed@amd.com>
Tested-by: Mark Broadworth <mark.broadworth@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: DCN42 RMCM and MCM 3DLUT support

[WHY & HOW]
Providing hardware programming for the RMCM and MCM IPs for 3DLUT in DCN42.

Reviewed-by: Charlene Liu <charlene.liu@amd.com>
Signed-off-by: Yihan Zhu <Yihan.Zhu@amd.com>
Signed-off-by: Zaeem Mohamed <zaeem.mohamed@amd.com>
Tested-by: Mark Broadworth <mark.broadworth@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: DCN32 null data check

[WHY & HOW]
Avoid null curve data structure used in the cm block for the potential issue.

Reviewed-by: Charlene Liu <charlene.liu@amd.com>
Signed-off-by: Yihan Zhu <Yihan.Zhu@amd.com>
Signed-off-by: Zaeem Mohamed <zaeem.mohamed@amd.com>
Tested-by: Mark Broadworth <mark.broadworth@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Force full update in gpu reset

[Why]
While system undergoing gpu reset always do full update
to sync the dc state before and after reset.

[How]
Return true in should_reset_plane() if gpu reset detected

Reviewed-by: Aurabindo Pillai <aurabindo.pillai@amd.com>
Reviewed-by: Mario Limonciello <mario.limonciello@amd.com>
Signed-off-by: Roman Li <Roman.Li@amd.com>
Signed-off-by: Zaeem Mohamed <zaeem.mohamed@amd.com>
Tested-by: Mark Broadworth <mark.broadworth@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Fix gpu reset in multidisplay config

[Why]
The indexing of stream_status in dm_gpureset_commit_state() is incorrect.
That leads to asserts in multi-display configuration after gpu reset.

[How]
Adjust the indexing logic to align stream_status with surface_updates.

Fixes: cdaae8371aa9 ("drm/amd/display: Handle GPU reset for DC block")
Closes: https://gitlab.freedesktop.org/drm/amd/-/issues/3808
Reviewed-by: Aurabindo Pillai <aurabindo.pillai@amd.com>
Reviewed-by: Mario Limonciello <mario.limonciello@amd.com>
Signed-off-by: Roman Li <Roman.Li@amd.com>
Signed-off-by: Zaeem Mohamed <zaeem.mohamed@amd.com>
Tested-by: Mark Broadworth <mark.broadworth@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Move Mode Support Prefetch Checks To Its Own Function

[Why]
Large stack size observed in DCN4 mode support when compiling with clang.
Additional instrumentation added by compiler adds to stack size.
dml_core_mode_support ends up going over the stack size limit
due to the size of the function.

[How]
Move checks and calculations for prefetch to its own function.

Reviewed-by: Aurabindo Pillai <aurabindo.pillai@amd.com>
Signed-off-by: Austin Zheng <Austin.Zheng@amd.com>
Signed-off-by: Zaeem Mohamed <zaeem.mohamed@amd.com>
Tested-by: Mark Broadworth <mark.broadworth@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: Don't pin VRAM without DMABUF_MOVE_NOTIFY

Pinning of VRAM is for peer devices that don't support dynamic attachment
and move notifiers. But it requires that all such peer devices are able to
access VRAM via PCIe P2P. Any device without P2P access requires migration
to GTT, which fails if the memory is already pinned for another peer
device.

Sharing between GPUs should not require pinning in VRAM. However, if
DMABUF_MOVE_NOTIFY is disabled in the kernel build, even DMABufs shared
between GPUs must be pinned, which can lead to failures and functional
regressions on systems where some peer GPUs are not P2P accessible.

Disable VRAM pinning if move notifiers are disabled in the kernel build
to fix regressions when sharing BOs between GPUs.

Signed-off-by: Felix Kuehling <felix.kuehling@amd.com>
Tested-by: Hao (Claire) Zhou <hao.zhou@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Move desync error counter operation up.

[Why & How]
Move desync error counter operation up to prevent
it from being skipped by force disable desync
error.

Reviewed-by: Robin Chen <robin.chen@amd.com>
Reviewed-by: Aric Cyr <aric.cyr@amd.com>
Signed-off-by: Jack Chang <jack.chang@amd.com>
Signed-off-by: Zaeem Mohamed <zaeem.mohamed@amd.com>
Tested-by: Mark Broadworth <mark.broadworth@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Avoid divide by zero by initializing dummy pitch to 1

[Why]
If the dummy values in `populate_dummy_dml_surface_cfg()` aren't updated
then they can lead to a divide by zero in downstream callers like
CalculateVMAndRowBytes()

[How]
Initialize dummy value to a value to avoid divide by zero.

Reviewed-by: Alex Hung <alex.hung@amd.com>
Signed-off-by: Mario Limonciello <mario.limonciello@amd.com>
Signed-off-by: Zaeem Mohamed <zaeem.mohamed@amd.com>
Tested-by: Mark Broadworth <mark.broadworth@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Implement HDMI Read Request

[Why]
Read Request provides alterative method to polling to
the HDMI sinks that support it.

[How]
Implement Read Request where interrupt can be generated
by the sink.

Reviewed-by: Joshua Aberback <joshua.aberback@amd.com>
Signed-off-by: Chris Park <chris.park@amd.com>
Signed-off-by: Zaeem Mohamed <zaeem.mohamed@amd.com>
Tested-by: Mark Broadworth <mark.broadworth@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: To apply the adjusted DP ref clock for DP devices

[Why]
For some pixel clock margin sensitive external monitor,
we could not keep original DP ref clock for the ASICs
supported SSC DP ref clock.

[How]
From slicon design team's comment,
we have to apply the adjusted DP ref clock for
DP devices.
DP 128b (DP2) signals uses the DTBCLK not DP ref.

Reviewed-by: Nicholas Kazlauskas <nicholas.kazlauskas@amd.com>
Signed-off-by: Yiling Chen <yi-ling.chen2@amd.com>
Signed-off-by: Zaeem Mohamed <zaeem.mohamed@amd.com>
Tested-by: Mark Broadworth <mark.broadworth@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu/gfx12: add support for TMZ queues to mqd_init

Set up TMZ for queues.

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

drm/amdgpu/gfx11: add support for TMZ queues to mqd_init

Set up TMZ for queues.

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

drm/amdgpu: Use allowed_domains for pinning dmabufs

When determining the domains for pinning DMABufs, filter allowed_domains
and fail with a warning if VRAM is forbidden and GTT is not an allowed
domain.

Fixes: f5e7fabd1f5c ("drm/amdgpu: allow pinning DMA-bufs into VRAM if all importers can do P2P")
Suggested-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Felix Kuehling <felix.kuehling@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: add tmz queue parameter to mqd props

Use this to track the whether we want TMZ for queues.

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

drm/amdgpu/userq: add UAPI for setting up secure queues

If the queues needs to access TMZ surfaces, it must
be set up as secure.

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

drm/amdgpu: Refine Cleaner Shader MEC firmware version for GFX10.1.x GPUs

Update the minimum firmware version for the Cleaner Shader in the
gfx_v10_0_sw_init function.

This change adjusts the minimum required firmware version for the MEC
firmware from 152 to 151, allowing for broader compatibility with
GFX10.1 GPUs.

Fixes: 25961bad9212 ("drm/amdgpu/gfx10: Add cleaner shader for GFX10.1.10")
Cc: Christian König <christian.koenig@amd.com>
Cc: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Srinivasan Shanmugam <srinivasan.shanmugam@amd.com>
Acked-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu:remove old sdma reset callback mechanism

This patch removes the deprecated SDMA reset callback mechanism, which was previously used to register pre-reset and post-reset callbacks for SDMA engine resets.
 The callback mechanism has been replaced with a more direct and efficient approach using `stop_queue` and `start_queue` functions in the ring's function table.

The SDMA reset callback mechanism allowed KFD and AMDGPU to register pre-reset and post-reset functions for handling SDMA engine resets.
However, this approach added unnecessary complexity and was no longer needed after the introduction of the `stop_queue` and `start_queue` functions in the ring's function table.

1. **Remove Callback Mechanism**:
   - Removed the `amdgpu_sdma_register_on_reset_callbacks` function and its associated data structures (`sdma_on_reset_funcs`).
   - Removed the callback registration logic from the SDMA v4.4.2 initialization code.

2. **Clean Up Related Code**:
   - Removed the `sdma_v4_4_2_set_engine_reset_funcs` function, which was used to register the callbacks.
   - Removed the `sdma_v4_4_2_engine_reset_funcs` structure, which contained the pre-reset and post-reset callback functions.

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/radeon: fix the warning for radeon_cs_parser_fini

Fix the below warning message.
radeon/radeon_cs.c:418: warning: Excess function parameter 'backoff' description in 'radeon_cs_parser_fini'

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

drm/amdgpu/userq: add context and seqno of the fence

Add context and seqno of the fence in error logging
rather than printing fence ptr.

Reviewed-by: Christian König <christian.koenig@amd.com>
Suggested-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Suggested-by: Tvrtko Ursulin <tvrtko.ursulin@igalia.com>
Signed-off-by: Sunil Khatri <sunil.khatri@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: optimize queue reset and stop logic for sdma_v5_2

This patch refactors the SDMA v5.2 queue reset and stop logic to improve
code readability, maintainability, and performance. The key changes include:

1. **Generalized `sdma_v5_2_gfx_stop` Function**:
   - Added an `inst_mask` parameter to allow stopping specific SDMA instances
     instead of all instances. This is useful for resetting individual queues.

2. **Simplified `sdma_v5_2_reset_queue` Function**:
   - Removed redundant loops and checks by directly using the `ring->me` field
     to identify the SDMA instance.
   - Reused the `sdma_v5_2_gfx_stop` function to stop the queue, reducing code
     duplication.

v1: The general coding style is to declare variables like "i" or "r" last. E.g. longest lines first and short lasts. (Chritian)

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/amdgpu: optimize queue reset and stop logic for sdma_v5_0

This patch refactors the SDMA v5.0 queue reset and stop logic to improve
code readability, maintainability, and performance. The key changes include:

1. **Generalized `sdma_v5_0_gfx_stop` Function**:
   - Added an `inst_mask` parameter to allow stopping specific SDMA instances
     instead of all instances. This is useful for resetting individual queues.

2. **Simplified `sdma_v5_0_reset_queue` Function**:
   - Removed redundant loops and checks by directly using the `ring->me` field
     to identify the SDMA instance.
   - Reused the `sdma_v5_0_gfx_stop` function to stop the queue, reducing code
     duplication.

v1: The general coding style is to declare variables like "i" or "r" last. E.g. longest lines first and short lasts. (Chritian)

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/amdgpu: Register the new sdma function pointers for sdma_v5_2

Register stop/start/soft_reset queue functions for SDMA IP versions v5.2.

Suggested-by: Alex Deucher <alexander.deucher@amd.com>
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/amdgpu/: drm/amdgpu: Register the new sdma function pointers for sdma_v5_0

Register stop/start/soft_reset queue functions for SDMA IP versions v5.0.

Suggested-by: Alex Deucher <alexander.deucher@amd.com>
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/amdgpu: Implement SDMA soft reset directly for v5.x

This patch introduces a new function `amdgpu_sdma_soft_reset` to handle SDMA soft resets directly,
rather than relying on the DPM interface.

1. **New `amdgpu_sdma_soft_reset` Function**:
   - Implements a soft reset for SDMA engines by directly writing to the hardware registers.
   - Handles SDMA versions 4.x and 5.x separately:
     - For SDMA 4.x, the existing `amdgpu_dpm_reset_sdma` function is used for backward compatibility.
     - For SDMA 5.x, the driver directly manipulates the `GRBM_SOFT_RESET` register to reset the specified SDMA instance.

2. **Integration into `amdgpu_sdma_reset_engine`**:
   - The `amdgpu_sdma_soft_reset` function is called during the SDMA reset process, replacing the previous call to `amdgpu_dpm_reset_sdma`.

v2: r should default to an error (Alex)

Suggested-by: Alex Deucher <alexander.deucher@amd.com>
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/amdgpu: switch amdgpu_sdma_reset_engine to use the new sdma function pointers

Replace old callback mechanism with direct calls to stop/start functions.

Suggested-by: Alex Deucher <alexander.deucher@amd.com>
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/amdgpu/userq: enable support for queue priorities

Enable users to create queues at different priority levels.
The highest level is restricted to drm master.

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

drm/amdgpu/userq/mes: handle user queue priority

Handle the queue priority set by the user.

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

drm/amdgpu/userq: add priorty to user queue structure

So we can track this when we create user queues.

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

drm/amdgpu/mes12: add conversion for priority levels

Convert driver priority levels to MES11 priority levels.
At the moment they are the same, but they may not always
be.

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

drm/amdgpu/mes11: add conversion for priority levels

Convert driver priority levels to MES11 priority levels.
At the moment they are the same, but they may not always
be.

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

drm/amdgpu/userq: add UAPI for setting queue priority

Allow the user to set a queue priority levels:
0 - normal low - most apps (maps to MES AMD_PRIORITY_LEVEL_NORMAL)
1 - low - background jobs (maps to MES AMD_PRIORITY_LEVEL_LOW)
2 - normal high - apps that need relative high (maps to MES AMD_PRIORITY_LEVEL_MEDIUM)
3 - high (admin only - for compositors) (maps to MES AMD_PRIORITY_LEVEL_HIGH)

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

drm/amdgpu: convert userq UAPI _pad to flags

Reuse the _pad field for flags.

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

drm/amd/display: Add error check for avi and vendor infoframe setup function

The function fill_stream_properties_from_drm_display_mode() calls the
function drm_hdmi_avi_infoframe_from_display_mode() and the
function drm_hdmi_vendor_infoframe_from_display_mode(), but does
not check its return value. Log the error messages to prevent silent
failure if either function fails.

Signed-off-by: Wentao Liang <vulab@iscas.ac.cn>
Reviewed-by: Alex Hung <alex.hung@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu/userq: integrate with enforce isolation

Enforce isolation serializes access to the GFX IP.  User
queues are isolated in the MES scheduler, but we still
need to serialize between kernel queues and user queues.
For enforce isolation, group KGD user queues with KFD user
queues.

v2: split out variable renaming, add config guards
v3: use new function names

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

drm/amdgpu: rename enforce isolation variables

Since they will be used for both KFD and KGD user queues,
rename them from kfd to userq.  No intended functional
change.

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

drm/amdgpu/userq: add helpers to start/stop scheduling

This will be used to stop/start user queue scheduling for
example when switching between kernel and user queues when
enforce isolation is enabled.

v2: use idx
v3: only stop compute/gfx queues

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

drm/amdgpu/userq: track the xcp_id associated with the queue

Track this to align with KFD for enforce isolation
handling.

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

drm/amdgpu: Clear overflow for SRIOV

For VF, it doesn't have the permission to clear overflow, clear the bit
by reset.

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

drm/amdgpu/userq: rework driver parameter

Replace disable_kq parameter with user_queue parameter.
The parameter has the following logic:
 -1 = auto (ASIC specific default)
  0 = user queues disabled
  1 = user queues enabled and kernel queues enabled (if supported)
  2 = user queues enabled and kernel queues disabled

The default behavior (-1) is currently the same as 0 for current
ASICs.  To enable user queues (in addition to kernel queues) set
user_queue=1. To enable user queues and disable kernel queues
(to make all resources available to user queues), set user_queue=2.

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

drm/amd/pm: Enable host limit metrics support

Enable host limit metrics support for smuv_13_0_12

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

drm/amdgpu/sdma7: properly reference trap interrupts for userqs

We need to take a reference to the interrupts to make
sure they stay enabled even if the kernel queues have
disabled them.

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

drm/amdgpu/sdma6: properly reference trap interrupts for userqs

We need to take a reference to the interrupts to make
sure they stay enabled even if the kernel queues have
disabled them.

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

drm/amd/pm: Enable host limit metrics support

Enable host limit metrics support for smuv_13_0_6

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

drm/amdgpu: Enable doorbell for JPEG5_0_1

Enable doorbell for JPEG5_0_1 and adjust index for VCN5_0_1.

Signed-off-by: Sathishkumar S <sathishkumar.sundararaju@amd.com>
Reviewed-by: Leo Liu <leo.liu@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: Update vcn doorbell range in NBIO 7.9

Increase vcn doorbell range for gfx950 to 11.

Signed-off-by: Shiwu Zhang <shiwu.zhang@amd.com>
Reviewed-by: Leo Liu <leo.liu@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu/gfx12: properly reference EOP interrupts for userqs

Regardless of whether we disable kernel queues, we need
to take an extra reference to the pipe interrupts for
user queues to make sure they stay enabled in case we
disable them for kernel queues.

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

drm/amdgpu/gfx11: properly reference EOP interrupts for userqs

Regardless of whether we disable kernel queues, we need
to take an extra reference to the pipe interrupts for
user queues to make sure they stay enabled in case we
disable them for kernel queues.

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

drm/amdkfd: fix a bug of smi event for superuser

rocm-smi with superuser permission doesn't show some
of smi events, i.e. page fault/migration, because the
condition of "(events & all)" is false. Superuser
should be able to detect all events, the condiiton of
"(events & all)" seems redundant, so removing it will
fix the issue.

Signed-off-by: Eric Huang <jinhuieric.huang@amd.com>
Reviewed-by: Kent Russell <kent.russell@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: add missing DCE6 to dce_version_to_string()

Missing DCE 6.0 6.1 and 6.4 are identified as UNKNOWN. Fix this.

Signed-off-by: Alexandre Demers <alexandre.f.demers@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: fix typo in bios_parser.c

Probably a cut and paste error from using get_integrated_info_v8's comment.
This has to be get_integrated_info_v9

Signed-off-by: Alexandre Demers <alexandre.f.demers@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: fix duplicated value setting in dce100_resource_construct()

i2c_speed_in_khz was set twice with the same values. Looking at other DCE
versions, we probably wanted to set the value for i2c_speed_in_khz_hdcp.

Signed-off-by: Alexandre Demers <alexandre.f.demers@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/radeon: fix typo in atombios.h

"aligned" not "aligend"

Signed-off-by: Alexandre Demers <alexandre.f.demers@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: fix typo in atombios.h

"aligned" not "aligend"

Signed-off-by: Alexandre Demers <alexandre.f.demers@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: add missing parameter name in dce110_clk_src_construct() declaration

While not needed per speaking, all the other parameters have names but
this one.

Signed-off-by: Alexandre Demers <alexandre.f.demers@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: rename function to follow naming convention in dce110

The prefix dce110 is used on all functions, but init_pipes() and
init_hw(). Under DCN, these sames functions are prefixed.

Let's keep thing coherent.

Signed-off-by: Alexandre Demers <alexandre.f.demers@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: Clean up error handling in amdgpu_userq_fence_driver_alloc()

1) Checkpatch complains if we print an error message for kzalloc()
   failure.  The kzalloc() failure already has it's own error messages
   built in.  Also this allocation is small enough that it is guaranteed
   to succeed.
2) Return directly instead of doing a goto free_fence_drv.  The
   "fence_drv" is already NULL so no cleanup is necessary.

Reviewed-by: Arvind Yadav <arvind.yadav@amd.com>
Signed-off-by: Dan Carpenter <dan.carpenter@linaro.org>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: Fix double free in amdgpu_userq_fence_driver_alloc()

The goto frees "fence_drv" so this is a double free bug.  There is no
need to call amdgpu_seq64_free(adev, fence_drv->va) since the seq64
allocation failed so change the goto to goto free_fence_drv.  Also
propagate the error code from amdgpu_seq64_alloc() instead of hard coding
it to -ENOMEM.

Fixes: e7cf21fbb277 ("drm/amdgpu: Few optimization and fixes for userq fence driver")
Reviewed-by: Arvind Yadav <arvind.yadav@amd.com>
Signed-off-by: Dan Carpenter <dan.carpenter@linaro.org>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu/userq: move runpm handling into core userq code

Pull it out of the MES code and into the generic code.
It's not MES specific and needs to be applied to all user
queues regardless of the backend.

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

drm/amdkfd: fix NULL check mistake for process smi event

The mistake will lead to NULL kernel oops, so fix it.

Fixes: 4172b556fd5b ("drm/amdkfd: add smi events for process start and end")
Signed-off-by: Eric Huang <jinhuieric.huang@amd.com>
Reviewed-by: Kent Russell <kent.russell@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu/sdma_v4: Register the new sdma function pointers

Register stop/start/soft_reset queue functions for sdma v4_4_2.

Suggested-by: Alex Deucher <alexander.deucher@amd.com>
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/amdgpu: Add the new sdma function pointers for amdgpu_sdma.h

This patch introduces new function pointers in the amdgpu_sdma structure
to handle queue stop, start and soft reset operations. These will replace
the older callback mechanism.

The new functions are:
- stop_kernel_queue: Stops a specific SDMA queue
- start_kernel_queue: Starts/Restores a specific SDMA queue
- soft_reset_kernel_queue: Performs soft reset on a specific SDMA queue

v2: Update stop_queue/start_queue function paramters to use ring pointer instead of device/instance(Chritian)
v3: move stop_queue/start_queue to struct amdgpu_sdma_instance and rename them. (Alex)
v4: rework the ordering a bit (Alex)

Suggested-by: Alex Deucher <alexander.deucher@amd.com>
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/amdgpu: don't swallow errors in amdgpu_userqueue_resume_all()

since we loop through the queues |= the errors.

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

drm/amdgpu/userq: handle system suspend and resume

Unmap user queues on suspend and map them on resume.

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

drm/amdgpu/userq: add suspend and resume helpers

Add helpers to unmap and map user queues on suspend and
resume.

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

drm/amdgpu/userq: properly clean up userq fence driver on failure

If userq creation fails, we need to properly unwind and free the
user queue fence driver.

v2: free idr as well (Sunil)

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

drm/amdgpu/userq: move some code around

Move some userq fence handling code into amdgpu_userq_fence.c.
This matches the other code in that file.

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

drm/amdgpu/userq: rework front end call sequence

Split out the queue map from the mqd create call and split
out the queue unmap from the mqd destroy call.  This splits
the queue setup and teardown with the actual enablement
in the firmware.

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

drm/amdgpu/userq: rename suspend/resume callbacks

Rename to map and umap to better align with what is happening
at the firmware level and remove the extra level of indirection
in the MES userq code.

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

drm/amdgpu/userq/mes: remove unused header

This is unused so remove it.

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

drm/amdkfd: Add rec SDMA engines support with limited XGMI

This patch adds recommended SDMA engines with limited XGMI SDMA engines.
It will help improve overall performance for device to device copies
with this optimization.

v2: Update the formatting issues and data type

Signed-off-by: Shane Xiao <shane.xiao@amd.com>
Suggested-by: Jonathan Kim <jonathan.kim@amd.com>
Reviewed-by: Jonathan Kim <jonathan.kim@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: Enhance Cleaner Shader Handling in GFX v9.0 Architecture v2

This commit modifies the gfx_v9_0_ring_emit_cleaner_shader function
to use a switch statement for cleaner shader emission based on the
specific GFX IP version.

The function now distinguishes between different IP versions, using
PACKET3_RUN_CLEANER_SHADER_9_0 for the versions 9.0.1, 9.1.0,
9.2.1, 9.2.2, 9.3.0, and 9.4.0, while retaining
PACKET3_RUN_CLEANER_SHADER for version 9.4.2.

v2: Simplify logic (Alex).

Cc: Christian König <christian.koenig@amd.com>
Cc: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Srinivasan Shanmugam <srinivasan.shanmugam@amd.com>
Suggested-by: Alex Deucher <alexander.deucher@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: Add PACKET3_RUN_CLEANER_SHADER_9_0 for Cleaner Shader execution

This commit introduces the PACKET3_RUN_CLEANER_SHADER_9_0 definition,
which is a command packet utilized to instruct the GPU to execute the
cleaner shader for the GFX9.0 graphics architecture.

The cleaner shader is a piece of GPU code that is responsible for
clearing or initializing essential GPU resources, such as Local Data
Share (LDS), Vector General Purpose Registers (VGPRs), and Scalar
General Purpose Registers (SGPRs). Properly clearing these resources  is
vital for ensuring data isolation and security between different
workloads executed on the GPU.

When the GPU receives this packet, it fetches and runs the cleaner
shader instructions from the specified location in the packet.  Thus by
preventing data leaks and ensuring that previous job states do not
interfere with subsequent 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/amdgpu: Fix out of bounds warning in amdgpu_hw_ip_info

Fix an array index out of bounds warning in the DMA IP case of
amdgpu_hw_ip_info() where it was incorrectly checking
adev->gfx.gfx_ring[i].no_user_submission instead of
adev->sdma.instance[i].ring.no_user_submission.

The mismatch caused UBSAN to report an array bounds violation since
it was accessing the GFX ring array with SDMA instance indices.

Fixes: 4310acd4464b ("drm/amdgpu: add ring flag for no user submissions")
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Jesse Zhang <jesse.zhang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdkfd: add smi events for process start and end

rocm-smi will be able to show the events for KFD process
start/end, it is the implementation of this feature.

Signed-off-by: Eric Huang <jinhuieric.huang@amd.com>
Reviewed-by: Kent Russell <kent.russell@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: Use the right function for hdp flush

There are a few prechecks made before HDP flush like a flush is not
required on APU bare metal. Using hdp callback directly bypasses those
checks. Use amdgpu_device_flush_hdp which takes care of prechecks.

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

drm/amdgpu: Direct ret in ras_reset_err_cnt on VF

With adding sriov_vf check, we directly return EOPNOTSUPP in
ras_reset_error_count as we should not do anything on VF to reset RAS error
count.

This also fixes the issue that loading guest driver causes register
violations.

Reviewed-by: Ahmad Rehman <Ahmad.Rehman@amd.com>
Signed-off-by: Ellen Pan <yunru.pan@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: Use generic hdp flush function

Except HDP v5.2 all use a common logic for HDP flush. Use a generic
function. HDP v5.2 forces NO_KIQ logic, revisit it later.

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

drm/amdgpu: Set RAS EEPROM table version to v3 for umc v12_5

Set RAS EEPROM table version to v3 for umc v12_5.

Signed-off-by: Candice Li <candice.li@amd.com>
Reviewed-by: Tao Zhou <tao.zhou1@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: Enable per-queue reset for SDMA v4.4.2 on IP v9.5.0

Add support for per-queue reset on SDMA v4.4.2 when running with:
1. MEC firmware version 17 or later
2. DPM indicates SDMA reset is supported

v2: Fixed supported firmware versions  (Lijo)

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>