drm/amdkfd: allow debug subscription to lds violations on gfx 1250

GFX 1250 allows the debugger to subcribe to LDS out-of-range read/write
memory violations.
Bump IOCTL minor version and flag KFD capabilities for enablement
hint.

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: enable gpu tlb flush for gfxhub

Enable gpu tlb flush for gfxhub without check gfx.is_poweron
as gfx is power on by default for gfx v12_1 ASIC.

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

drm/amd/include : Update MES v12 API header

Add LDS out of range reporting support in mes API

Signed-off-by: Shaoyun Liu <shaoyun.liu@amd.com>
Reviewed-by: Jonathan Kim <jonathan.kim@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: flush tlb properly for GMC v12.1 in early phase

Flush tlb properly for GMC v12.1

Signed-off-by: Le Ma <le.ma@amd.com>
Reviewed-by: Hawking Zhang <Hawking.Zhang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: Use AMDGPU_IS_GFXHUB to screen out GFXHUB for GMC v12.1

There're multiple gfxhubs on GMC v12.1

Signed-off-by: Le Ma <le.ma@amd.com>
Reviewed-by: Hawking Zhang <Hawking.Zhang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: only copy ucode for enabled xcc

Only copy ucode for enabled xcc instead of copy for all 8 xcc
for rlc autoload on gfx v12_1 to save time.

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

drm/amdgpu: fix issue when switch NPS1 to NPSX

fix the function execution sequence after removing
kgd2kfd_init_zone_device out of gpu full access region.

Fixes: c71980a3fc1d ("drm/amdgpu: reduce the full gpu access time in amdgpu_device_init.")
Signed-off-by: chong li <chongli2@amd.com>
Reviewed-by: Lijo Lazar <lijo.lazar@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Fix 64-bit state pointer passed as 32-bit GPINT response buffer

edp_pr_get_state() incorrectly casts a uint64_t * to uint32_t * when
calling dc_wake_and_execute_gpint(). The GPINT path writes only 32 bits,
leaving the upper 32 bits of the u64 output uninitialized. Replace the
cast with a u32 temporary and copy the result into the u64 pointer.

Fixes the below:
drivers/gpu/drm/amd/amdgpu/../display/dc/link/protocols/link_edp_panel_control.c
    1448 bool edp_pr_get_state(const struct dc_link *link, uint64_t *state)
                                                           ^^^^^^^^^^^^^^^
    1449 {

    ...

    1457         do {
    1458                 // Send gpint command and wait for ack
--> 1459                 if (!dc_wake_and_execute_gpint(dc->ctx, DMUB_GPINT__GET_REPLAY_STATE, panel_inst,
    1460                                                (uint32_t *)state, DM_DMUB_WAIT_TYPE_WAIT_WITH_REPLY)) {
                                                        ^^^^^^^^^^^^^^^^^

The dc_wake_and_execute_gpint() function doesn't take a u64, it takes a
u32.  It tries to initialize the state to zero at the start but that's
not going to work because of the type mismatch.  It suggests that
callers are allowed to pass uninitialized data to edp_pr_get_state() but
at present there are no callers so this is only a bug in the code but
doesn't affect runtime.

    1461                         // Return invalid state when GPINT times out
    1462                         *state = PR_STATE_INVALID;
    1463                 }

Fixes: 74ce00932e7e ("drm/amd/display: Refactor panel replay set dmub cmd flow")
Reported by: Dan Carpenter <dan.carpenter@linaro.org>
Cc: Robin Chen <robin.chen@amd.com>
Cc: Jack Chang <jack.chang@amd.com>
Cc: Leon Huang <Leon.Huang1@amd.com>
Cc: Alex Hung <alex.hung@amd.com>
Cc: Aurabindo Pillai <aurabindo.pillai@amd.com>
Cc: Roman Li <roman.li@amd.com>
Cc: Harry Wentland <harry.wentland@amd.com>
Cc: Tom Chung <chiahsuan.chung@amd.com>
Signed-off-by: Srinivasan Shanmugam <srinivasan.shanmugam@amd.com>
Reviewed-by: Aurabindo Pillai <aurabindo.pillai@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/include : Update MES v12 comments on RESET API

Added comments for the layout of contents that addressed by doorbell_offset_addr
in RESET API

Signed-off-by: Shaoyun Liu <shaoyun.liu@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: Set xcp id for mes ring

Set xcp id for mes ring

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

drm/amdgpu: Init partition_mode and xcc_mask for GFX_IMU_PARTITION_SWITCH

Set partition_mode and physical xcc mask fields in
GFX_IMU_PARTITION_SWITCH register

v2: cleanup (Alex)

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

drm/amdgpu: Initialize vram_info for gmc v12_1

Initialize vram_info for gmc v12_1

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

drm/amdgpu: Init compute partition mode for gfx v12_1

Init compute partition mode for gfx v12_1

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

drm/amdgpu: Initialize memory ranges for gmc v12_1

Initialize memory ranges for gmc v12_1

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

drm/amdgpu: Initialize memory partition callbacks for gmc v12_1

Initialize memory partition callbacks for gmv v12_1

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

drm/amdgpu: support rlc autoload for muti-xcc

Support rlc autload for muti-xcc on gfx v12_1.

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

drm/amdgpu: Enable atomics for all the available xcc

Apply TCP_UTCL0_CNTL1 settings to all the available
xcc

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

drm/amdgpu: Update MES VM_CNTX_CNTL for XNACK off for GFX 12.1

Currently, we do not turn off retry faults in VM_CONTEXT_CNTL value
when passing it to MES if XNACK is off. This creates a situation where
XNACK is disabled in SQ but enabled in UTCL2, which is not recommended.
As a result, turn off/on retry faults in both SQ and UTCL2 when passing
vm_context_cntl value to MES if XNACK is disabled/enabled.

Suggested-by: Jay Cornwall <jay.cornwall@amd.com>
Signed-off-by: Mukul Joshi <mukul.joshi@amd.com>
Reviewed-by: Philip Yang <Philip.Yang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdkfd: Enable per-process XNACK for GFX 12.1.0

GFX 12.1.0 will support enabling/disabling XNACK on a per-
process basis. This change enables the per process XNACK feature.

Signed-off-by: Mukul Joshi <mukul.joshi@amd.com>
Reviewed-by: Philip Yang <Philip.Yang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: Enable retry faults for GFX 12.1

Enable retry faults in both GCVM/MMVM Context1 Control
and L2_PROTECTION_FAULT_CNTL2 registers for GFX 12.1.

Signed-off-by: Mukul Joshi <mukul.joshi@amd.com>
Reviewed-by: Philip Yang <Philip.Yang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: Add IH node-id to XCC mapping

Add a generic function to map IH node-id to XCC instance.

Signed-off-by: Likun Gao <Likun.Gao@amd.com>
Signed-off-by: Mukul Joshi <mukul.joshi@amd.com>
Reviewed-by: Hawking Zhang <Hawking.Zhang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: Add interrupt handler for GFX 12.1.0

Add a separate interrupt handler for handling interrupts,
both retry and no-retry, for GFX 12.1.0.

Signed-off-by: Mukul Joshi <mukul.joshi@amd.com>
Reviewed-by: Philip Yang <Philip.Yang@amd.com>
Reviewed-by: Hawking Zhang <Hawking.Zhang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: Add UTCL2 Retry fault interrupt for GFX 12.1

Add the UTCL2 retry fault interrupt for both GCVM and MMVM for GFX 12.1.

Signed-off-by: Mukul Joshi <mukul.joshi@amd.com>
Reviewed-by: Philip Yang <Philip.Yang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu/sdma: add query for CSA size and alignment

Needed to query the CSA size and alignment for SDMA
user queues.

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

drm/amdgpu: fix mes packet params issue when flush hdp.

v4:
use func "amdgpu_gfx_get_hdp_flush_mask" to get ref_and_mask for
gfx9 through gfx12.

v3:
Unify the get_ref_and_mask function in amdgpu_gfx_funcs,
to support both GFX11 and earlier generations

v2:
place "get_ref_and_mask" in amdgpu_gfx_funcs instead of amdgpu_ring,
since this function only assigns the cp entry.

v1:
both gfx ring and mes ring use cp0 to flush hdp, cause conflict.

use function get_ref_and_mask to assign the cp entry.
reassign mes to use cp8 instead.

Signed-off-by: chong li <chongli2@amd.com>
Acked-by: Lijo Lazar <lijo.lazar@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu/gfx: add eop size and alignment to shadow info

This is used by firmware for compute user queues.

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

drm/amd/ras: Add vram_type to ras_ta_init_flags

Add vram_type to ras_ta_init_flags.

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: update sdma configuration for soc v1_0

Update SDMA instances/masks according to xcc num for
multi-xcc models on soc v1.0.

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

drm/amdgpu: Initialize xcp manager for soc v1_0

Initialize xcp manager for soc v1_0

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

drm/amdgpu: Add soc_v1_0_xcp_funcs

Implement xcp mgr callbacks for soc v1_0

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

drm/amdgpu: Export sdma_v7_1_xcp_funcs

To be used by soc v1_0 xcp manager

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

drm/amdgpu: Export gfx_v12_1_xcp_func

To be used by soc v1_0 xcp manager

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

drm/amdgpu: Add vram_type to ras init_flags

Add vram_type to ras init_flags.

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/amd/ras: Reduce stack usage in amdgpu_virt_ras_get_cper_records()

amdgpu_virt_ras_get_cper_records() was using a large stack array
of ras_log_info pointers. This contributed to the frame size
warning on this function.

Replace the fixed-size stack array:

    struct ras_log_info *trace[MAX_RECORD_PER_BATCH];

with a heap-allocated array using kcalloc().

We free the trace buffer together with out_buf on all exit paths.
If allocation of trace or out_buf fails, we return a generic RAS
error code.

This reduces stack usage and keeps the runtime behaviour
unchanged.

Fixes:
stack frame size: 1112 bytes (limit: 1024)

Cc: Tao Zhou <tao.zhou1@amd.com>
Cc: Hawking Zhang <Hawking.Zhang@amd.com>
Cc: Christian König <christian.koenig@amd.com>
Cc: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Srinivasan Shanmugam <srinivasan.shanmugam@amd.com>
Reviewed-by: Tao Zhou <tao.zhou1@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdkfd: Handle GPU reset and drain retry fault race

Only check and drain IH1 ring if CAM is not enabled.

If GPU is under reset, don't access IH to drain retry fault.

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

Revert "drm/amd/display: Fix pbn to kbps Conversion"

Deeply daisy chained DP/MST displays are no longer able to light
up. This reverts commit e0dec00f3d05 ("drm/amd/display: Fix pbn
to kbps Conversion")

Cc: Jerry Zuo <jerry.zuo@amd.com>
Reported-by: nat@nullable.se
Closes: https://gitlab.freedesktop.org/drm/amd/-/issues/4756
Signed-off-by: Mario Limonciello <mario.limonciello@amd.com>
Acked-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: Add switch_compute_partition callback for imu v12_1

To enable switching compute partition mode

v2: cleanup (Alex)

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

drm/amdgpu: Implement gfx_v12_1_get_xccs_per_xcp

Use gfx v12_1 callback to query the numbers of xccs
per xcp

v2: add todo (Alex)

Signed-off-by: Hawking Zhang <Hawking.Zhang@amd.com>
Reviewed-by: Le Ma <le.ma@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: Remove redundant check for async_gfx_ring

Remove the redundant check for async_gfx_ring,
as it is not required for gfx v12_1

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

drm/amdgpu: disable graphics doorbell range for gfx v12_1

Disable doorbell range for graphics engine on gfx v12_1.

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

drm/amdgpu: enable unmap doorbell handle for gfx v12_1

Enable unmapped doorbell handling for gfx v12_1.

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

drm/amdgpu: revision doorbel range for gfx v12_1

Revision doorbell range on muti-XCC mode for gfx v12_1.
Clean up doorbell range set for graphics engine.
V2: Remove doorbell range set from gfx_v12_1_xcc_kiq_init_register.

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

drm/amdkfd: disable shader message vgpr deallocation on gc 12.1

Shader messages to deallocate VGPRs prior to shader end can prevent
the trap handler from saving context, making debugging and core dumps
unreliable.

VGPR deallocations for performance gain is negligible.
GC 12.1 will NOP shader VGPR deallocation messages via HW
settings on driver boot.

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

drm/amdgpu: Remove redundant pmfw backdoor loading

PMFW is integrated into ifwi for gfx 12_1 adapter,
making PMFW backdoor loading unnecessary.

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

drm/amd: Fix unbind/rebind for VCN 4.0.5

Unbinding amdgpu has no problems, but binding it again leads to an
error of sysfs file already existing.  This is because it wasn't
actually cleaned up on unbind.  Add the missing cleanup step.

Fixes: 547aad32edac ("drm/amdgpu: add VCN4 ip block support")
Signed-off-by: Mario Limonciello (AMD) <superm1@kernel.org>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu/acpi: Reduce amdgpu_acpi_detect stack usage

amdgpu_acpi_detect() calls some helper functions it calls have large
local structures.  When the compiler inlines these helpers, their local
data adds to the amdgpu_acpi_detect() stack frame.

Mark the helpers with noinline_for_stack:
- amdgpu_atif_verify_interface()
- amdgpu_atif_get_notification_params()
- amdgpu_atif_query_backlight_caps()
- amdgpu_atcs_verify_interface()
- amdgpu_acpi_enumerate_xcc()

This keeps the large temporary objects inside the helper’s own stack
frame instead of being inlined into the caller, preventing the caller
from growing beyond the stack limit.

Fixes the below with gcc W=1:
drivers/gpu/drm/amd/amdgpu/amdgpu_acpi.c:1403:6: warning: stack frame size (1688) exceeds limit (1024) in 'amdgpu_acpi_detect' [-Wframe-larger-than]

Cc: Alex Deucher <alexander.deucher@amd.com>
Cc: Christian König <christian.koenig@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: pass the entity to use to ttm public functions

This way the caller can select the one it wants to use.

Signed-off-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Acked-by: Felix Kuehling <felix.kuehling@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: pass the entity to use to amdgpu_ttm_map_buffer

This way the caller can select the one it wants to use.

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

drm/amdgpu: fix error handling in amdgpu_copy_buffer

drm_sched_job_add_resv_dependencies can fail in amdgpu_ttm_prepare_job.
In this case we need to use amdgpu_job_free to release memory.

---
v4: moved job pointer clearing to a different patchset
---

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

drm/amdgpu: add amdgpu_ttm_job_submit helper

Deduplicate the IB padding code and will also be used
later to check locking.

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

drm/amdgpu: introduce amdgpu_ttm_buffer_entity

No functional change for now, but this struct will have more
fields added in the next commit.

This change would introduce synchronisation issue, because
dependencies between successive jobs are not taken care of
properly. For instance, amdgpu_ttm_clear_buffer uses
amdgpu_ttm_map_buffer then amdgpu_ttm_fill_mem which should
use different entities (default_entity then move/clear entity).
To prevent failures for this commit, we limit ourselves to
2 entities: default_entity (which replaces high_pr usages) and
clear_entity (which replaces low_pr usages).

The next commits will deal with these dependencies correctly,
and then we'll be able to use move_entity.

---
v2: renamed amdgpu_ttm_buffer_entity
v4: don't use move_entity in ttm yet
---

Signed-off-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com> (v3)
Acked-by: Felix Kuehling <felix.kuehling@amd.com> (v3)
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: add imu support for gc 12_1

Add IMU support for gc version 12.1.0.
Only support imu fw loading for imu 12.1.0.

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

drm/amdgpu: fix mes code error for muti-xcc

Fix some code error for muti-xcc on mes v12_1.

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

drm/amdgpu/gmc12: set MMHUBs based on aid_mask

Update number of mmhub and mid_mask via reuse aid_mask.

Signed-off-by: Likun Gao <Likun.Gao@amd.com>
Reviewed-by: Le Ma <le.ma@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: set gfxhub according xcc_mask for gfx 12_1

Set GFXHUB accodring to xcc_mask for gfx version 12_1.

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

drm/amdgpu: add xcc info for compute ring name

Add XCC id info for compute ring name on gfx version 12.1.

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

drm/amdkfd: bump minimum vgpr size for gfx1151

GFX1151 has 1.5x the number of available physical VGPRs per SIMD.
Bump total memory availability for acquire checks on queue creation.

Signed-off-by: Jonathan Kim <jonathan.kim@amd.com>
Reviewed-by: Mario Limonciello <mario.limonciello@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: Revert retry based thrashing prevention on GFX 12.1.0

Revert the change to enable retry based thrashing prevention on GFX 12.1.0
for now as its causing data mismatch and slowness issues with multiple HIP
tests.

Signed-off-by: Mukul Joshi <mukul.joshi@amd.com>
Reviewed-by: Harish Kasiviswanathan <Harish.Kasiviswanathan@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: Init single mes instance if xcc_mask is unset

Configure a single mes instance if the xcc_mask remains
uninitialized.

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

drm/amdgpu: Setup MTYPEs for GFX 12.1.0

For GFX 12.1.0, setup correct MTYPE for a BO depending on
its current location relative to the mapping GPU.

Signed-off-by: Mukul Joshi <mukul.joshi@amd.com>
Reviewed-by: Alex Sierra <alex.sierra@amd.com>
Reviewed-by: Harish Kasiviswanathan <Harish.Kasiviswanathan@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: update sh mem base offsets for gfx 12.1

Signed-off-by: Alex Sierra <alex.sierra@amd.com>
Reviewed-by: Philip Yang <Philip.Yang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/ras: Reduce stack usage in ras_umc_handle_bad_pages()

ras_umc_handle_bad_pages() function used a large local array:
  struct eeprom_umc_record records[MAX_ECC_NUM_PER_RETIREMENT];

Move this array off the stack by allocating it with kcalloc()
and freeing it before return.

This reduces the stack frame size of ras_umc_handle_bad_pages()
and avoids the frame size warning.

Fixes the below:
drivers/gpu/drm/amd/amdgpu/../ras/rascore/ras_umc.c:498:5: warning: stack frame size (1208) exceeds limit (1024) in 'ras_umc_handle_bad_pages' [-Wframe-larger-than]

v2: Removed the duplicate ras_umc_get_new_records() invocation. (Lijo)

Cc: Tao Zhou <tao.zhou1@amd.com>
Cc: Hawking Zhang <Hawking.Zhang@amd.com>
Cc: Christian König <christian.koenig@amd.com>
Cc: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Srinivasan Shanmugam <srinivasan.shanmugam@amd.com>
Reviewed-by: Tao Zhou <tao.zhou1@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: Fix SHMEM alignment mode for GFX 12.1.0

Alignment mode in SHMEM config register is only a single bit
value on GFX 12.1.0 instead of 2 bits in previous asics.
Add a new enum and use the correct value of SHMEM alignment mode
when programming the SHMEM config register.

Signed-off-by: Mukul Joshi <mukul.joshi@amd.com>
Reviewed-by: Hawking Zhang <Hawking.Zhang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: shrink struct members

On a 32-bit ARM system, the audio_decoder struct ends up being too large
for dp_retrain_link_dp_test.

link_dp_cts.c:157:1: error: the frame size of 1328 bytes is larger than
1280 bytes [-Werror=frame-larger-than=]

This is mitigated by shrinking the members of the struct and avoids
having to deal with dynamic allocation.

feed_back_divider is assigned but otherwise unused. Remove both.

pixel_repetition looks like it should be a bool since it's only ever
assigned to 1. But there are checks for 2 and 4. Reduce to uint8_t.

Remove ss_percentage_divider. Unused.

Shrink refresh_rate as it gets assigned to at most a 3 digit integer
value.

Signed-off-by: Rosen Penev <rosenp@gmail.com>
Reviewed-by: Alex Hung <alex.hung@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdkfd: Export the cwsr_size and ctl_stack_size to userspace

This is important for userspace to avoid hardcoding VGPR size.

Reviewed-by: Kent Russell <kent.russell@amd.com>
Signed-off-by: Mario Limonciello <mario.limonciello@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

amdkfd: Bump ABI to indicate presence of Trap handler support for expert scheduling

commit 0f0c8a6983db ("drm/amdkfd: Trap handler support for expert
scheduling mode") introduced support for a trap handler when expert
scheduling mode. However userspace needs to know whether or not a trap
handler support is present.

Bump the KFD IOCTL API so that userspace can key off this to decide.

Suggested-by: Stella Laurenzo <stella.laurenzo@amd.com>
Fixes: 423888879412 ("drm/amdkfd: Trap handler support for expert scheduling mode")
Reviewed-by: Kent Russell <kent.russell@amd.com>
Signed-off-by: Mario Limonciello <mario.limonciello@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Promote DC to 3.2.362

This version brings along the following updates:

 - Defer transitions from minimal state to final state
 - Remove periodic detection callbacks from dcn35+
 - Fixes for S0i3 exit
 - Refactor dml_core_mode_support to reduce stack frame
 - Add additional info from DML for DMU

Signed-off-by: Taimur Hassan <Syed.Hassan@amd.com>
Signed-off-by: Chenyu Chen <chen-yu.chen@amd.com>
Acked-by: Tom Chung <chiahsuan.chung@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Additional info from DML for DMU

[WHAT]
Add additional info from DML for DMU when applicable
on future platforms.

Reviewed-by: Alvin Lee <alvin.lee2@amd.com>
Signed-off-by: Nevenko Stupar <Nevenko.Stupar@amd.com>
Signed-off-by: Chenyu Chen <chen-yu.chen@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Refactor dml_core_mode_support to reduce stack frame

[WHAT]
When compiling Linux kernel with clang, the following warning / error
messages pops up:

drivers/gpu/drm/amd/amdgpu/../dal-dev/dc/dml2_0/display_mode_core.c:6853:12:
error: stack frame size (2120) exceeds limit (2056) in
'dml_core_mode_support' [-Werror,-Wframe-larger-than]
 6853 | dml_bool_t dml_core_mode_support(struct display_mode_lib_st
*mode_lib)

[HOW]
Refactoring CalculateVMRowAndSwath_params assignments to a new function
helps reduce the stack frame size in dml_core_mode_support.

Closes: https://gitlab.freedesktop.org/drm/amd/-/issues/4733
Reviewed-by: Austin Zheng <austin.zheng@amd.com>
Reviewed-by: Aurabindo Pillai <aurabindo.pillai@amd.com>
Signed-off-by: Alex Hung <alex.hung@amd.com>
Signed-off-by: Chenyu Chen <chen-yu.chen@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Fixes for S0i3 exit

[why & how]
Add debug flag "ignore_pg" to dcn32 PG functions.
Update default z10 support status.
Temp disable RFB features for ASIC.
Remove legacy code path.

Reviewed-by: Charlene Liu <charlene.liu@amd.com>
Signed-off-by: Ovidiu Bunea <ovidiu.bunea@amd.com>
Signed-off-by: Chenyu Chen <chen-yu.chen@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Remove periodic detection callbacks from dcn35+

[WHY&HOW]
These will not be needed going forward as DMU will communicate to the
driver when detection may be needed after a power saving event.

Reviewed-by: Nicholas Kazlauskas <nicholas.kazlauskas@amd.com>
Signed-off-by: Dillon Varone <Dillon.Varone@amd.com>
Signed-off-by: Chenyu Chen <chen-yu.chen@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Defer transitions from minimal state to final state

[Why]
In non-seamless pipe transitions, it can take several frames to process
a single flip. One of the reasons is the 2-step transition implementation
where first the minimal transition state is applied, then the final state
is applied, all within the same flip. This delay is noticeable to the user
in some video playback scenarios, which makes for a bad user experience.

[How]
 - in applicable non-seamless cases, complete the flip with the minimal
   state applied, start a counter, and create all new contexts as minimal
 - if another pipe transition occurs while counting, reset the counter
 - when the counter finishes, promote the current flip to a full update
   and restore creation of optimized contexts
 - when creating minimal states from new context, apply stream updates

Reviewed-by: Aric Cyr <aric.cyr@amd.com>
Signed-off-by: Joshua Aberback <joshua.aberback@amd.com>
Signed-off-by: Chenyu Chen <chen-yu.chen@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: don't attach the tlb fence for SI

SI hardware doesn't support pasids, user mode queues, or
KIQ/MES so there is no need for this.  Doing so results in
a segfault as these callbacks are non-existent for SI.

Closes: https://gitlab.freedesktop.org/drm/amd/-/issues/4744
Fixes: f3854e04b708 ("drm/amdgpu: attach tlb fence to the PTs update")
Reviewed-by: Timur Kristóf <timur.kristof@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Use GFP_ATOMIC in dc_create_plane_state()

This can get called from an atomic context.

Closes: https://gitlab.freedesktop.org/drm/amd/-/issues/4470
Reviewed-by: Harry Wentland <harry.wentland@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdkfd: Trap handler support for expert scheduling mode

The trap may be entered with dependency checking disabled.
Wait for dependency counters and save/restore scheduling mode.

v2:

Use ttmp1 instead of ttmp11. ttmp11 is not zero-initialized.
While the trap handler does zero this field before use, a user-mode
second-level trap handler could not rely on this being zero when
using an older kernel mode driver.

v3:

Use ttmp11 primarily but copy to ttmp1 before jumping to the
second level trap handler. ttmp1 is inspectable by a debugger.
Unexpected bits in the unused space may regress existing software.

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

drm/amdgpu/sdma_v7_1: Add missing inst_mask entry in sdma_v7_1_inst_gfx_resume()

The comment for sdma_v7_1_inst_gfx_resume() did not include the
inst_mask parameter, even though the function takes it as an argument.

Update the comment to document inst_mask as the mask of SDMA engine
instances to be enabled.

Fixes the below with gcc W=1:
drivers/gpu/drm/amd/amdgpu/sdma_v7_1.c:644 function parameter 'inst_mask' not described in 'sdma_v7_1_inst_gfx_resume'

Cc: Likun Gao <Likun.Gao@amd.com>
Cc: Le Ma <le.ma@amd.com>
Cc: Alex Deucher <alexander.deucher@amd.com>
Cc: Christian König <christian.koenig@amd.com>
Signed-off-by: Srinivasan Shanmugam <srinivasan.shanmugam@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/pm: Use common helper for smuv14.0.2 dpm

Use helper function to print smuv14.0.2 dpm tables to sysfs.

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/amd/pm: Use common helper for smuv13.0.7 dpm

Use helper function to print smuv13.0.7 dpm tables to sysfs.

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/amd/pm: Use common helper for smuv13.0.6 dpm

Use helper function to print navi10 dpm tables to sysfs.

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/amd/pm: Use common helper for smuv13.0.0 dpm

Use helper function to print smuv13.0.0 dpm tables to sysfs.

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/amd/pm: Use common helper for aldebaran dpm table

Use helper function to print aldebaran dpm tables to sysfs. Remove
unused functions.

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/amd/pm: Use common helper for sienna dpm table

Use helper function to print sienna cichlid dpm tables to sysfs.

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/amd/pm: Use common helper for navi10 dpm table

Use helper function to print navi10 dpm tables to sysfs. Also, remove
FCLK table as it is not supported in navi10.

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/amd/pm: Use common helper for arcturus dpm

Use the helper function to print DPM clock levels to sysfs.

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/amd/pm: Add a helper to show dpm table

Add a helper function to print clock and pcie dpm levels through sysfs.

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/amd/pm: Use generic pcie dpm table for SMUv14

Use smu_pcie_table for SMUv14 SOCs.

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/amd/pm: Use generic pcie dpm table for SMUv13

Use smu_pcie_table for SMUv13 SOCs.

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/amd/pm: Use generic pcie dpm table for SMUv11

Use smu_pcie_table for SMUv11 SOCs.

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/amd/pm: Add generic pcie dpm table

Add a generic pcie dpm table which contains the number of link clock
levels and link clock, pcie gen speed/width corresponding to each level.

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/amd/pm: Use generic dpm table for SMUv14 SOCs

Use the generic dpm table structure instead of SMUv14 specific table.

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/amd/pm: Use generic dpm table for SMUv13 SOCs

Use the generic dpm table structure instead of SMUv13 specific table.

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>

amdkfd: introduce new ioctl AMDKFD_IOC_CREATE_PROCESS

This commit implemetns a new ioctl AMDKFD_IOC_CREATE_PROCESS
that creates a new secondary kfd_progress on the FD.

To keep backward compatibility, userspace programs need to invoke
this ioctl explicitly on a FD to create a secondary
kfd_process which replacing its primary kfd_process.

This commit bumps ioctl minor version.

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

amdkfd: process debug trap ioctl only on a primary context

Set_debug_trap ioctl should work on a specific kfd_process
even when multiple contexts feature is implemented.

For consistency, this commit allow set_debug_trap ioctl only work on the
primary kfd process of a user space program

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

drm/amd/pm: Use generic dpm table for SMUv11 SOCs

Remove SMUv11 specific DPM table and use the generic dpm table
structure.

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/amd/pm: Add clock table structure

Add a common clock table structure to represent dpm levels for different
clocks.

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/amdkfd: Use huge page size to check split svm range alignment

When split svm ranges that have been mapped using huge page should use huge
page size(2MB) to check split range alignment, not prange->granularity that
means migration granularity.

Fixes: 7ef6b2d4b7e5 ("drm/amdkfd: remap unaligned svm ranges that have split")
Signed-off-by: Xiaogang Chen <xiaogang.chen@amd.com>
Reviewed-by: Philip Yang <Philip.Yang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: add a helper for processing recoverable GPUVM faults

Add a common helper to remove the repeated logic from each
gmc module.

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

drm/amd/pm: Deprecate print_clock_levels interface

Use emit_clock_levels instead of print_clock_levels

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

drm/amd/pm: Use emit_clock_levels in vega20

Move to emit_clock_levels from print_clock_levels

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

drm/amd/pm: Use emit_clock_levels in vega12

Move to emit_clock_levels from print_clock_levels

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