drm/amdgpu: allow fragment processing for invalid PTEs

That should improve the PRT performance on Vega quite a bit.

Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Felix Kuehling <Felix.Kuehling@amd.com>
Reviewed-by: Huang Rui <ray.huang@amd.com>
Acked-by: Junwei Zhang <Jerry.Zhang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: use the maximum possible fragment size on Vega/Raven

The fragment size controls only the L1 on Vega/Raven and we now don't
have any extra overhead any more because of larger fragments.

Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Huang Rui <ray.huang@amd.com>
Reviewed-by: Felix Kuehling <Felix.Kuehling@amd.com>
Acked-by: Junwei Zhang <Jerry.Zhang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: meld together VM fragment and huge page handling

This optimizes the generating of PTEs by walking the hierarchy only once
for a range and making changes as necessary.

It allows for both huge (2MB) as well giant (1GB) pages to be used on
Vega and Raven.

Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Felix Kuehling <Felix.Kuehling@amd.com>
Reviewed-by: Huang Rui <ray.huang@amd.com>
Acked-by: Junwei Zhang <Jerry.Zhang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: use leaf iterator for filling PTs

Less overhead and is the starting point for further cleanups and
improvements.

Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Felix Kuehling <Felix.Kuehling@amd.com>
Reviewed-by: Junwei Zhang <Jerry.Zhang@amd.com>
Reviewed-by: Huang Rui <ray.huang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: use the DFS iterator in amdgpu_vm_invalidate_pds v2

Less code and easier to maintain.

v2: rename the function as well

Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Huang Rui <ray.huang@amd.com>
Reviewed-by: Junwei Zhang <Jerry.Zhang@amd.com>
Reviewed-by: Felix Kuehling <Felix.Kuehling@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: use dfs iterator to free PDs/PTs

Allows us to free all PDs/PTs without recursion.

Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Felix Kuehling <Felix.Kuehling@amd.com>
Reviewed-by: Junwei Zhang <Jerry.Zhang@amd.com>
Reviewed-by: Huang Rui <ray.huang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: use leaf iterator for allocating PD/PT

Less code and allows for easier error handling.

Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Felix Kuehling <Felix.Kuehling@amd.com>
Reviewed-by: Junwei Zhang <Jerry.Zhang@amd.com>
Reviewed-by: Huang Rui <ray.huang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: add some VM PD/PT iterators v2

Both a leaf as well as dfs iterator to walk over all the PDs/PTs.

v2: update comments and fix for_each_amdgpu_vm_pt_dfs_safe

Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Huang Rui <ray.huang@amd.com>
Reviewed-by: Junwei Zhang <Jerry.Zhang@amd.com>
Reviewed-by: Felix Kuehling <Felix.Kuehling@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu:Add error message when register failed to reach expected value

Add error message when register failed to reach expected value, It will
help discover potential issue.

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

drm/amd/display: fix ptr_ret.cocci warnings

drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_debugfs.c:771:1-3: WARNING: PTR_ERR_OR_ZERO can be used

 Use PTR_ERR_OR_ZERO rather than if(IS_ERR(...)) + PTR_ERR

Generated by: scripts/coccinelle/api/ptr_ret.cocci

Fixes: e498eb713604 ("drm/amd/display: Add support for hw_state logging via debugfs")
CC: Nicholas Kazlauskas <nicholas.kazlauskas@amd.com>
Signed-off-by: kbuild test robot <fengguang.wu@intel.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd: Add DM DMCU support

DMCU (Display Microcontroller Unit) is a GPU chip involved in
eDP features like Adaptive Backlight Modulation and Panel Self
Refresh.

DC is already fully equipped to initialize DMCU as long as the
firmware is loaded.

At the moment only the raven firmware is available.

A single .bin file is loaded by the kernel's loading mechanism
and split into two ucodes according to the header.

DMCU is optional, so if the firmware is not found, no error or
warning is raised.

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

drm/amd: Add PSP DMCU support

DMCU (Display Microcontroller Unit) is a GPU chip involved in
eDP features like Adaptive Backlight Modulation and Panel Self
Refresh.

PSP is already equipped to handle DMCU firmware loading, all
that is needed is to translate between the new DMCU ucode ID and
the equivalent psp_gfx_fw_type.

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

drm/amd: Add ucode DMCU support

DMCU (Display Microcontroller Unit) is a GPU chip involved in
eDP features like Adaptive Backlight Modulation and Panel Self
Refresh.

DMCU has two pieces of firmware: the ERAM and the interrupt
vectors, which must be loaded seperately.

To this end, the DMCU firmware has a custom header and parsing
logic similar to MEC, to extract the two ucodes from a single
struct firmware.

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

drm/amdgpu: Move fault hash table to amdgpu vm

In stead of share one fault hash table per device, make it
per vm. This can avoid inter-process lock issue when fault
hash table is full.

Change-Id: I5d1281b7c41eddc8e26113e010516557588d3708
Signed-off-by: Oak Zeng <Oak.Zeng@amd.com>
Suggested-by: Christian Konig <Christian.Koenig@amd.com>
Suggested-by: Felix Kuehling <Felix.Kuehling@amd.com>
Reviewed-by: Christian Konig <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: move cs dependencies front a bit

cs dependencies handling doesn't need in vm resv

Signed-off-by: Chunming Zhou <david1.zhou@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: try allocating VRAM as power of two

Try to allocate VRAM in power of two sizes and only fallback to vram
split sizes if that fails.

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

drm/amdgpu: Fix SDMA TO after GPU reset v3

After GPU reset amdgpu_vm_clear_bo triggers VM flush
but job->vm_pd_addr is not set causing SDMA TO.

v2:
Per advise by Christian König avoid flushing VM for jobs where
job->vm_pd_addr wasn't explicitly set.

v3:
Shortcut vm_flush_needed early.

Fixes cbd5285 drm/amdgpu: move setting the GART addr into TTM.
Signed-off-by: Andrey Grodzovsky <andrey.grodzovsky@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: add amdgpu_vm_update_func

Add helper to call the update function for both BO and shadow.

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

drm/amdgpu: add amdgpu_vm_pt_parent helper

Add a function to get the parent of a PD/PT.

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

drm/amdgpu: fix error handling in amdgpu_cs_user_fence_chunk

Slowly leaking memory one page at a time :)

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

drm/amdkfd: Only add bi-directional iolink on GPU with XGMI or largebar (v2)

v2: compile fix

Signed-off-by: shaoyunl <Shaoyun.Liu@amd.com>
Reviewed-by: Felix Kuehling <Felix.Kuehling@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdkfd: Generate xGMI direct iolink

Generate xGMI iolink for upper level usage

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

drm/amdkfd: Add new iolink type defines

Update the iolink type defines according to the new thunk spec

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

drm/amdkfd: kfd expose the hive_id of the device through its node properties

Thunk will generate the XGMI topology information when necessary with the hive_id
for each specified device

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

drm/amdgpu: get_hive_id from amdgpu side (v2)

Retrieve hive_id from amdgpu device

v2: compile fix

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

drm/amd/include: Add get_hive_id interface in kfd2kgd

KFD need to get hive id from amdgpu to build up the XGMI topology

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

drm/amdgpu : Generate XGMI topology info from driver level

Driver will save an array of XGMI hive info, each hive will have a list of devices
that have the same hive ID.

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

drm/amdgpu: Add place holder functions for xgmi topology interface with psp

Add dummy function for xgmi function interface with psp

Signed-off-by: Shaoyun Liu <Shaoyun.Liu@amd.com>
Reviewed-by: Felix Kuehling <Felix.Kuehling@amd.com>
Reviewed-by: Huang Rui <ray.huang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu : Add psp function interfaces for XGMI support

Place holder for XGMI support

Signed-off-by: Shaoyun Liu <Shaoyun.Liu@amd.com>
Reviewed-by: Huang Rui <ray.huang@amd.com>
Acked-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu/gmc9: Adjust GART and AGP location with xgmi offset (v2)

On hives with xgmi enabled, the fb_location aperture is a size
which defines the total framebuffer size of all nodes in the
hive.  Each GPU in the hive has the same view via the fb_location
aperture.  GPU0 starts at offset (0 * segment size),
GPU1 starts at offset (1 * segment size), etc.

For access to local vram on each GPU, we need to take this offset into
account. This including on setting up GPUVM page table and GART table

v2: squash in "drm/amdgpu: Init correct fb region for none XGMI configuration"

Acked-by: Huang Rui <ray.huang@amd.com>
Acked-by: Slava Abramov <slava.abramov@amd.com>
Signed-off-by: Shaoyun Liu <Shaoyun.Liu@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Reviewed-by: Felix Kuehling <Felix.Kuehling@amd.com>
Acked-by: Huang Rui <ray.huang@amd.com>

drm/amdgpu/gmc9: add a new gfxhub 1.1 helper for xgmi

Used to populate the xgmi info on vega20.

v2: PF_MAX_REGION is val - 1 (Ray)

Acked-by: Huang Rui <ray.huang@amd.com>
Acked-by: Slava Abramov <slava.abramov@amd.com>
Reviewed-by :Shaoyun liu <Shaoyun.liu@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Acked-by: Christian König <christian.koenig@amd.com>
Signed-off-by :Shaoyun liu <Shaoyun.liu@amd.com>

drm/amdgpu/gmc: add initial xgmi structure to amdgpu_gmc structure

Initial pass at a structure to store xgmi info.  xgmi is a high
speed cross gpu interconnect.

Acked-by: Huang Rui <ray.huang@amd.com>
Acked-by: Slava Abramov <slava.abramov@amd.com>
Reviewed-by :Shaoyun liu <Shaoyun.liu@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Shaoyun Liu <Shaoyun.Liu@amd.com>

drm/amd/include: update the bitfield define for PF_MAX_REGION

Correct the definition based on vega20 register spec

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

drm/amdgpu: move PSP init prior to IH in gpu reset

since we use PSP to program IH regs now

Signed-off-by: Monk Liu <Monk.Liu@amd.com>
Acked-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Huang Rui <ray.huang@amd.com>
Signed-off-by: Emily Deng <Emily.Deng@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: set bulk_moveable to false when a per VM is released

Otherwise we might run into a use after free during bulk move.

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

drm/amdgpu: Fix SDMA hang in prt mode v2

Fix SDMA hang in prt mode, clear XNACK_WATERMARK in reg SDMA0_UTCL1_WATERMK to avoid the issue

Affected ASICs: VEGA10 VEGA12 RV1 RV2

v2: add reg clear for SDMA1

Signed-off-by: Tao Zhou <tao.zhou1@amd.com>
Tested-by: Yukun Li <yukun1.li@amd.com>
Reviewed-by: Hawking Zhang <Hawking.Zhang@amd.com>
Acked-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/powerplay: fix compile warning for wrong data type V2

do_div expects the 1st argument in 64bit instead of 32bit.
Drop the usage of do_div as it seems unnecessary.

V2: drop usage of do_div completely

Signed-off-by: Evan Quan <evan.quan@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: Fix warnings while make xmldocs

This patch fixes following warnings.

./drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c:3011:
warning: Excess function parameter 'dev' description
in 'amdgpu_vm_get_task_info'

./drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c:3012:
warning: Function parameter or member 'adev' not
described in 'amdgpu_vm_get_task_info'

./drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c:3012:
warning: Excess function parameter 'dev' description
in 'amdgpu_vm_get_task_info'

Signed-off-by: Masanari Iida <standby24x7@gmail.com>
Signed-off-by: Michel Dänzer <michel.daenzer@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: Fix compute VM BO params after rebase v2

The intent of two commits was lost in the last rebase:

810955b drm/amdgpu: Fix acquiring VM on large-BAR systems
b5d21aa drm/amdgpu: Don't use shadow BO for compute context

This commit restores the original behaviour:
* Don't set AMDGPU_GEM_CREATE_NO_CPU_ACCESS for page directories
  to allow them to be reused for compute VMs
* Don't create shadow BOs for page tables in compute VMs

v2: move more logic into amdgpu_vm_bo_param

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

drm/amdgpu: enable AGP aperture for GMC9 v2

Enable the old AGP aperture to avoid GART mappings.

v2: don't enable it for SRIOV

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

drm/amdgpu: Clean up KFD init and fini

Only initialize KFD once by moving amdgpu_amdkfd_init from
amdgpu_pci_probe to amdgpu_init. This fixes kernel oopses and hangs
when booting multi-GPU systems.

Also removed some vestiges of KFD being its own module.

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

drm/amd/display: Add invariant support instrumentation in driver

Existing debug dump are all invariant, new “low 32-bit of address”
dump is not invariant

Signed-off-by: Jun Lei <Jun.Lei@amd.com>
Reviewed-by: Eric Yang <eric.yang2@amd.com>
Acked-by: Bhawanpreet Lakha <Bhawanpreet.Lakha@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: clean code for transition event log.

[Why]
There are same purpose transition events.

[How]
remove the redundant event log.

Signed-off-by: Chiawen Huang <chiawen.huang@amd.com>
Reviewed-by: Tony Cheng <Tony.Cheng@amd.com>
Acked-by: Bhawanpreet Lakha <Bhawanpreet.Lakha@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Remove call to amdgpu_pm_compute_clocks

[Why]
The extraneous call to amdgpu_pm_compute_clocks is deprecated.

[How]
Remove it.

Signed-off-by: David Francis <David.Francis@amd.com>
Signed-off-by: Leo Li <sunpeng.li@amd.com>
Reviewed-by: David Francis <David.Francis@amd.com>
Acked-by: Bhawanpreet Lakha <Bhawanpreet.Lakha@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: use link type to decide stream enc acquisition

[Why]
Virtual sink is used when set mode happens on a disconnected display
to allow the mode set to proceed. This did not work with MST because
the logic for acquiring stream encoder uses stream signal to determine
the special handling is required, and stream signal is virtual instead
of DP in this case.

[How]
Use link type to decide instead.

Signed-off-by: Eric Yang <Eric.Yang2@amd.com>
Reviewed-by: Tony Cheng <Tony.Cheng@amd.com>
Acked-by: Bhawanpreet Lakha <Bhawanpreet.Lakha@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: dc 3.1.65

Signed-off-by: Tony Cheng <tony.cheng@amd.com>
Reviewed-by: Steven Chiu <Steven.Chiu@amd.com>
Acked-by: Bhawanpreet Lakha <Bhawanpreet.Lakha@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Reorder resource_pool to put i2c with aux

[Why]
The i2c and aux engines are similar, and should be placed
next to eachother for readability

[How]
Reorder the elements of the resource_pool struct

Signed-off-by: David Francis <David.Francis@amd.com>
Reviewed-by: Tony Cheng <Tony.Cheng@amd.com>
Acked-by: Bhawanpreet Lakha <Bhawanpreet.Lakha@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Use DRM helper for best_encoder

[Why]
Our implementation is functionally identical to DRM's

Note that instead of checking if the provided id is 0, the helper
follows through with the mode object search. However, It will still
return NULL, since 0 is not a valid object id, and missed searches
will return NULL.

[How]
Remove our implementation, and replace it with
drm_atomic_helper_best_encoder.

Signed-off-by: Leo Li <sunpeng.li@amd.com>
Reviewed-by: David Francis <David.Francis@amd.com>
Acked-by: Bhawanpreet Lakha <Bhawanpreet.Lakha@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: num of sw i2c/aux engines less than num of connectors

[why]
AMD Stoney reference board, there are only 2 pipes (not include
underlay), and 3 connectors. resource creation, only
2 I2C/AUX engines are created. Within dc_link_aux_transfer, when
pin_data_en =2, refer to enengines[ddc_pin->pin_data->en] = NULL.
NULL point is referred later causing system crash.

[how]
each asic design has fixed number of ddc engines at hw side.
for each ddc engine, create its i2x/aux engine at sw side.

Signed-off-by: Hersen Wu <hersenxs.wu@amd.com>
Reviewed-by: Tony Cheng <Tony.Cheng@amd.com>
Acked-by: Bhawanpreet Lakha <Bhawanpreet.Lakha@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: add aux transition event log.

[Why]
Enhance aux transition debugging information.

[How]
Added Aux request and reply event log.

Signed-off-by: Chiawen Huang <chiawen.huang@amd.com>
Reviewed-by: Tony Cheng <Tony.Cheng@amd.com>
Acked-by: Bhawanpreet Lakha <Bhawanpreet.Lakha@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: add disconnect_delay to dc_panel_patch

Some display need disconnect delay. Adding this parameter for future use

Signed-off-by: Derek Lai <Derek.Lai@amd.com>
Reviewed-by: Charlene Liu <Charlene.Liu@amd.com>
Acked-by: Bhawanpreet Lakha <Bhawanpreet.Lakha@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: remove unused clk_src code

Signed-off-by: Dmytro Laktyushkin <Dmytro.Laktyushkin@amd.com>
Reviewed-by: Charlene Liu <Charlene.Liu@amd.com>
Acked-by: Bhawanpreet Lakha <Bhawanpreet.Lakha@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: remove dead dc vbios code

Signed-off-by: Dmytro Laktyushkin <Dmytro.Laktyushkin@amd.com>
Reviewed-by: Charlene Liu <Charlene.Liu@amd.com>
Acked-by: Bhawanpreet Lakha <Bhawanpreet.Lakha@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Add driver-side parsing for CM

Although 4 unique register values exist for gamma modes, two are
actually the same (the two RAMs) It’s not possible for caller to
understand this HW specific behavior, so some parsing is necessary
in driver

Signed-off-by: Jun Lei <Jun.Lei@amd.com>
Reviewed-by: Wesley Chalmers <Wesley.Chalmers@amd.com>
Acked-by: Bhawanpreet Lakha <Bhawanpreet.Lakha@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Fix DAL217 tests modify DTN logs for other tests

[Why]Update Code to get DTN golden log check to pass for tests run after
DAL217 tests.
[How]Change how dcn10_log_hw_state function prints HW state info
(CM_GAMUT_REMAP_Cx_Cx registers) when GAMUT REMAP is in bypass mode.

Signed-off-by: Gary Kattan <gary.kattan@amd.com>
Reviewed-by: Dmytro Laktyushkin <Dmytro.Laktyushkin@amd.com>
Acked-by: Bhawanpreet Lakha <Bhawanpreet.Lakha@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Add DP YCbCr 4:2:0 support

[Why]
For supporting DP YCbCr 4:2:0 output.

[How]
Update mod_build_vsc_infopacket to support Pixel
Encoding/Colorimetry Format indication for VSC SDP rev5.

Signed-off-by: Eric Bernstein <eric.bernstein@amd.com>
Reviewed-by: Dmytro Laktyushkin <Dmytro.Laktyushkin@amd.com>
Acked-by: Bhawanpreet Lakha <Bhawanpreet.Lakha@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Build stream update and plane updates in dm

[Why]
We currently lock modeset by setting a boolean in dm. We want to lock
Based on what DC tells us.

[How]
Build stream_updates and plane_update based on what changed. Then we
call check_update_surfaces_for_stream() to get the update type
We lock only if update_type is not fast

Signed-off-by: Bhawanpreet Lakha <Bhawanpreet.Lakha@amd.com>
Reviewed-by: Harry Wentland <Harry.Wentland@amd.com>
Acked-by: Bhawanpreet Lakha <Bhawanpreet.Lakha@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: dc 3.1.64

Signed-off-by: Tony Cheng <tony.cheng@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: manually map the shadow BOs again

Otherwise we won't be able to use the AGP aperture.

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

drm/amdgpu: use the AGP aperture for system memory access v2

Start to use the old AGP aperture for system memory access.

v2: Move that to amdgpu_ttm_alloc_gart

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

drm/amdgpu: add amdgpu_gmc_agp_location v3

Helper to figure out the location of the AGP BAR.

v2: fix a couple of bugs
v3: correctly add one to vram_end

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

drm/amdgpu: correctly sign extend 48bit addresses v3

Correct sign extend the GMC addresses to 48bit.

v2: sign extending turned out easier than thought.
v3: clean up the defines and move them into amdgpu_gmc.h as well

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

drm/amdgpu: improve VM state machine documentation v2

Since we have a lot of FAQ on the VM state machine try to improve the
documentation by adding functions for each state move.

v2: fix typo in amdgpu_vm_bo_invalidated, use amdgpu_vm_bo_relocated in
    one more place as well.

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

drm/amdgpu: fix amdgpu_mn_unlock() in the CS error path

Avoid unlocking a lock we never locked.

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

drm/amdgpu: separate per VM BOs from normal in the moved state

Allows us to avoid taking the spinlock in more places.

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

drm/amdgpu: move size calculations to the front of the file again

amdgpu_vm_bo_* functions should come much later.

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

drm/amd/powerplay: correct data type to support under voltage

For under voltage, negative value will be applied to voltage
offset. Update the data type to cover this case.

Signed-off-by: Evan Quan <evan.quan@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/powerplay: added vega20 overdrive support V3

Added vega20 overdrive support based on existing OD sysfs
APIs. However, the OD logics are simplified on vega20. So,
the behavior will be a little different and works only on
some limited levels.

V2: fix typo
    fix commit description
    revise error logs
    add support for clock OD

V3: separate clock from voltage OD settings

Signed-off-by: Evan Quan <evan.quan@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu/gmc9: don't keep stolen memory on vega20

Vega20 does not appear to be affected by the same issue
as vega10.  Enable the full stolen memory handling on
vega20.  Reserve the appropriate size at init time to avoid
display artifacts and then free it at the end of init once
the new FB is up and running.

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

drm/amdgpu/gmc9: don't keep stolen memory on vega12

vega12 does not appear to be affected by the same issue
as vega10.  Enable the full stolen memory handling on
vega12.  Reserve the appropriate size at init time to avoid
display artifacts and then free it at the end of init once
the new FB is up and running.

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

drm/amdgpu/gmc9: don't keep stolen memory on Raven

Raven does not appear to be affected by the same issue
as vega10.  Enable the full stolen memory handling on
Raven.  Reserve the appropriate size at init time to avoid
display artifacts and then free it at the end of init once
the new FB is up and running.

Bug: https://bugs.freedesktop.org/show_bug.cgi?id=106639
Reviewed-by: Junwei Zhang <Jerry.Zhang@amd.com>
Acked-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu/gmc9: rework stolen vga memory handling

No functional change, just rework it in order to adjust the
behavior on a per asic level.  The problem is that on vega10,
something corrupts the lower 8 MB of vram on the second
resume from S3.  This does not seem to affect Raven, other
gmc9 based asics need testing.

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

drm/amdgpu: fix idle state and bulk_moveable flag

Add BOs to the idle state again and correctly clear the flag when
new BOs are added.

Signed-off-by: Christian König <christian.koenig@amd.com>
Tested-by: Michel Dänzer <michel.daenzer@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: fix "use bulk moves for efficient VM LRU handling" v2

First step to fix the LRU corruption, we accidentially tried to move things
on the LRU after dropping the lock.

Signed-off-by: Christian König <christian.koenig@amd.com>
Tested-by: Michel Dänzer <michel.daenzer@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/ttm: fix ttm_bo_bulk_move_helper

Staring at the function for six hours, just to essentially move one line
of code. The problem was that the first list_cut_position call could result
in list2 pointing to la-la-land.

Signed-off-by: Christian König <christian.koenig@amd.com>
Tested-by: Michel Dänzer <michel.daenzer@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: Use drm_dev_unplug in PCI .remove

This at least allows to fail any subsequent IOCTLs with -ENODEV
after the device is gone.
Still this operation is not supported yet in graphic mode
and will lead at least to page faults and other issues.

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

drm/amdgpu: Revert "kmap PDs/PTs in amdgpu_vm_update_directories"

This reverts commit a7f91061c60ad9cac2e6a03b642be6a4f88b3662.

Felix pointed out that we need to have the BOs mapped even before
amdgpu_vm_update_directories is called.

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

drm/amdgpu: put GART away from VRAM v2

Always try to put the GART away from where VRAM is.

v2: correctly handle the 4GB limitation

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

drm/amdgpu: add missing CHIP_HAINAN in amdgpu_ucode_get_load_type

This caused a confusing error message, but there is functionally
no problem since the default method is DIRECT.

Reviewed-by: Michel Dänzer <michel.daenzer@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Cc: stable@vger.kernel.org

drm/amdgpu: Relocate some definitions v2

Move some KFD-related (but used in amdgpu_drv.c) definitions from
kfd_priv.h to kgd_kfd_interface.h so we don't need to include kfd_priv.h
in amdgpu_drv.c. This fixes a build failure when AMDGPU is enabled but
MMU_NOTIFIER is not.
This patch also disables KFD-related module options when HSA_AMD is not
enabled.

v2: rebase (Alex)

Signed-off-by: Amber Lin <Amber.Lin@amd.com>
Reviewed-by: Felix Kuehling <Felix.Kuehling@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: remove redundant memset

kvmalloc_array uses __GFP_ZERO flag ensures that the returned address
is zeroed already, memset it to zero again afterwards is unnecessary,
and in this case buggy because we only clear the first entry.

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

drm/ttm: Initialize local lists in ttm_bo_bulk_move_helper

The first parameter of list_cut_position() must point to an initialized
list.

Noticed thanks to KASAN pointing out something's fishy here.

Fixes: "drm/ttm: add bulk move function on LRU"
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Michel Dänzer <michel.daenzer@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: use the smaller hole for GART

Instead of the larger one use the smaller hole in the MC address
space for the GART mappings.

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

Revert "drm/amdgpu: move PD/PT bos on LRU again"

This reverts commit 31625ccae4464b61ec8cdb9740df848bbc857a5b.

It triggered various badness on my development machine when running the
piglit gpu profile with radeonsi on Bonaire, looks like memory
corruption due to insufficiently protected list manipulations.

Signed-off-by: Michel Dänzer <michel.daenzer@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: distinct between allocated GART space and GMC addr

Most of the time we only need to know if the BO has a valid GMC addr.

Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Junwei Zhang <Jerry.Zhang@amd.com>
Acked-by: Felix Kuehling <Felix.Kuehling@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: stop using gart_start as offset for the GTT domain

Further separate GART and GTT domain.

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

drm/amdgpu: fix amdgpu_gmc_gart_location a little bit

Improve the VCE limitation handling.

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

drm/amdgpu: move amdgpu_device_(vram|gtt)_location

Move that into amdgpu_gmc.c since we are really deadling with GMC
address space here.

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

drm/amdgpu: remove amdgpu_bo_gpu_accessible

Not used any more.

Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Michel Dänzer <michel.daenzer@amd.com>
Reviewed-by: Felix Kuehling <Felix.Kuehling@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdkfd: Release an acquired process vm

For compute vm acquired from amdgpu, vm.pasid is managed
by kfd. Decouple pasid from such vm on process destroy
to avoid duplicate pasid release.

Signed-off-by: Oak Zeng <Oak.Zeng@amd.com>
Reviewed-by: Felix Kuehling <Felix.Kuehling@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: Set pasid for compute vm (v2)

To make a amdgpu vm to a compute vm, the old pasid will be freed and
replaced with a pasid managed by kfd. Kfd can't reuse original pasid
allocated by amdgpu because kfd uses different pasid policy with amdgpu.
For example, all graphic devices share one same pasid in a process.

v2: rebase (Alex)

Signed-off-by: Oak Zeng <Oak.Zeng@amd.com>
Reviewed-by: Felix Kuehling <Felix.Kuehling@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: Only retrieve GPU address of GART table after pinning it

Doing it earlier hits a WARN_ON_ONCE in amdgpu_bo_gpu_offset.

Fixes: "drm/amdgpu: remove gart.table_addr"
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Michel Dänzer <michel.daenzer@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: Need to set moved to true when evict bo

Fix the VMC page fault when the running sequence is as below:
1.amdgpu_gem_create_ioctl
2.ttm_bo_swapout->amdgpu_vm_bo_invalidate, as not called
amdgpu_vm_bo_base_init, so won't called
list_add_tail(&base->bo_list, &bo->va). Even the bo was evicted,
it won't set the bo_base->moved.
3.drm_gem_open_ioctl->amdgpu_vm_bo_base_init, here only called
list_move_tail(&base->vm_status, &vm->evicted), but not set the
bo_base->moved.
4.amdgpu_vm_bo_map->amdgpu_vm_bo_insert_map, as the bo_base->moved is
not set true, the function amdgpu_vm_bo_insert_map will call
list_move(&bo_va->base.vm_status, &vm->moved)
5.amdgpu_cs_ioctl won't validate the swapout bo, as it is only in the
moved list, not in the evict list. So VMC page fault occurs.

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

drm/amdgpu: move full access into amdgpu_device_ip_suspend

It will be more safe to make full-acess include both phase1 and phase2.
Then accessing special registeris wherever at phase1 or phase2 will not
block any shutdown and suspend process under virtualization.

Signed-off-by: Yintian Tao <yttao@amd.com>
Acked-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: remove extra newline when printing VM faults

Looks like a copy&paste error to me.

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

drm/amdgpu: Move KFD parameters to amdgpu (v3)

After merging KFD into amdgpu, move module parameters defined in KFD to
amdgpu_drv.c, where other module parameters are declared.

v2: add kernel-doc comments
v3: rebase and fix parameter variable name (Alex)

Signed-off-by: Amber Lin <Amber.Lin@amd.com>
Reviewed-by: Felix Kuehling <Felix.Kuehling@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: Remove CONFIG_HSA_AMD_MODULE

After amdkfd is merged to amdgpu, CONFIG_HSA_AMD_MODULE no longer exists.

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

drm/amdgpu: Merge amdkfd into amdgpu

Since KFD is only supported by single GPU driver, it makes sense to merge
amdgpu and amdkfd into one module. This patch is the initial step: merge
Kconfig and Makefile.

v2: also remove kfd from drm Kconfig

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

drm/amdgpu: Refine gmc9 VM fault print.

The fault reports the page number where the fault happend and not
the exact faulty address. Update the print message to reflect that.

Signed-off-by: Andrey Grodzovsky <andrey.grodzovsky@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Flatten unnecessary i2c functions

[Why]
The dce_i2c_hw code contained four funtcions that were only
called in one place and did not have a clearly delineated
purpose.

[How]
Inline these functions, keeping the same functionality.

This is not a functional change.

The functions disable_i2c_hw_engine and release_engine_dce_hw were
pulled into their respective callers.

The most interesting part of this change is the acquire functions.
dce_i2c_hw_engine_acquire_engine was pulled into
dce_i2c_engine_acquire_hw, and dce_i2c_engine_acquire_hw was pulled
into acquire_i2c_hw_engine.

Some notes to show that this change is not functional:
-Failure conditions in any function resulted in a cascade of calls that
ended in a 'return NULL'.
Those are replaced with a direct 'return NULL'.

-The variable result is the one from dce_i2c_hw_engine_acquire_engine.
The boolean result used as part of return logic was removed.

-As the second half of dce_i2c_hw_engine_acquire_engine is only executed
if that function is returning true and therefore exiting the do-while
loop in dce_i2c_engine_acquire_hw, those lines were moved outside
of the loop.

Signed-off-by: David Francis <David.Francis@amd.com>
Acked-by: Leo Li <sunpeng.li@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>