drm/amdgpu: handle denied inject error into critical regions v2

Changed from v1:
    remove unused brace

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

drm/amd/pm: add message smu to get ecc_table

support ECC TABLE message, this table include umc ras error count
and error address

V2:
Return after smu version check fail

V3:
Return -EOPNOTSUPP, if fail to get smc ver.

V4:
ECCTABLE typo corrected and sentence rephrased.

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

drm/amdgpu: add new query interface for umc_v8_7 block

add smu message query error information interface,

function name align with IP version number

V2:
Removed unused err cnt entry

Signed-off-by: mziya <Mohammadzafar.ziya@amd.com>
Acked-by: Evan Quan <evan.quan@amd.com>
Reviewed-by: Stanley.Yang <Stanley.Yang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/pm: Update smu driver interface for sienna cichlid

update smu driver if version to 0x40

V2:
Interface version append with sienna_cichlid
V3:
Aligned with latest driver interface.

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

drm/amd/display: Revert W/A for hard hangs on DCN20/DCN21

The WA from commit 2a50edbf10c8 ("drm/amd/display: Apply w/a for hard hang
on HPD") and commit 1bd3bc745e7f ("drm/amd/display: Extend w/a for hard
hang on HPD to dcn20") causes a regression in s0ix where the system will
fail to resume properly on many laptops.  Pull the workarounds out to
avoid that s0ix regression in the common case.  This HPD hang happens with
an external device in special circumstances and a new W/A will need to be
developed for this in the future.

Cc: Qingqing Zhuo <qingqing.zhuo@amd.com>
Reported-by: Scott Bruce <smbruce@gmail.com>
Reported-by: Chris Hixon <linux-kernel-bugs@hixontech.com>
Reported-by: spasswolf@web.de
Link: https://bugzilla.kernel.org/show_bug.cgi?id=215436
Link: https://gitlab.freedesktop.org/drm/amd/-/issues/1821
Link: https://gitlab.freedesktop.org/drm/amd/-/issues/1852
Fixes: 2a50edbf10c8 ("drm/amd/display: Apply w/a for hard hang on HPD")
Fixes: 1bd3bc745e7f ("drm/amd/display: Extend w/a for hard hang on HPD to dcn20")
Reviewed-by: Nicholas Kazlauskas <nicholas.kazlauskas@amd.com>
Signed-off-by: Mario Limonciello <mario.limonciello@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: drop flags check for CHIP_IP_DISCOVERY

Support for IP based discovery is in place now so this
check is no longer required.

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

drm/amdgpu: Fix rejecting Tahiti GPUs

eb4fd29afd4a ("drm/amdgpu: bind to any 0x1002 PCI diplay class device") added
generic bindings to amdgpu so that that it binds to all display class devices
with VID 0x1002 and then rejects those in amdgpu_pci_probe.

Unfortunately it reuses a driver_data value of 0 to detect those new bindings,
which is already used to denote CHIP_TAHITI ASICs.

The driver_data value given to those new bindings was changed in
dd0761fd24ea1 ("drm/amdgpu: set CHIP_IP_DISCOVERY as the asic type by default")
to CHIP_IP_DISCOVERY (=36), but it seems that the check in amdgpu_pci_probe
was forgotten to be changed. Therefore, it still rejects Tahiti GPUs.

Link: https://gitlab.freedesktop.org/drm/amd/-/issues/1860
Fixes: eb4fd29afd4a ("drm/amdgpu: bind to any 0x1002 PCI diplay class device")

Signed-off-by: Lukas Fink <lukas.fink1@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: don't do resets on APUs which don't support it

It can cause a hang.  This is normally not enabled for GPU
hangs on these asics, but was recently enabled for handling
aborted suspends.  This causes hangs on some platforms
on suspend.

Fixes: daf8de0874ab5b ("drm/amdgpu: always reset the asic in suspend (v2)")
Bug: https://gitlab.freedesktop.org/drm/amd/-/issues/1858
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: invert the logic in amdgpu_device_should_recover_gpu()

Rather than opting into GPU recovery support, default to on, and
opt out if it's not working on a particular GPU.  This avoids the
need to add new asics to this list since this is a core feature.

Reviewed-by: Evan Quan <evan.quan@amd.com>
Reviewed-by: Guchun Chen <guchun.chen@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: Enable recovery on yellow carp

Add yellow carp to devices which support recovery

Signed-off-by: CHANDAN VURDIGERE NATARAJ <chandan.vurdigerenataraj@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Remove redundant initialization of dpg_width

dpg_width is being initialized to width but this is never read
as dpg_width is overwritten later on. Remove the redundant
initialization.

Cleans up the following clang-analyzer warning:

drivers/gpu/drm/amd/amdgpu/../display/dc/core/dc_link_dp.c:6020:8:
warning: Value stored to 'dpg_width' during its initialization is never
read [clang-analyzer-deadcode.DeadStores].

Reported-by: Abaci Robot <abaci@linux.alibaba.com>
Signed-off-by: Jiapeng Chong <jiapeng.chong@linux.alibaba.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/pm: Replace one-element array with flexible-array member

There is a regular need in the kernel to provide a way to declare having
a dynamically sized set of trailing elements in a structure. Kernel code
should always use "flexible array members" for these cases. The older
style of one-element or zero-length arrays should no longer be used.
Reference:
https://www.kernel.org/doc/html/latest/process/deprecated.html#zero-length-and-one-element-arrays

Reported-by: Zeal Robot <zealci@zte.com.cn>
Signed-off-by: Changcheng Deng <deng.changcheng@zte.com.cn>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/pm: fix null ptr access

check null ptr first before access its element

v2: check adev->pm.dpm_enabled early in amdgpu_debugfs_pm_init()

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

drm/amdgpu: fix compile warning for ras_block_match_default

fix compile warning for ras_block_match_default

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

drm/amdgpu: Use ARRAY_SIZE to get array length

Use ARRAY_SIZE to get array length.

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

drm/amdgpu: clean up some inconsistent indenting

Eliminate the follow smatch warnings:
drivers/gpu/drm/amd/amdgpu/amdgpu_device.c:3504 amdgpu_device_init()
warn: inconsistent indenting
drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c:1716
amdgpu_ras_error_status_query() warn: if statement not indented
drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c:1058 amdgpu_ras_error_inject()
warn: inconsistent indenting

Reported-by: Abaci Robot <abaci@linux.alibaba.com>
Reviewed-by: Guchun Chen <guchun.chen@amd.com>
Signed-off-by: Yang Li <yang.lee@linux.alibaba.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: remove unneeded semicolon

Eliminate the following coccicheck warning:
./drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c:2725:16-17: Unneeded semicolon

Reported-by: Abaci Robot <abaci@linux.alibaba.com>
Reviewed-by: Guchun Chen <guchun.chen@amd.com>
Signed-off-by: Yang Li <yang.lee@linux.alibaba.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: No longer insert ras blocks into ras_list if it already exists in ras_list

No longer insert ras blocks into ras_list if it already exists in ras_list.

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

drm/amdgpu: Add ras supported check for register_ras_block

Add ras supported check for register_ras_block.

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

drm/amdgpu: Add interface to load SRIOV cap FW

- Add interface to load SRIOV cap FW. If the FW does not
  exist, simply skip this FW loading routine.
  This FW will only be loaded under SRIOV. Other driver
  configuration will not be affected.
  By adding this interface, it will make us easier to
  prepare SRIOV Linux guest driver for different users.

- Update sysfs interface to read cap FW version.

- Refactor PSP FW loading routine under SRIOV to use a
  unified SWITCH statement instead of using IF statement

- Remove redundant amdgpu_sriov_vf() check in FW loading
  routine

Acked-by: Monk Liu <monk.liu@amd.com>
Acked-by: Guchun Chen <guchun.chen@amd.com>
Signed-off-by: Bokun Zhang <Bokun.Zhang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdkfd: Fix indentation on switch statement

Cases should be same indentation as switch. Also fix string spanning
across multiple lines.

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

drm/amdgpu: cleanup ttm debug sdma vram access function

Some suggested cleanups to declutter ttm when doing debug VRAM access over
SDMA.

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

drm/amdgpu: improve debug VRAM access performance using sdma

For better performance during VRAM access for debugged processes, do
read/write copies over SDMA.

In order to fulfill post mortem debugging on a broken device, fallback to
stable MMIO access when gpu recovery is disabled or when job submission
time outs are set to max.  Failed SDMA access should automatically fall
back to MMIO access.

Use a pre-allocated GTT bounce buffer pre-mapped into GART to avoid
page-table updates and TLB flushes on access.

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: Removed redundant ras code

Removed redundant ras code.

Signed-off-by: yipechai <YiPeng.Chai@amd.com>
Reviewed-by: Hawking Zhang <Hawking.Zhang@amd.com>
Reviewed-by: John Clements <john.clements@amd.com>
Reviewed-by: Tao Zhou <tao.zhou1@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: Adjust error inject function code style in amdgpu_ras.c

1. Move xgmi special error inject function from amdgpu_ras.c to xgmi block.
2. Support to use psp_ras_trigger_error as default error inject function in amdgpu_ras.c. If .ras_error_inject isn't defined in ras block, default error inject function will take effect.

v2: squash in warning fix (Alex)

Signed-off-by: yipechai <YiPeng.Chai@amd.com>
Reviewed-by: Hawking Zhang <Hawking.Zhang@amd.com>
Reviewed-by: John Clements <john.clements@amd.com>
Reviewed-by: Tao Zhou <tao.zhou1@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: Modify mca block to fit for the unified ras block data and ops

1.Modify mca block to fit for the unified ras block data and ops.
2.Define special .ras_block_match function for mca block to identify itself.
3.Change amdgpu_mca_ras_funcs to amdgpu_mca_ras_block(amdgpu_mca_ras had been used), and the corresponding variable name remove _funcs suffix.
4.Remove the const flag of cma ras variable so that cma ras block can be able to be inserted into amdgpu device ras block link list.
5.Invoke amdgpu_ras_register_ras_block function to register cma ras block into amdgpu device ras block link list.
6.Remove the redundant code about cma in amdgpu_ras.c after using the unified ras block.

Signed-off-by: yipechai <YiPeng.Chai@amd.com>
Reviewed-by: Hawking Zhang <Hawking.Zhang@amd.com>
Reviewed-by: John Clements <john.clements@amd.com>
Reviewed-by: Tao Zhou <tao.zhou1@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: Modify sdma block to fit for the unified ras block data and ops

1.Modify sdma block to fit for the unified ras block data and ops.
2.Change amdgpu_sdma_ras_funcs to amdgpu_sdma_ras, and the corresponding variable name remove _funcs suffix.
3.Remove the const flag of sdma ras variable so that sdma ras block can be able to be inserted into amdgpu device ras block link list.
4.Invoke amdgpu_ras_register_ras_block function to register sdma ras block into amdgpu device ras block link list.
5.Remove the redundant code about sdma in amdgpu_ras.c after using the unified ras block.
6.Fill unified ras block .name .block .ras_late_init and .ras_fini for all of sdma versions. If .ras_late_init and .ras_fini had been defined by the selected sdma version, the defined functions will take effect; if not defined, default fill them with amdgpu_sdma_ras_late_init and amdgpu_sdma_ras_fini.

v2: squash in warning fix (Alex)

Signed-off-by: yipechai <YiPeng.Chai@amd.com>
Reviewed-by: Hawking Zhang <Hawking.Zhang@amd.com>
Reviewed-by: John Clements <john.clements@amd.com>
Reviewed-by: Tao Zhou <tao.zhou1@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: Modify umc block to fit for the unified ras block data and ops

1.Modify umc block to fit for the unified ras block data and ops.
2.Change amdgpu_umc_ras_funcs to amdgpu_umc_ras, and the corresponding variable name remove _funcs suffix.
3.Remove the const flag of umc ras variable so that umc ras block can be able to be inserted into amdgpu device ras block link list.
4.Invoke amdgpu_ras_register_ras_block function to register umc ras block into amdgpu device ras block link list.
5.Remove the redundant code about umc in amdgpu_ras.c after using the unified ras block.
6.Fill unified ras block .name .block .ras_late_init and .ras_fini for all of umc versions. If .ras_late_init and .ras_fini had been defined by the selected umc version, the defined functions will take effect; if not defined, default fill them with amdgpu_umc_ras_late_init and amdgpu_umc_ras_fini.

Signed-off-by: yipechai <YiPeng.Chai@amd.com>
Reviewed-by: Hawking Zhang <Hawking.Zhang@amd.com>
Reviewed-by: John Clements <john.clements@amd.com>
Reviewed-by: Tao Zhou <tao.zhou1@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: Modify nbio block to fit for the unified ras block data and ops

1.Modify nbio block to fit for the unified ras block data and ops.
2.Change amdgpu_nbio_ras_funcs to amdgpu_nbio_ras, and the corresponding variable name remove _funcs suffix.
3.Remove the const flag of mmhub ras variable so that nbio ras block can be able to be inserted into amdgpu device ras block link list.
4.Invoke amdgpu_ras_register_ras_block function to register nbio ras block into amdgpu device ras block link list.
5.Remove the redundant code about nbio in amdgpu_ras.c after using the unified ras block.

Signed-off-by: yipechai <YiPeng.Chai@amd.com>
Reviewed-by: Hawking Zhang <Hawking.Zhang@amd.com>
Reviewed-by: John Clements <john.clements@amd.com>
Reviewed-by: Tao Zhou <tao.zhou1@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: Modify mmhub block to fit for the unified ras block data and ops

1.Modify mmhub block to fit for the unified ras block data and ops.
2.Change amdgpu_mmhub_ras_funcs to amdgpu_mmhub_ras, and the corresponding variable name remove _funcs suffix.
3.Remove the const flag of mmhub ras variable so that mmhub ras block can be able to be inserted into amdgpu device ras block link list.
4.Invoke amdgpu_ras_register_ras_block function to register mmhub ras block into amdgpu device ras block link list. 5.Remove the redundant code about mmhub in amdgpu_ras.c after using the unified ras block.
5.Remove the redundant code about mmhub in amdgpu_ras.c after using the unified ras block.
6.Fill unified ras block .name .block .ras_late_init and .ras_fini for all of mmhub versions. If .ras_late_init and .ras_fini had been defined by the selected mmhub version, the defined functions will take effect; if not defined, default fill them with amdgpu_mmhub_ras_late_init and amdgpu_mmhub_ras_fini.

Signed-off-by: yipechai <YiPeng.Chai@amd.com>
Reviewed-by: Hawking Zhang <Hawking.Zhang@amd.com>
Reviewed-by: John Clements <john.clements@amd.com>
Reviewed-by: Tao Zhou <tao.zhou1@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: Modify hdp block to fit for the unified ras block data and ops

1.Modify hdp block to fit for the unified ras block data and ops.
2.Change amdgpu_hdp_ras_funcs to amdgpu_hdp_ras, and the corresponding variable name remove _funcs suffix.
3.Remove the const flag of hdp ras variable so that hdp ras block can be able to be inserted into amdgpu device ras block link list.
4.Invoke amdgpu_ras_register_ras_block function to register hdp ras block into amdgpu device ras block link list.
5.Remove the redundant code about hdp in amdgpu_ras.c after using the unified ras block.

Signed-off-by: yipechai <YiPeng.Chai@amd.com>
Reviewed-by: Hawking Zhang <Hawking.Zhang@amd.com>
Reviewed-by: John Clements <john.clements@amd.com>
Reviewed-by: Tao Zhou <tao.zhou1@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: Modify xgmi block to fit for the unified ras block data and ops

1.Modify gmc block to fit for the unified ras block data and ops.
2.Change amdgpu_xgmi_ras_funcs to amdgpu_xgmi_ras, and the corresponding variable name remove _funcs suffix.
3.Remove the const flag of gmc ras variable so that gmc ras block can be able to be inserted into amdgpu device ras block link list.
4.Invoke amdgpu_ras_register_ras_block function to register gmc ras block into amdgpu device ras block link list.
5.Remove the redundant code about gmc in amdgpu_ras.c after using the unified ras block.

Signed-off-by: yipechai <YiPeng.Chai@amd.com>
Reviewed-by: Hawking Zhang <Hawking.Zhang@amd.com>
Reviewed-by: John Clements <john.clements@amd.com>
Reviewed-by: Tao Zhou <tao.zhou1@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: Modify gfx block to fit for the unified ras block data and ops

1.Modify gfx block to fit for the unified ras block data and ops.
2.Change amdgpu_gfx_ras_funcs to amdgpu_gfx_ras, and the corresponding variable name remove _funcs suffix.
3.Remove the const flag of gfx ras variable so that gfx ras block can be able to be inserted into amdgpu device ras block link list.
4.Invoke amdgpu_ras_register_ras_block function to register gfx ras block into amdgpu device ras block link list.
5.Remove the redundant code about gfx in amdgpu_ras.c after using the unified ras block.
6.Fill unified ras block .name .block .ras_late_init and .ras_fini for all of gfx versions. If .ras_late_init and .ras_fini had been defined by the selected gfx version, the defined functions will take effect; if not defined, default fill with amdgpu_gfx_ras_late_init and amdgpu_gfx_ras_fini.

Signed-off-by: yipechai <YiPeng.Chai@amd.com>
Reviewed-by: Hawking Zhang <Hawking.Zhang@amd.com>
Reviewed-by: John Clements <john.clements@amd.com>
Reviewed-by: Tao Zhou <tao.zhou1@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: Modify the compilation failed problem when other ras blocks' .h include amdgpu_ras.h

Modify the compilation failed problem when other ras blocks' .h include amdgpu_ras.h.

v2: squash in forward declaration warning fix (Alex)

Signed-off-by: yipechai <YiPeng.Chai@amd.com>
Reviewed-by: Hawking Zhang <Hawking.Zhang@amd.com>
Reviewed-by: John Clements <john.clements@amd.com>
Reviewed-by: Tao Zhou <tao.zhou1@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: Unify ras block interface for each ras block

1. Define unified ops interface for each block.
2. Add ras_block_match function pointer in ops interface, each ras block can customize specail match function to identify itself.
3. Add amdgpu_ras_block_match_default new function. If a ras block doesn't define .ras_block_match, default execute amdgpu_ras_block_match_default to identify this ras block.
4. Define unified basic ras block data for each ras block.
5. Create dedicated amdgpu device ras block link list to manage all of the ras blocks.
6. Add amdgpu_ras_register_ras_block new function interface for each ras block to register itself to ras controlling block.

Signed-off-by: yipechai <YiPeng.Chai@amd.com>
Reviewed-by: Hawking Zhang <Hawking.Zhang@amd.com>
Reviewed-by: John Clements <john.clements@amd.com>
Reviewed-by: Tao Zhou <tao.zhou1@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/pm: correct the checks for fan attributes support

On functionality unsupported, -EOPNOTSUPP will be returned. And we rely
on that to determine the fan attributes support.

Fixes: 79c65f3fcbb128 ("drm/amd/pm: do not expose power implementation details to amdgpu_pm.c")

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

drm/amdgpu/pm: move additional logic into amdgpu_dpm_force_performance_level

This is part of the forced performance level.  Move it from
the sysfs handler into amdgpu_dpm_force_performance_level.

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

drm/amd/pm: drop those unrealistic thermal_type checks

As it's impossible the thermal sensor of KV is one of them.

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: wrap those atombios APIs used by SI under CONFIG_DRM_AMDGPU_SI

No need to compile them on CONFIG_DRM_AMDGPU_SI disabled.

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/pm: unified lock protections in amdgpu_dpm.c

As the only entry point, it's now safe and reasonable to
enforce the lock protections in amdgpu_dpm.c. And with
this, we can drop other internal used power locks.

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

drm/amd/pm: revise the performance level setting APIs

Avoid cross callings which make lock protection enforcement
on amdgpu_dpm_force_performance_level() impossible.

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

drm/amd/pm: drop unnecessary gfxoff controls

Those gfxoff controls added for some specific ASICs are unnecessary.
The functionalities are not affected without them. Also to align with
other ASICs, they should also be dropped.

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

drm/amd/pm: relocate the power related headers

Instead of centralizing all headers in the same folder. Separate them into
different folders and place them among those source files those who really
need them.

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

drm/amd/pm: do not expose the smu_context structure used internally in power

This can cover the power implementation details. And as what did for
powerplay framework, we hook the smu_context to adev->powerplay.pp_handle.

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

drm/amd/pm: drop redundant or unused APIs and data structures

Drop those unused APIs and data structures.

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

drm/amd/pm: move those code piece used by Stoney only to smu8_hwmgr.c

Instead of putting them in amdgpu_dpm.c.

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

drm/amd/pm: optimize the amdgpu_pm_compute_clocks() implementations

Drop cross callings and multi-function APIs. Also avoid exposing
internal implementations details.

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

drm/amd/pm: move pp_force_state_enabled member to amdgpu_pm structure

As it lables an internal pm state and amdgpu_pm structure is the more
proper place than amdgpu_device structure for it.

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

drm/amd/pm: create a new holder for those APIs used only by legacy ASICs(si/kv)

Those APIs are used only by legacy ASICs(si/kv). They cannot be
shared by other ASICs. So, we create a new holder for them.

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

drm/amd/pm: do not expose the API used internally only in kv_dpm.c

Move it to kv_dpm.c instead.

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

drm/amd/pm: do not expose those APIs used internally only in si_dpm.c

Move them to si_dpm.c instead.

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

drm/amd/pm: do not expose those APIs used internally only in amdgpu_dpm.c

Move them to amdgpu_dpm.c instead.

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

drm/amd/pm: do not expose power implementation details to display

Display is another client of our power APIs. It's not proper to spike
into power implementation details there.

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

drm/amd/pm: do not expose power implementation details to amdgpu_pm.c

amdgpu_pm.c holds all the user sysfs/hwmon interfaces. It's another
client of our power APIs. It's not proper to spike into power
implementation details there.

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

drm/amd/pm: do not expose implementation details to other blocks out of power

Those implementation details(whether swsmu supported, some ppt_funcs supported,
accessing internal statistics ...)should be kept internally. It's not a good
practice and even error prone to expose implementation details.

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

drm/radeon: use kernel is_power_of_2 rather than local version

Follow the amdgpu change made in
7611750784664db46d0db95631e322aeb263dde7
and replace local radeon function with is_power_of_2().

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

drm/amd/display: Enable Freesync Video Mode by default

[Why&How]
Freesync Video Mode is a experimental feature previously,
and need to be enabled by kernel parameter. We enable it
by default with removing module paramterter in amdgpu_dm.

v2: squash the patches together

Signed-off-by: Solomon Chiu <solomon.chiu@amd.com>
Reviewed-by: Aurabindo Jayamohanan Pillai <Aurabindo.Pillai@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu/swsmu: make sienna cichlid function static

Unused outside of the file.

Reported-by: kernel test robot <lkp@intel.com>
Reviewed-by: Evan Quan <evan.quan@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

Merge tag 'amd-drm-next-5.17-2022-01-12' of https://gitlab.freedesktop.org/agd5f/linux into drm-next

amd-drm-next-5.17-2022-01-12:

amdgpu:
- SR-IOV fixes
- Suspend/resume fixes
- Display fixes
- DMCUB fixes
- DP alt mode fixes
- RAS fixes
- UBSAN fix
- Navy Flounder VCN fix
- ttm resource manager cleanup
- default_groups change for kobj_type
- vkms fix
- Aldebaran fixes

amdkfd:
- SDMA ECC interrupt fix
- License clarification
- Pointer check fix
- DQM fixes for hawaii
- default_groups change for kobj_type
- Typo fixes

Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
From: Alex Deucher <alexander.deucher@amd.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20220113030537.5758-1-alexander.deucher@amd.com

Merge tag 'drm-misc-next-fixes-2022-01-14' of git://anongit.freedesktop.org/drm/drm-misc into drm-next

 * atomic helpers: Fix error messages
 * mipi-dbi: Fix buffer mapping

Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
From: Thomas Zimmermann <tzimmermann@suse.de>
Link: https://patchwork.freedesktop.org/patch/msgid/YeGHu7qU92pjuQOn@linux-uq9g

Merge tag 'drm-misc-fixes-2022-01-14' of git://anongit.freedesktop.org/drm/drm-misc into drm-next

Two DT bindings fixes for meson, a device refcounting fix for sun4i, a
probe fix for vga16fb, a locking fix for the CMA dma-buf heap and a
compilation fix for ttm.

Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
[danvet: I made sure I have exactly the same conflict resolution as
Linus in 8d0749b4f83b ("Merge tag 'drm-next-2022-01-07' of
git://anongit.freedesktop.org/drm/drm") to avoid further conflict fun.
From: Maxime Ripard <maxime@cerno.tech>
Link: https://patchwork.freedesktop.org/patch/msgid/20220114125454.zs46ny52lrxk3ljz@houat

drm/mipi-dbi: Fix source-buffer address in mipi_dbi_buf_copy

Set the source-buffer address after mapping the buffer into the
kernel's address space. Makes MIPI DBI helpers work again.

Signed-off-by: Thomas Zimmermann <tzimmermann@suse.de>
Fixes: c47160d8edcd ("drm/mipi-dbi: Remove dependency on GEM CMA helper library")
Reviewed-by: Noralf Trønnes <noralf@tronnes.org>
Reported-by: Noralf Trønnes <noralf@tronnes.org>
Cc: Thomas Zimmermann <tzimmermann@suse.de>
Cc: Daniel Vetter <daniel.vetter@ffwll.ch>
Cc: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Cc: Maxime Ripard <mripard@kernel.org>
Link: https://patchwork.freedesktop.org/patch/msgid/20220111132634.18302-1-tzimmermann@suse.de
(cherry picked from commit 9d31993451f6bb4059a9b9eec4856b2225e36df0)
Signed-off-by: Thomas Zimmermann <tzimmermann@suse.de>

drm: fix error found in some cases after the patch d1af5cd86997

The patch d1af5cd86997 ("drm: get rid of DRM_DEBUG_* log
calls in drm core, files drm_a*.c") fails when the drm_device
cannot be found in the parameter plane_state->crtc.
Fix it using plane_state->plane.

Reported-by: kernel test robot <oliver.sang@intel.com>
Fixes: 6e22dc358377 ("drm: get rid of DRM_DEBUG_* log calls in drm core, files drm_a*.c")
Signed-off-by: Claudio Suarez <cssk@net-c.es>
[danvet: fix Fixes: line]
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Link: https://patchwork.freedesktop.org/patch/msgid/YaiXEARd8z2C463h@gineta.localdomain
(cherry picked from commit 521d459b1935628a7caa0753429f880dae2dbfc9)
Signed-off-by: Thomas Zimmermann <tzimmermann@suse.de>

Merge tag 'drm-misc-next-fixes-2022-01-13' of git://anongit.freedesktop.org/drm/drm-misc into drm-next

 * Fix use of CRTC state's active vs enable in atomic helper

Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
From: Thomas Zimmermann <tzimmermann@suse.de>
Link: https://patchwork.freedesktop.org/patch/msgid/Yd/i/zj9vEHisSSB@linux-uq9g

Merge tag 'drm-intel-next-fixes-2022-01-13' of git://anongit.freedesktop.org/drm/drm-intel into drm-next

- Hold runtime PM wakelock during PXP unbind (Juston Li)
- Three fixes for the TTM backend fault handling (Matthew Auld)
- Make sure to unmap when purging in the TTM backend (Matthew Auld)

Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
From: Tvrtko Ursulin <tvrtko.ursulin@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/Yd/xzyCM87rfrwQT@tursulin-mobl2

drm/ttm: fix compilation on ARCH=um

Even if it's probably not really useful, it can get selected
by e.g. randconfig builds, and then failing to compile is an
annoyance. Unfortunately, it's hard to fix in Kconfig, since
DRM_TTM is selected by many things that don't really depend
on any specific architecture, and just depend on PCI (which
is indeed now available in ARCH=um via simulation/emulation).

Fix this in the code instead by just ifdef'ing the relevant
two lines that depend on "real X86".

Reported-by: Geert Uytterhoeven <geert@linux-m68k.org>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Link: https://patchwork.freedesktop.org/patch/msgid/20211220111519.a4c8c6eff702.Ie4cf4e68698f6a9f546b83379bc52c266504424f@changeid

dma-buf: cma_heap: Fix mutex locking section

Fix cma_heap_buffer mutex locking critical section to protect vmap_cnt
and vaddr.

Fixes: a5d2d29e24be ("dma-buf: heaps: Move heap-helper logic into the cma_heap implementation")
Signed-off-by: Weizhao Ouyang <o451686892@gmail.com>
Acked-by: John Stultz <john.stultz@linaro.org>
Signed-off-by: Sumit Semwal <sumit.semwal@linaro.org>
Link: https://patchwork.freedesktop.org/patch/msgid/20220104073545.124244-1-o451686892@gmail.com

video: vga16fb: Only probe for EGA and VGA 16 color graphic cards

The vga16fb framebuffer driver only supports Enhanced Graphics Adapter
(EGA) and Video Graphics Array (VGA) 16 color graphic cards.

But it doesn't check if the adapter is one of those or if a VGA16 mode
is used. This means that the driver will be probed even if a VESA BIOS
Extensions (VBE) or Graphics Output Protocol (GOP) interface is used.

This issue has been present for a long time but it was only exposed by
commit d391c5827107 ("drivers/firmware: move x86 Generic System
Framebuffers support") since the platform device registration to match
the {vesa,efi}fb drivers is done later as a consequence of that change.

All non-x86 architectures though treat orig_video_isVGA as a boolean so
only do the supported video mode check for x86 and not for other arches.

Link: https://bugzilla.kernel.org/show_bug.cgi?id=215001
Fixes: d391c5827107 ("drivers/firmware: move x86 Generic System Framebuffers support")
Reported-by: Kris Karas <bugs-a21@moonlit-rail.com>
Cc: <stable@vger.kernel.org> # 5.15.x
Signed-off-by: Javier Martinez Canillas <javierm@redhat.com>
Tested-by: Kris Karas <bugs-a21@moonlit-rail.com>
Acked-by: Maxime Ripard <maxime@cerno.tech>
Link: https://patchwork.freedesktop.org/patch/msgid/20220110095625.278836-3-javierm@redhat.com

drm/amdkfd: Fix ASIC name typos

Three misspelled ASICs in comments here, so fix the spelling

Signed-off-by: Kent Russell <kent.russell@amd.com>
Reviewed-by: Felix Kuehling <Felix.Kuehling@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdkfd: Fix DQM asserts on Hawaii

start_nocpsch would never set dqm->sched_running on Hawaii due to an
early return statement. This would trigger asserts in other functions
and end up in inconsistent states.

Bug: https://github.com/RadeonOpenCompute/ROCm/issues/1624
Signed-off-by: Felix Kuehling <Felix.Kuehling@amd.com>
Reviewed-by: Kent Russell <kent.russell@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: Use correct VIEWPORT_DIMENSION for DCN2

For some reason this file isn't using the appropriate register
headers for DCN headers, which means that on DCN2 we're getting
the VIEWPORT_DIMENSION offset wrong.

This means that we're not correctly carving out the framebuffer
memory correctly for a framebuffer allocated by EFI and
therefore see corruption when loading amdgpu before the display
driver takes over control of the framebuffer scanout.

Fix this by checking the DCE_HWIP and picking the correct offset
accordingly.

Long-term we should expose this info from DC as GMC shouldn't
need to know about DCN registers.

Cc: stable@vger.kernel.org
Signed-off-by: Harry Wentland <harry.wentland@amd.com>
Reviewed-by: Huang Rui <ray.huang@amd.com>
Acked-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/amd/pm: only send GmiPwrDnControl msg on master die (v3)

PMFW only returns 0 on master die and sends NACK back on other dies for
the message.

v2: only send GmiPwrDnControl msg on master die instead of all
dies.
v3: remove the pointer check for get_socket_id and get_die_id as they
should be present on Aldebaran.

Signed-off-by: Tao Zhou <tao.zhou1@amd.com>
Reviewed-by: Lijo Lazar <lijo.lazar@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: use spin_lock_irqsave to avoid deadlock by local interrupt

This is observed in SRIOV case with virtual KMS as display.

_raw_spin_lock_irqsave+0x37/0x40
drm_handle_vblank+0x69/0x350 [drm]
? try_to_wake_up+0x432/0x5c0
? amdgpu_vkms_prepare_fb+0x1c0/0x1c0 [amdgpu]
drm_crtc_handle_vblank+0x17/0x20 [drm]
amdgpu_vkms_vblank_simulate+0x4d/0x80 [amdgpu]
__hrtimer_run_queues+0xfb/0x230
hrtimer_interrupt+0x109/0x220
__sysvec_apic_timer_interrupt+0x64/0xe0
asm_call_irq_on_stack+0x12/0x20

Fixes: 84ec374bd580 ("drm/amdgpu: create amdgpu_vkms (v4)")
Signed-off-by: Guchun Chen <guchun.chen@amd.com>
Acked-by: Alex Deucher <alexander.deucher@amd.com>
Tested-by: Kelly Zytaruk <kelly.zytaruk@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: not return error on the init_apu_flags

In some APU project we needn't always assign flags to identify each other,
so we may not need return an error.

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

drm/amdkfd: Use prange->update_list head for remove_list

The remove_list head was only used for keeping track of existing ranges
that are to be removed from the svms->list. The update_list was used for
new or existing ranges that need updated attributes. These two cases are
mutually exclusive (i.e. the same range will never be on both lists).
Therefore we can use the update_list head to track the remove_list and
save another 16 bytes in the svm_range struct.

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

drm/amdkfd: Use prange->list head for insert_list

There are seven list_heads in struct svm_range: list, update_list,
remove_list, insert_list, svm_bo_list, deferred_list, child_list. This
patch and the next one remove two of them that are redundant.

The insert_list head was only used for new ranges that are not on the
svms->list yet. So we can use that list head for keeping track of
new ranges before they get added, and use list_move_tail to move them
to the svms->list when ready.

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

drm/amdkfd: make SPDX License expression more sound

Commit b5f57384805a ("drm/amdkfd: Add sysfs bitfields and enums to uAPI")
adds include/uapi/linux/kfd_sysfs.h with the "GPL-2.0 OR MIT WITH
Linux-syscall-note" SPDX-License expression.

The command ./scripts/spdxcheck.py warns:

  include/uapi/linux/kfd_sysfs.h: 1:48 Exception not valid for license MIT: Linux-syscall-note

For a uapi header, the file under GPLv2 License must be combined with the
Linux-syscall-note, but combining the MIT License with the
Linux-syscall-note makes no sense, as the note provides an exception for
GPL-licensed code, not for permissively licensed code.

So, reorganize the SPDX expression to only combine the note with the GPL
License condition. This makes spdxcheck happy again.

Fixes: b5f57384805a ("drm/amdkfd: Add sysfs bitfields and enums to uAPI")
Signed-off-by: Lukas Bulwahn <lukas.bulwahn@gmail.com>
Reviewed-by: kstewart@linuxfoundation.org
Reviewed-by: Felix Kuehling <Felix.Kuehling@amd.com>
Signed-off-by: Felix Kuehling <Felix.Kuehling@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdkfd: Check for null pointer after calling kmemdup

As the possible failure of the allocation, kmemdup() may return NULL
pointer.
Therefore, it should be better to check the 'props2' in order to prevent
the dereference of NULL pointer.

Fixes: 3a87177eb141 ("drm/amdkfd: Add topology support for dGPUs")
Signed-off-by: Jiasheng Jiang <jiasheng@iscas.ac.cn>
Reviewed-by: Felix Kuehling <Felix.Kuehling@amd.com>
Signed-off-by: Felix Kuehling <Felix.Kuehling@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: invalid parameter check in dmub_hpd_callback

The function performs a check on the "adev" input parameter, however, it
is used before the check.

Initialize the "dev" variable after the sanity check to avoid a possible
NULL pointer dereference.

Fixes: e27c41d5b0681 ("drm/amd/display: Support for DMUB HPD interrupt handling")
Addresses-Coverity-ID: 1493909 ("Null pointer dereference")
Reviewed-by: Harry Wentland <harry.wentland@amd.com>
Signed-off-by: José Expósito <jose.exposito89@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

Revert "drm/amdgpu: Don't inherit GEM object VMAs in child process"

This reverts commit fbcdbfde87509d523132b59f661a355c731139d0.

Reviewed-by: Felix Kuehling <Felix.Kuehling@amd.com>
Signed-off-by: Rajneesh Bhardwaj <rajneesh.bhardwaj@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: reset dcn31 SMU mailbox on failures

Otherwise future commands may fail as well leading to downstream
problems that look like they stemmed from a timeout the first time
but really didn't.

Signed-off-by: Mario Limonciello <mario.limonciello@amd.com>
Reviewed-by: Nicholas Kazlauskas <nicholas.kazlauskas@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdkfd: use default_groups in kobj_type

There are currently 2 ways to create a set of sysfs files for a
kobj_type, through the default_attrs field, and the default_groups
field.  Move the amdkfd sysfs code to use default_groups field which has
been the preferred way since aa30f47cf666 ("kobject: Add support for
default attribute groups to kobj_type") so that we can soon get rid of
the obsolete default_attrs field.

Cc: Felix Kuehling <Felix.Kuehling@amd.com>
Cc: Alex Deucher <alexander.deucher@amd.com>
Cc: "Christian König" <christian.koenig@amd.com>
Cc: "Pan, Xinhui" <Xinhui.Pan@amd.com>
Cc: David Airlie <airlied@linux.ie>
Cc: Daniel Vetter <daniel@ffwll.ch>
Cc: amd-gfx@lists.freedesktop.org
Cc: dri-devel@lists.freedesktop.org
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: use default_groups in kobj_type

There are currently 2 ways to create a set of sysfs files for a
kobj_type, through the default_attrs field, and the default_groups
field.  Move the amdgpu sysfs code to use default_groups field which has
been the preferred way since aa30f47cf666 ("kobject: Add support for
default attribute groups to kobj_type") so that we can soon get rid of
the obsolete default_attrs field.

Cc: Alex Deucher <alexander.deucher@amd.com>
Cc: "Christian König" <christian.koenig@amd.com>
Cc: "Pan, Xinhui" <Xinhui.Pan@amd.com>
Cc: David Airlie <airlied@linux.ie>
Cc: Daniel Vetter <daniel@ffwll.ch>
Cc: Hawking Zhang <Hawking.Zhang@amd.com>
Cc: John Clements <john.clements@amd.com>
Cc: Felix Kuehling <Felix.Kuehling@amd.com>
Cc: Jonathan Kim <jonathan.kim@amd.com>
Cc: Kevin Wang <kevin1.wang@amd.com>
Cc: shaoyunl <shaoyun.liu@amd.com>
Cc: Tao Zhou <tao.zhou1@amd.com>
Cc: amd-gfx@lists.freedesktop.org
Cc: dri-devel@lists.freedesktop.org
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/amdgpu: Add pcie indirect support to amdgpu_mm_wreg_mmio_rlc()

The function amdgpu_mm_wreg_mmio_rlc() is used by debugfs to write to
MMIO registers.  It didn't support registers beyond the BAR mapped MMIO
space.  This adds pcie indirect write support.

Signed-off-by: Tom St Denis <tom.stdenis@amd.com>
Acked-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: recover gart table at resume

Get rid off pin/unpin of gart BO at resume/suspend and
instead pin only once and try to recover gart content
at resume time. This is much more stable in case there
is OOM situation at 2nd call to amdgpu_device_evict_resources()
while evicting GART table.

v3: remove gart recovery from other places
v2: pin gart at amdgpu_gart_table_vram_alloc()

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

drm/amdgpu: do not pass ttm_resource_manager to vram_mgr

Do not allow exported amdgpu_vram_mgr_*() to accept
any ttm_resource_manager pointer. Also there is no need
to force other module to call a ttm function just to
eventually call vram_mgr functions.

v2: pass adev's vram_mgr instead of adev

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

drm/amdkfd: remove unused function

Remove unused amdgpu_amdkfd_get_vram_usage()

CC: Felix.Kuehling@amd.com
Signed-off-by: Nirmoy Das <nirmoy.das@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Fixes: dfcbe6d5f4a340 ("drm/amdgpu: Remove unused function pointers")

drm/amdgpu: do not pass ttm_resource_manager to gtt_mgr

Do not allow exported amdgpu_gtt_mgr_*() to accept
any ttm_resource_manager pointer. Also there is no need
to force other module to call a ttm function just to
eventually call gtt_mgr functions.

v4: remove unused adev.
v3: upcast mgr from ttm resopurce manager instead of
getting it from adev.
v2: pass adev's gtt_mgr instead of adev.

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

drm/amdgpu: Unmap MMIO mappings when device is not unplugged

Patch: 3efb17ae7e92 ("drm/amdgpu: Call amdgpu_device_unmap_mmio() if device
is unplugged to prevent crash in GPU initialization failure") makes call to
amdgpu_device_unmap_mmio() conditioned on device unplugged. This patch unmaps
MMIO mappings even when device is not unplugged.

v2: Add condition of drm_dev_enter() to deleted unmaps in patch
"drm/amdgpu: Unmap all MMIO mappings"

Signed-off-by: Leslie Shi <Yuliang.Shi@amd.com>
Reviewed-by: Andrey Grodzovsky <andrey.grodzovsky@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: Enable second VCN for certain Navy Flounder.

Certain Navy Flounder cards have 2 VCNs, enable it.

Signed-off-by: Peng Ju Zhou <PengJu.Zhou@amd.com>
Acked-by: Alex Deucher <alexander.deucher@amd.com>
Reviewed-by: Guchun Chen <guchun.chen@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: explicitly set is_dsc_supported to false before use

When UBSAN is enabled a case is shown on unplugging the display that
this variable hasn't been initialized by `update_dsc_caps`, presumably
when the display was unplugged it wasn't copied from the DPCD.

Link: https://bugs.launchpad.net/ubuntu/+source/linux/+bug/1956497
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: Clear garbage data in err_data before usage

Memory of err_data should be cleaned before usage
when there're multiple entry in ras ih.
Otherwise garbage data from last loop will be used.

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

drm/atomic: Check new_crtc_state->active to determine if CRTC needs disable in self refresh mode

Actual hardware state of CRTC is controlled by the member 'active' in
struct drm_crtc_state instead of the member 'enable', according to the
kernel doc of the member 'enable'.  In fact, the drm client modeset
and atomic helpers are using the member 'active' to do the control.

Referencing the member 'enable' of new_crtc_state, the function
crtc_needs_disable() may fail to reflect if CRTC needs disable in
self refresh mode, e.g., when the framebuffer emulation will be blanked
through the client modeset helper with the next commit, the member
'enable' of new_crtc_state is still true while the member 'active' is
false, hence the relevant potential encoder and bridges won't be disabled.

So, let's check new_crtc_state->active to determine if CRTC needs disable
in self refresh mode instead of new_crtc_state->enable.

Fixes: 1452c25b0e60 ("drm: Add helpers to kick off self refresh mode in drivers")
Cc: Sean Paul <seanpaul@chromium.org>
Cc: Rob Clark <robdclark@chromium.org>
Cc: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Cc: Maxime Ripard <mripard@kernel.org>
Cc: Thomas Zimmermann <tzimmermann@suse.de>
Cc: David Airlie <airlied@linux.ie>
Cc: Daniel Vetter <daniel@ffwll.ch>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Liu Ying <victor.liu@nxp.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20211230040626.646807-1-victor.liu@nxp.com

drm/sun4i: dw-hdmi: Fix missing put_device() call in sun8i_hdmi_phy_get

The reference taken by 'of_find_device_by_node()' must be released when
not needed anymore.
Add the corresponding 'put_device()' in the error handling path.

Fixes: 9bf3797796f5 ("drm/sun4i: dw-hdmi: Make HDMI PHY into a platform device")
Signed-off-by: Miaoqian Lin <linmq006@gmail.com>
Signed-off-by: Maxime Ripard <maxime@cerno.tech>
Link: https://patchwork.freedesktop.org/patch/msgid/20220107083633.20843-1-linmq006@gmail.com

drm/i915/ttm: ensure we unmap when purging

Purging can happen during swapping out, or directly invoked with the
madvise ioctl. In such cases this doesn't involve a ttm move, which
skips umapping the object.

v2(Thomas):
- add ttm_truncate helper, and just call into i915_ttm_move_notify() to
  handle the unmapping step

Fixes: 213d50927763 ("drm/i915/ttm: Introduce a TTM i915 gem object backend")
Signed-off-by: Matthew Auld <matthew.auld@intel.com>
Cc: Thomas Hellström <thomas.hellstrom@linux.intel.com>
Reviewed-by: Thomas Hellström <thomas.hellstrom@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20220106174910.280616-4-matthew.auld@intel.com
(cherry picked from commit ab4911b7d411ab2ef3b38322178b9138e156c393)
Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>

drm/i915/ttm: add unmap_virtual callback

Ensure we call ttm_bo_unmap_virtual when releasing the pages.
Importantly this should now handle the ttm swapping case, and all other
places that already call into i915_ttm_move_notify().

v2: fix up the selftest

Fixes: cf3e3e86d779 ("drm/i915: Use ttm mmap handling for ttm bo's.")
Signed-off-by: Matthew Auld <matthew.auld@intel.com>
Cc: Thomas Hellström <thomas.hellstrom@linux.intel.com>
Reviewed-by: Thomas Hellström <thomas.hellstrom@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20220106174910.280616-3-matthew.auld@intel.com
(cherry picked from commit 903e0387270eef14a711c0feb23b7bf62d2480df)
Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>

drm/i915/ttm: only fault WILLNEED objects

Don't attempt to fault and re-populate purged objects. By some fluke
this passes the dontneed-after-mmap IGT, but for the wrong reasons.

Fixes: cf3e3e86d779 ("drm/i915: Use ttm mmap handling for ttm bo's.")
Signed-off-by: Matthew Auld <matthew.auld@intel.com>
Cc: Thomas Hellström <thomas.hellstrom@linux.intel.com>
Reviewed-by: Thomas Hellström <thomas.hellstrom@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20220106174910.280616-2-matthew.auld@intel.com
(cherry picked from commit f3cb4a2de5410147b53e53416a3af0ffe26b5f4e)
Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>

drm/i915: don't call free_mmap_offset when purging

The TTM backend is in theory the only user here(also purge should only
be called once we have dropped the pages), where it is setup at object
creation and is only removed once the object is destroyed. Also
resetting the node here might be iffy since the ttm fault handler
uses the stored fake offset to determine the page offset within the pages
array.

This also blows up in the dontneed-before-mmap test, since the
expectation is that the vma_node will live on, until the object is
destroyed:

<2> [749.062902] kernel BUG at drivers/gpu/drm/i915/gem/i915_gem_ttm.c:943!
<4> [749.062923] invalid opcode: 0000 [#1] PREEMPT SMP NOPTI
<4> [749.062928] CPU: 0 PID: 1643 Comm: gem_madvise Tainted: G     U  W         5.16.0-rc8-CI-CI_DRM_11046+ #1
<4> [749.062933] Hardware name: Gigabyte Technology Co., Ltd. GB-Z390 Garuda/GB-Z390 Garuda-CF, BIOS IG1c 11/19/2019
<4> [749.062937] RIP: 0010:i915_ttm_mmap_offset.cold.35+0x5b/0x5d [i915]
<4> [749.063044] Code: 00 48 c7 c2 a0 23 4e a0 48 c7 c7 26 df 4a a0 e8 95 1d d0 e0 bf 01 00 00 00 e8 8b ec cf e0 31 f6 bf 09 00 00 00 e8 5f 30 c0 e0 <0f> 0b 48 c7 c1 24 4b 56 a0 ba 5b 03 00 00 48 c7 c6 c0 23 4e a0 48
<4> [749.063052] RSP: 0018:ffffc90002ab7d38 EFLAGS: 00010246
<4> [749.063056] RAX: 0000000000000240 RBX: ffff88811f2e61c0 RCX: 0000000000000006
<4> [749.063060] RDX: 0000000000000000 RSI: 0000000000000000 RDI: 0000000000000009
<4> [749.063063] RBP: ffffc90002ab7e58 R08: 0000000000000001 R09: 0000000000000001
<4> [749.063067] R10: 000000000123d0f8 R11: ffffc90002ab7b20 R12: ffff888112a1a000
<4> [749.063071] R13: 0000000000000004 R14: ffff88811f2e61c0 R15: ffff888112a1a000
<4> [749.063074] FS:  00007f6e5fcad500(0000) GS:ffff8884ad600000(0000) knlGS:0000000000000000
<4> [749.063078] CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
<4> [749.063081] CR2: 00007efd264e39f0 CR3: 0000000115fd6005 CR4: 00000000003706f0
<4> [749.063085] Call Trace:
<4> [749.063087]  <TASK>
<4> [749.063089]  __assign_mmap_offset+0x41/0x300 [i915]
<4> [749.063171]  __assign_mmap_offset_handle+0x159/0x270 [i915]
<4> [749.063248]  ? i915_gem_dumb_mmap_offset+0x70/0x70 [i915]
<4> [749.063325]  drm_ioctl_kernel+0xae/0x140
<4> [749.063330]  drm_ioctl+0x201/0x3d0
<4> [749.063333]  ? i915_gem_dumb_mmap_offset+0x70/0x70 [i915]
<4> [749.063409]  ? do_user_addr_fault+0x200/0x670
<4> [749.063415]  __x64_sys_ioctl+0x6d/0xa0
<4> [749.063419]  do_syscall_64+0x3a/0xb0
<4> [749.063423]  entry_SYSCALL_64_after_hwframe+0x44/0xae
<4> [749.063428] RIP: 0033:0x7f6e5f100317

Testcase: igt/gem_madvise/dontneed-before-mmap
Fixes: cf3e3e86d779 ("drm/i915: Use ttm mmap handling for ttm bo's.")
Signed-off-by: Matthew Auld <matthew.auld@intel.com>
Cc: Thomas Hellström <thomas.hellstrom@linux.intel.com>
Reviewed-by: Thomas Hellström <thomas.hellstrom@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20220106174910.280616-1-matthew.auld@intel.com
(cherry picked from commit 658a0c632625e1db51837ff754fe18a6a7f2ccf8)
Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>

drm/i915/pxp: Hold RPM wakelock during PXP unbind

Similar to commit b8d8436840ca ("drm/i915/gt: Hold RPM wakelock during
PXP suspend") but to fix the same warning for unbind during shutdown:

------------[ cut here ]------------
RPM wakelock ref not held during HW access
WARNING: CPU: 0 PID: 4139 at drivers/gpu/drm/i915/intel_runtime_pm.h:115
gen12_fwtable_write32+0x1b7/0
Modules linked in: 8021q ccm rfcomm cmac algif_hash algif_skcipher
af_alg uinput snd_hda_codec_hdmi vf industrialio iwl7000_mac80211
cros_ec_sensorhub lzo_rle lzo_compress zram iwlwifi cfg80211 joydev
CPU: 0 PID: 4139 Comm: halt Tainted: G     U  W
5.10.84 #13 344e11e079c4a03940d949e537eab645f6
RIP: 0010:gen12_fwtable_write32+0x1b7/0x200
Code: 48 c7 c7 fc b3 b5 89 31 c0 e8 2c f3 ad ff 0f 0b e9 04 ff ff ff c6
05 71 e9 1d 01 01 48 c7 c7 d67
RSP: 0018:ffffa09ec0bb3bb0 EFLAGS: 00010246
RAX: 12dde97bbd260300 RBX: 00000000000320f0 RCX: ffffffff89e60ea0
RDX: 0000000000000000 RSI: 00000000ffffdfff RDI: ffffffff89e60e70
RBP: ffffa09ec0bb3bd8 R08: 0000000000000000 R09: ffffa09ec0bb3950
R10: 00000000ffffdfff R11: ffffffff89e91160 R12: 0000000000000000
R13: 0000000028121969 R14: ffff9515c32f0990 R15: 0000000040000000
FS:  0000790dcf225740(0000) GS:ffff951737800000(0000) knlGS:0000000000000000
CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
CR2: 000058b25efae147 CR3: 0000000133ea6001 CR4: 0000000000770ef0
DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000
DR3: 0000000000000000 DR6: 00000000ffff07f0 DR7: 0000000000000400
PKRU: 55555554
Call Trace:
 intel_pxp_fini_hw+0x2f/0x39
 i915_pxp_tee_component_unbind+0x1c/0x42
 component_unbind+0x32/0x48
 component_unbind_all+0x80/0x9d
 take_down_master+0x24/0x36
 component_master_del+0x56/0x70
 mei_pxp_remove+0x2c/0x68
 mei_cl_device_remove+0x35/0x68
 device_release_driver_internal+0x100/0x1a1
 mei_cl_bus_remove_device+0x21/0x79
 mei_cl_bus_remove_devices+0x3b/0x51
 mei_stop+0x3b/0xae
 mei_me_shutdown+0x23/0x58
 device_shutdown+0x144/0x1d3
 kernel_power_off+0x13/0x4c
 __se_sys_reboot+0x1d4/0x1e9
 do_syscall_64+0x43/0x55
 entry_SYSCALL_64_after_hwframe+0x44/0xa9
RIP: 0033:0x790dcf316273
Code: 64 89 01 48 83 c8 ff c3 66 2e 0f 1f 84 00 00 00 00 00 0f 1f 44 00
00 89 fa be 69 19 12 28 bf ad8
RSP: 002b:00007ffca0df9198 EFLAGS: 00000202 ORIG_RAX: 00000000000000a9
RAX: ffffffffffffffda RBX: 000000004321fedc RCX: 0000790dcf316273
RDX: 000000004321fedc RSI: 0000000028121969 RDI: 00000000fee1dead
RBP: 00007ffca0df9200 R08: 0000000000000007 R09: 0000563ce8cd8970
R10: 0000000000000000 R11: 0000000000000202 R12: 00007ffca0df9308
R13: 0000000000000001 R14: 0000000000000000 R15: 0000000000000003
---[ end trace 2f501b01b348f114 ]---
ACPI: Preparing to enter system sleep state S5
reboot: Power down

Changes since v1:
 - Rebase to latest drm-tip

Fixes: 0cfab4cb3c4e ("drm/i915/pxp: Enable PXP power management")
Suggested-by: Lee Shawn C <shawn.c.lee@intel.com>
Signed-off-by: Juston Li <juston.li@intel.com>
Reviewed-by: Daniele Ceraolo Spurio <daniele.ceraolospurio@intel.com>
Signed-off-by: John Harrison <John.C.Harrison@Intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20220106200236.489656-2-juston.li@intel.com
(cherry picked from commit 57ded5fc98b11d76dae505ca3591b61c9dbbbda7)
Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>

drm/amd/display: Add version check before using DP alt query interface

[Why]
To maintain compatibility with firmware older than 4.0.11.

Those firmware may have interrmittent hangs with RDCSPIPE or the PHY,
but we shouldn't regress their previous behavior.

[How]
Use the new path if firmware is development or 4.0.11 or newer. Use the
legacy path otherwise.

Fixes: 41f91315b5be5a ("drm/amd/display: Query DMCUB for dp alt status")

Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Reviewed-by: Hansen Dsouza <hansen.dsouza@amd.com>
Acked-by: Rodrigo Siqueira <Rodrigo.Siqueira@amd.com>
Signed-off-by: Nicholas Kazlauskas <nicholas.kazlauskas@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>