drm/amdgpu: add shadow bo support V2

shadow bo is the shadow of a bo, which is always in GTT,
which can be used to backup the original bo.
V2:
reference shadow parent, shadow bo will be freed by who allocted him.

Signed-off-by: Chunming Zhou <[email protected]>
Reviewed-by: Christian König <[email protected]>
Signed-off-by: Alex Deucher <[email protected]>

drm/amd/amdgpu: UVD v6 register cleanup

Signed-off-by: Tom St Denis <[email protected]>
Reviewed-by: Christian König <[email protected]>
Signed-off-by: Alex Deucher <[email protected]>

drm/amd/amdgpu: VCE v2 register cleanup

Signed-off-by: Tom St Denis <[email protected]>
Reviewed-by: Christian König <[email protected]>
Signed-off-by: Alex Deucher <[email protected]>

drm/amd/amdgpu: add mutex in check_soft for VCE v3

Signed-off-by: Tom St Denis <[email protected]>
Reviewed-by: Christian König <[email protected]>
Signed-off-by: Alex Deucher <[email protected]>

drm/amd/amdgpu: Cleanup register access in VCE v3

Signed-off-by: Tom St Denis <[email protected]>
Reviewed-by: Christian König <[email protected]>
Signed-off-by: Alex Deucher <[email protected]>

Fixing copy-paste errors and removing unneeded newlines

Reviewed-by: Christian König <[email protected]>
Signed-off-by: Alexandre Demers <[email protected]>
Signed-off-by: Alex Deucher <[email protected]>

drm: Add DRM_MODE_PAGE_FLIP_TARGET_ABSOLUTE/RELATIVE flags v2

These flags allow userspace to explicitly specify the target vertical
blank period when a flip should take effect.

v2:
* Add new struct drm_mode_crtc_page_flip_target instead of modifying
  struct drm_mode_crtc_page_flip, to make sure all existing userspace
  code keeps compiling (Daniel Vetter)

Acked-by: Christian König <[email protected]>
Reviewed-by: Daniel Vetter <[email protected]>
Signed-off-by: Michel Dänzer <[email protected]>
Signed-off-by: Alex Deucher <[email protected]>

drm/radeon: Set MASTER_UPDATE_MODE to 0 again

With the previous change, it's safe to let page flips take effect
anytime during a vertical blank period.

This can avoid delaying a flip by a frame in some cases where we get to
radeon_flip_work_func -> adev->mode_info.funcs->page_flip during a
vertical blank period.

Acked-by: Christian König <[email protected]>
Signed-off-by: Michel Dänzer <[email protected]>
Signed-off-by: Alex Deucher <[email protected]>

drm/radeon: Provide page_flip_target hook

Now we can program a flip during a vertical blank period, if it's the
one targeted by the flip (or a later one). This allows simplifying
radeon_flip_work_func considerably.

Acked-by: Christian König <[email protected]>
Signed-off-by: Michel Dänzer <[email protected]>
Signed-off-by: Alex Deucher <[email protected]>

drm/amdgpu: Set MASTER_UPDATE_MODE to 0 again

With the previous change, it's safe to let page flips take effect
anytime during a vertical blank period.

This can avoid delaying a flip by a frame in some cases where we get to
amdgpu_flip_work_func -> adev->mode_info.funcs->page_flip during a
vertical blank period.

Acked-by: Christian König <[email protected]>
Signed-off-by: Michel Dänzer <[email protected]>
Signed-off-by: Alex Deucher <[email protected]>

drm/amdgpu: Provide page_flip_target hook

Now we can program a flip during a vertical blank period, if it's the
one targeted by the flip (or a later one). This allows simplifying
amdgpu_flip_work_func considerably.

agd: update dce_virtual.c as well.

Acked-by: Christian König <[email protected]>
Signed-off-by: Michel Dänzer <[email protected]>
Signed-off-by: Alex Deucher <[email protected]>

drm: Add page_flip_target CRTC hook v2

Mostly the same as the existing page_flip hook, but takes an additional
parameter specifying the target vertical blank period when the flip
should take effect.

v2:
* Add curly braces around else statement corresponding to an if block
  with curly braces (Alex Deucher)
* Call drm_crtc_vblank_put in the error case (Daniel Vetter)
* Clarify entry point documentation comment (Daniel Vetter)

Acked-by: Christian König <[email protected]>
Reviewed-by: Alex Deucher <[email protected]>
Reviewed-by: Daniel Vetter <[email protected]>
Signed-off-by: Michel Dänzer <[email protected]>
Signed-off-by: Alex Deucher <[email protected]>

gpu: drm: radeon: radeon_i2c: don't print error when adding adapter fails

The core will do this for us now.

Signed-off-by: Wolfram Sang <[email protected]>
Signed-off-by: Alex Deucher <[email protected]>

gpu: drm: amd: amdgpu: amdgpu_i2c: don't print error when adding adapter fails

The core will do this for us now.

Signed-off-by: Wolfram Sang <[email protected]>
Signed-off-by: Alex Deucher <[email protected]>

drm/amdgpu: use more than 64KB fragment size if possible

We align to 64KB, but when userspace aligns even more we can easily use more.

Signed-off-by: Christian König <[email protected]>
Reviewed-by: Alex Deucher <[email protected]>
Signed-off-by: Alex Deucher <[email protected]>

drm/amdgpu: flip frag_ptes and update_pts

We can add the fragment params before we split the update for the page tables.
That should save a few CPU cycles for larger updates.

Signed-off-by: Christian König <[email protected]>
Reviewed-by: Alex Deucher <[email protected]>
Signed-off-by: Alex Deucher <[email protected]>

drm/amdgpu: add adev to the pte_update_params

No need to carry that forward as a separate parameter.

Signed-off-by: Christian König <[email protected]>
Reviewed-by: Alex Deucher <[email protected]>
Signed-off-by: Alex Deucher <[email protected]>

drm/amdgpu: rename amdgpu_vm_update_params

Well those are actually page table entry parameters.
This also makes the variable names used a bit shorter.

Signed-off-by: Christian König <[email protected]>
Reviewed-by: Alex Deucher <[email protected]>
Signed-off-by: Alex Deucher <[email protected]>

drm/amdgpu: cleanup VM fragment defines

We can actually do way more than just the 64KB we currently used as default.

Signed-off-by: Christian König <[email protected]>
Reviewed-by: Alex Deucher <[email protected]>
Signed-off-by: Alex Deucher <[email protected]>

drm/amdgpu: remove unused VM defines

Not used for a long time.

Signed-off-by: Christian König <[email protected]>
Reviewed-by: Alex Deucher <[email protected]>
Signed-off-by: Alex Deucher <[email protected]>

drm/amd/amdgpu: Simplify bitfield operations in gfx v8

This patch introduces a new macro WREG32_FIELD which is used
to write to a register with a new value in a field.  It's designed
to replace the pattern:

tmp = RREG32(mmFoo);
tmp &= ~REG__FIELD_MASK;
tmp |= new_value << REG__FIELD__SHIFT;
WREG32(mmFoo, tmp)

with:

WREG32_FIELD(Foo, FIELD, new_value);

Unlike WREG32_P() it understands offsets/masks and doesn't
require the caller to shift the value (or mask properly).

It's applied where suitable in the gfx_v8_0.c driver to start
with.

Signed-off-by: Tom St Denis <[email protected]>
Reviewed-by: Christian König <[email protected]>
Signed-off-by: Alex Deucher <[email protected]>

drm/amd/amdgpu: Simplify various gfx v8 functions

Signed-off-by: Tom St Denis <[email protected]>
Reviewed-by: Christian König <[email protected]>
Signed-off-by: Alex Deucher <[email protected]>

drm/amd/amdgpu: Correct whitespace in GFX v8

Fix various whitespace issues in gfx v8 driver.

Signed-off-by: Tom St Denis <[email protected]>
Reviewed-by: Christian König <[email protected]>
Signed-off-by: Alex Deucher <[email protected]>

drm/amdgpu: Change the virtual_display type from int to char*.

For virtual display feature, as there may be multiple GPUs,
for user could choose whiche GPU need to enable this feature, change
the type of virtual_display from int to char*. The variable will be set
like this virtual_display="xxxx:xx:xx.x;xxxx:xx:xx.x;".

Signed-off-by: Emily Deng <[email protected]>
Reviewed-by: Alex Deucher <[email protected]>
Signed-off-by: Alex Deucher <[email protected]>

drm/amdgpu: add virtual dce support for iceland

Signed-off-by: Alex Deucher <[email protected]>

drm/amdgpu/virtual_dce: add case for topaz for disable_dce

This asic has no DCE block.  Also clarify the error message
for unmatched chips.

Signed-off-by: Alex Deucher <[email protected]>

drm/amdgpu: move vsync_timer_enabled setup to dce virtual early_init

Put it in one place.

Signed-off-by: Alex Deucher <[email protected]>

drm/amdgpu: Set ip_blocks according variable amdgpu_virtual_display.

For virtual display feature, if user set the option "amdgpu.virtual_display=1"
when load amdgpu.ko. Then need to set the ip_blocks with virtual display ip
blocks. And when enable virtual display, the amdgpu_dal need to be set to zero.

Signed-off-by: Emily Deng <[email protected]>
Reviewed-by: Alex Deucher <[email protected]>
Signed-off-by: Alex Deucher <[email protected]>

drm/amdgpu: Define one variable for virtual display.

For virtual display feature, define on variable in amdgpu.ko. When want to
enable virtual display feature, need set the option "amdgpu.virtual_display=1".
And then disable vga render and crtc if have DCE engine.

Signed-off-by: Emily Deng <[email protected]>
Reviewed-by: Alex Deucher <[email protected]>
Signed-off-by: Alex Deucher <[email protected]>

drm/amdgpu: Define virtual display ip blocks.

For virtual display feature, define virtual display ip blocks, and set
dce_virtual_ip_funcs to DCE block.

Signed-off-by: Emily Deng <[email protected]>
Reviewed-by: Alex Deucher <[email protected]>
Signed-off-by: Alex Deucher <[email protected]>

drm/amdgpu: Add DRM_MODE_CONNECTOR_VIRTUAL connector in amdgpu_connector_add.

For virtual display feature, add one connector type in amdgpu_connector_add.

Signed-off-by: Emily Deng <[email protected]>
Reviewed-by: Alex Deucher <[email protected]>
Signed-off-by: Alex Deucher <[email protected]>

drm/amdgpu: Call pageflip irq funtion when receiced vsync interrupt.

For virtual display feature, as there is no dce engine, so no pageflip irq
generated. So directly call pageflip irq funtion when received vysn interrupt.

Signed-off-by: Emily Deng <[email protected]>
Reviewed-by: Alex Deucher <[email protected]>
Signed-off-by: Alex Deucher <[email protected]>

drm/amdgpu: Use software timer to generate vsync interrupt.

For virtual display feature, use the software timer to
simulate the vsync interrupt.

Signed-off-by: Emily Deng <[email protected]>
Reviewed-by: Alex Deucher <[email protected]>
Signed-off-by: Alex Deucher <[email protected]>

drm/amdgpu: Disable VGA render and crtc when init GMC.

For virtual display feature, when the GPU has DCE engine, need to disable
the VGA render and CRTC, or it will hang when initialize GMC.

Signed-off-by: Emily Deng <[email protected]>
Reviewed-by: Alex Deucher <[email protected]>
Signed-off-by: Alex Deucher <[email protected]>

drm/amdgpu: To define whether the GPU has DCE engine.

For virtual display feature, when the GPU has DCE engine, need to disable
the VGA render and CRTC, or it will hang when initialize GMC. So first detect
whether the GPU has DCE engine.

Signed-off-by: Emily Deng <[email protected]>
Reviewed-by: Alex Deucher <[email protected]>
Signed-off-by: Alex Deucher <[email protected]>

drm/amdgpu: Initialize dce_virtual_crtc_funcs. (v2)

For virtual display feature, initialize dce_virtual_crtc_funcs.

v2: agd: rebase on upstream

Signed-off-by: Emily Deng <[email protected]>
Reviewed-by: Alex Deucher <[email protected]>
Signed-off-by: Alex Deucher <[email protected]>

drm/amdgpu: Initialize dce_virtual_crtc_helper_funcs

For virtual display feature, initialize dce_virtual_crtc_helper_funcs.

Signed-off-by: Emily Deng <[email protected]>
Reviewed-by: Alex Deucher <[email protected]>
Signed-off-by: Alex Deucher <[email protected]>

drm/amdgpu: Initialize crtc, pageflip irq funcs (v2)

For virtual display feature, initialize dce_virtual_crtc_irq_funcs,
dce_virtual_pageflip_irq_funcs. As it has no dce engine, so the
pageflip interrupt won't be generated, and the vsync interrupt will
be generated by smu's periodic timer or software timer which will
be implemented later.

v2: agd: rebase on upstream

Signed-off-by: Emily Deng <[email protected]>
Reviewed-by: Alex Deucher <[email protected]>
Signed-off-by: Alex Deucher <[email protected]>

drm/amdgpu: Initialize dce_virtual_display_funcs.

For virtual display feature, initialize dce_virtual_display_funcs,
which will be used in function dce_virtual_set_display_funcs.

Signed-off-by: Emily Deng <[email protected]>
Reviewed-by: Alex Deucher <[email protected]>
Signed-off-by: Alex Deucher <[email protected]>

drm/amdgpu: Initialize dce_virtual_ip_funcs

For virtual display feature, first need to initialize dce_virtual_ip_funcs,
which will be used when set ip blocks.

Signed-off-by: Emily Deng <[email protected]>
Reviewed-by: Alex Deucher <[email protected]>
Signed-off-by: Alex Deucher <[email protected]>

drm/amdgpu: Add virtual connector and encoder macros.

For virtual display feature, add virtual connector and encoder macros.

Signed-off-by: Emily Deng <[email protected]>
Reviewed-by: Alex Deucher <[email protected]>
Signed-off-by: Alex Deucher <[email protected]>

drm/amdgpu: Don't retry 7 times in amdgpu_atombios_dp_get_dpcd()

When this code was written, we didn't retry DP aux transactions on any
error, which required retrying important transactions like this in
individual drivers. Since that's no longer the case, retrying here is
not necessary. As well, we retry any aux transaction on any error 32
times. 7 * 32 = 224, which means this loop causes us to retry grabbing
the dpcd 224 times. This is definitely far more then we actually need to
do.

Reviewed-by: Christian König <[email protected]>
Signed-off-by: Lyude <[email protected]>
Signed-off-by: Alex Deucher <[email protected]>

drm/amdgpu: Don't print error on aux transaction timeouts

Since it's normal for DRM to retry our aux transaction helpers multiple
times in a row, up to 32 times for each attempted transaction, we're
making a lot of noise that is no longer necessary now that DRM will just
print the return code we give it.

Reviewed-by: Christian König <[email protected]>
Signed-off-by: Lyude <[email protected]>
Signed-off-by: Alex Deucher <[email protected]>

drm/radeon: Don't retry 7 times in radeon_dp_dpcd()

When this code was written, we didn't retry DP aux transactions on any
error, which required retrying important transactions like this in
individual drivers. Since that's no longer the case, retrying here is
not necessary. As well, we retry any aux transaction on any error 32
times. 7 * 32 = 224, which means this loop causes us to retry grabbing
the dpcd 224 times. This is definitely far more then we actually need to
do.

Reviewed-by: Christian König <[email protected]>
Signed-off-by: Lyude <[email protected]>
Signed-off-by: Alex Deucher <[email protected]>

drm/radeon: Don't print error on aux transaction timeouts

Since it's normal for DRM to retry our aux transaction helpers multiple
times in a row, up to 32 times for each attempted transaction, we're
making a lot of noise that is no longer necessary now that DRM will just
print the return code we give it.

Reviewed-by: Christian König <[email protected]>
Signed-off-by: Lyude <[email protected]>
Signed-off-by: Alex Deucher <[email protected]>

drm/ttm: Remove unused parameter evict from ttm_bo_move_memcpy

Reviewed-by: Christian König <[email protected]>
Signed-off-by: Michel Dänzer <[email protected]>
Signed-off-by: Alex Deucher <[email protected]>

drm/ttm: Remove unused parameter evict from ttm_bo_move_ttm

Reviewed-by: Christian König <[email protected]>
Signed-off-by: Michel Dänzer <[email protected]>
Signed-off-by: Alex Deucher <[email protected]>

drm/ttm: Wait for a BO to become idle before unbinding it from GTT

Fixes hangs under memory pressure, e.g. running the piglit test
tex3d-maxsize concurrently with other tests.

Fixes: 17d33bc9d6ef ("drm/ttm: drop waiting for idle in ttm_bo_evict.")
Reviewed-by: Christian König <[email protected]>
Signed-off-by: Michel Dänzer <[email protected]>
Signed-off-by: Alex Deucher <[email protected]>

drm/amdgpu: move all Kconfig options to amdgpu/Kconfig

For consistency.

Reviewed-by: Harry Wentland <[email protected]>
Signed-off-by: Alex Deucher <[email protected]>

drm/amd/powerplay: enable powerplay as default on CZ/ST

Enable powerplay as default on Carrizo and Stoney. And it can be
disabled with amdgpu.powerplay=0.

Reviewed-by: Alex Deucher <[email protected]>
Signed-off-by: Huang Rui <[email protected]>
Cc: Rex Zhu <[email protected]>
Cc: Flora Cui <[email protected]>
Cc: Ken Wang <[email protected]>
Signed-off-by: Alex Deucher <[email protected]>

drm/amd/powerplay: change structure variable name.

Signed-off-by: Rex Zhu <[email protected]>
Reviewed-by: Alex Deucher <[email protected]>
Signed-off-by: Alex Deucher <[email protected]>

drm/amd/powerplay: delete code set/unset deep_sleep/power_containment.

they were controled by module parameter.

Signed-off-by: Rex Zhu <[email protected]>
Reviewed-by: Ken Wang <[email protected]>
Signed-off-by: Alex Deucher <[email protected]>

drm/amdgpu: enable VCE soft reset

Signed-off-by: Chunming Zhou <[email protected]>
Reviewed-by: Christian König <[email protected]>
Signed-off-by: Alex Deucher <[email protected]>

drm/amdgpu: remove RB1_BUSY bit checking

This is a workaround to let VCE soft reset work.
RB1_BUSY bit is always set, so remove its checking now, and we
will depend on RB0_BUSY currently.
After we find the root cause of RB1_BUSY, we can add it back.

Signed-off-by: Rex Zhu <[email protected]>
Reviewed-by: Christian König <[email protected]>
Signed-off-by: Alex Deucher <[email protected]>

drm/amdgpu: fix harvest config checking in vce3 check_soft_reset

Signed-off-by: Chunming Zhou <[email protected]>
Reviewed-by: Christian König <[email protected]>
Signed-off-by: Alex Deucher <[email protected]>

drm/amdgpu: irq resume should be immediately after gpu resume

Signed-off-by: Chunming Zhou <[email protected]>
Reviewed-by: Christian König <[email protected]>
Signed-off-by: Alex Deucher <[email protected]>

drm/amd/amdgpu: Partially revert change to UVD v3 CG

Partially undo changes made by commit:

drm/amd/amdgpu: don't track state in UVD clockgating

To keep bypass even if CG flags are not set.

Signed-off-by: Tom St Denis <[email protected]>
Reviewed-by: Alex Deucher <[email protected]>
Signed-off-by: Alex Deucher <[email protected]>

drm/amd/amdgpu: Enable VCE MGCG for Stoney

Signed-off-by: Tom St Denis <[email protected]>
Reviewed-by: Alex Deucher <[email protected]>
Signed-off-by: Alex Deucher <[email protected]>

drm/amd/amdgpu: Enable VCE MGCG for Carrizo

Signed-off-by: Tom St Denis <[email protected]>
Reviewed-by: Alex Deucher <[email protected]>
Signed-off-by: Alex Deucher <[email protected]>

drm/amd/amdgpu: Don't set VCE CG in startup

It's handled by DPM/PP properly.

Signed-off-by: Tom St Denis <[email protected]>
Reviewed-by: Alex Deucher <[email protected]>
Signed-off-by: Alex Deucher <[email protected]>

drm/amd/amdgpu: Fix VCE CG order and resume defaults

CG was being enabled in reverse sense from dpm/powerplay.
Also fix the default CLK_EN signal to enable all of the blocks.

Signed-off-by: Tom St Denis <[email protected]>
Reviewed-by: Alex Deucher <[email protected]>
Signed-off-by: Alex Deucher <[email protected]>

drm/amd/amdgpu: fix indentation in vce3 CG

Reviewed-by: Christian König <[email protected]>
Signed-off-by: Tom St Denis <[email protected]>
Reviewed-by: Alex Deucher <[email protected]>
Signed-off-by: Alex Deucher <[email protected]>

drm/amdgpu: remove the check for sessions being closed

This will make clock and power gated when no block decoded, for example
when paused during the playback.

Signed-off-by: Leo Liu <[email protected]>
Reviewed-by: Christian König <[email protected]>
Signed-off-by: Alex Deucher <[email protected]>

drm/amdgpu: print more accurate error messages on IB submission failure

It's useful for debugging.

Signed-off-by: Marek Olšák <[email protected]>
Reviewed-by: Alex Deucher <[email protected]>
Reviewed-by: Christian König <[email protected]>
Signed-off-by: Alex Deucher <[email protected]>

drm/amdgpu: use the kernel zone memory size as the max remote memory in amdgpu

Signed-off-by: Ken Wang <[email protected]>
Reviewed-by: Alex Deucher <[email protected]>
Signed-off-by: Alex Deucher <[email protected]>

drm/ttm: Add interface to export kernel_zone max memory size in ttm

Signed-off-by: Ken Wang <[email protected]>
Reviewed-by: Alex Deucher <[email protected]>
Reviewed-by: Christian König <[email protected]>
Signed-off-by: Alex Deucher <[email protected]>

drm/amd/powrplay: delete code no longer in use on Polaris.

Signed-off-by: Rex Zhu <[email protected]>
Reviewed-by: Alex Deucher <[email protected]>
Signed-off-by: Alex Deucher <[email protected]>

drm/amd/powrplay: workaround Memory EDC Error for certain partner boards.

Signed-off-by: Rex Zhu <[email protected]>
Reviewed-by: Alex Deucher <[email protected]>
Signed-off-by: Alex Deucher <[email protected]>

drm/amdgpu: AMD SW workaround for certain partner boards

Signed-off-by: Rex Zhu <[email protected]>
Reviewed-by: Alex Deucher <[email protected]>
Signed-off-by: Alex Deucher <[email protected]>

drm/amd/powerplay: pass sub_device_id and sub_vendor_id to powerplay.

Signed-off-by: Rex Zhu <[email protected]>
Reviewed-by: Alex Deucher <[email protected]>
Signed-off-by: Alex Deucher <[email protected]>

drm/amd/powerplay: delete useless code in hwmgr.

Signed-off-by: Rex Zhu <[email protected]>
Reviewed-by: Alex Deucher <[email protected]>
Signed-off-by: Alex Deucher <[email protected]>

drm/amdgpu: expose AMDGPU_GEM_CREATE_VRAM_CLEARED to user space

V2: fix the return value for fill failure and validate bo before
filling data

Reviewed-by: Christian König <[email protected]>
Signed-off-by: Flora Cui <[email protected]>
Reviewed-by: Chunming Zhou <[email protected]>
Reviewed-by: Alex Deucher <[email protected]>
Signed-off-by: Alex Deucher <[email protected]>

drm/amdgpu: implement amdgpu_fill_buffer()

so that bo could be set to some pattern

Signed-off-by: Flora Cui <[email protected]>
Reviewed-by: Chunming Zhou <[email protected]>
Reviewed-by: Christian König <[email protected]>
Signed-off-by: Alex Deucher <[email protected]>

drm/amdgpu/gfx8: remove stale function declaration

This got leftover somehow when I cleaned this up.

Reviewed-by: Christian König <[email protected]>
Reviewed-by: Chunming Zhou <[email protected]>
Signed-off-by: Alex Deucher <[email protected]>

drm/amd/amdgpu: change pptable output format from ASCII to binary

Reviewed-by: Edward O'Callaghan <[email protected]>
Reviewed-by: Alex Deucher <[email protected]>
Signed-off-by: Eric Huang <[email protected]>
Signed-off-by: Alex Deucher <[email protected]>

drm/amdgpu/ci: add mullins to default case for smc ucode

It's already covered by the default case, but add it for
consistency.

Reviewed-by: Alexandre Demers <[email protected]>
Reviewed-by: Edward O'Callaghan <[email protected]>
Signed-off-by: Alex Deucher <[email protected]>

drm/amdgpu/gmc7: add missing mullins case

Looks like this got missed when we ported the code from radeon.

Reviewed-by: Edward O'Callaghan <[email protected]>
Signed-off-by: Alex Deucher <[email protected]>
Cc: [email protected]

drm/amdgpu: use modules parameter to ctrl deep sleep feature in dpm

Signed-off-by: Rex Zhu <[email protected]>
Reviewed-by: Alex Deucher <[email protected]>
Signed-off-by: Alex Deucher <[email protected]>

drm/amdgpu: add module parameters to ctrl powerplay feature

Signed-off-by: Rex Zhu <[email protected]>
Reviewed-by: Alex Deucher <[email protected]>
Signed-off-by: Alex Deucher <[email protected]>

drm/amd/powerplay: fix typos of volt/millivolt symbols in comment.

Signed-off-by: Rex Zhu <[email protected]>
Reviewed-by: Alexandre Demers <[email protected]>
Reviewed-by: Alex Deucher <[email protected]>
Signed-off-by: Alex Deucher <[email protected]>

drm/amd/amdgpu: Enable stoney VCE PG

Signed-off-by: Tom St Denis <[email protected]>
Reviewed-by: Rex Zhu <[email protected]>
Signed-off-by: Alex Deucher <[email protected]>

drm/amd/amdgpu: Enable stoney UVD PG

Signed-off-by: Tom St Denis <[email protected]>
Reviewed-by: Rex Zhu <[email protected]>
Signed-off-by: Alex Deucher <[email protected]>

drm/amd/amdgpu: Enable stoney GFX PG

Signed-off-by: Tom St Denis <[email protected]>
Reviewed-by: Rex Zhu <[email protected]>
Signed-off-by: Alex Deucher <[email protected]>

drm/amd/amdgpu: Enable carrizo VCE PG

Signed-off-by: Tom St Denis <[email protected]>
Reviewed-by: Rex Zhu <[email protected]>
Signed-off-by: Alex Deucher <[email protected]>

drm/amd/amdgpu: Enable carrizo UVD PG

Signed-off-by: Tom St Denis <[email protected]>
Reviewed-by: Rex Zhu <[email protected]>
Signed-off-by: Alex Deucher <[email protected]>

drm/amd/amdgpu: Enable carrizo GFX PG

Signed-off-by: Tom St Denis <[email protected]>
Reviewed-by: Rex Zhu <[email protected]>
Signed-off-by: Alex Deucher <[email protected]>

drm/amd/amdgpu: Add error messages to UVD PG in DPM

Signed-off-by: Tom St Denis <[email protected]>
Reviewed-by: Rex Zhu <[email protected]>
Signed-off-by: Alex Deucher <[email protected]>

drm/amd/amdgpu: enable PG_EN bit in powergating UVD

Enable the PG_EN bit just before the SMU would be tasked
with the PG transition.

Signed-off-by: Tom St Denis <[email protected]>
Reviewed-by: Rex Zhu <[email protected]>
Signed-off-by: Alex Deucher <[email protected]>

drm/amd/amdgpu: don't track state in UVD clockgating

There's no need to track CG state anymore.

Signed-off-by: Tom St Denis <[email protected]>
Reviewed-by: Rex Zhu <[email protected]>
Signed-off-by: Alex Deucher <[email protected]>

drm/amd/amdgpu: don't set clockgating in uvd_v6_0_start()

This is handled properly by both DPM and PP externally.

Signed-off-by: Tom St Denis <[email protected]>
Reviewed-by: Rex Zhu <[email protected]>
Signed-off-by: Alex Deucher <[email protected]>

drm/amd/amdgpu: add pm lock to debugfs mmio entry

Adds support for PM locks around access to registers that might
have race conditions on PG transistions.

Signed-off-by: Tom St Denis <[email protected]>
Reviewed-by: Rex Zhu <[email protected]>
Signed-off-by: Alex Deucher <[email protected]>

drm/amd/amdgpu: add mutex locking for both DPM and PP based powergating for UVD/VCE

This adds a mutex lock for both DPM/PP around the changes in
power gating state so that userspace can poll registers without
a race condition on power state.

Signed-off-by: Tom St Denis <[email protected]>
Reviewed-by: Rex Zhu <[email protected]>
Signed-off-by: Alex Deucher <[email protected]>

drm/amdgpu/powerplay: enable powerplay by default on TOPAZ

Now that the implementation is complete.

Acked-by: Huang Rui <[email protected]>
Signed-off-by: Alex Deucher <[email protected]>

drm/amd/powerplay: add DPM running checking back

This patch adds DPM running checking back, because the DPM issue is
fixed.

Signed-off-by: Huang Rui <[email protected]>
Reviewed-by: Ken Wang <[email protected]>
Signed-off-by: Alex Deucher <[email protected]>

drm/amd/powerplay: add enabling voltage controller back

Signed-off-by: Huang Rui <[email protected]>
Reviewed-by: Ken Wang <[email protected]>
Signed-off-by: Alex Deucher <[email protected]>

drm/amd/powerplay: set the platform capability flags for iceland

Signed-off-by: Huang Rui <[email protected]>
Reviewed-by: Ken Wang <[email protected]>
Signed-off-by: Alex Deucher <[email protected]>

drm/amd/powerplay: add deep sleep initialization

This patch adds the deep sleep initialization at DPM, it needs send a
message to SMC to enable this feature before enable voltage controller.

Signed-off-by: Huang Rui <[email protected]>
Reviewed-by: Ken Wang <[email protected]>
Signed-off-by: Alex Deucher <[email protected]>

drm/amd/powerplay: fix the incorrect checking condition

Signed-off-by: Huang Rui <[email protected]>
Reviewed-by: Ken Wang <[email protected]>
Signed-off-by: Alex Deucher <[email protected]>

drm/amdgpu: enable iceland powerplay manually

It's able to enable iceland powerplay manually via the module
parameter. The default state is disabled.

Signed-off-by: Huang Rui <[email protected]>
Reviewed-by: Alex Deucher <[email protected]>
Signed-off-by: Alex Deucher <[email protected]>

drm/amd/powerplay: add iceland HW manager

This patch introduces the iceland HW manager of powerplay which
includes HW manager, clockpowergating, thermal, and powertune.

Signed-off-by: Huang Rui <[email protected]>
Reviewed-by: Alex Deucher <[email protected]>
Signed-off-by: Alex Deucher <[email protected]>