]> git.proxmox.com Git - mirror_ubuntu-kernels.git/log
mirror_ubuntu-kernels.git
4 years agodrm/amdgpu: enable TMZ bit in sdma copy pkt for sdma v4
Aaron Liu [Tue, 15 Oct 2019 08:47:44 +0000 (16:47 +0800)]
drm/amdgpu: enable TMZ bit in sdma copy pkt for sdma v4

Enable sdma TMZ mode via setting TMZ bit in sdma copy pkt
for sdma v4

Signed-off-by: Aaron Liu <aaron.liu@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Huang Rui <ray.huang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
4 years agodrm/amdgpu: expand amdgpu_copy_buffer interface with tmz parameter
Aaron Liu [Tue, 15 Oct 2019 07:45:23 +0000 (15:45 +0800)]
drm/amdgpu: expand amdgpu_copy_buffer interface with tmz parameter

This patch expands amdgpu_copy_buffer interface with tmz parameter.

Signed-off-by: Aaron Liu <aaron.liu@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Huang Rui <ray.huang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
4 years agodrm/amdgpu: expand sdma copy_buffer interface with tmz parameter
Aaron Liu [Tue, 15 Oct 2019 07:37:48 +0000 (15:37 +0800)]
drm/amdgpu: expand sdma copy_buffer interface with tmz parameter

This patch expands sdma copy_buffer interface with tmz parameter.

Signed-off-by: Aaron Liu <aaron.liu@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Huang Rui <ray.huang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
4 years agodrm/amdgpu: fix up for amdgpu_tmz.c and removal of drm/drmP.h
Stephen Rothwell [Wed, 16 Oct 2019 00:22:07 +0000 (11:22 +1100)]
drm/amdgpu: fix up for amdgpu_tmz.c and removal of drm/drmP.h

Signed-off-by: Stephen Rothwell <sfr@canb.auug.org.au>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
4 years agodrm/amdgpu: set TMZ bits in PTEs for secure BO (v4)
Alex Deucher [Thu, 8 Aug 2019 03:32:46 +0000 (22:32 -0500)]
drm/amdgpu: set TMZ bits in PTEs for secure BO (v4)

If a buffer object is secure, i.e. created with
AMDGPU_GEM_CREATE_ENCRYPTED, then the TMZ bit of
the PTEs that belong the buffer object should be
set.

v1: design and draft the skeletion of TMZ bits setting on PTEs (Alex)
v2: return failure once create secure BO on non-TMZ platform  (Ray)
v3: amdgpu_bo_encrypted() only checks the BO (Luben)
v4: move TMZ flag setting into amdgpu_vm_bo_update  (Christian)

Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Reviewed-by: Huang Rui <ray.huang@amd.com>
Signed-off-by: Huang Rui <ray.huang@amd.com>
Signed-off-by: Luben Tuikov <luben.tuikov@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
4 years agodrm/amdgpu: job is secure iff CS is secure (v5)
Huang Rui [Thu, 8 Aug 2019 12:05:15 +0000 (20:05 +0800)]
drm/amdgpu: job is secure iff CS is secure (v5)

Mark a job as secure, if and only if the command
submission flag has the secure flag set.

v2: fix the null job pointer while in vmid 0
submission.
v3: Context --> Command submission.
v4: filling cs parser with cs->in.flags
v5: move the job secure flag setting out of amdgpu_cs_submit()

Signed-off-by: Huang Rui <ray.huang@amd.com>
Signed-off-by: Luben Tuikov <luben.tuikov@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
4 years agodrm/amdgpu: expand the context control interface with trust flag
Huang Rui [Wed, 22 Apr 2020 21:49:17 +0000 (17:49 -0400)]
drm/amdgpu: expand the context control interface with trust flag

This patch expands the context control function to support trusted flag while we
want to set command buffer in trusted mode.

Signed-off-by: Huang Rui <ray.huang@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Acked-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
4 years agodrm/amdgpu: expand the emit tmz interface with trusted flag
Huang Rui [Mon, 9 Mar 2020 17:53:37 +0000 (13:53 -0400)]
drm/amdgpu: expand the emit tmz interface with trusted flag

This patch expands the emit_tmz function to support trusted flag while we want
to set command buffer in trusted mode.

Signed-off-by: Huang Rui <ray.huang@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Acked-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
4 years agodrm/amdgpu: add tmz bit in frame control packet
Huang Rui [Thu, 8 Aug 2019 09:00:16 +0000 (17:00 +0800)]
drm/amdgpu: add tmz bit in frame control packet

This patch adds tmz bit in frame control pm4 packet, and it will used in future.

Signed-off-by: Huang Rui <ray.huang@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
4 years agodrm/amdgpu: add function to check tmz capability (v4)
Huang Rui [Thu, 19 Mar 2020 20:46:10 +0000 (16:46 -0400)]
drm/amdgpu: add function to check tmz capability (v4)

Add a function to check tmz capability with kernel parameter and ASIC type.

v2: use a per device tmz variable instead of global amdgpu_tmz.
v3: refine the comments for the function. (Luben)
v4: add amdgpu_tmz.c/h for future use.

Signed-off-by: Huang Rui <ray.huang@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
4 years agodrm/amdgpu: add amdgpu_tmz data structure
Huang Rui [Tue, 14 Jan 2020 15:10:51 +0000 (10:10 -0500)]
drm/amdgpu: add amdgpu_tmz data structure

This patch to add amdgpu_tmz structure which stores all tmz related fields.

Signed-off-by: Huang Rui <ray.huang@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
4 years agodrm/amdgpu: add tmz feature parameter (v2)
Huang Rui [Sun, 4 Aug 2019 08:33:16 +0000 (16:33 +0800)]
drm/amdgpu: add tmz feature parameter (v2)

This patch adds tmz parameter to enable/disable
the feature in the amdgpu kernel module. Nomally,
by default, it should be auto (rely on the
hardware capability).

But right now, it need to set "off" to avoid
breaking other developers' work because it's not
totally completed.

Will set "auto" till the feature is stable and
completely verified.

v2: add "auto" option for future use.

Signed-off-by: Huang Rui <ray.huang@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
4 years agodrm/amdgpu: define the TMZ bit for the PTE
Alex Deucher [Thu, 8 Aug 2019 03:31:50 +0000 (22:31 -0500)]
drm/amdgpu: define the TMZ bit for the PTE

Define the TMZ (encryption) bit in the page table entry (PTE) for
Raven and newer asics.

Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Reviewed-by: Huang Rui <ray.huang@amd.com>
4 years agodrm/amdgpu: add UAPI to create secure commands (v3)
Luben Tuikov [Mon, 23 Sep 2019 23:02:41 +0000 (19:02 -0400)]
drm/amdgpu: add UAPI to create secure commands (v3)

Add a flag to the command submission IOCTL
structure which when present indicates that this
command submission should be treated as
secure. The kernel driver uses this flag to
determine whether the engine should be
transitioned to secure or unsecure, or the work
can be submitted to a secure queue depending on
the IP.

v3: the flag is now at command submission IOCTL

Signed-off-by: Luben Tuikov <luben.tuikov@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
4 years agodrm/amdgpu: add UAPI for creating encrypted buffers
Alex Deucher [Thu, 8 Aug 2019 02:43:24 +0000 (21:43 -0500)]
drm/amdgpu: add UAPI for creating encrypted buffers

Add a flag to the GEM_CREATE ioctl to create encrypted buffers.
Buffers with this flag set will be created with the TMZ bit set
in the PTEs or engines accessing them.  This is required in order
to properly access the data from the engines.

Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Reviewed-by: Huang Rui <ray.huang@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
4 years agodrm/amd/display: 3.2.83.1
Aric Cyr [Wed, 22 Apr 2020 22:08:05 +0000 (18:08 -0400)]
drm/amd/display: 3.2.83.1

Update firmware blobs

Signed-off-by: Aric Cyr <aric.cyr@amd.com>
Acked-by: Aurabindo Pillai <aurabindo.pillai@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
4 years agodrm/amd/display: 3.2.83
Aric Cyr [Wed, 22 Apr 2020 22:08:04 +0000 (18:08 -0400)]
drm/amd/display: 3.2.83

Signed-off-by: Aric Cyr <aric.cyr@amd.com>
Acked-by: Aurabindo Pillai <aurabindo.pillai@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
4 years agodrm/amd/display: Use cursor locking to prevent flip delays
Aric Cyr [Wed, 22 Apr 2020 22:08:03 +0000 (18:08 -0400)]
drm/amd/display: Use cursor locking to prevent flip delays

[Why]
Current locking scheme for cursor can result in a flip missing
its vsync, deferring it for one or more vsyncs.  Result is a
potential for stuttering when cursor is moved.

[How]
Use cursor update lock so that flips are not blocked while cursor
is being programmed.

Signed-off-by: Aric Cyr <aric.cyr@amd.com>
Reviewed-by: Nicholas Kazlauskas <Nicholas.Kazlauskas@amd.com>
Acked-by: Aurabindo Pillai <aurabindo.pillai@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
4 years agodrm/amd/display: Internal refactoring to abstract color caps
Krunoslav Kovac [Wed, 22 Apr 2020 22:08:02 +0000 (18:08 -0400)]
drm/amd/display: Internal refactoring to abstract color caps

[Why&How]
modules/color calculates various colour operations which are translated
to abstracted HW. DCE 5-12 had almost no important changes, but
starting with DCN1, every new generation comes with fairly major
differences in color pipeline.
We would hack it with some DCN checks, but a better approach is to
abstract color pipe capabilities so modules/DM can decide mapping to
HW block based on logical capabilities,

Signed-off-by: Krunoslav Kovac <Krunoslav.Kovac@amd.com>
Reviewed-by: Aric Cyr <Aric.Cyr@amd.com>
Acked-by: Anthony Koo <Anthony.Koo@amd.com>
Acked-by: Aurabindo Pillai <aurabindo.pillai@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
4 years agodrm/amd/display: Add set backlight to hw sequencer.
Yongqiang Sun [Wed, 22 Apr 2020 22:08:01 +0000 (18:08 -0400)]
drm/amd/display: Add set backlight to hw sequencer.

[Why & How]
Add set backlight to hw sequencer, dmu communication will
be handled in hw sequencer for new asics.

Signed-off-by: Yongqiang Sun <yongqiang.sun@amd.com>
Reviewed-by: Anthony Koo <Anthony.Koo@amd.com>
Acked-by: Aurabindo Pillai <aurabindo.pillai@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
4 years agodrm/amd/display: Add dummy p-state latency bounding box override
Joshua Aberback [Wed, 22 Apr 2020 22:08:00 +0000 (18:08 -0400)]
drm/amd/display: Add dummy p-state latency bounding box override

[Why]

For debugging, it can be useful to be able to modify the dummy
p-state latency, this will make it easier to do so.

Signed-off-by: Joshua Aberback <joshua.aberback@amd.com>
Reviewed-by: Wesley Chalmers <Wesley.Chalmers@amd.com>
Acked-by: Aurabindo Pillai <aurabindo.pillai@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
4 years agodrm/amd/display: Add panel cntl id for set backlight level.
Yongqiang Sun [Wed, 22 Apr 2020 22:07:59 +0000 (18:07 -0400)]
drm/amd/display: Add panel cntl id for set backlight level.

[Why & How]
Add panel cntl instance when calling set backlight.

Signed-off-by: Yongqiang Sun <yongqiang.sun@amd.com>
Reviewed-by: Anthony Koo <Anthony.Koo@amd.com>
Acked-by: Aurabindo Pillai <aurabindo.pillai@amd.com>
Acked-by: Tony Cheng <Tony.Cheng@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
4 years agodrm/amd/display: Pass command instead of header into DMUB service
Nicholas Kazlauskas [Wed, 22 Apr 2020 22:07:58 +0000 (18:07 -0400)]
drm/amd/display: Pass command instead of header into DMUB service

[Why]
We read memory that we shouldn't be touching if the struct isn't
a full union dmub_rb_cmd.

[How]
Fix up all the callers and functions that take in the dmub_cmd_header
to use the dmub_rb_cmd instead.

Signed-off-by: Nicholas Kazlauskas <nicholas.kazlauskas@amd.com>
Reviewed-by: Tony Cheng <Tony.Cheng@amd.com>
Acked-by: Aurabindo Pillai <aurabindo.pillai@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
4 years agodrm/amd/display: Update downspread percent to match spreadsheet for DCN2.1
Sung Lee [Wed, 22 Apr 2020 22:07:57 +0000 (18:07 -0400)]
drm/amd/display: Update downspread percent to match spreadsheet for DCN2.1

[WHY]
The downspread percentage was copied over from a previous version
of the display_mode_lib spreadsheet. This value has been updated,
and the previous value is too high to allow for such modes as
4K120hz. The new value is sufficient for such modes.

[HOW]
Update the value in dcn21_resource to match the spreadsheet.

Signed-off-by: Sung Lee <sung.lee@amd.com>
Reviewed-by: Yongqiang Sun <yongqiang.sun@amd.com>
Acked-by: Aurabindo Pillai <aurabindo.pillai@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
4 years agodrm/amd/display: Defer cursor update around VUPDATE for all ASIC
Nicholas Kazlauskas [Wed, 22 Apr 2020 22:07:56 +0000 (18:07 -0400)]
drm/amd/display: Defer cursor update around VUPDATE for all ASIC

[Why]
Fixes the following scenario:

- Flip has been prepared sometime during the frame, update pending
- Cursor update happens right when VUPDATE would happen
- OPTC lock acquired, VUPDATE is blocked until next frame
- Flip is delayed potentially infinitely

With the igt@kms_cursor_legacy cursor-vs-flip-legacy test we can
observe nearly *13* frames of delay for some flips on Navi.

[How]
Apply the Raven workaround generically. When close enough to VUPDATE
block cursor updates from occurring from the dc_stream_set_cursor_*
helpers.

This could perhaps be a little smarter by checking if there were
pending updates or flips earlier in the frame on the HUBP side before
applying the delay, but this should be fine for now.

This fixes the kms_cursor_legacy test.

Signed-off-by: Nicholas Kazlauskas <nicholas.kazlauskas@amd.com>
Reviewed-by: Aric Cyr <Aric.Cyr@amd.com>
Acked-by: Aurabindo Pillai <aurabindo.pillai@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
4 years agodrm/amd/display: Change viewport limit to 12 for DCN2
Sung Lee [Wed, 22 Apr 2020 22:07:55 +0000 (18:07 -0400)]
drm/amd/display: Change viewport limit to 12 for DCN2

[WHY & HOW]
Viewport limit was set to 16 pixels due to an issue with MPO
on small viewports. This restriction does not apply and the
viewport limit can now be lowered.

Signed-off-by: Sung Lee <sung.lee@amd.com>
Reviewed-by: Dmytro Laktyushkin <Dmytro.Laktyushkin@amd.com>
Acked-by: Aurabindo Pillai <aurabindo.pillai@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
4 years agodrm/amd/display: Fail validation if building scaling params fails
Sung Lee [Wed, 22 Apr 2020 22:07:54 +0000 (18:07 -0400)]
drm/amd/display: Fail validation if building scaling params fails

[WHY & HOW]
If building scaling parameters fails, validation
should also fail.

Signed-off-by: Sung Lee <sung.lee@amd.com>
Reviewed-by: Dmytro Laktyushkin <Dmytro.Laktyushkin@amd.com>
Acked-by: Aurabindo Pillai <aurabindo.pillai@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
4 years agodrm/amd/display: fix rn soc bb update
Dmytro Laktyushkin [Wed, 22 Apr 2020 22:07:53 +0000 (18:07 -0400)]
drm/amd/display: fix rn soc bb update

Currently RN SOC bounding box update assumes we will get at least
2 clock states from SMU. This isn't always true and because of special
casing on first clock state we end up with low disp, dpp, dsc and phy
clocks.

This change removes the special casing allowing the first state to
acquire correct clocks.

Signed-off-by: Dmytro Laktyushkin <Dmytro.Laktyushkin@amd.com>
Reviewed-by: Eric Yang <eric.yang2@amd.com>
Acked-by: Aurabindo Pillai <aurabindo.pillai@amd.com>
Acked-by: Tony Cheng <Tony.Cheng@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
4 years agodrm/amd/display: check if REFCLK_CNTL register is present
Dmytro Laktyushkin [Wed, 22 Apr 2020 22:07:52 +0000 (18:07 -0400)]
drm/amd/display: check if REFCLK_CNTL register is present

Check before programming the register since it isn't present on
all IPs using this code.

Signed-off-by: Dmytro Laktyushkin <Dmytro.Laktyushkin@amd.com>
Reviewed-by: Eric Bernstein <Eric.Bernstein@amd.com>
Acked-by: Aurabindo Pillai <aurabindo.pillai@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
4 years agodrm/amd/display: Add DML variable for future asics
Joshua Aberback [Wed, 22 Apr 2020 22:07:51 +0000 (18:07 -0400)]
drm/amd/display: Add DML variable for future asics

Signed-off-by: Joshua Aberback <joshua.aberback@amd.com>
Reviewed-by: Dmytro Laktyushkin <Dmytro.Laktyushkin@amd.com>
Reviewed-by: Jun Lei <Jun.Lei@amd.com>
Acked-by: Aurabindo Pillai <aurabindo.pillai@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
4 years agodrm/amd/display: clean up some header paths
Anthony Koo [Wed, 22 Apr 2020 22:07:50 +0000 (18:07 -0400)]
drm/amd/display: clean up some header paths

[Why]
Some include paths don't need to have relative paths
And some types missing

[How]
make some changes to headers and modify include path

Signed-off-by: Anthony Koo <Anthony.Koo@amd.com>
Reviewed-by: Tony Cheng <Tony.Cheng@amd.com>
Acked-by: Aurabindo Pillai <aurabindo.pillai@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
4 years agodrm/amd/display: Fix DMUB meta offset for new load method
Nicholas Kazlauskas [Wed, 22 Apr 2020 22:07:49 +0000 (18:07 -0400)]
drm/amd/display: Fix DMUB meta offset for new load method

[Why]
The new metadata offset is located at the end of the firmware binary
without any additional padding.

Firmware state is currently larger than 1024 bytes so new firmware state
will hang when trying to access any data above 1024 bytes.

[How]
Specify the correct offset based on legacy vs new loading method.

Signed-off-by: Nicholas Kazlauskas <nicholas.kazlauskas@amd.com>
Reviewed-by: Yongqiang Sun <yongqiang.sun@amd.com>
Acked-by: Aurabindo Pillai <aurabindo.pillai@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
4 years agodrm/amd/display: Do not disable pipe split if mode is not supported
Sung Lee [Wed, 22 Apr 2020 22:07:48 +0000 (18:07 -0400)]
drm/amd/display: Do not disable pipe split if mode is not supported

[WHY]
If mode is not supported, pipe split should not be disabled.
This may cause more modes to fail.

[HOW]
Check for mode support before disabling pipe split.

This commit was previously reverted as it was thought to
have problems, but those issues have been resolved.

Signed-off-by: Sung Lee <sung.lee@amd.com>
Reviewed-by: Yongqiang Sun <yongqiang.sun@amd.com>
Acked-by: Aurabindo Pillai <aurabindo.pillai@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
4 years agodrm/amd/display: 3.2.82
Aric Cyr [Wed, 22 Apr 2020 22:07:47 +0000 (18:07 -0400)]
drm/amd/display: 3.2.82

Signed-off-by: Aric Cyr <aric.cyr@amd.com>
Acked-by: Aurabindo Pillai <aurabindo.pillai@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
4 years agodrm/amdgpu: bump version for invalidate L2 before SDMA IBs
Marek Olšák [Mon, 27 Apr 2020 20:25:39 +0000 (16:25 -0400)]
drm/amdgpu: bump version for invalidate L2 before SDMA IBs

This fixes GPU hangs due to cache coherency issues.
Bump the driver version. Split out from the original patch.

Signed-off-by: Marek Olšák <marek.olsak@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Tested-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
4 years agodrm/amdgpu: invalidate L2 before SDMA IBs (v2)
Marek Olšák [Mon, 27 Apr 2020 19:59:22 +0000 (15:59 -0400)]
drm/amdgpu: invalidate L2 before SDMA IBs (v2)

This fixes GPU hangs due to cache coherency issues.

v2: Split the version bump to a separate patch

Signed-off-by: Marek Olšák <marek.olsak@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Tested-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
4 years agodrm/amdgpu: add tiling flags from Mesa
Marek Olšák [Sat, 18 Apr 2020 00:50:30 +0000 (20:50 -0400)]
drm/amdgpu: add tiling flags from Mesa

DCC_INDEPENDENT_128B is needed for displayble DCC on gfx10.
SCANOUT is not needed by the kernel, but Mesa uses it.

Signed-off-by: Marek Olšák <marek.olsak@amd.com>
Acked-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
4 years agodrm/amdgpu/vcn2.5: wait for tiles off after unpause
James Zhu [Mon, 27 Apr 2020 18:49:37 +0000 (14:49 -0400)]
drm/amdgpu/vcn2.5: wait for tiles off after unpause

Wait for tiles off after unpause to fix transcode timeout issue.
It is a work around.

Signed-off-by: James Zhu <James.Zhu@amd.com>
Reviewed-by: Leo Liu <leo.liu@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
4 years agodrm/amd/powerplay: avoid using pm_en before it is initialized revised
Tiecheng Zhou [Sun, 26 Apr 2020 11:03:17 +0000 (19:03 +0800)]
drm/amd/powerplay: avoid using pm_en before it is initialized revised

hwmgr->pm_en is initialized at hwmgr_hw_init.

during amdgpu_device_init, there is amdgpu_asic_reset that calls to
soc15_asic_reset (for V320 usecase, Vega10 asic), in which:
1) soc15_asic_reset_method calls to pp_get_asic_baco_capability (pm_en)
2) soc15_asic_baco_reset calls to pp_set_asic_baco_state (pm_en)

pm_en is used in the above two cases while it has not yet been initialized

So avoid using pm_en in the above two functions for V320 passthrough.

Reviewed-by: Evan Quan <evan.quan@amd.com>
Signed-off-by: Tiecheng Zhou <Tiecheng.Zhou@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
4 years agoRevert "drm/amd/powerplay: avoid using pm_en before it is initialized"
Tiecheng Zhou [Mon, 27 Apr 2020 01:44:38 +0000 (09:44 +0800)]
Revert "drm/amd/powerplay: avoid using pm_en before it is initialized"

This reverts commit c5207876232649ca5e5ddd6f966d2da75ffded8f.

The commit being reverted changed the wrong place, it should have
changed in func get_asic_baco_capability.

Reviewed-by: Evan Quan <evan.quan@amd.com>
Signed-off-by: Tiecheng Zhou <Tiecheng.Zhou@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
4 years agodrm/amdkfd: Put ASIC revision into HSA capability
Joseph Greathouse [Thu, 16 Apr 2020 19:08:59 +0000 (14:08 -0500)]
drm/amdkfd: Put ASIC revision into HSA capability

In order to surface the ASIC revision to user level, we want
to put it into the HSA topology. This can be because different
ASIC revisions may require user-level software to do different
things (e.g. patch code for things that are changed in later
hardware revisions).

The ASIC revision from the hardware is maximum of 4 bits at this
time, so put it into 4 of the open bits in the HSA capability.
Then user-level software can use this capability information to
know -- for each ASIC -- what revision-based things must be done.

Signed-off-by: Joseph Greathouse <Joseph.Greathouse@amd.com>
Reviewed-by: Felix Kuehling <Felix.Kuehling@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
4 years agodrm/amd/display: remove conversion to bool in dc_link_ddc.c
Jason Yan [Mon, 27 Apr 2020 06:37:26 +0000 (14:37 +0800)]
drm/amd/display: remove conversion to bool in dc_link_ddc.c

The '>' expression itself is bool, no need to convert it to bool again.
This fixes the following coccicheck warning:

drivers/gpu/drm/amd/display/dc/core/dc_link_ddc.c:602:28-33: WARNING:
conversion to bool not needed here

Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Jason Yan <yanaijie@huawei.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
4 years agodrm/amd/display: remove conversion to bool in dcn20_mpc.c
Jason Yan [Mon, 27 Apr 2020 06:37:14 +0000 (14:37 +0800)]
drm/amd/display: remove conversion to bool in dcn20_mpc.c

The '==' expression itself is bool, no need to convert it to bool again.
This fixes the following coccicheck warning:

drivers/gpu/drm/amd/display/dc/dcn20/dcn20_mpc.c:455:70-75: WARNING:
conversion to bool not needed here

Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Jason Yan <yanaijie@huawei.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
4 years agodrm/amdgpu: remove conversion to bool in amdgpu_device.c
Jason Yan [Mon, 27 Apr 2020 06:36:55 +0000 (14:36 +0800)]
drm/amdgpu: remove conversion to bool in amdgpu_device.c

The '>' expression itself is bool, no need to convert it to bool again.
This fixes the following coccicheck warning:

drivers/gpu/drm/amd/amdgpu/amdgpu_device.c:3004:68-73: WARNING:
conversion to bool not needed here

Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Jason Yan <yanaijie@huawei.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
4 years agodrm/amdgpu: decouple EccErrCnt query and clear operation
Guchun Chen [Sun, 26 Apr 2020 09:04:31 +0000 (17:04 +0800)]
drm/amdgpu: decouple EccErrCnt query and clear operation

Due to hardware bug that when RSMU UMC index is disabled,
clear EccErrCnt at the first UMC instance will clean up all other
EccErrCnt registes from other instances at the same time. This
will break the correctable error count log in EccErrCnt register
once querying it. So decouple both to make error count query workable.

Signed-off-by: Guchun Chen <guchun.chen@amd.com>
Reviewed-by: Tao Zhou <tao.zhou1@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
4 years agodrm/amdgpu: switch to SMN interface to operate RSMU index mode
Guchun Chen [Sun, 26 Apr 2020 08:44:16 +0000 (16:44 +0800)]
drm/amdgpu: switch to SMN interface to operate RSMU index mode

This makes consistent with other regsiters' access in this module.

Signed-off-by: Guchun Chen <guchun.chen@amd.com>
Reviewed-by: Tao Zhou <tao.zhou1@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
4 years agodrm/amdgpu: drop redundant cg/pg ungate on runpm enter
Evan Quan [Fri, 24 Apr 2020 07:36:22 +0000 (15:36 +0800)]
drm/amdgpu: drop redundant cg/pg ungate on runpm enter

CG/PG ungate is already performed in ip_suspend_phase1. Otherwise,
the CG/PG ungate will be performed twice. That will cause gfxoff
disablement is performed twice also on runpm enter while gfxoff
enablemnt once on rump exit. That will put gfxoff into disabled
state.

Signed-off-by: Evan Quan <evan.quan@amd.com>
Acked-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
4 years agodrm/amdgpu: move kfd suspend after ip_suspend_phase1
Evan Quan [Fri, 24 Apr 2020 07:32:41 +0000 (15:32 +0800)]
drm/amdgpu: move kfd suspend after ip_suspend_phase1

This sequence change should be safe as what did in ip_suspend_phase1
is to suspend DCE only. And this is a prerequisite for coming
redundant cg/pg ungate dropping.

Signed-off-by: Evan Quan <evan.quan@amd.com>
Acked-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
4 years agodrm/amdgpu: sw pstate switch should only be for vega20
Jonathan Kim [Fri, 24 Apr 2020 22:16:06 +0000 (18:16 -0400)]
drm/amdgpu: sw pstate switch should only be for vega20

Driver steered p-state switching is designed for Vega20 only.
Also simplify early return for temporary disable due to SMU FW
bug.

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>
4 years agodrm/amdgpu: Remove unneeded semicolon
Zheng Bin [Fri, 24 Apr 2020 07:56:20 +0000 (15:56 +0800)]
drm/amdgpu: Remove unneeded semicolon

Fixes coccicheck warning:

drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c:2534:2-3: Unneeded semicolon

Reported-by: Hulk Robot <hulkci@huawei.com>
Signed-off-by: Zheng Bin <zhengbin13@huawei.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
4 years agoamdgpu/dc: remove redundant assignment to variable 'option'
Colin Ian King [Fri, 24 Apr 2020 11:12:26 +0000 (12:12 +0100)]
amdgpu/dc: remove redundant assignment to variable 'option'

The variable option is being initialized with a value that is
never read and it is being updated later with a new value.  The
initialization is redundant and can be removed.

Addresses-Coverity: ("Unused value")
Signed-off-by: Colin Ian King <colin.king@canonical.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
4 years agodrm/amdgpu/gmc: Use consistent variable on unlocks
Colin Ian King [Fri, 24 Apr 2020 12:56:40 +0000 (13:56 +0100)]
drm/amdgpu/gmc: Use consistent variable on unlocks

Currently the error returns paths are unlocking lock kiq->ring_lock
however it seems this should be dev->gfx.kiq.ring_lock as this
is the lock that is being locked and unlocked around the ring
operations.  This looks like a bug, but it's not.  The kiq is just
a local variable pointing to the same structure.  Make it consistent.

Signed-off-by: Colin Ian King <colin.king@canonical.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
4 years agodrm/amd/display: remove redundant assignment to variable ret
Colin Ian King [Thu, 23 Apr 2020 14:17:28 +0000 (15:17 +0100)]
drm/amd/display: remove redundant assignment to variable ret

The variable ret is being initialized with a value that is never read
and it is being updated later with a new value. The initialization is
redundant and can be removed.

Addresses-Coverity: ("Unused value")
Signed-off-by: Colin Ian King <colin.king@canonical.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
4 years agodrm/amdgpu: protect ring overrun
Yintian Tao [Thu, 23 Apr 2020 04:05:54 +0000 (12:05 +0800)]
drm/amdgpu: protect ring overrun

Wait for the oldest sequence on the ring
to be signaled in order to make sure there
will be no command overrun.

v2: fix coding stype and remove abs operation
v3: remove the initialization of variable r

Signed-off-by: Yintian Tao <yttao@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
4 years agodrm/amdgpu: extent threshold of waiting FLR_COMPLETE
Monk Liu [Tue, 21 Apr 2020 10:04:50 +0000 (18:04 +0800)]
drm/amdgpu:  extent threshold of waiting FLR_COMPLETE

to 5s to satisfy WHOLE GPU reset which need 3+ seconds to
finish

Signed-off-by: Monk Liu <Monk.Liu@amd.com>
Acked-by: Yintian Tao <yttao@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
4 years agodrm/amdgpu: for nv12 always need smu ip
Monk Liu [Wed, 22 Apr 2020 04:09:16 +0000 (12:09 +0800)]
drm/amdgpu: for nv12 always need smu ip

because nv12 SRIOV support one vf mode

Signed-off-by: Monk Liu <Monk.Liu@amd.com>
Acked-by: Yintian Tao <yttao@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
4 years agodrm/amdgpu: skip sysfs node not belong to one vf mode
Monk Liu [Wed, 22 Apr 2020 04:19:13 +0000 (12:19 +0800)]
drm/amdgpu: skip sysfs node not belong to one vf mode

Signed-off-by: Monk Liu <Monk.Liu@amd.com>
Acked-by: Yintian Tao <yttao@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
4 years agodrm/amdgpu: enable one vf mode for nv12
Monk Liu [Wed, 22 Apr 2020 11:28:48 +0000 (19:28 +0800)]
drm/amdgpu: enable one vf mode for nv12

Signed-off-by: Monk Liu <Monk.Liu@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
4 years agodrm/amdgpu: clear the messed up checking logic
Monk Liu [Thu, 23 Apr 2020 06:57:40 +0000 (14:57 +0800)]
drm/amdgpu: clear the messed up checking logic

for ARCTURUS+ ASICS, we always support SW_SMU for bare-metal
and for SRIOV one_vf_mode

Signed-off-by: Monk Liu <Monk.Liu@amd.com>
Acked-by: Yintian Tao <yttao@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
4 years agodrm/amdgpu: provide RREG32_SOC15_NO_KIQ, will be used later
Monk Liu [Wed, 22 Apr 2020 11:27:44 +0000 (19:27 +0800)]
drm/amdgpu: provide RREG32_SOC15_NO_KIQ, will be used later

Signed-off-by: Monk Liu <Monk.Liu@amd.com>
Acked-by: Yintian Tao <yttao@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
4 years agodrm/amdgpu: sriov is forbidden to call disable DPM
Monk Liu [Wed, 22 Apr 2020 04:15:23 +0000 (12:15 +0800)]
drm/amdgpu: sriov is forbidden to call disable DPM

Signed-off-by: Monk Liu <Monk.Liu@amd.com>
Acked-by: Yintian Tao <yttao@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
4 years agodrm/amdgpu: skip cg/pg set for SRIOV
Monk Liu [Thu, 23 Apr 2020 03:09:23 +0000 (11:09 +0800)]
drm/amdgpu: skip cg/pg set for SRIOV

Signed-off-by: Monk Liu <Monk.Liu@amd.com>
Acked-by: Yintian Tao <yttao@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
4 years agodrm/amdgpu: ignore TA ucode for SRIOV
Monk Liu [Wed, 22 Apr 2020 03:42:37 +0000 (11:42 +0800)]
drm/amdgpu: ignore TA ucode for SRIOV

Signed-off-by: Monk Liu <Monk.Liu@amd.com>
Acked-by: Yintian Tao <yttao@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
4 years agodrm/amdgpu: retire legacy vega10 sos version check
Hawking Zhang [Mon, 20 Apr 2020 10:04:54 +0000 (18:04 +0800)]
drm/amdgpu: retire legacy vega10 sos version check

retired those early sos version used in vega10 bring up
phase

Signed-off-by: Hawking Zhang <Hawking.Zhang@amd.com>
Reviewed-by: Guchun Chen <guchun.chen@amd.com>
Reviewed-by: John Clements <john.clements@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
4 years agodrm/amdgpu: switch to helper function to init sos ucode
Hawking Zhang [Mon, 20 Apr 2020 10:01:33 +0000 (18:01 +0800)]
drm/amdgpu: switch to helper function to init sos ucode

call common helper function to init sos ucode, instead
of duplicate codes per ip version

Signed-off-by: Hawking Zhang <Hawking.Zhang@amd.com>
Reviewed-by: Guchun Chen <guchun.chen@amd.com>
Reviewed-by: John Clements <john.clements@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
4 years agodrm/amdgpu: add helper function to init sos ucode
Hawking Zhang [Mon, 20 Apr 2020 09:47:53 +0000 (17:47 +0800)]
drm/amdgpu: add helper function to init sos ucode

driver already had psp_firmware_header struture to
deal with different layout of sos ucode. the sos
micorcode initialization could be common one.

Signed-off-by: Hawking Zhang <Hawking.Zhang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
4 years agodrm/amdgpu: switch to helper function to init asd ucode
Hawking Zhang [Mon, 20 Apr 2020 09:35:20 +0000 (17:35 +0800)]
drm/amdgpu: switch to helper function to init asd ucode

call common helper function to initialize asd ucode

Signed-off-by: Hawking Zhang <Hawking.Zhang@amd.com>
Reviewed-by: Guchun Chen <guchun.chen@amd.com>
Reviewed-by: John Clements <john.clements@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
4 years agodrm/amdgpu: add helper function to init asd ucode
Hawking Zhang [Mon, 20 Apr 2020 09:15:07 +0000 (17:15 +0800)]
drm/amdgpu: add helper function to init asd ucode

asd is unified ucode across asic. it is not necessary
to keep its software structure to be ip specific one

Signed-off-by: Hawking Zhang <Hawking.Zhang@amd.com>
Reviewed-by: Guchun Chen <guchun.chen@amd.com>
Reviewed-by: John Clements <john.clements@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
4 years agodrm/amdgpu: retire unused check_fw_loading status
Hawking Zhang [Mon, 20 Apr 2020 08:26:15 +0000 (16:26 +0800)]
drm/amdgpu: retire unused check_fw_loading status

The driver can't access UCODE_DATA/ADDR registers on production boards.

Signed-off-by: Hawking Zhang <Hawking.Zhang@amd.com>
Reviewed-by: Guchun Chen <guchun.chen@amd.com>
Reviewed-by: John Clements <john.clements@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
4 years agodrm/amdgpu: remove unnecessary tOS version check
Hawking Zhang [Mon, 20 Apr 2020 07:57:26 +0000 (15:57 +0800)]
drm/amdgpu: remove unnecessary tOS version check

tOS version is available through debugfs interface

Signed-off-by: Hawking Zhang <Hawking.Zhang@amd.com>
Reviewed-by: Guchun Chen <guchun.chen@amd.com>
Reviewed-by: John Clements <john.clements@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
4 years agodrm/amdgpu: retire support_vmr_ring interface
Hawking Zhang [Mon, 20 Apr 2020 08:01:11 +0000 (16:01 +0800)]
drm/amdgpu: retire support_vmr_ring interface

vmr ring is dedicated for sriov vf (i.e.guest driver
in sriov), which is general communication interface
between driver and psp fw accross all ip version.
it is not correct to make it as ip specific callback.
it is even worse to check specific tOS version per IP
version (like psp_v11/v12).

Signed-off-by: Hawking Zhang <Hawking.Zhang@amd.com>
Reviewed-by: Guchun Chen <guchun.chen@amd.com>
Reviewed-by: John Clements <john.clements@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
4 years agodrm/amdgpu: shrink critical section in amdgpu_amdkfd_gpuvm_free_memory_of_gpu
Bernard Zhao [Tue, 21 Apr 2020 06:48:18 +0000 (23:48 -0700)]
drm/amdgpu: shrink critical section in amdgpu_amdkfd_gpuvm_free_memory_of_gpu

Reduce the mem->lock`s protected code area, no need to protect pr_debug.
This also simplifies error handling.

Signed-off-by: Bernard Zhao <bernard@vivo.com>
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>
4 years agodrm/amdgpu: Init data to avoid oops while reading pp_num_states.
limingyu [Wed, 22 Apr 2020 09:01:19 +0000 (17:01 +0800)]
drm/amdgpu: Init data to avoid oops while reading pp_num_states.

For chip like CHIP_OLAND with si enabled(amdgpu.si_support=1),
the amdgpu will expose pp_num_states to the /sys directory.
In this moment, read the pp_num_states file will excute the
amdgpu_get_pp_num_states func. In our case, the data hasn't
been initialized, so the kernel will access some ilegal
address, trigger the segmentfault and system will reboot soon:

    uos@uos-PC:~$ cat /sys/devices/pci0000\:00/0000\:00\:00.0/0000\:01\:00
    .0/pp_num_states

    Message from syslogd@uos-PC at Apr 22 09:26:20 ...
     kernel:[   82.154129] Internal error: Oops: 96000004 [#1] SMP

This patch aims to fix this problem, avoid that reading file
triggers the kernel sementfault.

Signed-off-by: limingyu <limingyu@uniontech.com>
Signed-off-by: zhoubinbin <zhoubinbin@uniontech.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
4 years agodrm/amdgpu: remove set but not used variable 'priority'
YueHaibing [Tue, 21 Apr 2020 02:41:25 +0000 (10:41 +0800)]
drm/amdgpu: remove set but not used variable 'priority'

drivers/gpu/drm/amd/amdgpu/amdgpu_job.c: In function amdgpu_job_submit:
drivers/gpu/drm/amd/amdgpu/amdgpu_job.c:148:26: warning: variable priority set but not used [-Wunused-but-set-variable]

commit 33abcb1f5a17 ("drm/amdgpu: set compute queue priority at mqd_init")
left behind this, remove it.

Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: YueHaibing <yuehaibing@huawei.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
4 years agodrm: amdgpu: fix kernel-doc struct warning
Randy Dunlap [Mon, 20 Apr 2020 01:50:15 +0000 (18:50 -0700)]
drm: amdgpu: fix kernel-doc struct warning

Fix a kernel-doc warning of missing struct field desription:

../drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c:92: warning: Function parameter or member 'vm' not described in 'amdgpu_vm_eviction_lock'

Fixes: a269e44989f3 ("drm/amdgpu: Avoid reclaim fs while eviction lock")
Signed-off-by: Randy Dunlap <rdunlap@infradead.org>
Cc: Signed-off-by: Alex Sierra <alex.sierra@amd.com>
Cc: Felix Kuehling <Felix.Kuehling@amd.com>
Cc: Christian König <christian.koenig@amd.com>
Cc: Alex Deucher <alexander.deucher@amd.com>
Cc: David (ChunMing) Zhou <David1.Zhou@amd.com>
Cc: amd-gfx@lists.freedesktop.org
Reviewed-by: Harry Wentland <harry.wentland@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
4 years agodrm: amd/display: fix Kconfig help text
Randy Dunlap [Tue, 21 Apr 2020 23:34:00 +0000 (16:34 -0700)]
drm: amd/display: fix Kconfig help text

Fix help text: indent one tab + 2 spaces; end a sentence with a
period; and collapse short lines of text to one line.

Fixes: 23c61b4599c4 ("drm/amd: Fix Kconfig indentation")
Fixes: 4562236b3bc0 ("drm/amd/dc: Add dc display driver (v2)")
Signed-off-by: Randy Dunlap <rdunlap@infradead.org>
Cc: Harry Wentland <harry.wentland@amd.com>
Cc: Alex Deucher <alexander.deucher@amd.com>
Cc: Krzysztof Kozlowski <krzk@kernel.org>
Reviewed-by: Harry Wentland <harry.wentland@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
4 years agodrm/amdgpu: request reg_val_offs each kiq read reg
Yintian Tao [Wed, 22 Apr 2020 11:58:22 +0000 (19:58 +0800)]
drm/amdgpu: request reg_val_offs each kiq read reg

According to the current kiq read register method,
there will be race condition when using KIQ to read
register if multiple clients want to read at same time
just like the expample below:
1. client-A start to read REG-0 throguh KIQ
2. client-A poll the seqno-0
3. client-B start to read REG-1 through KIQ
4. client-B poll the seqno-1
5. the kiq complete these two read operation
6. client-A to read the register at the wb buffer and
   get REG-1 value

Therefore, use amdgpu_device_wb_get() to request reg_val_offs
for each kiq read register.

v2: fix the error remove
v3: fix the print typo
v4: remove unused variables

Signed-off-by: Yintian Tao <yttao@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
4 years agodrm/amdgpu: change how we update mmRLC_SPM_MC_CNTL
Christian König [Tue, 21 Apr 2020 14:20:26 +0000 (16:20 +0200)]
drm/amdgpu: change how we update mmRLC_SPM_MC_CNTL

In pp_one_vf mode avoid the extra overhead and read/write the
registers without the KIQ.

Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Monk Liu <monk.liu@amd.com>
Acked-by: Yintian Tao <yintian.tao@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
4 years agodrm/amdgpu: set error query ready after all IPs late init
Dennis Li [Wed, 22 Apr 2020 04:22:54 +0000 (12:22 +0800)]
drm/amdgpu: set error query ready after all IPs late init

If set error query ready in amdgpu_ras_late_init, which will
cause some IP blocks aren't initialized, but their error query
is ready.

Signed-off-by: Dennis Li <Dennis.Li@amd.com>
Reviewed-by: Guchun Chen <guchun.chen@amd.com>
Reviewed-by: Hawking Zhang <Hawking.Zhang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
4 years agodrm/amdgpu: code cleanup around gpu reset
Evan Quan [Thu, 16 Apr 2020 04:39:04 +0000 (12:39 +0800)]
drm/amdgpu: code cleanup around gpu reset

Make code more readable.

Signed-off-by: Evan Quan <evan.quan@amd.com>
Reviewed-by: Andrey Grodzovsky <andrey.grodzovsky@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
4 years agodrm/amdgpu: optimize the gpu reset for XGMI setup V2
Evan Quan [Thu, 16 Apr 2020 04:27:28 +0000 (12:27 +0800)]
drm/amdgpu: optimize the gpu reset for XGMI setup V2

This is basically just some code cosmetic. The current design
for XGMI setup gput reset is to operate on current device(adev)
first and then on other devices from the hive(by another 'for' loop).
But actually we can do some sort to the device list(to put current
device 1st position) and handle all the devices in a single 'for'
loop.

V2: added missing hive->hive_lock protection

Signed-off-by: Evan Quan <evan.quan@amd.com>
Reviewed-by: Andrey Grodzovsky <andrey.grodzovsky@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
4 years agodrm/amdgpu: correct cancel_delayed_work_sync on gpu reset
Evan Quan [Thu, 16 Apr 2020 04:20:38 +0000 (12:20 +0800)]
drm/amdgpu: correct cancel_delayed_work_sync on gpu reset

As for XGMI setup, it should be performed on other devices
from the hive also.

Signed-off-by: Evan Quan <evan.quan@amd.com>
Reviewed-by: Andrey Grodzovsky <andrey.grodzovsky@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
4 years agodrm/amdgpu: correct fbdev suspend on gpu reset
Evan Quan [Thu, 16 Apr 2020 04:15:31 +0000 (12:15 +0800)]
drm/amdgpu: correct fbdev suspend on gpu reset

As for XGMI setup, it needs to be performed on
all the devices from the same hive.

Signed-off-by: Evan Quan <evan.quan@amd.com>
Acked-by: Andrey Grodzovsky <andrey.grodzovsky@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
4 years agodrm/amdgpu: cleanup coding style in amdkfd a bit
Bernard Zhao [Wed, 22 Apr 2020 01:46:18 +0000 (18:46 -0700)]
drm/amdgpu: cleanup coding style in amdkfd a bit

Make the code a bit more readable by using a common
error handling pattern.

Signed-off-by: Bernard Zhao <bernard@vivo.com>
Reviewed-by: Christian König <christian.koenig@amd.com>.
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>
4 years agodrm/amdgpu: clean up unused variable about ring lru
Kevin Wang [Mon, 20 Apr 2020 15:08:14 +0000 (23:08 +0800)]
drm/amdgpu: clean up unused variable about ring lru

clean up unused variable:
1. ring_lru_list
2. ring_lru_list_lock

related-commit:
drm/amdgpu: remove ring lru handling

Signed-off-by: Kevin Wang <kevin1.wang@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
4 years agodrm/amdgpu: replace DRM prefix with PCI device info for gfx/mmhub
Dennis Li [Sat, 18 Apr 2020 04:08:20 +0000 (12:08 +0800)]
drm/amdgpu: replace DRM prefix with PCI device info for gfx/mmhub

Prefix RAS message printing in gfx/mmhub with PCI device info,
which assists the debug in multiple GPU case.

Reviewed-by: Guchun Chen <guchun.chen@amd.com>
Signed-off-by: Dennis Li <Dennis.Li@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
4 years agodrm/amd/powerplay: limit smu support to Arcturus for onevf
Jiansong Chen [Mon, 20 Apr 2020 07:36:53 +0000 (15:36 +0800)]
drm/amd/powerplay: limit smu support to Arcturus for onevf

Under onevf mode the smu support to other chips is not well
verified yet.

Signed-off-by: Jiansong Chen <Jiansong.Chen@amd.com>
Reviewed-by: Evan Quan <evan.quan@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
4 years agodrm/amdgpu: disble vblank when unloading sriov driver
Jiawei [Fri, 17 Apr 2020 11:06:52 +0000 (19:06 +0800)]
drm/amdgpu: disble vblank when unloading sriov driver

disble vblank in dce_vitual_crtc_commit(), which is skipped
under sriov before

Reviewed-by: Emily Deng <Emily.Deng@amd.com>
Signed-off-by: Jiawei <Jiawei.Gu@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
4 years agodrm/amdgpu: Print CU information by default during initialization
Yong Zhao [Fri, 17 Apr 2020 20:11:48 +0000 (16:11 -0400)]
drm/amdgpu: Print CU information by default during initialization

This is convenient for multiple teams to obtain the information. Also,
add device info by using dev_info().

Signed-off-by: Yong Zhao <Yong.Zhao@amd.com>
Reviewed-by: Dennis Li <Dennis.Li@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
4 years agodrm/amd/powerplay: update smu12_driver_if.h to align with pmfw
Prike Liang [Fri, 3 Apr 2020 06:47:53 +0000 (14:47 +0800)]
drm/amd/powerplay: update smu12_driver_if.h to align with pmfw

Update the smu12_driver_if.h header to follow the pmfw release.

Signed-off-by: Prike Liang <Prike.Liang@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Acked-by: Huang Rui <ray.huang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
4 years agodrm/amdgpu/powerplay:avoid to show invalid DPM table info
Yuxian Dai [Wed, 15 Apr 2020 11:03:52 +0000 (19:03 +0800)]
drm/amdgpu/powerplay:avoid to show invalid DPM table info

for different ASIC support different the number of DPM levels,
we should avoid to show the invalid level value.
v1 -> v2:
follow the suggestion,clarifiy the description for this
change

Reviewed-by: Huang Rui <ray.huang@amd.com>
Signed-off-by: Yuxian Dai <Yuxian.Dai@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
4 years agodrm/amdgpu: Adjust the SDMA doorbell info printing
Yong Zhao [Fri, 17 Apr 2020 20:16:57 +0000 (16:16 -0400)]
drm/amdgpu: Adjust the SDMA doorbell info printing

Turn off the printing by default because it is not very useful, while
adding more details.

Signed-off-by: Yong Zhao <Yong.Zhao@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
4 years agodrm/amdkfd: Adjust three kfd dmesg printings during initialization
Yong Zhao [Fri, 17 Apr 2020 20:06:59 +0000 (16:06 -0400)]
drm/amdkfd: Adjust three kfd dmesg printings during initialization

Delete two printings which are not very useful, and change one from
pr_info() to pr_debug().

Signed-off-by: Yong Zhao <Yong.Zhao@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
4 years agodrm/amd/display: fix bug in the logic for panel power control
Anthony Koo [Fri, 17 Apr 2020 22:26:47 +0000 (18:26 -0400)]
drm/amd/display: fix bug in the logic for panel power control

[Why]
there's a bug in the new logic for panel power control.  the check is
wrong, and will skip panel power control under the wrong conditions.

[How]
fix to check for NULL panel_cntl

Signed-off-by: Anthony Koo <Anthony.Koo@amd.com>
Reviewed-by: Ashley Thomas <Ashley.Thomas2@amd.com>
Acked-by: Rodrigo Siqueira <Rodrigo.Siqueira@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
4 years agodrm/amd/display: Convert memory from cpu to fw endianness correctly
Wyatt Wood [Fri, 17 Apr 2020 18:10:23 +0000 (14:10 -0400)]
drm/amd/display: Convert memory from cpu to fw endianness correctly

[Why]
Current code does not guarantee the correct endianness of memory being
copied to fw, specifically in the case where cpu isn't little endian.

[How]
Windows and Diags are always little endian, so we define a macro that
does nothing. Linux already defines this macro and will do the correct
endianness conversion.

Signed-off-by: Wyatt Wood <wyatt.wood@amd.com>
Reviewed-by: Harry Wentland <Harry.Wentland@amd.com>
Acked-by: Anthony Koo <Anthony.Koo@amd.com>
Acked-by: Rodrigo Siqueira <Rodrigo.Siqueira@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
4 years agodrm/amd/display: Fix green screen issue after suspend
Rodrigo Siqueira [Wed, 22 Apr 2020 21:43:26 +0000 (17:43 -0400)]
drm/amd/display: Fix green screen issue after suspend

[why]
We have seen a green screen after resume from suspend in a Raven system
connected with two displays (HDMI and DP) on X based system. We noticed
that this issue is related to bad DCC metadata from user space which may
generate hangs and consequently an underflow on HUBP. After taking a
deep look at the code path we realized that after resume we try to
restore the commit with the DCC enabled framebuffer but the framebuffer
is no longer valid.

[how]
This problem was only reported on Raven based system and after suspend,
for this reason, this commit adds a new parameter on
fill_plane_dcc_attributes() to give the option of disabling DCC
programmatically. In summary, for disabling DCC we first verify if is a
Raven system and if it is in suspend; if both conditions are true we
disable DCC temporarily, otherwise, it is enabled.

Co-developed-by: Nicholas Kazlauskas <nicholas.kazlauskas@amd.com>
Signed-off-by: Nicholas Kazlauskas <nicholas.kazlauskas@amd.com>
Signed-off-by: Rodrigo Siqueira <Rodrigo.Siqueira@amd.com>
Reviewed-by: Nicholas Kazlauskas <Nicholas.Kazlauskas@amd.com>
Acked-by: Rodrigo Siqueira <Rodrigo.Siqueira@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
4 years agodrm/amd/display: Adjust refactored dm for color management only
Stylon Wang [Thu, 9 Apr 2020 14:37:47 +0000 (22:37 +0800)]
drm/amd/display: Adjust refactored dm for color management only

[Why]
Commit cdde482caabf2adad47d23f0d1e235db2c4f2b7d is causing regression
from changing the order of call sequence.

[How]
Keep the call sequence and take in extra dm state only if plane-level
color management is enabled.

Fixes: cdde482caabf2a ("drm/amd/display: Refactor color management to take dm plane state")
Signed-off-by: Stylon Wang <stylon.wang@amd.com>
Reviewed-by: Nicholas Kazlauskas <Nicholas.Kazlauskas@amd.com>
Acked-by: Rodrigo Siqueira <Rodrigo.Siqueira@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
4 years agodrm/amd/display: access ABM from stream resource.
Yongqiang Sun [Thu, 9 Apr 2020 00:40:02 +0000 (20:40 -0400)]
drm/amd/display: access ABM from stream resource.

[Why]
Since ABM resource is mapped to stream res, all the ABM access should
via stream res.

[How]
Get ABM instance from stream res instead of resource pool.

Signed-off-by: Yongqiang Sun <yongqiang.sun@amd.com>
Reviewed-by: Tony Cheng <Tony.Cheng@amd.com>
Acked-by: Rodrigo Siqueira <Rodrigo.Siqueira@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
4 years agodrm/amd/display: blank dp stream before re-train the link
Xiaodong Yan [Thu, 9 Apr 2020 09:37:40 +0000 (17:37 +0800)]
drm/amd/display: blank dp stream before re-train the link

[Why]
When link loss happened, monitor can not light up if only re-train the
link.

[How]
Blank all the DP streams on this link before re-train the link, and then
unblank the stream

Signed-off-by: Xiaodong Yan <Xiaodong.Yan@amd.com>
Reviewed-by: Tony Cheng <Tony.Cheng@amd.com>
Acked-by: Rodrigo Siqueira <Rodrigo.Siqueira@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
4 years agodrm/amd/display: DispalyPort: Write OUI only if panel supports it
Aurabindo Pillai [Tue, 7 Apr 2020 14:22:27 +0000 (10:22 -0400)]
drm/amd/display: DispalyPort: Write OUI only if panel supports it

[why]
Organizational Unit Identifier register is optional, and its
presence is published via Down Stream Port Count register.
Writing this register when not available will result in errors

[how]
Read this register and continue writing OUI only if the panel
has the support advertised.

Signed-off-by: Aurabindo Pillai <aurabindo.pillai@amd.com>
Reviewed-by: Aric Cyr <Aric.Cyr@amd.com>
Acked-by: Rodrigo Siqueira <Rodrigo.Siqueira@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>