drm/amdgpu: Clear RAS interrupt status on aldebaran

resolve register address issue for detecting/clearing RAS interrupt

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

drm/amdgpu: Add support for RAS XGMI err query

Update XGMI RAS to support error query on aldebaran

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

drm/amdkfd: Account for SH/SE count when setting up cu masks.

On systems with multiple SH per SE compute_static_thread_mgmt_se#
is split into independent masks, one for each SH, in the upper and
lower 16 bits.  We need to detect this and apply cu masking to each
SH.  The cu mask bits are assigned first to each SE, then to
alternate SHs, then finally to higher CU id.  This ensures that
the maximum number of SPIs are engaged as early as possible while
balancing CU assignment to each SH.

v2: Use max SH/SE rather than max SH in cu_per_sh.

v3: Fix comment blocks, ensure se_mask is initially zero filled,
    and correctly assign se.sh.cu positions to unset bits in cu_mask.

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

drm/amdgpu: rename amdgpu_bo_get_preferred_pin_domain

amdgpu_bo_get_preferred_pin_domain is used for page tables
creation, which is not involved with page pinning. And it is used in
more cases than display scanout, modify its documentation as well.

Signed-off-by: Yifan Zhang <yifan1.zhang@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: drop redundant cancel_delayed_work_sync call

As those _sw_fini() APIs follow just after _suspend() APIs.
And the cancel_delayed_work_sync was already called in latter.

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

drm/amdgpu: add missing cleanups for more ASICs on UVD/VCE suspend

This is a supplement for commit below:
"drm/amdgpu: add missing cleanups for Polaris12 UVD/VCE on suspend".

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

drm/amdgpu: add missing cleanups for Polaris12 UVD/VCE on suspend

Perform proper cleanups on UVD/VCE suspend: powergate enablement,
clockgating enablement and dpm disablement. This can fix some hangs
observed on suspending when UVD/VCE still using(e.g. issue
"pm-suspend" when video is still playing).

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

drm/amdkfd: map SVM range with correct access permission

Restore retry fault or prefetch range, or restore svm range after
eviction to map range to GPU with correct read or write access
permission.

Range may includes multiple VMAs, update GPU page table with offset of
prange, number of pages for each VMA according VMA access permission.

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

drm/amdkfd: check access permisson to restore retry fault

Check range access permission to restore GPU retry fault, if GPU retry
fault on address which belongs to VMA, and VMA has no read or write
permission requested by GPU, failed to restore the address. The vm fault
event will pass back to user space.

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

drm/amdgpu: Update RAS XGMI Error Query

Resolve bug querying error on unsupported ASIC

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

drm/amdgpu: Add driver infrastructure for MCA RAS

Add MCA specific IP blocks targetting RAS features

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

drm/amd/display: Add Logging for HDMI color depth information

[Why]
Recent HDMI2.0 HF1-1 V-Swing testing showed that logging deep color
status helps in validation of testcase.

[How]
Add logging based on various color depths and pixel encoding
formats.

Signed-off-by: Praful Swarnakar <Praful.Swarnakar@amd.com>
Reviewed-by: Hersen Wu <hersenwu@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/amdgpu: consolidate PSP TA init shared buf functions

Signed-off-by: Candice Li <candice.li@amd.com>
Reviewed-by: John Clements <john.clements@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/amdgpu: add name field back to ras_common_if

Adding name field back to ras_common_if to work around error
injection failure with amdgpuras tool.

Signed-off-by: Candice Li <candice.li@amd.com>
Reviewed-by: John Clements <john.clements@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: Fix build with missing pm_suspend_target_state module export

Building a randconfig here triggered:

  ERROR: modpost: "pm_suspend_target_state" [drivers/gpu/drm/amd/amdgpu/amdgpu.ko] undefined!

because the module export of that symbol happens in
kernel/power/suspend.c which is enabled with CONFIG_SUSPEND.

The ifdef guards in amdgpu_acpi_is_s0ix_supported(), however, test for
CONFIG_PM_SLEEP which is defined like this:

  config PM_SLEEP
          def_bool y
          depends on SUSPEND || HIBERNATE_CALLBACKS

and that randconfig has:

  # CONFIG_SUSPEND is not set
  CONFIG_HIBERNATE_CALLBACKS=y

leading to the module export missing.

Change the ifdeffery to depend directly on CONFIG_SUSPEND.

Fixes: 5706cb3c910c ("drm/amdgpu: fix checking pmops when PM_SLEEP is not enabled")
Reviewed-by: Lijo Lazar <lijo.lazar@amd.com>
Signed-off-by: Borislav Petkov <bp@suse.de>
Link: https://lkml.kernel.org/r/YSP6Lv53QV0cOAsd@zn.tnic
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Cc: stable@vger.kernel.org

drm/radeon: switch from 'pci_' to 'dma_' API

The wrappers in include/linux/pci-dma-compat.h should go away.

The patch has been generated with the coccinelle script below.

It has been compile tested.

@@
@@
-    PCI_DMA_BIDIRECTIONAL
+    DMA_BIDIRECTIONAL

@@
@@
-    PCI_DMA_TODEVICE
+    DMA_TO_DEVICE

@@
@@
-    PCI_DMA_FROMDEVICE
+    DMA_FROM_DEVICE

@@
@@
-    PCI_DMA_NONE
+    DMA_NONE

@@
expression e1, e2, e3;
@@
-    pci_alloc_consistent(e1, e2, e3)
+    dma_alloc_coherent(&e1->dev, e2, e3, GFP_)

@@
expression e1, e2, e3;
@@
-    pci_zalloc_consistent(e1, e2, e3)
+    dma_alloc_coherent(&e1->dev, e2, e3, GFP_)

@@
expression e1, e2, e3, e4;
@@
-    pci_free_consistent(e1, e2, e3, e4)
+    dma_free_coherent(&e1->dev, e2, e3, e4)

@@
expression e1, e2, e3, e4;
@@
-    pci_map_single(e1, e2, e3, e4)
+    dma_map_single(&e1->dev, e2, e3, e4)

@@
expression e1, e2, e3, e4;
@@
-    pci_unmap_single(e1, e2, e3, e4)
+    dma_unmap_single(&e1->dev, e2, e3, e4)

@@
expression e1, e2, e3, e4, e5;
@@
-    pci_map_page(e1, e2, e3, e4, e5)
+    dma_map_page(&e1->dev, e2, e3, e4, e5)

@@
expression e1, e2, e3, e4;
@@
-    pci_unmap_page(e1, e2, e3, e4)
+    dma_unmap_page(&e1->dev, e2, e3, e4)

@@
expression e1, e2, e3, e4;
@@
-    pci_map_sg(e1, e2, e3, e4)
+    dma_map_sg(&e1->dev, e2, e3, e4)

@@
expression e1, e2, e3, e4;
@@
-    pci_unmap_sg(e1, e2, e3, e4)
+    dma_unmap_sg(&e1->dev, e2, e3, e4)

@@
expression e1, e2, e3, e4;
@@
-    pci_dma_sync_single_for_cpu(e1, e2, e3, e4)
+    dma_sync_single_for_cpu(&e1->dev, e2, e3, e4)

@@
expression e1, e2, e3, e4;
@@
-    pci_dma_sync_single_for_device(e1, e2, e3, e4)
+    dma_sync_single_for_device(&e1->dev, e2, e3, e4)

@@
expression e1, e2, e3, e4;
@@
-    pci_dma_sync_sg_for_cpu(e1, e2, e3, e4)
+    dma_sync_sg_for_cpu(&e1->dev, e2, e3, e4)

@@
expression e1, e2, e3, e4;
@@
-    pci_dma_sync_sg_for_device(e1, e2, e3, e4)
+    dma_sync_sg_for_device(&e1->dev, e2, e3, e4)

@@
expression e1, e2;
@@
-    pci_dma_mapping_error(e1, e2)
+    dma_mapping_error(&e1->dev, e2)

@@
expression e1, e2;
@@
-    pci_set_dma_mask(e1, e2)
+    dma_set_mask(&e1->dev, e2)

@@
expression e1, e2;
@@
-    pci_set_consistent_dma_mask(e1, e2)
+    dma_set_coherent_mask(&e1->dev, e2)

Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Christophe JAILLET <christophe.jaillet@wanadoo.fr>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: switch from 'pci_' to 'dma_' API

The wrappers in include/linux/pci-dma-compat.h should go away.

The patch has been generated with the coccinelle script below.

It has been compile tested.

@@
@@
-    PCI_DMA_BIDIRECTIONAL
+    DMA_BIDIRECTIONAL

@@
@@
-    PCI_DMA_TODEVICE
+    DMA_TO_DEVICE

@@
@@
-    PCI_DMA_FROMDEVICE
+    DMA_FROM_DEVICE

@@
@@
-    PCI_DMA_NONE
+    DMA_NONE

@@
expression e1, e2, e3;
@@
-    pci_alloc_consistent(e1, e2, e3)
+    dma_alloc_coherent(&e1->dev, e2, e3, GFP_)

@@
expression e1, e2, e3;
@@
-    pci_zalloc_consistent(e1, e2, e3)
+    dma_alloc_coherent(&e1->dev, e2, e3, GFP_)

@@
expression e1, e2, e3, e4;
@@
-    pci_free_consistent(e1, e2, e3, e4)
+    dma_free_coherent(&e1->dev, e2, e3, e4)

@@
expression e1, e2, e3, e4;
@@
-    pci_map_single(e1, e2, e3, e4)
+    dma_map_single(&e1->dev, e2, e3, e4)

@@
expression e1, e2, e3, e4;
@@
-    pci_unmap_single(e1, e2, e3, e4)
+    dma_unmap_single(&e1->dev, e2, e3, e4)

@@
expression e1, e2, e3, e4, e5;
@@
-    pci_map_page(e1, e2, e3, e4, e5)
+    dma_map_page(&e1->dev, e2, e3, e4, e5)

@@
expression e1, e2, e3, e4;
@@
-    pci_unmap_page(e1, e2, e3, e4)
+    dma_unmap_page(&e1->dev, e2, e3, e4)

@@
expression e1, e2, e3, e4;
@@
-    pci_map_sg(e1, e2, e3, e4)
+    dma_map_sg(&e1->dev, e2, e3, e4)

@@
expression e1, e2, e3, e4;
@@
-    pci_unmap_sg(e1, e2, e3, e4)
+    dma_unmap_sg(&e1->dev, e2, e3, e4)

@@
expression e1, e2, e3, e4;
@@
-    pci_dma_sync_single_for_cpu(e1, e2, e3, e4)
+    dma_sync_single_for_cpu(&e1->dev, e2, e3, e4)

@@
expression e1, e2, e3, e4;
@@
-    pci_dma_sync_single_for_device(e1, e2, e3, e4)
+    dma_sync_single_for_device(&e1->dev, e2, e3, e4)

@@
expression e1, e2, e3, e4;
@@
-    pci_dma_sync_sg_for_cpu(e1, e2, e3, e4)
+    dma_sync_sg_for_cpu(&e1->dev, e2, e3, e4)

@@
expression e1, e2, e3, e4;
@@
-    pci_dma_sync_sg_for_device(e1, e2, e3, e4)
+    dma_sync_sg_for_device(&e1->dev, e2, e3, e4)

@@
expression e1, e2;
@@
-    pci_dma_mapping_error(e1, e2)
+    dma_mapping_error(&e1->dev, e2)

@@
expression e1, e2;
@@
-    pci_set_dma_mask(e1, e2)
+    dma_set_mask(&e1->dev, e2)

@@
expression e1, e2;
@@
-    pci_set_consistent_dma_mask(e1, e2)
+    dma_set_coherent_mask(&e1->dev, e2)

Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Christophe JAILLET <christophe.jaillet@wanadoo.fr>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdkfd: CWSR with sw scheduler on Aldebaran and Arcturus

Program trap handler settings to enable CWSR with software scheduler
on Aldebaran and Arcturus.

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

drm/amdgpu/OLAND: clip the ref divider max value

This patch limits the ref_div_max value to 100, during the
calculation of PLL feedback reference divider. With current
value (128), the produced fb_ref_div value generates unstable
output at particular frequencies. Radeon driver limits this
value at 100.

On Oland, when we try to setup mode 2048x1280@60 (a bit weird,
I know), it demands a clock of 221270 Khz. It's been observed
that the PLL calculations using values 128 and 100 are vastly
different, and look like this:

+------------------------------------------+
|Parameter    |AMDGPU        |Radeon       |
|             |              |             |
+-------------+----------------------------+
|Clock feedback              |             |
|divider max  |  128         |   100       |
|cap value    |              |             |
|             |              |             |
|             |              |             |
+------------------------------------------+
|ref_div_max  |              |             |
|             |  42          |  20         |
|             |              |             |
|             |              |             |
+------------------------------------------+
|ref_div      |  42          |  20         |
|             |              |             |
+------------------------------------------+
|fb_div       |  10326       |  8195       |
+------------------------------------------+
|fb_div       |  1024        |  163        |
+------------------------------------------+
|fb_dev_p     |  4           |  9          |
|frac fb_de^_p|              |             |
+----------------------------+-------------+

With ref_div_max value clipped at 100, AMDGPU driver can also
drive videmode 2048x1280@60 (221Mhz) and produce proper output
without any blanking and distortion on the screen.

PS: This value was changed from 128 to 100 in Radeon driver also, here:
https://github.com/freedesktop/drm-tip/commit/4b21ce1b4b5d262e7d4656b8ececc891fc3cb806

V1:
Got acks from:
Acked-by: Alex Deucher <alexander.deucher@amd.com>
Acked-by: Christian König <christian.koenig@amd.com>
V2:
- Restricting the changes only for OLAND, just to avoid any regression
  for other cards.
- Changed unsigned -> unsigned int to make checkpatch quiet.

V3: Apply the change on SI family (not only oland) (Christian)

Cc: Alex Deucher <Alexander.Deucher@amd.com>
Cc: Christian König <christian.koenig@amd.com>
Cc: Eddy Qin <Eddy.Qin@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Shashank Sharma <shashank.sharma@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: refactor riommu invalidation wa

[Why]
A cleaner solution, only done once on boot.

[How]
Remove previous workaround and configure an extra
vmid one time on boot

Reviewed-by: Kazlauskas Nicholas <Nicholas.Kazlauskas@amd.com>
Acked-by: Solomon Chiu <solomon.chiu@amd.com>
Signed-off-by: Eric Yang <Eric.Yang2@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: Cancel delayed work when GFXOFF is disabled

schedule_delayed_work does not push back the work if it was already
scheduled before, so amdgpu_device_delay_enable_gfx_off ran ~100 ms
after the first time GFXOFF was disabled and re-enabled, even if GFXOFF
was disabled and re-enabled again during those 100 ms.

This resulted in frame drops / stutter with the upcoming mutter 41
release on Navi 14, due to constantly enabling GFXOFF in the HW and
disabling it again (for getting the GPU clock counter).

To fix this, call cancel_delayed_work_sync when the disable count
transitions from 0 to 1, and only schedule the delayed work on the
reverse transition, not if the disable count was already 0. This makes
sure the delayed work doesn't run at unexpected times, and allows it to
be lock-free.

v2:
* Use cancel_delayed_work_sync & mutex_trylock instead of
  mod_delayed_work.
v3:
* Make amdgpu_device_delay_enable_gfx_off lock-free (Christian König)
v4:
* Fix race condition between amdgpu_gfx_off_ctrl incrementing
  adev->gfx.gfx_off_req_count and amdgpu_device_delay_enable_gfx_off
  checking for it to be 0 (Evan Quan)

Cc: stable@vger.kernel.org
Reviewed-by: Evan Quan <evan.quan@amd.com>
Reviewed-by: Lijo Lazar <lijo.lazar@amd.com> # v3
Acked-by: Christian König <christian.koenig@amd.com> # v3
Signed-off-by: Michel Dänzer <mdaenzer@redhat.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: use the preferred pin domain after the check

For some reason we run into an use case where a BO is already pinned
into GTT, but should be pinned into VRAM|GTT again.

Handle that case gracefully as well.

Reviewed-by: Shashank Sharma <Shashank.sharma@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Cc: stable@vger.kernel.org

drm/amd/pm: a quick fix for "divided by zero" error

Considering Arcturus is a dedicated ASIC for computing, it
will be more proper to drop the support for fan speed reading
and setting. That's on the TODO list.

Signed-off-by: Evan Quan <evan.quan@amd.com>
Reported-by: Rui Teng <rui.teng@amd.com>
Reviewed-by: Guchun Chen <guchun.chen@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm: amdgpu: remove obsolete reference to config CHASH

Commit 04ed8459f334 ("drm/amdgpu: remove chash") removes the chash
architecture and its corresponding config CHASH.

There is still a reference to CHASH in the config DRM_AMDGPU in
./drivers/gpu/drm/Kconfig.

Remove this obsolete reference to config CHASH.

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

drm/amd/pm: Fix spelling mistake "firwmare" -> "firmware"

There is a spelling mistake in a dev_err error message. Fix it.

Signed-off-by: Colin Ian King <colin.king@canonical.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/amdgpu:flush ttm delayed work before cancel_sync

[Why]
In some cases when we unload driver, warning call trace
will show up in vram_mgr_fini which claims that LRU is not empty, caused
by the ttm bo inside delay deleted queue.

[How]
We should flush delayed work to make sure the delay deleting is done.

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

drm/amd: consolidate TA shared memory structures

Signed-off-by: Candice Li <candice.li@amd.com>
Reviewed-by: John Clements <john.clements@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: increase max xgmi physical node for aldebaran

aldebaran supports up to 16 xgmi physical nodes.

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

drm/amdgpu: disable BACO support for 699F:C7 polaris12 SKU temporarily

We have a S3 issue on that SKU with BACO enabled. Will bring back this
when that root caused.

Signed-off-by: Evan Quan <evan.quan@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: Use DCN30 watermark calc for DCN301

[why]
dcn301_calculate_wm_and_dl() causes flickering when external monitor is
connected.

This issue has been fixed before by commit 0e4c0ae59d7e
("drm/amdgpu/display: drop dcn301_calculate_wm_and_dl for now"), however
part of the fix was gone after commit 2cbcb78c9ee5 ("Merge tag 'amd-drm-next-5.13-2021-03-23' of https://gitlab.freedesktop.org/agd5f/linux into drm-next").

[how]
Use dcn30_calculate_wm_and_dlg() instead as in the original fix.

Fixes: 2cbcb78c9ee5 ("Merge tag 'amd-drm-next-5.13-2021-03-23' of https://gitlab.freedesktop.org/agd5f/linux into drm-next")

Signed-off-by: Nikola Cornij <nikola.cornij@amd.com>
Reviewed-by: Zhan Liu <zhan.liu@amd.com>
Tested-by: Zhan Liu <zhan.liu@amd.com>
Tested-by: Oliver Logush <oliver.logush@amd.com>
Signed-off-by: Zhan Liu <zhan.liu@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: correct MMSCH 1.0 version

MMSCH 1.0 doesn't have major/minor version, only verison.

Signed-off-by: Zhigang Luo <zhigang.luo@amd.com>
Reviewed by Shaoyun.liu <Shaoyun.liu@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: get extended xgmi topology data

The TA has a limit to the amount of data that can be retrieved from
GET_TOPOLOGY.  For setups that exceed this limit, the xGMI topology
needs to be re-initialized and data needs to be re-fetched from the
extended link records by setting a flag in the shared command buffer.

The number of hops and the number of links must be accumulated by the
driver. Other data points are all fetched from the first request.
Because the TA has already exceeded its link record limit, it
cannot hold bidirectional information.  Otherwise the driver would
have to do more than two fetches so the driver has to reflect the
topology information in the opposite direction.

v2: squashed with internal reviewed fix

Signed-off-by: Jonathan Kim <jonathan.kim@amd.com>
Reviewed-by: Hawking Zhang <hawking.zhang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/radeon: Add break to switch statement in radeonfb_create_pinned_object()

Clang + -Wimplicit-fallthrough warns:

drivers/gpu/drm/radeon/radeon_fb.c:170:2: warning: unannotated
fall-through between switch labels [-Wimplicit-fallthrough]
        default:
        ^
drivers/gpu/drm/radeon/radeon_fb.c:170:2: note: insert 'break;' to avoid
fall-through
        default:
        ^
        break;
1 warning generated.

Clang's version of this warning is a little bit more pedantic than
GCC's. Add the missing break to satisfy it to match what has been done
all over the kernel tree.

Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Nathan Chancellor <nathan@kernel.org>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: 3.2.149

This version brings along following fixes:
- Ensure DCN save init registers after VM setup
- Fix multi-display support for idle opt workqueue
- Use vblank control events for PSR enable/disable
- Create default dc_sink when fail reading EDID under MST

Acked-by: Wayne Lin <wayne.lin@amd.com>
Signed-off-by: Aric Cyr <aric.cyr@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: [FW Promotion] Release 0.0.79

Acked-by: Wayne Lin <wayne.lin@amd.com>
Signed-off-by: Anthony Koo <Anthony.Koo@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Guard vblank wq flush with DCN guards

[Why]
Compilation of the workqueue fails if not building with the DCN config
option set.

[How]
Guard calls to the flush with the DCN config option to fix the build.

Reviewed-by: Roman Li <Roman.Li@amd.com>
Acked-by: Wayne Lin <wayne.lin@amd.com>
Signed-off-by: Nicholas Kazlauskas <nicholas.kazlauskas@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Ensure DCN save after VM setup

[Why]
DM initializes VM context after DMCUB initialization.
This results in loss of DCN_VM_CONTEXT registers after z10.

[How]
Notify DMCUB when VM setup is complete, and have DMCUB
save init registers.

v2: squash in CONFIG_DRM_AMD_DC_DCN3_1 fix

Reviewed-by: Nicholas Kazlauskas <nicholas.kazlauskas@amd.com>
Acked-by: Wayne Lin <wayne.lin@amd.com>
Signed-off-by: Jake Wang <haonan.wang2@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Use vblank control events for PSR enable/disable

[Why]
PSR can disable the HUBP along with the OTG when PSR is active.

We'll hit a pageflip timeout when the OTG is disable because we're no
longer updating the CRTC vblank counter and the pflip high IRQ will
not fire on the flip.

In order to flip the page flip timeout occur we should modify the
enter/exit conditions to match DRM requirements.

[How]
Use our deferred handlers for DRM vblank control to notify DMCU(B)
when it can enable or disable PSR based on whether vblank is disabled or
enabled respectively.

We'll need to pass along the stream with the notification now because
we want to access the CRTC state while the CRTC is locked to get the
stream state prior to the commit.

Retain a reference to the stream so it remains safe to continue to
access and release that reference once we're done with it.

Enable/disable logic follows what we were previously doing in
update_planes.

The workqueue has to be flushed before programming streams or planes
to ensure that we exit out of idle optimizations and PSR before
these events occur if necessary.

To keep the skip count logic the same to avoid FBCON PSR enablement
requires copying the allow condition onto the DM IRQ parameters - a
field that we can actually access from the worker.

Reviewed-by: Roman Li <Roman.Li@amd.com>
Acked-by: Wayne Lin <wayne.lin@amd.com>
Signed-off-by: Nicholas Kazlauskas <nicholas.kazlauskas@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Fix multi-display support for idle opt workqueue

[Why]
The current implementation for idle optimization support only has a
single work item that gets reshuffled into the system workqueue
whenever we receive an enable or disable event.

We can have mismatched events if the work hasn't been processed or if
we're getting control events from multiple displays at once.

This fixes this issue and also makes the implementation usable for
PSR control - which will be addressed in another patch.

[How]
We need to be able to flush remaining work out on demand for driver stop
and psr disable so create a driver specific workqueue instead of using
the system one. The workqueue will be single threaded to guarantee the
ordering of enable/disable events.

Refactor the queue to allocate the control work and deallocate it
after processing it.

Pass the acrtc directly to make it easier to handle psr enable/disable
in a later patch.

Rename things to indicate that it's not just MALL specific.

Reviewed-by: Roman Li <Roman.Li@amd.com>
Acked-by: Wayne Lin <wayne.lin@amd.com>
Signed-off-by: Nicholas Kazlauskas <nicholas.kazlauskas@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Create dc_sink when EDID fail

[Why]
While reading remote EDID via Startech 1-to-4 hub, occasionally we
won't get response in time and won't light up corresponding monitor.

Ideally, we can still add generic modes for userspace to choose to try
to light up the monitor and which is done in
drm_helper_probe_single_connector_modes(). So the main problem here is
that we fail .mode_valid since we don't create remote dc_sink for this
case.

[How]
Also add default dc_sink if we can't get the EDID.

Reviewed-by: Nicholas Kazlauskas <nicholas.kazlauskas@amd.com>
Signed-off-by: Wayne Lin <Wayne.Lin@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/pm: correct the address of Arcturus fan related registers

These registers have different address from other SMU V11 ASICs.

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 manual mode check

As the fan control was guarded under manual mode before fan speed
RPM/PWM setting. Thus the extra check is totally redundant.

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 the unnecessary intermediate percent-based transition

Currently, the readout of fan speed pwm is transited into percent-based
and then pwm-based. However, the transition into percent-based is totally
unnecessary and make the final output less accurate.

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: correct the fan speed RPM retrieving

The relationship "PWM = RPM / smu->fan_max_rpm" between fan speed
PWM and RPM is not true for SMU11 ASICs. So, we need a new way to
retrieving the fan speed RPM.

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: correct the fan speed PWM retrieving

The relationship "PWM = RPM / smu->fan_max_rpm" between fan speed
PWM and RPM is not true for SMU11 ASICs. So, we need a new way to
retrieving the fan speed PWM.

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: record the RPM and PWM based fan speed settings

As the relationship "PWM = RPM / smu->fan_max_rpm" between fan speed
PWM and RPM is not true for SMU11 ASICs. So, both the RPM and PWM
settings need to be saved.

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: correct the fan speed RPM setting

The relationship "PWM = RPM / smu->fan_max_rpm" between fan speed
PWM and RPM is not true for SMU11 ASICs. So, we need a new way to
perform the fan speed RPM setting.

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/amdgpu: remove unnecessary RAS context field

Delete ras_if->name in the RAS ctx structure and remove related lines.

Signed-off-by: Candice Li <candice.li@amd.com>
Reviewed-by: John Clements <john.clements@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdkfd: fix random KFDSVMRangeTest.SetGetAttributesTest test failure

KFDSVMRangeTest.SetGetAttributesTest randomly fails in stress test.

Note: Google Test filter = KFDSVMRangeTest.*
[==========] Running 18 tests from 1 test case.
[----------] Global test environment set-up.
[----------] 18 tests from KFDSVMRangeTest
[ RUN      ] KFDSVMRangeTest.BasicSystemMemTest
[       OK ] KFDSVMRangeTest.BasicSystemMemTest (30 ms)
[ RUN      ] KFDSVMRangeTest.SetGetAttributesTest
[          ] Get default atrributes
/home/yifan/brahma/libhsakmt/tests/kfdtest/src/KFDSVMRangeTest.cpp:154: Failure
Value of: expectedDefaultResults[i]
  Actual: 4294967295
Expected: outputAttributes[i].value
Which is: 0
/home/yifan/brahma/libhsakmt/tests/kfdtest/src/KFDSVMRangeTest.cpp:154: Failure
Value of: expectedDefaultResults[i]
  Actual: 4294967295
Expected: outputAttributes[i].value
Which is: 0
/home/yifan/brahma/libhsakmt/tests/kfdtest/src/KFDSVMRangeTest.cpp:152: Failure
Value of: expectedDefaultResults[i]
  Actual: 4
Expected: outputAttributes[i].type
Which is: 2
[          ] Setting/Getting atrributes
[  FAILED  ]

the root cause is that svm work queue has not finished when svm_range_get_attr is called, thus
some garbage svm interval tree data make svm_range_get_attr get wrong result. Flush work queue before
iterate svm interval tree.

Signed-off-by: Yifan Zhang <yifan1.zhang@amd.com>
Reviewed-by: Felix Kuehling <Felix.Kuehling@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/pm: change the workload type for some cards

change the workload type for some cards as it is needed.

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

Revert "drm/amd/pm: fix workload mismatch on vega10"

This reverts commit 0979d43259e13846d86ba17e451e17fec185d240.
Revert this because it does not apply to all the cards.

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

drm/amd/amdgpu: consolidate PSP TA context

Signed-off-by: Candice Li <candice.li@amd.com>
Reviewed-by: John Clements <john.clements@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: Add MB_REQ_MSG_READY_TO_RESET response when VF get FLR notification.

When guest received FLR notification from host, it would
lock adapter into reset state. There will be no more
job submission and hardware access after that.

Then it should send a response to host that it has prepared
for host reset.

Signed-off-by: Jiange Zhao <Jiange.Zhao@amd.com>
Signed-off-by: Peng Ju Zhou <PengJu.Zhou@amd.com>
Reviewed-by: Emily.Deng <Emily.Deng@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/pm: change pp_dpm_sclk/mclk/fclk attribute is RO for aldebaran

the following clock is only support voltage DPM, change attribute to RO:
1. pp_dpm_sclk
2. pp_dpm_mclk
3. pp_dpm_fclk

Signed-off-by: Kevin Wang <kevin1.wang@amd.com>
Reviewed-by: Lijo Lazar <lijo.lazar@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/pm: change smu msg's attribute to allow working under sriov

the following message is allowed in sriov mode:
1. GetEnabledSmuFeaturesLow
2. GetEnabledSmuFeaturesHigh

Signed-off-by: Kevin Wang <kevin1.wang@amd.com>
Reviewed-by: Lijo Lazar <lijo.lazar@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/pm: change return value in aldebaran_get_power_limit()

v1:
1. change return value to avoid smu driver probe fails when FEATURE_PPT is
not enabled.
2. if FEATURE_PPT is not enabled, set power limit value to 0.

v2:
instead dev_err with dev_warn

Signed-off-by: Kevin Wang <kevin1.wang@amd.com>
Reviewed-by: Lijo Lazar <lijo.lazar@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/pm: skip to load smu microcode on sriov for aldebaran

v1:
1. skip to load smu firmware in sriov mode for aldebaran chip
2. using vbios pptable if in sriov mode.

v2:
clean up smu driver code in sriov code path

Signed-off-by: Kevin Wang <kevin1.wang@amd.com>
Reviewed-by: Lijo Lazar <lijo.lazar@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/pm: correct DPM_XGMI/VCN_DPM feature name

the following feature is wrong, it will cause sysnode of pp_features show error:
1. DPM_XGMI
2. VCN_DPM

Signed-off-by: Kevin Wang <kevin1.wang@amd.com>
Reviewed-by: Lijo Lazar <lijo.lazar@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/amdgpu embed hw_fence into amdgpu_job

Why: Previously hw fence is alloced separately with job.
It caused historical lifetime issues and corner cases.
The ideal situation is to take fence to manage both job
and fence's lifetime, and simplify the design of gpu-scheduler.

How:
We propose to embed hw_fence into amdgpu_job.
1. We cover the normal job submission by this method.
2. For ib_test, and submit without a parent job keep the
legacy way to create a hw fence separately.
v2:
use AMDGPU_FENCE_FLAG_EMBED_IN_JOB_BIT to show that the fence is
embedded in a job.
v3:
remove redundant variable ring in amdgpu_job
v4:
add tdr sequence support for this feature. Add a job_run_counter to
indicate whether this job is a resubmit job.
v5
add missing handling in amdgpu_fence_enable_signaling

Signed-off-by: Jingwen Chen <Jingwen.Chen2@amd.com>
Signed-off-by: Jack Zhang <Jack.Zhang7@hotmail.com>
Reviewed-by: Andrey Grodzovsky <andrey.grodzovsky@amd.com>
Reviewed by: Monk Liu <monk.liu@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/display: fix possible null-pointer dereference in dcn10_set_clock()

The variable dc->clk_mgr is checked in:
  if (dc->clk_mgr && dc->clk_mgr->funcs->get_clock)

This indicates dc->clk_mgr can be NULL.
However, it is dereferenced in:
    if (!dc->clk_mgr->funcs->get_clock)

To fix this null-pointer dereference, check dc->clk_mgr and the function
pointer dc->clk_mgr->funcs->get_clock earlier, and return if one of them
is NULL.

Reported-by: TOTE Robot <oslab@tsinghua.edu.cn>
Signed-off-by: Tuo Li <islituo@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

gpu: drm: amd: amdgpu: amdgpu_i2c: fix possible uninitialized-variable access in amdgpu_i2c_router_select_ddc_port()

The variable val is declared without initialization, and its address is
passed to amdgpu_i2c_get_byte(). In this function, the value of val is
accessed in:
  DRM_DEBUG("i2c 0x%02x 0x%02x read failed\n",
       addr, *val);

Also, when amdgpu_i2c_get_byte() returns, val may remain uninitialized,
but it is accessed in:
  val &= ~amdgpu_connector->router.ddc_mux_control_pin;

To fix this possible uninitialized-variable access, initialize val to 0 in
amdgpu_i2c_router_select_ddc_port().

Reported-by: TOTE Robot <oslab@tsinghua.edu.cn>
Signed-off-by: Tuo Li <islituo@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdkfd: CWSR with software scheduler

This patch adds support to program trap handler settings
when loading driver with software scheduler (sched_policy=2).

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

drm/amdkfd: AIP mGPUs best prefetch location for xnack on

For xnack on, if range ACCESS or ACCESS_IN_PLACE (AIP) by single GPU, or
range is ACCESS_IN_PLACE by mGPUs and all mGPUs connection on XGMI same
hive, the best prefetch location is prefetch_loc GPU. Otherwise, the best
prefetch location is always CPU because GPU does not have coherent
mapping VRAM of other GPUs even with large-BAR PCIe connection.

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

drm/amd/pm: graceful exit on restore fan mode failure (v2)

Attempt od settings restore and disable restore flag on restore fan mode
failure.

v2: Update fan mode to auto and fan speed to zero (Lijo)

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

drm/amd/pm: restore fan_mode AMD_FAN_CTRL_NONE on resume (v2)

Adds missing edge case to smu_restore_dpm_user_profile.

v2: Don't restore fan mode auto (Lijo)

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

amdgpu/pm: Replace amdgpu_pm usage of sprintf with sysfs_emit

 replacing printfs with sysfs_emit

=== Test ===
AMDGPU_PCI_ADDR=`lspci -nn | grep "VGA\|Display" | cut -d " " -f 1`
AMDGPU_HWMON=`ls -la /sys/class/hwmon | grep $AMDGPU_PCI_ADDR | awk '{print $9}'`
HWMON_DIR=/sys/class/hwmon/${AMDGPU_HWMON}
LOGFILE=pp_printf.test.log

lspci -nn | grep "VGA\|Display"  > $LOGFILE
FILES="pwm1_enable
pwm1_min
pwm1_max
pwm1
fan1_input
fan1_target
fan1_enable
power1_cap_min "

for f in $FILES
do
  echo === $f === >> $LOGFILE
  cat $HWMON_DIR/device/$f >> $LOGFILE
done
cat $LOGFILE

Signed-off-by: Darren Powell <darren.powell@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

amdgpu/pm: Replace hwmgr smu usage of sprintf with sysfs_emit

 replacing printfs with sysfs_emit
 minor smu7 change to remove compiler warning comparison of int and uint32_t
 minor smu8 change to remove compiler warning comparison of int and uint32_t

=== Test ===
AMDGPU_PCI_ADDR=`lspci -nn | grep "VGA\|Display" | cut -d " " -f 1`
AMDGPU_HWMON=`ls -la /sys/class/hwmon | grep $AMDGPU_PCI_ADDR | awk '{print $9}'`
HWMON_DIR=/sys/class/hwmon/${AMDGPU_HWMON}
LOGFILE=pp_printf.test.log

lspci -nn | grep "VGA\|Display"  > $LOGFILE
FILES="pp_dpm_sclk
pp_power_profile_mode "

for f in $FILES
do
  echo === $f === >> $LOGFILE
  cat $HWMON_DIR/device/$f >> $LOGFILE
done
cat $LOGFILE

Signed-off-by: Darren Powell <darren.powell@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

amdgpu/pm: Replace vega12,20 usage of sprintf with sysfs_emit

=== Test ===
AMDGPU_PCI_ADDR=`lspci -nn | grep "VGA\|Display" | cut -d " " -f 1`
AMDGPU_HWMON=`ls -la /sys/class/hwmon | grep $AMDGPU_PCI_ADDR | awk '{print $9}'`
HWMON_DIR=/sys/class/hwmon/${AMDGPU_HWMON}
LOGFILE=pp_printf.test.log

lspci -nn | grep "VGA\|Display"  > $LOGFILE
FILES="pp_dpm_sclk
pp_features
pp_power_profile_mode "

for f in $FILES
do
  echo === $f === >> $LOGFILE
  cat $HWMON_DIR/device/$f >> $LOGFILE
done
cat $LOGFILE

Signed-off-by: Darren Powell <darren.powell@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

amdgpu/pm: Replace vega10 usage of sprintf with sysfs_emit

 initial modification of files
  vega10_hwmgr.c

=== Test ===
AMDGPU_PCI_ADDR=`lspci -nn | grep "VGA\|Display" | cut -d " " -f 1`
AMDGPU_HWMON=`ls -la /sys/class/hwmon | grep $AMDGPU_PCI_ADDR | awk '{print $9}'`
HWMON_DIR=/sys/class/hwmon/${AMDGPU_HWMON}
LOGFILE=pp_printf.test.log

lspci -nn | grep "VGA\|Display"  > $LOGFILE
FILES="pp_dpm_sclk
pp_features
pp_power_profile_mode "

for f in $FILES
do
  echo === $f === >> $LOGFILE
  cat $HWMON_DIR/device/$f >> $LOGFILE
done
cat $LOGFILE

Signed-off-by: Darren Powell <darren.powell@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

amdgpu/pm: Replace smu12/13 usage of sprintf with sysfs_emit

 initial modification of files
  renoir_ppt.c
  aldebaran_ppt.c
  yellow_carp_ppt.c

=== Test ===
AMDGPU_PCI_ADDR=`lspci -nn | grep "VGA\|Display" | cut -d " " -f 1`
AMDGPU_HWMON=`ls -la /sys/class/hwmon | grep $AMDGPU_PCI_ADDR | awk '{print $9}'`
HWMON_DIR=/sys/class/hwmon/${AMDGPU_HWMON}
LOGFILE=pp_printf.test.log

lspci -nn | grep "VGA\|Display"  > $LOGFILE
FILES="pp_dpm_sclk
pp_power_profile_mode "

for f in $FILES
do
  echo === $f === >> $LOGFILE
  cat $HWMON_DIR/device/$f >> $LOGFILE
done
cat $LOGFILE

Signed-off-by: Darren Powell <darren.powell@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

amdgpu/pm: Replace smu11 usage of sprintf with sysfs_emit

 modification of smu11 files
  arcturus_ppt.c
  sienna_cichlid_ppt.c
  vangogh_ppt.c

=== Test ===
AMDGPU_PCI_ADDR=`lspci -nn | grep "VGA\|Display" | cut -d " " -f 1`
AMDGPU_HWMON=`ls -la /sys/class/hwmon | grep $AMDGPU_PCI_ADDR | awk '{print $9}'`
HWMON_DIR=/sys/class/hwmon/${AMDGPU_HWMON}
LOGFILE=pp_printf.test.log

lspci -nn | grep "VGA\|Display"  > $LOGFILE
FILES="pp_dpm_sclk
pp_power_profile_mode "

for f in $FILES
do
  echo === $f === >> $LOGFILE
  cat $HWMON_DIR/device/$f >> $LOGFILE
done
cat $LOGFILE

Signed-off-by: Darren Powell <darren.powell@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

amdgpu/pm: Replace navi10 usage of sprintf with sysfs_emit

 initial modification of files
  smu_cmn.c
  navi10_ppt.c

=== Test ===
AMDGPU_PCI_ADDR=`lspci -nn | grep "VGA\|Display" | cut -d " " -f 1`
AMDGPU_HWMON=`ls -la /sys/class/hwmon | grep $AMDGPU_PCI_ADDR | awk '{print $9}'`
HWMON_DIR=/sys/class/hwmon/${AMDGPU_HWMON}
LOGFILE=pp_printf.test.log

lspci -nn | grep "VGA\|Display"  > $LOGFILE
FILES="pp_dpm_sclk
pp_sclk_od
pp_mclk_od
pp_dpm_pcie
pp_od_clk_voltage
pp_power_profile_mode "

for f in $FILES
do
  echo === $f === >> $LOGFILE
  cat $HWMON_DIR/device/$f >> $LOGFILE
done
cat $LOGFILE

Signed-off-by: Darren Powell <darren.powell@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: handle VCN instances when harvesting (v2)

There may be multiple instances and only one is harvested.

v2: fix typo in commit message

Fixes: 83a0b8639185 ("drm/amdgpu: add judgement when add ip blocks (v2)")
Bug: https://gitlab.freedesktop.org/drm/amd/-/issues/1673
Reviewed-by: Guchun Chen <guchun.chen@amd.com>
Reviewed-by: James Zhu <James.Zhu@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: Removed unnecessary if statement

There was an "if" statement that did nothing so it was removed.

Signed-off-by: Sergio Miguéns Iglesias <sergio@lony.xyz>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: fix kernel-doc warnings on non-kernel-doc comments

Don't use "begin kernel-doc notation" (/**) for comments that are
not kernel-doc. This eliminates warnings reported by the 0day bot.

drivers/gpu/drm/amd/amdgpu/gfx_v9_4_2.c:89: warning: This comment starts with '/**', but isn't a kernel-doc comment. Refer Documentation/doc-guide/kernel-doc.rst
    * This shader is used to clear VGPRS and LDS, and also write the input
drivers/gpu/drm/amd/amdgpu/gfx_v9_4_2.c:209: warning: This comment starts with '/**', but isn't a kernel-doc comment. Refer Documentation/doc-guide/kernel-doc.rst
    * The below shaders are used to clear SGPRS, and also write the input
drivers/gpu/drm/amd/amdgpu/gfx_v9_4_2.c:301: warning: This comment starts with '/**', but isn't a kernel-doc comment. Refer Documentation/doc-guide/kernel-doc.rst
    * This shader is used to clear the uninitiated sgprs after the above

Fixes: 0e0036c7d13b ("drm/amdgpu: fix no full coverage issue for gprs initialization")
Signed-off-by: Randy Dunlap <rdunlap@infradead.org>
Reported-by: kernel test robot <lkp@intel.com>
Cc: Alex Deucher <alexander.deucher@amd.com>
Cc: Christian König <christian.koenig@amd.com>
Cc: "Pan, Xinhui" <Xinhui.Pan@amd.com>
Cc: Dennis Li <Dennis.Li@amd.com>
Cc: amd-gfx@lists.freedesktop.org
Cc: dri-devel@lists.freedesktop.org
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: use do-while-0 for DC_TRACE_LEVEL_MESSAGE()

Building with W=1 complains about an empty 'else' statement, so use the
usual do-nothing-while-0 loop to quieten this warning.

../drivers/gpu/drm/amd/amdgpu/../display/dc/dce/dmub_psr.c:113:53: warning: suggest braces around empty body in an 'else' statement [-Wempty-body]
  113 |                                 *state, retry_count);

Fixes: b30eda8d416c ("drm/amd/display: Add ETW log to dmub_psr_get_state")
Signed-off-by: Randy Dunlap <rdunlap@infradead.org>
Cc: Wyatt Wood <wyatt.wood@amd.com>
Cc: Alex Deucher <alexander.deucher@amd.com>
Cc: Christian König <christian.koenig@amd.com>
Cc: "Pan, Xinhui" <Xinhui.Pan@amd.com>
Cc: Harry Wentland <harry.wentland@amd.com>
Cc: Leo Li <sunpeng.li@amd.com>
Cc: amd-gfx@lists.freedesktop.org
Cc: dri-devel@lists.freedesktop.org
Cc: David Airlie <airlied@linux.ie>
Cc: Daniel Vetter <daniel@ffwll.ch>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: use GFP_ATOMIC in amdgpu_dm_irq_schedule_work

Replace GFP_KERNEL with GFP_ATOMIC as amdgpu_dm_irq_schedule_work
can't sleep.

BUG: sleeping function called from invalid context at include/linux/sched/mm.h:196
in_atomic(): 1, irqs_disabled(): 1, non_block: 0, pid: 253, name: kworker/6:1H
CPU: 6 PID: 253 Comm: kworker/6:1H Tainted: G        W  OE     5.11.0-promotion_2021_06_07-18_36_28_prelim_revert_retrain #8
Hardware name: System manufacturer System Product Name/PRIME X570-PRO, BIOS 3405 02/01/2021
Workqueue: events_highpri dm_irq_work_func [amdgpu]
Call Trace:
 <IRQ>
 dump_stack+0x5e/0x74
 ___might_sleep.cold+0x87/0x98
 __might_sleep+0x4b/0x80
 kmem_cache_alloc_trace+0x390/0x4f0
 amdgpu_dm_irq_handler+0x171/0x230 [amdgpu]
 amdgpu_irq_dispatch+0xc0/0x1e0 [amdgpu]
 amdgpu_ih_process+0x81/0x100 [amdgpu]
 amdgpu_irq_handler+0x26/0xa0 [amdgpu]
 __handle_irq_event_percpu+0x49/0x190
 ? __hrtimer_get_next_event+0x4d/0x80
 handle_irq_event_percpu+0x33/0x80
 handle_irq_event+0x33/0x60
 handle_edge_irq+0x82/0x190
 asm_call_irq_on_stack+0x12/0x20
 </IRQ>
 common_interrupt+0xbb/0x140
 asm_common_interrupt+0x1e/0x40
RIP: 0010:amdgpu_device_rreg.part.0+0x44/0xf0 [amdgpu]
Code: 53 48 89 fb 4c 3b af c8 08 00 00 73 6d 83 e2 02 75 0d f6 87 40 62 01 00 10 0f 85 83 00 00 00 4c 03 ab d0 08 00 00 45 8b 6d 00 <8b> 05 3e b6 52 00 85 c0 7e 62 48 8b 43 08 0f b7 70 3e 65 8b 05 e3
RSP: 0018:ffffae7740fff9e8 EFLAGS: 00000286
RAX: ffffffffc05ee610 RBX: ffff8aaf8f620000 RCX: 0000000000000006
RDX: 0000000000000000 RSI: 0000000000005430 RDI: ffff8aaf8f620000
RBP: ffffae7740fffa08 R08: 0000000000000001 R09: 000000000000000a
R10: 0000000000000001 R11: 0000000000000001 R12: 0000000000005430
R13: 0000000071000000 R14: 0000000000000001 R15: 0000000000005430
 ? amdgpu_cgs_write_register+0x20/0x20 [amdgpu]
 amdgpu_device_rreg+0x17/0x20 [amdgpu]
 amdgpu_cgs_read_register+0x14/0x20 [amdgpu]
 dm_read_reg_func+0x38/0xb0 [amdgpu]
 generic_reg_wait+0x80/0x160 [amdgpu]
 dce_aux_transfer_raw+0x324/0x7c0 [amdgpu]
 dc_link_aux_transfer_raw+0x43/0x50 [amdgpu]
 dm_dp_aux_transfer+0x87/0x110 [amdgpu]
 drm_dp_dpcd_access+0x72/0x110 [drm_kms_helper]
 drm_dp_dpcd_read+0xb7/0xf0 [drm_kms_helper]
 drm_dp_get_one_sb_msg+0x349/0x480 [drm_kms_helper]
 drm_dp_mst_hpd_irq+0xc5/0xe40 [drm_kms_helper]
 ? drm_dp_mst_hpd_irq+0xc5/0xe40 [drm_kms_helper]
 dm_handle_hpd_rx_irq+0x184/0x1a0 [amdgpu]
 ? dm_handle_hpd_rx_irq+0x184/0x1a0 [amdgpu]
 handle_hpd_rx_irq+0x195/0x240 [amdgpu]
 ? __switch_to_asm+0x42/0x70
 ? __switch_to+0x131/0x450
 dm_irq_work_func+0x19/0x20 [amdgpu]
 process_one_work+0x209/0x400
 worker_thread+0x4d/0x3e0
 ? cancel_delayed_work+0xa0/0xa0
 kthread+0x124/0x160
 ? kthread_park+0x90/0x90
 ret_from_fork+0x22/0x30

Reviewed-by: Aurabindo Jayamohanan Pillai <Aurabindo.Pillai@amd.com>
Acked-by: Anson Jacob <Anson.Jacob@amd.com>
Signed-off-by: Anson Jacob <Anson.Jacob@amd.com>
Cc: stable@vger.kernel.org
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Increase timeout threshold for DMCUB reset

[Why]
If we're backdoor loading the DMCUB performs more work than just
the PHY reset so we can end up resetting before the cleanup has fully
finished.

[How]
Increase timeout, add udelay between spins to guarantee a minimum.

Reviewed-by: Roy Chan <roy.chan@amd.com>
Reviewed-by: Eric Yang <Eric.Yang2@amd.com>
Acked-by: Anson Jacob <Anson.Jacob@amd.com>
Signed-off-by: Nicholas Kazlauskas <nicholas.kazlauskas@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Clear GPINT after DMCUB has reset

[Why]
Otherwise we can end up processing whatever was left in the register
if the DMCUB was previously reset.

If DMCUB gets force reset too early from another client then we might
not have even acked the disable yet - causing DMCUB instantly shutdown
if the command was 10020000.

[How]
Move the GPINT clear outside of the reset loop and do it unconditionally
after the DMCUB has been properly reset.

Reviewed-by: Roy Chan <roy.chan@amd.com>
Reviewed-by: Eric Yang <Eric.Yang2@amd.com>
Acked-by: Anson Jacob <Anson.Jacob@amd.com>
Signed-off-by: Nicholas Kazlauskas <nicholas.kazlauskas@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: 3.2.148

This version brings along following fixes:
- Fix memory allocation in dm IRQ context to use GFP_ATOMIC
- Increase timeout threshold for DMCUB reset
- Clear GPINT after DMCUB has reset
- Add AUX I2C tracing
- Fix code commenting style
- Some refactoring
- Remove invalid assert for ODM + MPC case

Reviewed-by: Wyatt Wood <Wyatt.Wood@amd.com>
Acked-by: Anson Jacob <Anson.Jacob@amd.com>
Signed-off-by: Anthony Koo <Anthony.Koo@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: [FW Promotion] Release 0.0.78

Acked-by: Anson Jacob <Anson.Jacob@amd.com>
Signed-off-by: Anthony Koo <Anthony.Koo@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: add authentication_complete in hdcp output

[why]
DM needs to be notified when hdcp module has completed
authentication attempt.

Reviewed-by: Bhawanpreet Lakha <bhawanpreet.lakha@amd.com>
Reviewed-by: George Shen <George.Shen@amd.com>
Acked-by: Anson Jacob <Anson.Jacob@amd.com>
Signed-off-by: Wenjing Liu <wenjing.liu@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Add AUX I2C tracing.

[Why]
Developers can find it useful if the driver can produce
AUX traces without special equipment.

[How]
Add AUX tracing.

Reviewed-by: Zhan Liu <zhan.liu@amd.com>
Acked-by: Anson Jacob <Anson.Jacob@amd.com>
Signed-off-by: Ashley Thomas <Ashley.Thomas2@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Correct comment style

Acked-by: Anson Jacob <Anson.Jacob@amd.com>
Signed-off-by: Roy Chan <roy.chan@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: fix incorrect CM/TF programming sequence in dwb

[How]
the programming sequeune was for old asic.
the correct programming sequeunce should be similar to the one
used in mpc. the fix is copied from the mpc programming sequeunce.

Reviewed-by: Anthony Koo <Anthony.Koo@amd.com>
Acked-by: Anson Jacob <Anson.Jacob@amd.com>
Signed-off-by: Roy Chan <roy.chan@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: refactor the cursor programing codes

Acked-by: Anson Jacob <Anson.Jacob@amd.com>
Signed-off-by: Roy Chan <roy.chan@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: refactor the codes to centralize the stream/pipe checking logic

Acked-by: Anson Jacob <Anson.Jacob@amd.com>
Signed-off-by: Roy Chan <roy.chan@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: fix missing writeback disablement if plane is removed

[Why]
If the plane has been removed, the writeback disablement logic
doesn't run

[How]
fix the logic order

Acked-by: Anson Jacob <Anson.Jacob@amd.com>
Signed-off-by: Roy Chan <roy.chan@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Remove invalid assert for ODM + MPC case

Reviewed-by: Dmytro Laktyushkin <Dmytro.Laktyushkin@amd.com>
Acked-by: Anson Jacob <Anson.Jacob@amd.com>
Signed-off-by: Eric Bernstein <eric.bernstein@amd.com>
Cc: stable@vger.kernel.org
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/amdgpu: skip locking delayed work if not initialized.

When init failed in early init stage, amdgpu_object has
not been initialized, so hasn't the ttm delayed queue functions.

Signed-off-by: YuBiao Wang <YuBiao.Wang@amd.com>
Reviewed-by: Emily.Deng <Emily.Deng@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: Extend full access wait time in guest

- Extend wait time and add retry, currently 6s * 2times
- Change timing algorithm

Signed-off-by: Victor Zhao <Victor.Zhao@amd.com>
Signed-off-by: Peng Ju Zhou <PengJu.Zhou@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdkfd: Allow querying SVM attributes that are clear

Currently the SVM get_attr call allows querying, which flags are set
in the entire address range. Add the opposite query, which flags are
clear in the entire address range. Both queries can be combined in a
single get_attr call, which allows answering questions such as, "is
this address range coherent, non-coherent, or a mix of both"?

Proposed userspace for UAPI:
https://github.com/RadeonOpenCompute/ROCR-Runtime/tree/memory_model_queries

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

drm/amd/display: Remove redundant initialization of variable eng_id

The variable eng_id is being initialized with a value that is never
read, it is being re-assigned on the next statment. The assignment
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>

drm/amdgpu: check for allocation failure in amdgpu_vkms_sw_init()

Check whether the kcalloc() fails and return -ENOMEM if it does.

Fixes: 84ec374bd58036 ("drm/amdgpu: create amdgpu_vkms (v4)")
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/pm: bug fix for the runtime pm BACO

In some systems only MACO is supported. This is to fix the problem
that runtime pm is enabled but BACO is not supported. MACO will be
handled seperately.

Signed-off-by: Kenneth Feng <kenneth.feng@amd.com>
Reviewed-by: Jack Gui <Jack.Gui@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: Add preferred mode in modeset when freesync video mode's enabled.

[Why]
With kernel module parameter "freesync_video" is enabled, if the mode
is changed to preferred mode(the mode with highest rate), then Freesync
fails because the preferred mode is treated as one of freesync video
mode, and then be configurated as freesync video mode(fixed refresh
rate).

[How]
Skip freesync fixed rate configurating when modeset to preferred mode.

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

drm/amd/pm: Fix a memory leak in an error handling path in 'vangogh_tables_init()'

'watermarks_table' must be freed instead 'clocks_table', because
'clocks_table' is known to be NULL at this point and 'watermarks_table' is
never freed if the last kzalloc fails.

Fixes: c98ee89736b8 ("drm/amd/pm: add the fine grain tuning function for vangogh")
Signed-off-by: Christophe JAILLET <christophe.jaillet@wanadoo.fr>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

DRM: gpu: radeon: Fixed coding style issues

Fixed braces, an unnecessary if statement and added a missing space.

Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Sergio Miguéns Iglesias <sergio@lony.xyz>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/radeon: Update pitch for page flip

When primary bo is updated, crtc's pitch may
have not been updated, this will lead to show
disorder content when user changes display mode,
we update crtc's pitch in page flip to avoid
this bug.
This refers to amdgpu's pageflip.

v1->v2:
Update all of the pitch in all of the page_flip functions
in radeon rather than just the evergreen one.

v2->v3:
Update pitch set method for r100 according to
radeon_legacy_crtc.c

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
Cc: linux-kernel@vger.kernel.org
Signed-off-by: Zhenneng Li <lizhenneng@kylinos.cn>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdkfd: Expose GFXIP engine version to sysfs

Add u32 gfx_target_version field to kfd_node_properties and
kfd_device_info. Populate <asic>_device_info structs accordingly and
expose to sysfs.

This allows eliminating device-ID-based lookup tables in user mode for
future ASICs.

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