drm/nouveau/gr/gf100-: virtualise tpc_per_gpc

GM20B now also shares the same code, as NVGPU shows it doesn't need
special treatment.

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>

drm/nouveau/gr/gf100-: virtualise sm_id/tpc_nr

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>

drm/nouveau/gr/gf100-: virtualise patch_ltc, noting missing init

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>

drm/nouveau/gr/gf100-: support firmware-provided sw_ctx everywhere

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>

drm/nouveau/gr/gf100-: delete duplicated init code

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>

drm/nouveau/gr/gf100-: virtualise init_400054

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>

drm/nouveau/gr/gf100-: apply be exception fixes from traces

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>

drm/nouveau/gr/gf100-: virtualise init_shader_exceptions

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>

drm/nouveau/gr/gf100-: virtualise init_504430

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>

drm/nouveau/gr/gf100-: virtualise init_tex_hww_esr

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>

drm/nouveau/gr/gf100-: virtualise init_ppc_exceptions

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>

drm/nouveau/gr/gf100-: virtualise init_419c9c + apply fixes from traces

Deliberately removed from non-GP100, as RM doesn't touch it.

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>

drm/nouveau/gr/gf100-: virtualise init_419eb4 + apply fixes from traces

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>

drm/nouveau/gr/gf100-: virtualise init_419cc0 + apply fixes from traces

Pulled some init out of main per-GPC/TPC loops to match RM.

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>

drm/nouveau/gr/gf100-: virtualise init_sked_hww_esr

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>

drm/nouveau/gr/gf100-: virtualise init_40601c

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>

drm/nouveau/gr/gf100-: virtualise init_ds_hww_esr_2

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>

drm/nouveau/gr/gf100-: virtualise init_fecs_exceptions + apply fixes from traces

The value for GF100 has changed here, but it matches RM now.

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>

drm/nouveau/gr/gf100: write 0x400124 during init

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>

drm/nouveau/gr/gf100-: virtualise init_swdx_pes_mask

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>

drm/nouveau/gr/gf100-: implement another chunk of bios-provided init

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>

drm/nouveau/gr/gf100-: virtualise init_rop_active_fbps

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>

drm/nouveau/gr/gf100-: virtualise init_num_active_ltcs

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>

drm/nouveau/gr/gf100-: virtualise init_zcull

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>

drm/nouveau/gr/gf100-: virtualise init_vsc_stream_master

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>

drm/nouveau/gr/gf100-: virtualise init_bios

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>

drm/nouveau/gr/gf100-: support clkgate_pack everywhere

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>

drm/nouveau/gr/gf100-: virtualise r405a14

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>

drm/nouveau/gr/gf100-: support firmware-provided sw_nonctx everywhere

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>

drm/nouveau/gr/gf100-: virtualise init_gpc_mmu + apply fixes from traces

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>

drm/nouveau/fifo/gp100-: force individual channels into a channel group

RM does this for some reason, and is enforced in HW on Volta.

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>

drm/nouveau/fifo/gm107-: write instance address in channel runlist entry

RM does this for some reason.

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>

drm/nouveau/fifo/gk208-: write pbdma timeout regs during initialisation

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>

drm/nouveau/fifo/gk110-: support writing channel group runlist entries

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>

drm/nouveau/fifo/gk104-: poll for runlist update completion

Newer HW doesn't appear to send this event, which will cause long delays
in runlist updates if they don't complete immediately.

RM doesn't use these events anywhere, and an NVGPU commit message notes
that polling is the preferred method even on HW that supports the event.

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>

drm/nouveau/fifo/gk104-: add interfaces to support different runlist layouts

This will be required to support features on newer hardware.

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>

drm/nouveau/fifo/gk104-: simplify definition of channel classes

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>

drm/nouveau/fifo/gk104-: require explicit runlist selection for channel allocation

We didn't used to be aware that runlist/engine IDs weren't the same thing,
or that there was such variability in configuration between GPUs.

By exposing this information to a client, and giving it explicit control
of which runlist it's allocating a channel on, we're able to make better
choices.

The immediate effect of this is that on GPUs where CE0 is the "GRCE", we
will now be allocating a copy engine running asynchronously to GR for BO
migrations - as intended.

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>

drm/nouveau/fifo/gk104-: support querying engines available on each runlist

Will be used to improve channel runlist selection.

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>

drm/nouveau/fifo/gk104-: allow fault recovery code to be called by other subdevs

This will be required to support Volta.

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>

drm/nouveau/fifo/gk104-: accept engine contexts for CE3 and up

These can exist on GP100 and newer.

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>

drm/nouveau/fifo: support channel count query

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>

drm/nouveau/device: support querying available engines of a specific type

Will be used for fifo runlist selection.

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>

drm/nouveau/device: implement a generic method to query device-specific properties

We have a need to fetch data from GPU-specific sub-devices that is not
tied to any particular engine object.

This commit provides the framework to support such queries.

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>

drm/nouveau/disp/nv50-: pass nvkm_memory objects for channel push buffers

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>

drm/nouveau/disp/nv50-: add channel interfaces to control error interrupts

This will be required to support Volta, but also allows us to remove code
that's duplicated for each channel type already.

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>

drm/nouveau/disp/nv50-: add channel interfaces to determine the user area

This will be required to support Volta.

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>

drm/nouveau/disp/nv50-: merge handling of pio and dma channels

Unnecessarily complicated, and a barrier to cleanly supporting Volta.

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>

drm/nouveau/disp/nv50-: simplify definiton of core channels

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>

drm/nouveau/disp/nv50-: simplify definition of cursor channels

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>

drm/nouveau/disp/nv50-: simplify definition of base channels

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>

drm/nouveau/disp/nv50-: simplify definition of overlay immediate channels

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>

drm/nouveau/disp/nv50-: simplify definition of overlay channels

Introduces a new method of defining channels available from the display,
common to all channel types, allowing for more flexibility in available
channel types/counts, and reducing the amount of boiler-plate required.

This will be required to support Volta.

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>

drm/nouveau/disp/nv50-: replace user object with engine pointer in channels

More simplification.

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>

drm/nouveau/disp/nv50-: initialise from the engine, rather than the user object

Engines are initialised on an as-needed basis, so this results in the
same behaviour, whilst allowing us to simplify things a bit.

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>

drm/nouveau/disp/nv50-: fetch mask of available piors during oneinit

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>

drm/nouveau/disp/nv50-: fetch mask of available sors during oneinit

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>

drm/nouveau/disp/nv50-: fetch mask of available dacs during oneinit

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>

drm/nouveau/disp/nv50-: fetch mask of available heads during oneinit

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>

drm/nouveau/disp/nv50-: delay subunit construction until oneinit

We should be reading registers to determine which subunits are really
present on a given board, and this needs to be done after DEVINIT.

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>

drm/nouveau/fb/gm200-: fix overwriting of big page setting

Likely a rebase bug.  Should have no impact in default configuration due
to using per-instance setting by default.

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>

drm/nouveau/fb/gf100-: bump size of mmu debug buffers to match big page size

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>

drm/nouveau/fault/gp100: implement replayable fault buffer initialisation

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>

drm/nouveau/fault: add infrastructure to support fault buffers

GPU-specific support will be added separately.

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>

drm/nouveau/mc/gp100-: route fault buffer interrupts to FAULT

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>

drm/nouveau/core: define FAULT subdev

This will be responsible for the handling of MMU fault buffers on GPUs
that support them.

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>

drm/nouveau/secboot: remove VLA usage

In preparation to enabling -Wvla, remove VLA. In this particular
case directly use macro NVKM_MSGQUEUE_CMDLINE_SIZE instead of local
variable cmdline_size. Also, remove cmdline_size as it is not
actually useful anymore.

The use of stack Variable Length Arrays needs to be avoided, as they
can be a vector for stack exhaustion, which can be both a runtime bug
or a security flaw. Also, in general, as code evolves it is easy to
lose track of how big a VLA can get. Thus, we can end up having runtime
failures that are hard to debug.

Also, fixed as part of the directive to remove all VLAs from
the kernel: https://lkml.org/lkml/2018/3/7/621

Signed-off-by: Gustavo A. R. Silva <gustavo@embeddedor.com>
Reviewed-by: Thierry Reding <treding@nvidia.com>
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>

drm/nouveau: nouveau: use larger buffer in nvif_vmm_map

gcc points out a buffer that is clearly too small to be used
in a meaningful way, as the 'sizeof(*args) + argc > sizeof(stack)'
will always fail:

In function 'memcpy',
    inlined from 'nvif_vmm_map' at drivers/gpu/drm/nouveau/nvif/vmm.c:55:2:
include/linux/string.h:353:9: error: '__builtin_memcpy' offset 40 is out of the bounds [0, 16] of object 'stack' with type 'u8[16]' {aka 'unsigned char[16]'} [-Werror=array-bounds]
  return __builtin_memcpy(p, q, size);
         ^~~~~~~~~~~~~~~~~~~~~~~~~~~~
drivers/gpu/drm/nouveau/nvif/vmm.c: In function 'nvif_vmm_map':
drivers/gpu/drm/nouveau/nvif/vmm.c:40:5: note: 'stack' declared here

This makes the buffer large enough so it should serve the purpose
that the author presumably had in mind. Alternatively we could
just get rid of it completely and simplify the code at the cost
of always doing the kmalloc (as we do in the current version).

Fixes: 920d2b5ef215 ("drm/nouveau/mmu: define user interfaces to mmu vmm opertaions")
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>

Merge drm-fixes-for-v4.17-rc6-urgent into drm-next

Need to backmerge some nouveau fixes to reduce
the nouveau -next conflicts a lot.

Signed-off-by: Dave Airlie <airlied@redhat.com>

drm: rcar-du: Fix build failure

Commit 75a07f399cd4 ("drm: rcar-du: Zero-out sg_tables when duplicating
plane state") introduced a reference to the alpha field of struct
rcar_du_vsp_plane_state that got removed in commit 301a9b8d5456
("drm/rcar-du: Convert to the new generic alpha property"). The issue
stems from the merge of the two commits through separate branches and
breaks compilation of the driver. Fix it.

Fixes: 75a07f399cd4 ("drm: rcar-du: Zero-out sg_tables when duplicating plane state")
Signed-off-by: Laurent Pinchart <laurent.pinchart+renesas@ideasonboard.com>
Reviewed-by: Kieran Bingham <kieran.bingham+renesas@ideasonboard.com>
Tested-by: Kieran Bingham <kieran.bingham+renesas@ideasonboard.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20180515174752.28954-1-laurent.pinchart+renesas@ideasonboard.com

drm: rcar-du: Fix rcar_du_of_init() stub

The rcar_du_of_init() function is supposed to be defined as a stub when
CONFIG_DRM_RCAR_LVDS is disabled as the rcar_du_of.c file isn't compiled
in that case. However, a bug in the configuration option check makes it
a stub when CONFIG_DRM_RCAR_LVDS=m as well, which prevents legacy DTs
from being fixed at boot time. Fix the configuration option check by
using IS_ENABLED.

Fixes: 81c0e3dd8292 ("drm: rcar-du: Fix legacy DT to create LVDS encoder nodes")
Signed-off-by: Laurent Pinchart <laurent.pinchart+renesas@ideasonboard.com>
Reviewed-by: Kieran Bingham <kieran.bingham+renesas@ideasonboard.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20180515155736.3379-1-laurent.pinchart+renesas@ideasonboard.com

drm/amdgpu: include pagemap.h for release_pages()

Fixes: 5ae0283e831a ("drm/amdgpu: Add userptr support for KFD"
Cc: Felix Kuehling <Felix.Kuehling@amd.com>
Cc: Oded Gabbay <oded.gabbay@gmail.com>
Signed-off-by: Stephen Rothwell <sfr@canb.auug.org.au>
Signed-off-by: Dave Airlie <airlied@redhat.com>

Merge branch 'drm-next-4.18' of git://people.freedesktop.org/~agd5f/linux into drm-next

Main changes for 4.18.  I'd like to do a separate pull for vega20 later
this week or next.  Highlights:
- Reserve pre-OS scanout buffer during init for seemless transition from
  console to driver
- VEGAM support
- Improved GPU scheduler documentation
- Initial gfxoff support for raven
- SR-IOV fixes
- Default to non-AGP on PowerPC for radeon
- Fine grained clock voltage control for vega10
- Power profiles for vega10
- Further clean up of powerplay/driver interface
- Underlay fixes
- Display link bw updates
- Gamma fixes
- Scatter/Gather display support on CZ/ST
- Misc bug fixes and clean ups

[airlied: fixup v3d vs scheduler API change]

Link: https://patchwork.freedesktop.org/patch/msgid/20180515185450.1113-1-alexander.deucher@amd.com
Signed-off-by: Dave Airlie <airlied@redhat.com>

Merge tag 'drm-intel-next-2018-05-14' of git://anongit.freedesktop.org/drm/drm-intel into drm-next

Last drm/i915 changes for v4.18:

- NV12 enabling (Chandra, Maarten)
- ICL workarounds (Oscar)
- ICL basic DPLL enabling (Paulo)
- GVT updates
- DP link config refactoring (Jani)
- Module parameter to override DMC firmware (Jani)
- PSR updates (José, DK, Daniel, Ville)
- ICL DP vswing programming (Manasi)
- ICL DBuf slice updates (Mahesh)
- Selftest fixes and updates (Chris, Matthew, Oscar)
- Execlist fixes and updates (Chris)
- Stolen memory first 4k fix (Hans de Goede)
- wait_for fixes (Mika)
- Tons of GEM improvements (Chris)
- Plenty of other fixes and improvements (Everyone)
- Crappy changelog (Me)

Signed-off-by: Dave Airlie <airlied@redhat.com>
# gpg: Signature made Mon 14 May 2018 11:04:24 PM AEST
# gpg:                using RSA key D398079D26ABEE6F
# gpg: Good signature from "Jani Nikula <jani.nikula@intel.com>"
# gpg: WARNING: This key is not certified with a trusted signature!
# gpg:          There is no indication that the signature belongs to the owner.
# Primary key fingerprint: 1565 A65B 77B0 632E 1124  E59C D398 079D 26AB EE6F

# Conflicts:
# drivers/gpu/drm/i915/intel_lrc.c
# drivers/gpu/drm/i915/intel_sprite.c
Link: https://patchwork.freedesktop.org/patch/msgid/87k1s51bvw.fsf@intel.com

drm/scheduler: remove unused parameter

this patch also effect the amdgpu and etnaviv drivers which
use the function drm_sched_entity_init

Signed-off-by: Nayan Deshmukh <nayan26deshmukh@gmail.com>
Suggested-by: Christian König <christian.koenig@amd.com>
Acked-by: Lucas Stach <l.stach@pengutronix.de>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

Add SPDX idenitifier and clarify license

This is dual licensed under GPL-2.0 or MIT.

Signed-off-by: Dirk Hohndel (VMware) <dirk@hohndel.org>
Signed-off-by: Thomas Hellstrom <thellstrom@vmware.com>
Acked-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: print the BO flags in the gem debugfs entry

Quite useful to know.

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

drm/amdgpu: drop printing the BO offset in the gem debugfs (v2)

It is meaningless anyway.

v2: remove unused variable (Alex)

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

drm/amdgpu: Add support to change mtype for 2nd part of gart BOs on GFX9

This change prepares for a workaround in amdkfd for a GFX9 HW bug. It
requires the control stack memory of compute queues, which is allocated
from the second page of MQD gart BOs, to have mtype NC, rather than
the default UC.

Signed-off-by: Yong Zhao <yong.zhao@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/powerplay: Add notify PWE function to SMU10

Functionality to message smc to enable pwe after gpu suspense.
It is used in case when display resumes from S3 and wants to start
audio driver by enabling pwe.

Signed-off-by: Mikita Lipski <mikita.lipski@amd.com>
Acked-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: remove need of modeset flag for overlay planes (V2)

This patch is in continuation to the
"843e3c7 drm/amd/display: defer modeset check in dm_update_planes_state"
where we started to eliminate the dependency on
DRM_MODE_ATOMIC_ALLOW_MODESET to be set by the user space,
which as such is not mandatory.

After deferring, this patch eliminates the dependency on the flag
for overlay planes.

This has to be done in stages as its a pretty complex and requires thorough
testing before we free primary planes as well from dependency on modeset
flag.

V2: Simplified the plane type check.

Signed-off-by: Shirish S <shirish.s@amd.com>
Reviewed-by: Harry Wentland <harry.wentland@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: fix null pointer for bo unmap trace function

fix crash in trace.

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

drm/amdgpu/gmc9: remove unused register defs

These got moved to the new df module so no longer
used in this file.

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

drm/amdgpu: add HDP flush dummy for UVD 6/7

The UVD firmware doesn't seem to like the HDP flush here.

This worked for years without HDP flush, so just skip it.

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

drm/amdgpu: set ttm bo priority before initialization

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

drm/ttm: remove priority hard code when initializing ttm bo

Then priority could be set before initialization.
By default, it requires to kzalloc ttm bo. In fact, we always do so.

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

drm/amd/pp: Fix performance drop on Fiji

The performance drop if the default TDP more than 256 Watt

Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Reviewed-by: Junwei Zhang <Jerry.Zhang@amd.com>
Signed-off-by: Rex Zhu <Rex.Zhu@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/pp: Implement force_clock_level for RV

under manual dpm mode, user can set gfx/mem clock
through sysfs pp_dpm_sclk/mclk on Rv.

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

drm/amd/powerplay: add PME smu message for raven

Used for working around an audio bug on some platforms.

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

drm/amd/display: clean up assignment of amdgpu_crtc

The declaration of pointer amdgpu_crtc has a redundant assignment to
amdgpu_crtc. Clean this up by removing it.

Detected by CoverityScan, CID#1460299 ("Evaluation order violation")

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

drm/amd/powerplay: fix spelling mistake: "contruct" -> "construct"

Trivial fix to spelling mistake in PP_ASSERT_WITH_CODE message text

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

drm/amd/pp: Refine the output of pp_power_profile_mode on VI

In order to keep consist with Vega,
the output format of the pp_power_profile_mode would be
<integer><mode name string>< “*” for current profile>:"detail settings"
and remove the "CURRENT" mode line.

for example:
NUM        MODE_NAME     SCLK_UP_HYST   SCLK_DOWN_HYST SCLK_ACTIVE_LEVEL     MCLK_UP_HYST   MCLK_DOWN_HYST MCLK_ACTIVE_LEVEL
  0   3D_FULL_SCREEN:        0              100               30                0              100               10
  1     POWER_SAVING:       10                0               30                -                -                -
  2            VIDEO:        -                -                -               10               16               31
  3               VR:        0               11               50                0              100               10
  4          COMPUTE:        0                5               30                -                -                -
  5         CUSTOM *:        0                5               30                0              100               10
NUM        MODE_NAME     SCLK_UP_HYST   SCLK_DOWN_HYST SCLK_ACTIVE_LEVEL     MCLK_UP_HYST   MCLK_DOWN_HYST MCLK_ACTIVE_LEVEL
  0   3D_FULL_SCREEN:        0              100               30                0              100               10
  1   POWER_SAVING *:       10                0               30                0              100               10
  2            VIDEO:        -                -                -               10               16               31
  3               VR:        0               11               50                0              100               10
  4          COMPUTE:        0                5               30                -                -                -
  5           CUSTOM:        -                -                -                -                -                -

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

drm/amdgpu: invalidate parent bo when shadow bo was invalidated

Shadow BO is located on GTT and its parent (PT and PD) BO could located on VRAM.
In some case, the BO on GTT could be evicted but the parent did not. This may
cause the shadow BO not be put in the evict list and could not be invalidate
correctly.
v2: suggested by Christian

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

drm/amdgpu: abstract bo_base init function

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

drm/amd/amdgpu: Add some documentation to the debugfs entries

Signed-off-by: Tom St Denis <tom.stdenis@amd.com>
Reviewed-by: Huang Rui <ray.huang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/amdgpu: vcn10 Add callback for emit_reg_write_reg_wait

The callback .emit_reg_write_reg_wait was missing for vcn decode
which resulted in a kernel oops.

Signed-off-by: Tom St Denis <tom.stdenis@amd.com>
Reviewed-by: Andrey Grodzovsky <andrey.grodzovsky@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: Fix display corruption on CI with dpm enabled

with dpm enabled, need to get active crtcs in dc/no-dc mode.

caused by
'commit ebb649667a31 ("drm/amdgpu: Set pm_display_cfg in non-dc mode")'

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

drm/amd/pp: Skip fan attributes if fan not present

With powerplay enabled, also need to skip fan attributes
if no fan present.

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

drm/amdgpu: Switch to interruptable wait to recover from ring hang.

v2:
Use dma_fence_wait instead of dma_fence_wait_timeout(...,MAX_SCHEDULE_TIMEOUT)
Avoid printing error message for ERESTARTSYS

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

drm/ttm: Use GFP_TRANSHUGE_LIGHT for allocating huge pages

GFP_TRANSHUGE tries very hard to allocate huge pages, which can result
in long delays with high memory pressure. I have observed firefox
freezing for up to around a minute due to this while restic was taking
a full system backup.

Since we don't really need huge pages, use GFP_TRANSHUGE_LIGHT |
__GFP_NORETRY instead, in order to fail quickly when there are no huge
pages available.

Set __GFP_KSWAPD_RECLAIM as well, in order for huge pages to be freed
up in the background if necessary.

With these changes, I'm no longer seeing freezes during a restic backup.

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