drm/amdkfd: Use kfd fd to mmap mmio

TTM doesn't support CPU mapping of sg type bo (under which
mmio bo is created). Switch mmaping of mmio page to kfd
device file.

Signed-off-by: Oak Zeng <Oak.Zeng@amd.com>
Acked-by: Christian Konig <christian.koenig@amd.com>
Reviewed-by: Felix Kuehling <Felix.Kuehling@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: fix unload driver fail

dc_destroy should be called amdgpu_cgs_destroy_device,
as it will use cgs context to read or write registers.

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

drm/amd/doc: Add RAS documentation to guide

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

drm/amd/doc: Add XGMI sysfs documentation

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

drm/amdgpu: sort probed modes before adding common modes

[Why]
There are monitors which can have more than one preferred mode
set. There are chances in these monitors that if common modes are
added in function amdgpu_dm_connector_add_common_modes(), these
common modes can be calculated with different preferred mode than
the one used in function decide_crtc_timing_for_drm_display_mode().
The preferred mode can be different because after common modes
are added, the mode list is sorted and this changes the order of
preferred modes in the list. The first mode in the list with
preferred flag set is selected as preferred mode. Due to this the
preferred mode selected varies.
If same preferred mode is not selected in common mode calculation
and crtc timing, then during mode set instead of setting preferred
timing, common mode timing will be applied which can cause "out of
range" message in the monitor with monitor blanking out.

[How]
Sort the modes before adding common modes. The same sorting function
is called during common mode addition and deciding crtc timing.

Signed-off-by: Yogesh Mohan Marimuthu <yogesh.mohanmarimuthu@amd.com>
Reviewed-by: Nicholas Kazlauskas <nicholas.kazlauskas@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Use new connector state when getting color depth

[Why]
The current state on the connector is queried when getting the max bpc
rather than the new state. This means that a new max bpc value can only
currently take effect on the commit *after* it changes.

The new state should be passed in instead.

[How]
Pass down the dm_state as drm state to where we do color depth lookup.

The passed in state can still be NULL when called from
amdgpu_dm_connector_mode_valid, so make sure that we have reasonable
defaults in place. That should probably be addressed at some point.

This change now (correctly) causes a modeset to occur when changing the
max bpc for a connector.

v2: Drop extra TODO.

Cc: Leo Li <sunpeng.li@amd.com>
Cc: Harry Wentland <harry.wentland@amd.com>
Signed-off-by: Nicholas Kazlauskas <nicholas.kazlauskas@amd.com>
Acked-by: Alex Deucher <alexander.deucher@amd.com>
Reviewed-by: Harry Wentland <harry.wentland@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Switch the custom "max bpc" property to the DRM prop

[Why]
The custom "max bpc" property was added to limit color depth while the
DRM one was still being merged. It's been a few kernel versions since
then and this TODO was still sticking around.

[How]
Attach the DRM max bpc property to the connector and drop all of our
custom property management. Set the max bpc to 8 by default since
DRM defaults to the max in the range which would be 16 in this case.

No behavioral changes are intended with this patch, it should just be
a refactor.

v2: Don't force 8bpc when no state is given

Cc: Leo Li <sunpeng.li@amd.com>
Cc: Harry Wentland <harry.wentland@amd.com>
Signed-off-by: Nicholas Kazlauskas <nicholas.kazlauskas@amd.com>
Acked-by: Alex Deucher <alexander.deucher@amd.com>
Reviewed-by: Harry Wentland <harry.wentland@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/powerplay: Fix code error for translating int type to bool type correctly

Fix code error to support value < 0 or > 1.

Signed-off-by: Chengming Gui <Jack.Gui@amd.com>
Reviewed-by: Hawking Zhang <hawking.zhang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: Add Unique Identifier sysfs file unique_id v2

Add a file that provides a Unique ID for the GPU.
This will persist across machines and is guaranteed to be unique.
This is only available for GFX9 and newer, so older ASICs will not
have this file in the sysfs pool

v2: Store it in adev for ASICs that don't have a hwmgr

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

drm/amdkfd: Add domain number into gpu_id

A multi-socket server can have multiple PCIe segments so BFD is not enough
to distingush each GPU. Also add domain number into account when generating
gpu_id.

Signed-off-by: Amber Lin <Amber.Lin@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>

drm/amdkfd: Add VegaM support

Add the VegaM information to KFD

Signed-off-by: Kent Russell <kent.russell@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>

drm/amdkfd: Fix a circular lock dependency

Fix a circular lock dependency exposed under userptr memory pressure.
The DQM lock is the only one taken inside the MMU notifier. We need
to make sure that no reclaim is done under this lock, and that
no other locks are taken under which reclaim is possible.

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

drm/amdkfd: Delete alloc_format field from map_queue struct

Alloc format was never really supported by MEC FW. FW always
does one per pipe allocation.

Signed-off-by: Oak Zeng <Oak.Zeng@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>

drm/amdkfd: Expose sdma engine numbers to topology

Expose available numbers of both SDMA queue types in the topology.

Signed-off-by: Oak Zeng <Oak.Zeng@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>

drm/amdkfd: Introduce XGMI SDMA queue type

Existing QUEUE_TYPE_SDMA means PCIe optimized SDMA queues.
Introduce a new QUEUE_TYPE_SDMA_XGMI, which is optimized
for non-PCIe transfer such as XGMI.

Signed-off-by: Oak Zeng <Oak.Zeng@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>

drm/amdkfd: Fix sdma queue map issue

Previous codes assumes there are two sdma engines.
This is not true e.g., Raven only has 1 SDMA engine.
Fix the issue by using sdma engine number info in
device_info.

Signed-off-by: Oak Zeng <Oak.Zeng@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>

drm/amdkfd: Move sdma_queue_id calculation into allocate_sdma_queue()

This avoids duplicated code.

Signed-off-by: Yong Zhao <Yong.Zhao@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>

drm/amdkfd: Allocate hiq and sdma mqd from mqd trunk

Instead of allocat hiq and sdma mqd from sub-allocator, allocate
them from a mqd trunk pool. This is done for all asics

Signed-off-by: Oak Zeng <ozeng@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>

drm/amdkfd: Move non-sdma mqd allocation out of init_mqd

This is preparation work to introduce more mqd allocation
scheme

Signed-off-by: Oak Zeng <ozeng@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>

drm/amdkfd: Fix a potential memory leak

Free mqd_mem_obj it GTT buffer allocation for MQD+control stack fails.

Signed-off-by: Oak Zeng <ozeng@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>

drm/amdkfd: Allocate MQD trunk for HIQ and SDMA

MEC FW for some new asic requires all SDMA MQDs to be in a continuous
trunk of memory right after HIQ MQD. Add a field in device queue manager
to hold the HIQ/SDMA MQD memory object and allocate MQD trunk on device
queue manager initialization.

Signed-off-by: Oak Zeng <ozeng@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>

drm/amdkfd: Add mqd size in mqd manager struct

Also initialize mqd size on mqd manager initialization

Signed-off-by: Oak Zeng <ozeng@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>

drm/amdkfd: Init mqd managers in device queue manager init

Previously mqd managers was initialized on demand. As there
are only a few type of mqd managers, the on demand initialization
doesn't save too much memory. Initialize them on device
queue initialization instead and delete the get_mqd_manager
interface. This makes codes more organized for future changes.

Signed-off-by: Oak Zeng <ozeng@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>

drm/amdkfd: Introduce DIQ type mqd manager

With introduction of new mqd allocation scheme for HIQ,
DIQ and HIQ use different mqd allocation scheme, DIQ
can't reuse HIQ mqd manager

Signed-off-by: Oak Zeng <ozeng@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>

drm/amdkfd: Introduce asic-specific mqd_manager_init function

Global function mqd_manager_init just calls asic-specific functions and it
is not necessary. Delete it and introduce a mqd_manager_init interface in
dqm for asic-specific mqd manager init. Call mqd_manager_init interface
directly to initialize mqd manager

Signed-off-by: Oak Zeng <ozeng@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>

drm/amdgpu: Improve error handling for HMM

Use unsigned long for number of pages.

Check that pfns are valid after hmm_vma_fault. If they are not,
return an error instead of continuing with invalid page pointers and
PTEs.

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

drm/amdgpu: more descriptive message if HMM not enabled

If using old kernel config file, CONFIG_ZONE_DEVICE is not selected,
so CONFIG_HMM and CONFIG_HMM_MIRROR is not enabled, the current driver
error message "Failed to register MMU notifier" is not clear. Inform
user with more descriptive message on how to fix the missing kernel
config option.

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=109808

Signed-off-by: Philip Yang <Philip.Yang@amd.com>
Reviewed-by: Michel Dänzer <michel.daenzer@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: support userptr cross VMAs case with HMM

userptr may cross two VMAs if the forked child process (not call exec
after fork) malloc buffer, then free it, and then malloc larger size
buf, kerenl will create new VMA adjacent to old VMA which was cloned
from parent process, some pages of userptr are in the first VMA, the
rest pages are in the second VMA.

HMM expects range only have one VMA, loop over all VMAs in the address
range, create multiple ranges to handle this case. See
is_mergeable_anon_vma in mm/mmap.c for details.

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: support concurrent userptr update for HMM

Userptr restore may have concurrent userptr invalidation after
hmm_vma_fault adds the range to the hmm->ranges list, needs call
hmm_vma_range_done to remove the range from hmm->ranges list first,
then reschedule the restore worker. Otherwise hmm_vma_fault will add
same range to the list, this will cause loop in the list because
range->next point to range itself.

Add function untrack_invalid_user_pages to reduce code duplication.

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: fix HMM config dependency issue

Only select HMM_MIRROR will get kernel config dependency warnings
if CONFIG_HMM is missing in the config. Add depends on HMM will
solve the issue.

Add conditional compilation to fix compilation errors if HMM_MIRROR
is not enabled as HMM config is not enabled.

Remove unused function amdgpu_ttm_tt_mark_user_pages.

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: replace get_user_pages with HMM mirror helpers

Use HMM helper function hmm_vma_fault() to get physical pages backing
userptr and start CPU page table update track of those pages. Then use
hmm_vma_range_done() to check if those pages are updated before
amdgpu_cs_submit for gfx or before user queues are resumed for kfd.

If userptr pages are updated, for gfx, amdgpu_cs_ioctl will restart
from scratch, for kfd, restore worker is rescheduled to retry.

HMM simplify the CPU page table concurrent update check, so remove
guptasklock, mmu_invalidations, last_set_pages fields from
amdgpu_ttm_tt struct.

HMM does not pin the page (increase page ref count), so remove related
operations like release_pages(), put_page(), mark_page_dirty().

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

drm/amdkfd: avoid HMM change cause circular lock

There is circular lock between gfx and kfd path with HMM change:
lock(dqm) -> bo::reserve -> amdgpu_mn_lock

To avoid this, move init/unint_mqd() out of lock(dqm), to remove nested
locking between mmap_sem and bo::reserve. The locking order
is: bo::reserve -> amdgpu_mn_lock(p->mn)

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

drm/amdgpu: use HMM callback to replace mmu notifier

Replace our MMU notifier with hmm_mirror_ops.sync_cpu_device_pagetables
callback. Enable CONFIG_HMM and CONFIG_HMM_MIRROR as a dependency in
DRM_AMDGPU_USERPTR Kconfig.

It supports both KFD userptr and gfx userptr paths.

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

drm/amdgpu: Use heavy weight for tlb invalidation on xgmi configuration

There is a bug found in vml2 xgmi logic:
mtype is always sent as NC on the VMC to TC interface for a page walk,
regardless of whether the request is being sent to local or remote GPU.
NC means non-coherent and will cause the VMC return data to be cached
in the TCC (versus UC – uncached will not cache the data). Since the
page table updates are being done by SDMA/HDP, then TCC will never be
updated and the GC VML2 will continue to hit on the TCC and never get
the updated page tables and result in a fault.
Heave weigh tlb invalidation does a WB/INVAL of the L1/L2 GL data
caches so TCC will not be hit on next request

Signed-off-by: shaoyunl <Shaoyun.Liu@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>

drm/amdkfd: Preserve ttmp[4:5] instead of ttmp[14:15]

ttmp[4:5] is initialized by the SPI with SPI_GDBG_TRAP_DATA* values.
These values are more useful to the debugger than ttmp[14:15], which
carries dispatch_scratch_base*. There are too few registers to
preserve both.

Signed-off-by: Jay Cornwall <Jay.Cornwall@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>

drm/amdkfd: Fix gfx9 XNACK state save/restore

SQ_WAVE_IB_STS.RCNT grew from 4 bits to 5 in gfx9. Do not truncate
when saving in the high bits of TTMP1.

Signed-off-by: Jay Cornwall <Jay.Cornwall@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>

drm/amdkfd: Preserve wave state after instruction fetch MEM_VIOL

If instruction fetch fails the wave cannot be halted and returned to
the shader without raising MEM_VIOL again. Currently the wave is
terminated if this occurs, but this loses information about the cause
of the fault. The debugger would prefer the faulting wave state to be
context-saved.

Poll inside the trap handler until TRAPSTS.SAVECTX indicates context
save is ready. Exit the poll loop and complete the remainder of the
exception handler, then return to the shader. The next instruction
fetch will be from the trap handler and not the faulting PC. Context
save will then deschedule the wave and save its state.

Signed-off-by: Jay Cornwall <Jay.Cornwall@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>

drm/amdkfd: Fix gfx8 MEM_VIOL exception handler

When MEM_VIOL is asserted the context save handler rewinds the
program counter. This is incorrect for any source of the exception.
MEM_VIOL may be raised in normal operation by out-of-bounds access
to LDS or GDS and does not require special handling.

Remove PC adjustment when MEM_VIOL has been raised.

Signed-off-by: Jay Cornwall <Jay.Cornwall@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>

drm/amdkfd: Fix compute profile switching

Fix compute profile switching on process termination.

Add a dedicated reference counter to keep track of entry/exit to/from
compute profile. This enables switching compute profiles for other
reasons than process creation or termination.

Signed-off-by: Harish Kasiviswanathan <Harish.Kasiviswanathan@amd.com>
Signed-off-by: Eric Huang <JinhuiEric.Huang@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>

drm/amdkfd: Shift sdma_engine_id and sdma_queue_id in mqd

FW of some new ASICs requires sdma mqd size to be not more than
128 dwords. Repurpose the last 2 reserved fields of sdma mqd for
driver internal use, so the total mqd size is no bigger than 128
dwords

Signed-off-by: Oak Zeng <ozeng@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>

drm/amdkfd: Differentiate b/t sdma_id and sdma_queue_id

sdma_queue_id is sdma queue index inside one sdma engine.
sdma_id is sdma queue index among all sdma engines. Use
those two names properly.

Signed-off-by: Oak Zeng <ozeng@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>

drm/amdkfd: Add sdma allocation debug message

Add debug messages during SDMA queue allocation.

Signed-off-by: Oak Zeng <Oak.Zeng@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>

drm/amdkfd: Use 64 bit sdma_bitmap

Maximumly support 64 sdma queues

Signed-off-by: Oak Zeng <Oak.Zeng@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>

drm/amd/powerplay: enable ppfeaturemask module parameter support on Vega20

Support DPM/DS/ULV related bitmasks of ppfeaturemask module parameter.

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

drm/amdgpu: use pcie_bandwidth_available rather than open coding it

It does the same thing we were doing already.  I though it needed
work for gen3/4 speeds, but that seems to be covered already.

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

gpu: fix typos in code comments

fix eror to error

Signed-off-by: Weitao Hou <houweitaoo@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/powerplay: fix sw SMU wrong UVD/VCE powergate setting

The UVD/VCE bits are set wrongly. This causes the UVD/VCE clocks
are not brought back correctly on needed.

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

drm/amdgpu: use div64_ul for 32-bit compatibility v1

v1: replace casting to unsigned long with div64_ul

Acked-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Slava Abramov <slava.abramov@amd.com>
Tested-by: Slava Abramov <slava.abramov@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/powerplay: Enable "disable dpm" feature to support swSMU debug (v2)

add pm_enabled to control the dpm off/on.
v2: Directly return 0 to replace return ret and merge some check code.

Signed-off-by: Chengming Gui <Jack.Gui@amd.com>
Reviewed-by: Hawking Zhang <hawking.zhang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/powerplay: Fix maybe-uninitialized in get_ppfeature_status

This fixes the warning below

error: ‘feature_mask’ may be used uninitialized in this function [-Werror=maybe-uninitialized]
  *features_enabled = ((((uint64_t)feature_mask[0] << SMU_FEATURES_LOW_SHIFT) & SMU_FEATURES_LOW_MASK) |
                      ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~^
    (((uint64_t)feature_mask[1] << SMU_FEATURES_HIGH_SHIFT) & SMU_FEATURES_HIGH_MASK));
    ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

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

drm/amdgpu: fix spelling mistake "retrived" -> "retrieved"

There is a spelling mistake in a DRM_ERROR error message. Fix this.

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

drm/amdgpu/vega20: use mode1 reset for RAS and XGMI

If RAS or XGMI are enabled, you have to use mode1 reset rather
than BACO.

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

drm/amd/display: 3.2.31

Signed-off-by: Aric Cyr <aric.cyr@amd.com>
Reviewed-by: Aric Cyr <Aric.Cyr@amd.com>
Acked-by: Bhawanpreet Lakha <Bhawanpreet.Lakha@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Disable ABM before destroy ABM struct

[Why]
When disable driver, OS will set backlight optimization
then do stop device.  But this flag will cause driver to
enable ABM when driver disabled.

[How]
Send ABM disable command before destroy ABM construct

Signed-off-by: Paul Hsieh <paul.hsieh@amd.com>
Reviewed-by: Anthony Koo <Anthony.Koo@amd.com>
Acked-by: Bhawanpreet Lakha <Bhawanpreet.Lakha@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: dont set  otg offset

move the update of otg instance outside of hw programming logic,
since this is sw state, it should always be updated and should
never be optimized away.

Signed-off-by: Jun Lei <Jun.Lei@amd.com>
Reviewed-by: Eric Yang <eric.yang2@amd.com>
Acked-by: Bhawanpreet Lakha <Bhawanpreet.Lakha@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Explicitly specify update type per plane info change

[Why]
The bit for flip addr is being set causing the determination for
FAST vs MEDIUM to always return MEDIUM when plane info is provided
as a surface update. This causes extreme stuttering for the typical
atomic update path on Linux.

[How]
Don't use update_flags->raw for determining FAST vs MEDIUM. It's too
fragile to changes like this.

Explicitly specify the update type per update flag instead. It's not
as clever as checking the bits itself but at least it's correct.

Signed-off-by: Nicholas Kazlauskas <nicholas.kazlauskas@amd.com>
Reviewed-by: Harry Wentland <Harry.Wentland@amd.com>
Acked-by: Bhawanpreet Lakha <Bhawanpreet.Lakha@amd.com>
Acked-by: Eryk Brol <Eryk.Brol@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Rename EDID_BLOCK_SIZE to DC_EDID_BLOCK_SIZE

Signed-off-by: Joshua Aberback <joshua.aberback@amd.com>
Reviewed-by: Abdoulaye Berthe <Abdoulaye.Berthe@amd.com>
Acked-by: Bhawanpreet Lakha <Bhawanpreet.Lakha@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Use VCP for extended colorimetry

DPRX should send the VCP extended colorimetry packet if the
sink supports DPCD rev1.4 and reports the extended colorimetry
bit.

Signed-off-by: Aric Cyr <aric.cyr@amd.com>
Reviewed-by: Anthony Koo <Anthony.Koo@amd.com>
Acked-by: Bhawanpreet Lakha <Bhawanpreet.Lakha@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Use DCN functions instead of DCE

[WHY]
DCN code should make as few references to DCE as possible

[HOW]
Copy DCE110 implementation of find_first_free_match_stream_enc_for_link
into DCN10

Signed-off-by: Wesley Chalmers <Wesley.Chalmers@amd.com>
Reviewed-by: Tony Cheng <Tony.Cheng@amd.com>
Acked-by: Bhawanpreet Lakha <Bhawanpreet.Lakha@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Engine-specific encoder allocation

[WHY]
From DCE110 onward, we have the ability to assign DIG BE and FE
separately for any display connector type; before, we could only do this
for DP.

Signed-off-by: Wesley Chalmers <Wesley.Chalmers@amd.com>
Reviewed-by: Tony Cheng <Tony.Cheng@amd.com>
Acked-by: Bhawanpreet Lakha <Bhawanpreet.Lakha@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: enabling stream after HPD low to high happened

[Why]
1. No real HPD plug in/out but HPD happens,
 the driver notifies OS connection changed.
2. No display in target.

When HPD goes low to high,
the driver should regard as HPD and enter setmode flow.

[How]
In this case, even stream didn't change but still retrain.

Signed-off-by: Chiawen Huang <chiawen.huang@amd.com>
Signed-off-by: Charlene Liu <charlene.liu@amd.com>
Reviewed-by: Tony Cheng <Tony.Cheng@amd.com>
Acked-by: Anthony Koo <Anthony.Koo@amd.com>
Acked-by: Bhawanpreet Lakha <Bhawanpreet.Lakha@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Program VTG params after programming Global Sync

[Why]
VTG has a parameter FP2, which is defined as:
    if VSTARTUP is before VSYNC:
        FP2 = number of lines in between VSTARTUP and VSYNC
    else
        FP2 = 0
Currently, FP2 is only programmed during "program_timing". However, the
position of VSTARTUP is affected by the prefetching requirements on all pipes,
so the position might change when we do memory request control on another pipe, so we need
to make sure that FP2 stays up-to-date whenever we adjust VSTARTUP.

[How]
 - refactor VTG_CONTROL programming into a new function "set_vtg_params"
 - call it after calling "program_global_sync"
   - make sure it's called after because it relies on the cached dlg params

Signed-off-by: Joshua Aberback <joshua.aberback@amd.com>
Reviewed-by: Tony Cheng <Tony.Cheng@amd.com>
Acked-by: Bhawanpreet Lakha <Bhawanpreet.Lakha@amd.com>
Acked-by: Jun Lei <Jun.Lei@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: add null checks and set update flags

* add plane state null checks
* add and set update surface flags

Signed-off-by: Dmytro Laktyushkin <Dmytro.Laktyushkin@amd.com>
Reviewed-by: Eric Bernstein <Eric.Bernstein@amd.com>
Acked-by: Bhawanpreet Lakha <Bhawanpreet.Lakha@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: 3.2.30

Signed-off-by: Aric Cyr <aric.cyr@amd.com>
Reviewed-by: Aric Cyr <Aric.Cyr@amd.com>
Acked-by: Bhawanpreet Lakha <Bhawanpreet.Lakha@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Reuse MPC OGRAM for 1D blender

[Why & How]
Reuse existent code path and in order to do that apply de gamma
in 1D blender LUT and re use MPC OGAM.
Follow up is required.

Signed-off-by: Vitaly Prosyak <vitaly.prosyak@amd.com>
Reviewed-by: Aric Cyr <Aric.Cyr@amd.com>
Acked-by: Bhawanpreet Lakha <Bhawanpreet.Lakha@amd.com>
Acked-by: Krunoslav Kovac <Krunoslav.Kovac@amd.com>
Acked-by: Vitaly Prosyak <Vitaly.Prosyak@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Remove DPMS state dependency for fast boot

[Why]
The DPMS state of a display should not impact whether we want to enable fast boot.
Currently fast boot is not enabled when resuming from S4 because of this.

[How]
Remove check for DPMS state when determining if fast boot
can be applied.

Signed-off-by: SivapiriyanKumarasamy <sivapiriyan.kumarasamy@amd.com>
Reviewed-by: Anthony Koo <Anthony.Koo@amd.com>
Acked-by: Bhawanpreet Lakha <Bhawanpreet.Lakha@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Hook up CRC capture support for dce120

[Why]
Many IGT tests require CRC capture in order to confirm that the output
is visually correct.

These skip on dce120 because configure_crc and get_crc aren't set.

[How]
Hook up is_tg_enabled, configure_crc and get_crc functions on dce120's
timing generator.

The logic should be the same as DCE and DCN with some minor register
naming differences.

Signed-off-by: Nicholas Kazlauskas <nicholas.kazlauskas@amd.com>
Reviewed-by: David Francis <David.Francis@amd.com>
Acked-by: Bhawanpreet Lakha <Bhawanpreet.Lakha@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: define v_total_min and max parameters

add these parameters for future use

Signed-off-by: Charlene Liu <charlene.liu@amd.com>
Reviewed-by: Dmytro Laktyushkin <Dmytro.Laktyushkin@amd.com>
Acked-by: Bhawanpreet Lakha <Bhawanpreet.Lakha@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/powerplay: support sw smu hotspot and memory temperature retrieval

Support hotspot and memory temperature retrieval on sw smu routine.

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

drm/amd/powerplay: support uclk activity retrieve on sw smu routine

Support realtime uclk activity report.

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

drm/amd/powerplay: honor hw limit on fetching metrics data

Request too frequently may get corrupt data.

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

drm/amd/powerplay: support ppfeatures sysfs interface on sw smu routine

Support ppfeatures sysfs interface on Vega20 sw smu routine.

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

drm/amd/powerplay: drop redundant smu call

smu_get_clk_info_from_vbios() was called repeatedly. It
seems a merge error.

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

drm/amd/powerplay: drop unnecessary sw smu check

There is already sw smu check on IP block adding.

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

drm/amd/display: Drop DCN1_01 guards

[WHY]
These were only needed for bringup. They're not needed anymore.

Signed-off-by: Harry Wentland <harry.wentland@amd.com>
Reviewed-by: Nicholas Kazlauskas <nicholas.kazlauskas@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Don't load DMCU for Raven 1 (v2)

[WHY]
Some early Raven boards had a bad SBIOS that doesn't play nicely with
the DMCU FW. We thought the issues were fixed by ignoring errors on DMCU
load but that doesn't seem to be the case. We've still seen reports of
users unable to boot their systems at all.

[HOW]
Disable DMCU load on Raven 1. Only load it for Raven 2 and Picasso.

v2: Fix ifdef (Alex)

Signed-off-by: Harry Wentland <harry.wentland@amd.com>
Reviewed-by: Nicholas Kazlauskas <nicholas.kazlauskas@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Add ASICREV_IS_PICASSO

[WHY]
We only want to load DMCU FW on Picasso and Raven 2, not on Raven 1.

Signed-off-by: Harry Wentland <harry.wentland@amd.com>
Reviewed-by: Nicholas Kazlauskas <nicholas.kazlauskas@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Cc: stable@vger.kernel.org

drm/amdgpu: Report firmware versions with sysfs v2

Firmware versions can be found as separate sysfs files at:
/sys/class/drm/cardX/device/fw_version (where X is the card number)
The firmware versions are displayed in hexadecimal.
v2: Moved sysfs files to subfolder

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

drm/amdgpu: make VCN DPG pause mode detached from general VCN

It should be attached to VCN 1.0

Signed-off-by: Leo Liu <leo.liu@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: move the VCN DPG mode read and write to VCN

Since this is VCN specific and only used by VCN

Signed-off-by: Leo Liu <leo.liu@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu/sriov: Need to initialize the HDP_NONSURFACE_BAStE

it requires to initialize HDP_NONSURFACE_BASE, so as to avoid
using the value left by a previous VM under sriov scenario.

v2: it should not hurt baremetal, generalize it for both sriov
and baremetal

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

drm/amdgpu: suppress repeating tmo report

only report once per TMO job and the timer would
be restarted upon the job finished if it's just slow.

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

drm/amdgpu: remove static GDS, GWS and OA allocation

As far as we know this was never used by userspace and so should be removed.

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

drm/amd/powerplay: update Vega10 ACG Avfs Gb parameters

Update Vega10 ACG Avfs GB parameters.

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>

drm/amd/powerplay: force to update all clock tables on OD reset

On OD reset, the clock tables in SMU need to be reset to default.

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>

drm/amd/powerplay: update Vega10 power state on OD

Update Vega10 top performance level power state accordingly
on OD.

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>

drm/amd/powerplay: avoid repeat AVFS enablement/disablement

No need to enable or disable AVFS if it's already in wanted
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>

drm/amd/powerplay: valid Vega10 DPMTABLE_OD_UPDATE_VDDC settings V2

With user specified voltage(DPMTABLE_OD_UPDATE_VDDC), the AVFS
will be disabled. However, the buggy code makes this actually not
working as expected.

- V2: clear all OD flags excpet DPMTABLE_OD_UPDATE_VDDC

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>

drm/amd/powerplay: fix Vega10 mclk/socclk voltage link setup

This may affects the Vega10 MCLK OD functionality.

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>

drm/amdgpu: check no_user_fence flag for engines

To replace checking ring type and make them generic

Signed-off-by: Leo Liu <leo.liu@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>

drm/amdgpu/VCN: set no_user_fence flag to true

There is no user fence support for VCN

Signed-off-by: Leo Liu <leo.liu@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>

drm/amdgpu/VCE: set no_user_fence flag to true

There is no user fence support for VCE

Signed-off-by: Leo Liu <leo.liu@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>

drm/amdgpu/UVD: set no_user_fence flag to true

There is no user fence support for UVD

Signed-off-by: Leo Liu <leo.liu@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>

drm/amdgpu: add no_user_fence flag to ring funcs

So we can generalize the no user fence supported engine

Signed-off-by: Leo Liu <leo.liu@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>

drm/amdgpu: sdma handle ras resume

During S3/S4 bootloader will re-init ras state behind us.
Resume might fail or raise a gpu reset.

Signed-off-by: xinhui pan <xinhui.pan@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Reviewed-by: James Zhu <James.Zhu@amd.com>
Tested-by: James Zhu <James.Zhu@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: gfx handle ras resume

During S3/S4 bootloader will re-init ras state behind us.
Resume might fail or raise a gpu reset.

Signed-off-by: xinhui pan <xinhui.pan@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Reviewed-by: James Zhu <James.Zhu@amd.com>
Tested-by: James Zhu <James.Zhu@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: gmc handle ras resume

During S3/S4 bootloader will re-init ras state behind us.
Resume might fail or raise a gpu reset.

Signed-off-by: xinhui pan <xinhui.pan@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Reviewed-by: James Zhu <James.Zhu@amd.com>
Tested-by: James Zhu <James.Zhu@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: enable ras suspend/resume

suspend/resume will change ras state behind us. Let driver get notified.

Signed-off-by: xinhui pan <xinhui.pan@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Reviewed-by: James Zhu <James.Zhu@amd.com>
Tested-by: James Zhu <James.Zhu@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: ras support suspend/resume

add ras suspend function. rename ras_post_init to amdgpu_ras_resume.

Signed-off-by: xinhui pan <xinhui.pan@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Reviewed-by: James Zhu <James.Zhu@amd.com>
Tested-by: James Zhu <James.Zhu@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: add badpages sysfs interafce

add badpages node.
it will output badpages list in format
gpu pfn : gpu page size : flags

example
0x00000000 : 0x00001000 : R
0x00000001 : 0x00001000 : R
0x00000002 : 0x00001000 : R
0x00000003 : 0x00001000 : R
0x00000004 : 0x00001000 : R
0x00000005 : 0x00001000 : R
0x00000006 : 0x00001000 : R
0x00000007 : 0x00001000 : P
0x00000008 : 0x00001000 : P
0x00000009 : 0x00001000 : P

flags can be one of below characters
R: reserved.
P: pending for reserve.
F: failed to reserve for some reasons.

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