drm/amdgpu: Set GTT_USWC flag to enable freesync v2

This patch sets 'AMDGPU_GEM_CREATE_CPU_GTT_USWC' as input
parameter flag, during object creation of an imported DMA
buffer.

In absence of this flag:
1. Function amdgpu_display_supported_domains() doesn't add
   AMDGPU_GEM_DOMAIN_GTT as supported domain.
2. Due to which, Function amdgpu_display_user_framebuffer_create()
   refuses to create framebuffer for imported DMA buffers.
3. Due to which, AddFB() IOCTL fails.
4. Due to which, amdgpu_present_check_flip() check fails in DDX
5. Due to which DDX driver doesn't allow flips (goes to blitting)
6. Due to which setting Freesync/VRR property fails for PRIME buffers.

So, this patch finally enables Freesync with PRIME buffer offloading.

v2 (chk): instead of just checking the flag we copy it over if the
          exporter is an amdgpu device as well.

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

drm/amdgpu: clean-up unused variable

Variable 'bp' seems to be unused residue from previous
logic, and is not required anymore.

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

Revert freesync video patches temporarily

This temporarily reverts freesync video patches since it causes regression with
eDP displays. This patch is a squashed revert of the following patches:

6f59f229f8ed ("drm/amd/display: Skip modeset for front porch change")
d10cd527f5e5 ("drm/amd/display: Add freesync video modes based on preferred modes")
0eb1af2e8205 ("drm/amd/display: Add module parameter for freesync video mode")

Signed-off-by: Aurabindo Pillai <aurabindo.pillai@amd.com>
Reviewed-by: Anson Jacob <anson.jacob@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdkfd: Fix UBSAN shift-out-of-bounds warning

If get_num_sdma_queues or get_num_xgmi_sdma_queues is 0, we end up
doing a shift operation where the number of bits shifted equals
number of bits in the operand. This behaviour is undefined.

Set num_sdma_queues or num_xgmi_sdma_queues to ULLONG_MAX, if the
count is >= number of bits in the operand.

Bug: https://gitlab.freedesktop.org/drm/amd/-/issues/1472

Reported-by: Lyude Paul <lyude@redhat.com>
Signed-off-by: Anson Jacob <Anson.Jacob@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Reviewed-by: Felix Kuehling <Felix.Kuehling@amd.com>
Tested-by: Lyude Paul <lyude@redhat.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: Increase PSP runtime TMR region size

Aldebaran uses more than 4M runtime TMR. The current
hard coded 4M TMR is not big enough for Aldebaran.
Increase it to 8M.

v2: Only do 8M size for ALDEBARAN (Hawking)

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

drm/amdkfd: apply uncached flag for aldebaran

The flag is only applied on fine-grained memory.

Signed-off-by: Eric Huang <jinhuieric.huang@amd.com>
Reviewed-by: Oak Zeng <Oak.Zeng@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: set snoop bit in pde/pte entries for A+A

Page tables in vram mapping to cpu is changed from uncached to
cached in A+A, the snoop bit in VM_CONTEXTx_PAGE_TABLE_BASE_ADDR/
PDE0s/PDE1s/PDE2s/PTE.TFs has to be set so gpuvm walker snoop
page table data out of CPU cache.

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

drm/amdgpu: set CPU mapping of vram as cached for A+A mode

New A+A HW supports cached vram mapped to cpu.

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

drm/amdgpu: harvest edc status when connected to host via xGMI

When connected to a host via xGMI, system fatal errors may trigger
warm reset, driver has no change to query edc status before reset.
Therefore in this case, driver should harvest previous error loging
registers during boot, instead of only resetting them.

v2:
1. IP's ras_manager object is created when its ras feature is enabled,
so change to query edc status after amdgpu_ras_late_init called

2. change to enable watchdog timer after finishing gfx edc init

Signed-off-by: Dennis Li <Dennis.Li@amd.com>
Reivewed-by: Hawking Zhang <hawking.zhang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: Make noretry the default on Aldebaran

This is needed for best machine learning performance. XNACK can still
be enabled per-process if needed.

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

drm/amdgpu: update default timeout of Aldebaran SQ watchdog

Signed-off-by: Harish Kasiviswanathan <Harish.Kasiviswanathan@amd.com>
Reivewed-by: Hawking Zhang <hawking.zhang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/pm: add new data in metrics table

Export new data in the metrics table for gfx and memory
utilization counter, and each hbm temperature as well.

v2:
change the metrics table version to v1.1

v3:
fix the coding style
v4:
rebase against latest kernel

Signed-off-by: Kenneth Feng <kenneth.feng@amd.com>
Reviewed-by: Kevin Wang <kevin1.wang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: add psp RAP L0 check support

add PSP RAP L0 check when RAP TA is loaded.

Signed-off-by: Kevin Wang <kevin1.wang@amd.com>
Reviewed-by: Hawking Zhang <Hawking.Zhang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: change psp_rap_invoke() function return value

RAP TA is an optional firmware. if it doesn’t exist,
the driver should bypass psp_rap_invoke() function.

1. bypass psp_rap_invoke() when RAP TA is not loaded.
2. add new parameter (status) to query RAP TA status.
   (the status value is different with psp_ta_invoke(),
3. fix the 'rap_status' MThread critical problem.
   (used without lock)

Signed-off-by: Kevin Wang <kevin1.wang@amd.com>
Reviewed-by: Hawking Zhang <Hawking.Zhang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/pm: add aldebaran serial number support

add aldebaran serial number support.
(serial number from metrics table)

Signed-off-by: Kevin Wang <kevin1.wang@amd.com>
Reviewed-by: Hawking Zhang <Hawking.Zhang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: Let KFD use more VMIDs on Aldebaran

When there is no graphics support, KFD can use more of the VMIDs. Graphics
VMIDs are only used for video decoding/encoding and post processing. With
two VCE engines, there is no reason to reserve more than 2 VMIDs for that.

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

drm/amdgpu: enable watchdog feature for SQ of aldebaran

SQ's watchdog timer monitors forward progress, a mask of which waves
caused the watchdog timeout is recorded into ras status registers and
then trigger a system fatal error event.

v2:
1. change *query_timeout_status to *query_sq_timeout_status.
2. move query_sq_timeout_status into amdgpu_ras_do_recovery.
3. add module parameters to enable/disable fatal error event and modify
the watchdog timer.

v3:
1. remove unused parameters of *enable_watchdog_timer

Signed-off-by: Dennis Li <Dennis.Li@amd.com>
Reviewed-by: Hawking Zhang <Hawking.Zhang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: refine ras codes for GC utc of aldebaran

The bank number of both VML2 and ATCL2 are changed to 8, so refine
related codes to avoid defining long name arrays.

Signed-off-by: Dennis Li <Dennis.Li@amd.com>
Reviewed-by: Hawking Zhang <Hawking.Zhang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: add ras support for gfx of aldebaran

add edc counter/status reset and query functions for gfx block of
aldebaran.

v2: change to clear edc counter explicitly
aldebaran hardware will not clear edc counter after driver reading them,
so driver should clear them explicitly.

Signed-off-by: Dennis Li <Dennis.Li@amd.com>
Reviewed-by: Hawking Zhang <Hawking.Zhang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: add gc powerbrake support (v2)

add GC power brake feature support for Aldebaran.

v2: squash in fixes (Alex)

Signed-off-by: Kevin Wang <kevin1.wang@amd.com>
Reviewed-by: Kenneth Feng <kenneth.feng@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: update TCP_CHAN_STEER_1 golden value for aldebaran

The golden setting was changed recently. update to
the latest one

Signed-off-by: Hawking Zhang <Hawking.Zhang@amd.com>
Reviewed-by: Kevin Wang <kevin1.wang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: add common gc golden settings for aldebaran

golden settings that should be applied

Signed-off-by: Hawking Zhang <Hawking.Zhang@amd.com>
Reviewed-by: Kevin Wang <kevin1.wang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: apply gc v9_4_2 golden settings for aldebaran

Those registers should be programmed as one-time initialization

Signed-off-by: Hawking Zhang <Hawking.Zhang@amd.com>
Reviewed-by: Kevin Wang <kevin1.wang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: restore aldebaran save ttmp and trap config on init (v2)

Initialization of TRAP_DATA0/1 is still required for the debugger to detect
new waves on Aldebaran.  Also, per-vmid global trap enablement may be
required outside of debugger scope so move to init phase.

v2: just add the gfx 9.4.2 changes (Alex)

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

drm/amdkfd: add aldebaran kfd2kgd callbacks to kfd device (v2)

Create dedicated Aldebaran kfd2kgd callbacks to prepare
for new per-vmid register instructions for debug trap
setting functions and sending host traps.

v2: rebase (Alex)

Signed-off-by: Jonathan Kim <Jonathan.Kim@amd.com>
Reviewed-by: Oak Zeng <Oak.Zeng@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdkfd: Check HIQ's MQD for queue preemption status

MEC firmware can silently fail the queue preemption request
without time out. In this case, HIQ's MQD's queue_doorbell_id
will be set. Check this field to see whether last queue preemption
was successful or not.

Signed-off-by: Oak Zeng <Oak.Zeng@amd.com>
Suggested-by: Jay Cornwall <Jay.Cornwall@amd.com>
Acked-by: Felix Kuehling <Felix.Kuehling@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdkfd: Add kernel parameter to stop queue eviction on vm fault

This is to keep wavefront context for debug purpose

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

drm/amdgpu: allow use psp to load firmware (v2)

Match existing asics.

v2: rebase (Alex)

Signed-off-by: Hawking Zhang <Hawking.Zhang@amd.com>
Reviewed-by: Kevin Wang <kevin1.wang@amd.com>
Reviewed-by: Le Ma <Le.Ma@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/pm: Enable user min/max gfxclk on aldebaran

Aldebaran has fine grained DPM for GFXCLK. Instead of a discrete level,
user can specify a min/max range of GFXCLK for any profiling/tuning
purpose.This option is available only in manual performance level mode.
Select "manual" as power_dpm_force_performance_level and specify the
min/max range using pp_dpm_sclk sysfs node. User cannot specify a min/max
range outside of the default min/max range of the ASIC. If specified
outside the range, values will be bound by the default min/max range.

Ex: To use gfxclk min = 600MHz and max = 900MHz

echo manual > /sys/bus/pci/devices/.../power_dpm_force_performance_level
echo min 600 max 900 > /sys/bus/pci/devices/.../pp_dpm_sclk

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

drm/amd/pm: remove aldebaran serial number support

the following message is not supported.

PPSMC_MSG_ReadSerialNumTop32
PPSMC_MSG_ReadSerialNumBottom32

Signed-off-by: Kevin Wang <kevin1.wang@amd.com>
Reviewed-by: Kenneth Feng <kenneth.feng@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: use pd addr based on gart level page table

With a recent gart page table re-construction, the gart page
table is now 2-level for some ASICs: PDB0->PTB.
In the case of 2-level gart page table, the page_table_base
of vmid0 should point to PDB0 instead of PTB.

Signed-off-by: Alex Sierra <alex.sierra@amd.com>
Reviewed-by: Felix Kuehling <Felix.Kuehling@amd.com>
Reviewed-by: Oak Zeng <Oak.Zeng@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: Fix the comment in amdgpu_gmc.h

More accurate words are used to address a
code review feedback

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

drm/amdgpu: Fix GART page table s-bit

For the new 2-level GART table, the last PDE0 points
to PTB. Since PTB is in vram and right now we are
runing under s=0 mode (vram is treated as FB carveout),
so the s bit of this PDE0 should be set to 0.

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

drm/amdgpu: update mmhub client ids for Aldebaran

update mmhub client id table for Aldebaran.

Signed-off-by: Alex Sierra <alex.sierra@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: enable sram initialization for aldebaran

Aldebaran can share the same initializing shader code witn
arcturus.

Signed-off-by: Dennis Li <Dennis.Li@amd.com>
Reviewed-by: Hawking Zhang <Hawking.Zhang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: workaround the TMR MC address issue (v2)

With the 2-level gart page table,  vram is squeezed into gart aperture
and FB aperture is disabled. Therefore all VRAM virtual addresses are
 in the GART aperture. However currently PSP requires TMR addresses
in FB aperture. So we need some design change at PSP FW level to support
this 2-level gart table driver change. Right now this PSP FW support
doesn't exist. To workaround this issue temporarily, FB aperture is
added back and the gart aperture address is converted back to FB aperture
for this PSP TMR address.

Will revert it after we get a fix from PSP FW.

v2: squash in tmr fix for other asics (Kevin)

Signed-off-by: Oak Zeng <Oak.Zeng@amd.com>
Reviewed-by: Felix Kuehling <felix.kuehling@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: HW setup of 2-level vmid0 page table

Set up HW for 2-level vmid0 page table: 1. Set up
PAGE_TABLE_START/END registers. Currently only plan
to do 2-level page table for ALDEBARAN, so only gfxhub1.0
and mmhub1.7 is changed. 2. Set page table base register.
For 2-level page table, the page table base should point
to PDB0. 3. Disable AGP and FB aperture as they are not
used.

Signed-off-by: Oak Zeng <Oak.Zeng@amd.com>
Reviewed-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: Set up vmid0 PDB0

If use gart for FB translation, allocate and fill
PDB0.

Signed-off-by: Oak Zeng <Oak.Zeng@amd.com>
Reviewed-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: Add function to allocate and fill PDB0

Add functions to allocate PDB0, map it for CPU access,
and fill it.

Those functions are only used for 2-level vmid0 page
table construction

Signed-off-by: Oak Zeng <Oak.Zeng@amd.com>
Reviewed-by: Felix Kuehling <felix.kuehling@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: Use different gart table parameters for 2-level gart table

If use gart for FB translation, we will squeeze vram into
sysvm aperture. This requires 2 level gart table. Add
page table depth and page table block size parameters
to gmc. This is prepare work to 2-level gart table
construction

Signed-off-by: Oak Zeng <Oak.Zeng@amd.com>
Reviewed-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: Placement of gart and vram in sysvm aperture

If use GART for FB translation, place both vram and gart to sysvm
aperture. AGP aperture is not set up in this case because it
is not used

Signed-off-by: Oak Zeng <Oak.Zeng@amd.com>
Reviewed-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: Modify comments of vram_start/end

Modify the comment to reflect the fact that, if
use GART for vram address translation for vmid0,
[vram_start, vram_end] will be placed inside SYSVM
aperture, together with GART.

Signed-off-by: Oak Zeng <Oak.Zeng@amd.com>
Reviewed-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: Moved gart_size calculation to mc_init functions

In amdgpu_gmc_gart_location function, gart_size is adjusted
by a smu_prv_buffer_size. This logic shouldn't belong to
this function. Move the logic to the mc_init functions

Signed-off-by: Oak Zeng <Oak.Zeng@amd.com>
Reviewed-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: Use physical translation mode to access page table

On A+A platform, CPU write page directory and page table in cached
mode. So it is necessary for page table walker to snoop CPU cache.
This setting is necessary for page walker to snoop page directory
and page table data out of CPU cache.

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: Don't reserve vram as WC for A+A

On A+A platform, vram can be mapped as WB. Not necessarily
to always map vram as WC on such platform.

Calling function arch_io_reserve_memtype_wc will mark the
whole vram region as WC. So don't call it for A+A platform.

Signed-off-by: Oak Zeng <Oak.Zeng@amd.com>
Suggested-by: Alex Deucher <alexander.deucher@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/amd/pm: Correct msg status check for powerlimit

Status 0 indicates success, fix the check before using PPTable limit

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

drm/amd/pm: Enable performance determinism on aldebaran

Performance Determinism is a new mode in Aldebaran where PMFW tries to
maintain sustained performance level. It can be enabled on a per-die
basis on aldebaran. To guarantee that it remains within the power cap,
a max GFX frequency needs to be specified in this mode. A new
power_dpm_force_performance_level, "perf_determinism", is defined to enable
this mode in amdgpu. The max frequency (in MHz) can be specified through
pp_dpm_sclk. The mode will be disabled once any other performance level
is chosen.

Ex: To enable perf determinism at 900Mhz max gfx clock

echo perf_determinism > /sys/bus/pci/devices/.../power_dpm_force_performance_level
echo max 900 > /sys/bus/pci/devices/.../pp_dpm_sclk

Signed-off-by: Lijo Lazar <lijo.lazar@amd.com>
Reviewed-by: Kenneth Feng <kenneth.feng@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/pm: Add DCBTC support for aldebaran

On aldebaran DCBTC should be run after enabling DPM. DCBTC won't be run
if support is not enabled in PPTable. Without PPTable support the message
is dummy and will return success always.

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

drm/amd/pm: Fix power limit query on aldebaran

Aldebaran doesn't have AC/DC power limits. Separate the implementation
from SMU13. Max power limit is queried from PPTable.

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

drm/amdgpu: mask the xgmi number of hops reported from psp to kfd

The psp supplies the link type in the upper 2 bits of the psp xgmi node
information num_hops field.  With a new link type, Aldebaran has these
bits set to a non-zero value (1 = xGMI3) so the KFD topology will report
the incorrect IO link weights without proper masking.
The actual number of hops is located in the 3 least significant bits of
this field so mask if off accordingly before passing it to the KFD.

Signed-off-by: Jonathan Kim <jonathan.kim@amd.com>
Reviewed-by: Amber Lin <amber.lin@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: enable 48-bit IH timestamp counter

By default this timestamp is 32 bit counter. It gets
overflowed in around 10 minutes.

Signed-off-by: Alex Sierra <alex.sierra@amd.com>
Reviewed-by: Felix Kuehling <Felix.Kuehling@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: enable retry fault wptr overflow

If xnack is on, VM retry fault interrupt send to IH ring1, and ring1
will be full quickly. IH cannot receive other interrupts, this causes
deadlock if migrating buffer using sdma and waiting for sdma done while
handling retry fault.

Remove VMC from IH storm client, enable ring1 write pointer overflow,
then IH will drop retry fault interrupts and be able to receive other
interrupts while driver is handling retry fault.

IH ring1 write pointer doesn't writeback to memory by IH, and ring1
write pointer recorded by self-irq is not updated, so always read
the latest ring1 write pointer from register.

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

drm/amdgpu: Use free system memory size for kfd memory accounting

With the current kfd memory accounting scheme, kfd applications
can use up to 15/16 of total system memory. For system which
has small total system memory size it leaves small system memory
for OS. For example, if the system has totally 16GB of system
memory, this scheme leave OS and non-kfd applications only 1GB
of system memory. In many cases, this leads to OOM killer.

This patch changed the KFD system memory accounting scheme.
15/16 of free system memory when kfd driver load. This deduct
the system memory that OS already use.

Signed-off-by: Oak Zeng <Oak.Zeng@amd.com>
Suggested-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: apply new pmfw loading sequence to arcturus and onwards

Arcturus and onwards products should follow the same sequence
that have pmfw loading ahead of tmr setup

Signed-off-by: Hawking Zhang <Hawking.Zhang@amd.com>
Reviewed-by: Kevin Wang <kevin1.wang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: Fix aldebaran MMHUB CG/LS logic

Aldebaran MMHUB CG/LS logic is controlled by VBIOS. Enable the state
change logic only if driver is used for control.

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

drm/amdgpu: Enable CP idle interrupts

v1: The interrupts need to be enabled to move to DS clocks.
v2: Don't enable GFX IDLE interrupts if there are no GFX rings.

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

drm/amdgpu/pm: Remove redundant generic message index

Remove SMU_MSG_GfxDriverReset generic index.
Always use SMU_MSG_GfxDeviceDriverReset as the generic index for reset.

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

drm/amdgpu/pm: Fix reset message mapping on aldebaran

Use the correct mapping for mode-reset messages on aldebaran

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

drm/amdgpu/pm: Remove unsupported MP1 messages from aldebaran

PrepareMp1Reset and SoftReset messages are not supported on aldebaran.

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

drm/amdgpu: Add clock gating support for aldebaran

Aldebaran clock gating support for GFX,SDMA,IH blocks
VCN/JPEG blocks are excluded in this patch, to be enabled later

Signed-off-by: Lijo Lazar <lijo.lazar@amd.com>
Acked-by: Feifei Xu <Feifei.Xu@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: add mmhub client ids for aldebaran

Add the mmhub client id table for aldebaran.

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

drm/amdgpu: enable dpg indirect sram mode on aldebaran

Enable dpg indirect sram mode on aldebaran.

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

drm/amdgpu: enable vcn dpg mode on aldebaran

Enable vcn dpg mode on aldebaran

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

drm/amdgpu: enable vcn and jpeg on aldebaran

Enable vcn and jpeg 2.6 on aldebaran.

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

drm/amdgpu: Enable swsmu block on aldebaran

Enable smu13 block on aldebaran

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

drm/amdgpu: switch to cached noretry setting for aldebaran

global noretry setting now is cached to gmc.noretry

Signed-off-by: Hawking Zhang <Hawking.Zhang@amd.com>
Reviewed-by: Kevin Wang <kevin1.wang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdkfd: Fix saving the ACC vgprs for Aldebaran

get_num_acc_vgprs does not set status.scc if the number of acc vgprs
is 0, so use an and instruction to set the condition code.

The Aldebaran handler binary was not based on the latest version of
the sources, so this update to the binary is the minimal change only
adding two instructions to set the condition code.

A newer version of the handler should be generated and tested in
another commit.

Signed-off-by: Laurent Morichetti <laurent.morichetti@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/pm: Set no fan control flag as needed.

For GPUs that don't support fan control, set the no fan control flag so
that they don't appear in hwmon sensors.

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

drm/amdgpu: bypass hdp read cache invalidation for aldebaran (v2)

hdp read cache is removed in aldebaran. don't issue
an mmio write or write data packet to hardware.

v2: rebase

Signed-off-by: Hawking Zhang <Hawking.Zhang@amd.com>
Reviewed-by: Feifei Xu <Feifei.Xu@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: Aldebaran doesn't use semaphore

Simplify all Aldebaran DIDs into one ASIC type.

Signed-off-by: Amber Lin <Amber.Lin@amd.com>
Reviewed-by: Kevin Wang <kevin1.wang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: UTLC1 RB SDMA timeout on Aldebaran

[Why]
This causes infinite retries on the UTCL1 RB, preventing
higher priority RB such as paging RB.

[How]
Set to one the SDMAx_UTLC1_TIMEOUT registers for all SDMAs.

Signed-off-by: Alex Sierra <alex.sierra@amd.com>
Reviewed-by: Felix Kuehling <Felix.Kuehling@amd.com>
Reviewed-by: Hawking Zhang <Hawking.Zhang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdpgu: add ATOM_DGPU_VRAM_TYPE_HBM2E vram type

0x61 is assigned to HBM2E in atom_dgpu_vram_type.

Signed-off-by: Feifei Xu <Feifei.Xu@amd.com>
Reviewed-by: Hawking Zhang <Hawking.Zhang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: retire aldebaran gpu_info firmware

driver should use the gfx_info atomfirmware interface

Signed-off-by: Hawking Zhang <Hawking.Zhang@amd.com>
Reviewed-by: Feifei Xu <Feifei.Xu@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: query aldebaran gfx_config through atomfirmware i/f

For ASICs that don't support ip discovery feature, query
gfx configuration through atomfirmware interface, rather
than gpu_info firmware.

Signed-off-by: Hawking Zhang <Hawking.Zhang@amd.com>
Reviewed-by: Feifei Xu <Feifei.Xu@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/pm: Remove CPU virtual address notification in aldebaran

PPSMC_MSG_SetSystemVirtualDramAddrHigh/Low messages are not handled by
PMFW in aldebaran

Signed-off-by: Lijo Lazar <Lijo.Lazar@amd.com>
Reviewed-by: Kenneth Feng <Kenneth.Feng@amd.com>
Reviewed-by: Feifei Xu <Feifei.Xu@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/pm: Add support to override pptable id for aldebaran

Temporarily force to use BU PPTable defined in VBIOS. Add support to
override PPTable defined by module parameter.Add FW reported version to
kernel log.

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

drm/amd/amdgpu: Add smu_pptable module parameter

Temporarily add smu_pptable module parameter for aldebaran.This is used
to force soft PPTable use overriding any VBIOS PPTable.

Signed-off-by: Lijo Lazar <Lijo.Lazar@amd.com>
Reviewed-by: Kenneth Feng <kenneth.feng@amd.com>
Reviewed-by: Kevin Wang <kevin1.wang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/pm: Add atom_smc_dpm_info_v4_10 for aldebaran

Add atom_smc_dpm_info_v4_10 that defines board parameters for aldebaran

Signed-off-by: Lijo Lazar <Lijo.Lazar@amd.com>
Reviewed-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/amdgpu: Don't do FB resize under A+A config

Disable PCIe BAR resizing on A+A config. It's not needed because we won't use the
PCIe BAR, but it breaks the PCI BAR configuration with the current SBIOS.

Error message of FB BAR resize failure under A+A:

[  154.913731] [drm:amdgpu_device_resize_fb_bar [amdgpu]] *ERROR* Problem resizing BAR0 (-22).

Signed-off-by: Oak Zeng <Oak.Zeng@amd.com>
Reviewed-by: Amber Lin <Amber.Lin@amd.com>
Reviewed-by: Felix Kuehling <Felix.kuehling@amd.com>
Reviewed-by: Christian Koenig <Christian.Koenig@amd.com>
Tested-by: Amber Lin <Amber.Lin@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: pre-map device buffer as cached for A+A config

For A+A configuration, device memory is supposed to be mapped as
cachable from CPU side. For kernel pre-map gpu device memory using
ioremap_cache

Signed-off-by: Oak Zeng <Oak.Zeng@amd.com>
Reviewed-by: Christian Koenig <Christian.Koenig@amd.com>
Tested-by: Amber Lin <Amber.Lin@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: update atom_firmware_info_v3_4 (v2)

v1: Added some pspbl parameters
v2: fix fallthrough issue

Signed-off-by: Feifei Xu <Feifei.Xu@amd.com>
Reviewed-by: Hawking Zhang <Hawking.Zhang@amd.com>
Reviewed-by: Kevin Wang <kevin1.wang@amd.com>
Reviewed-by: Lazar Lijo <Lijo.Lazar@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/pm:add aldebaran support for getting bootup values

for SMU config.

Signed-off-by: Feifei Xu <Feifei.Xu@amd.com>
Reviewed-by: Hawking Zhang <Hawking.Zhang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: disallow use semaphore on aldebaran

shall revisit the change later

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

drm/amdgpu: switch to vega20 ih block for aldebaran

replace vega10 ih block with vega20 ih block for
aldebaran.

Signed-off-by: Hawking Zhang <Hawking.Zhang@amd.com>
Acked-by: Christian König <christian.koenig@amd.com>
Acked-by: Felix Kuehling <Felix.Kuehling@amd.com>
Reviewed-by: Dennis Li <Dennis.Li@amd.com>
Reviewed-by: Feifei Xu <Feifei.Xu@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: correct IH_CHICKEN programming for aldebaran

For aldebaran, psp firmware won't program IH_CHICKEN.
it now depends on driver to program it properly so
either bus address or gpu virtual address is just
working for ih ring.

Signed-off-by: Hawking Zhang <Hawking.Zhang@amd.com>
Acked-by: Christian König <christian.koenig@amd.com>
Acked-by: Felix Kuehling <Felix.Kuehling@amd.com>
Reviewed-by: Dennis Li <Dennis.Li@amd.com>
Reviewed-by: Feifei Xu <Feifei.Xu@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: add mmhub error status query callback for aldebaran

The callback will be invoked to query mmea error
status when needed.

Signed-off-by: Hawking Zhang <Hawking.Zhang@amd.com>
Reviewed-by: Dennis Li<Dennis.Li@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: add mmhub ras error reset callback for aldebaran

The callback will be invoked to reset mmhub ras error
counters when needed.

Signed-off-by: Hawking Zhang <Hawking.Zhang@amd.com>
Reviewed-by: Dennis Li<Dennis.Li@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: add mmhub ras error query callback for aldebaran

The callback will be invoked to harvest all kinds
of mmhub ras error

Signed-off-by: Hawking Zhang <Hawking.Zhang@amd.com>
Reviewed-by: Dennis Li<Dennis.Li@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: add sdma ras error reset callback for aldebaran

The callback will be invoked to reset sdma ras error
counters when needed.

Signed-off-by: Hawking Zhang <Hawking.Zhang@amd.com>
Reviewed-by: Dennis Li<Dennis.Li@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: add sdma ras error query callback for aldebaran

The callback will be invoked to harvest all kinds
of sdma ras error

Signed-off-by: Hawking Zhang <Hawking.Zhang@amd.com>
Reviewed-by: Dennis Li<Dennis.Li@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: add sdma v4_4 ras function

sdma ras function is the main structure to support
sdma ras on aldebaran. the patch initializes late_init
late_fini callbacks.

Signed-off-by: Hawking Zhang <Hawking.Zhang@amd.com>
Reviewed-by: Dennis Li<Dennis.Li@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: apply sdma golden settings for aldebaran

perform one-time initialization for sdma registers

Signed-off-by: Hawking Zhang <Hawking.Zhang@amd.com>
Reviewed-by: Kevin Wang <kevin1.wang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: use physical_node_id to calculate aper_base

Similar as xgmi connected gpu nodes, physical_node_id
* segment_size should be used to calculate the offset
of aper_base.

The asic type check is redundant. once physical_node_id
and segment_size are initialized, it should be count
on.

Signed-off-by: Hawking Zhang <Hawking.Zhang@amd.com>
Reviewed-by: Kevin Wang <kevin1.wang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: skip gds ras workaround for aldebaran

there won't be any gds useage in either kernel or
pm4 anymore for aldebaran.

Signed-off-by: Hawking Zhang <Hawking.Zhang@amd.com>
Reviewed-by: Feifei Xu <Feifei.Xu@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: init gds for aldebaran

aldebaran removed gds internal memory for atomic usage.
it only supports gws opcode in kernel like barrier,
semaphore.etc. there won't be usage of gds in either
kernel or pm4 packet. max_wave_id should also be marked
as deprecated for aldebaran.

Signed-off-by: Hawking Zhang <Hawking.Zhang@amd.com>
Reviewed-by: Feifei Xu <Feifei.Xu@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: correct vram_info for HBM2E

correct atom_vram_info_header_v2_6 and its vram_module.

Signed-off-by: Feifei Xu <Feifei.Xu@amd.com>
Reviewed-by: Hawking Zhang <Hawking.Zhang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: support get_vram_info atomfirmware i/f for aldebaran

Query vram_type, channel_num, channel_width
information through atomfirmware i/f

Signed-off-by: Hawking Zhang <Hawking.Zhang@amd.com>
Reviewed-by: Feifei Xu <Feifei.Xu@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu:return true for mode1_reset_support on aldebaran

Will remove once validation finished.

Signed-off-by: Feifei Xu <Feifei.Xu@amd.com>
Reviewed-by: Hawking Zhang <Hawking.Zhang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu:add smu mode1/2 support for aldebaran

Use MSG_GfxDriverReset for mode reset and retire MSG_Mode1Reset.
Centralize soc15_asic_mode1_reset() and nv_asic_mode1_reset()functions.
Add mode2_reset_is_support() for smu->ppt_funcs.

Signed-off-by: Feifei Xu <Feifei.Xu@amd.com>
Reviewed-by: Hawking Zhang <Hawking.Zhang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: Add DID for aldebaran

Add 0x7408,0x740C,0x740F in pciidlist.

Signed-off-by: Feifei Xu <Feifei.Xu@amd.com>
Reviewed-by: Hawking Zhang <Hawking.Zhang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>