For some reason this file isn't using the appropriate register
headers for DCN headers, which means that on DCN2 we're getting
the VIEWPORT_DIMENSION offset wrong.
This means that we're not correctly carving out the framebuffer
memory correctly for a framebuffer allocated by EFI and
therefore see corruption when loading amdgpu before the display
driver takes over control of the framebuffer scanout.
Fix this by checking the DCE_HWIP and picking the correct offset
accordingly.
Long-term we should expose this info from DC as GMC shouldn't
need to know about DCN registers.
Cc: stable@vger.kernel.org
Signed-off-by: Harry Wentland <harry.wentland@amd.com>
Reviewed-by: Huang Rui <ray.huang@amd.com>
Acked-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Get rid off pin/unpin of gart BO at resume/suspend and
instead pin only once and try to recover gart content
at resume time. This is much more stable in case there
is OOM situation at 2nd call to amdgpu_device_evict_resources()
while evicting GART table.
v3: remove gart recovery from other places
v2: pin gart at amdgpu_gart_table_vram_alloc()
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Nirmoy Das <nirmoy.das@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Driver needs to call get_xgmi_info() before ip_init
to determine whether it needs to handle a pending hive reset.
Signed-off-by: Victor Skvortsov <victor.skvortsov@amd.com>
Reviewed-by: David Nieto <david.nieto@amd.com>
Reviewed by: shaoyun.liu <Shaoyun.lui@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Modify GC register access from MMIO to RLCG if the
indirect flag is set
v2: Replaced ternary operator with if-else for better
readability
Signed-off-by: Victor Skvortsov <victor.skvortsov@amd.com>
Reviewed-by: David Nieto <david.nieto@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
[Why]
gmc bo will be pinned during loading amdgpu and reset in SRIOV while
only unpinned in unload amdgpu
[How]
add amdgpu_in_reset and sriov judgement to skip pin bo
v2: fix wrong judgement
Signed-off-by: Jingwen Chen <Jingwen.Chen2@amd.com>
Reviewed-by: Horace Chen <horace.chen@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Pair the operations did in GMC ->hw_init and ->hw_fini. That
can help to maintain correct cached state for GMC and avoid
unintention gate operation dropping due to wrong cached state.
BugLink: https://gitlab.freedesktop.org/drm/amd/-/issues/1828
Signed-off-by: Evan Quan <evan.quan@amd.com>
Acked-by: Guchun Chen <guchun.chen@amd.com>
Reviewed-by: Mario Limonciello <mario.limonciello@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
return value form directly instead of
taking this in another redundant variable.
Reported-by: Zeal Robot <zealci@zte.com.cm>
Signed-off-by: chiminghao <chi.minghao@zte.com.cn>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
in case they are not avaiable in early phase
Signed-off-by: Hawking Zhang <Hawking.Zhang@amd.com>
Reviewed-by: Le Ma <Le.Ma@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
IH ring1 is used to process GPU retry fault, overflow is enabled to
drain retry fault because we want receive other interrupts while
handling retry fault to recover range. There is no overflow flag set
when wptr pass rptr. Use timestamp of rptr and wptr to handle overflow
and drain retry fault.
If fault timestamp goes backward, the fault is filtered and should not
be processed. Drain fault is finished if processed_timestamp is equal to
or larger than checkpoint timestamp.
Add amdgpu_ih_functions interface decode_iv_ts for different chips to
get timestamp from IV entry with different iv size and timestamp offset.
amdgpu_ih_decode_iv_ts_helper is used for vega10, vega20, navi10.
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>
Use IP versions rather than asic_type to differentiate
IP version specific features.
Acked-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
gmc_v{9,10}_0_gart_disable() isn't called matched with
correspoding gart_enbale function in SRIOV case. This will
lead to gart.bo pin_count leak on driver unload.
Cc: Hawking Zhang <Hawking.Zhang@amd.com>
Signed-off-by: Leslie Shi <Yuliang.Shi@amd.com>
Signed-off-by: Guchun Chen <guchun.chen@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Check range access permission to restore GPU retry fault, if GPU retry
fault on address which belongs to VMA, and VMA has no read or write
permission requested by GPU, failed to restore the address. The vm fault
event will pass back to user space.
Signed-off-by: Philip Yang <Philip.Yang@amd.com>
Reviewed-by: Felix Kuehling <Felix.Kuehling@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Add MCA specific IP blocks targetting RAS features
Reviewed-by: Hawking Zhang <Hawking.Zhang@amd.com>
Signed-off-by: John Clements <john.clements@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Fix the channel_index table layout to fetch the correct
channel_index when calculating physical address from
normalized address during page retirement.
Also, fix the number of UMC instances and number of channels
within each UMC instance for Aldebaran.
Signed-off-by: Mukul Joshi <mukul.joshi@amd.com>
Reviewed-By: John Clements <john.clements@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Use correct channel and instance values
Reviewed-by: Hawking Zhang <Hawking.Zhang@amd.com>
Signed-off-by: John Clements <john.clements@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
support umc ras function initialization for aldebaran
v2: squash in compile fix
Signed-off-by: John Clements <john.clements@amd.com>
Signed-off-by: Stanley.Yang <Stanley.Yang@amd.com>
Reviewed-by: Guchun Chen <guchun.chen@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
host driver programmed fb location registers for vf, no need to
check anymore.
Signed-off-by: Zhigang Luo <zhigang.luo@amd.com>
Reviewed-by: Hawking Zhang <Hawking.Zhang@amd.com>
Reviewed-By : Shaoyun.liu <shaoyunl@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Problem:
Handle all DMA IOMMU group related dependencies before the
group is removed. Those manifest themself in that when IOMMU
enabled DMA map/unmap is dependent on the presence of IOMMU
group the device belongs to but, this group is released once
the device is removed from PCI topology.
Fix:
Expedite all such unmap operations to pci remove driver callback.
v5: Drop IOMMU notifier and switch to lockless call to ttm_tt_unpopulate
v6: Drop the BO unamp list
v7:
Drop amdgpu_gart_fini
In amdgpu_ih_ring_fini do uncinditional check (!ih->ring)
to avoid freeing uniniitalized rings.
Call amdgpu_ih_ring_fini unconditionally.
v8: Add deatiled explanation
Signed-off-by: Andrey Grodzovsky <andrey.grodzovsky@amd.com>
Reviewed-by: Felix Kuehling <Felix.Kuehling@amd.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20210517143851.475058-1-andrey.grodzovsky@amd.com
Only clear RAS error counters if perestent EDC harvesting is not supported
Reviewed-by: Hawking Zhang <Hawking.Zhang@amd.com>
Signed-off-by: John Clements <john.clements@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Rename,
ras_hw_supported --> ras_hw_enabled, and
ras_features --> ras_enabled,
to show that ras_enabled is a subset of
ras_hw_enabled, which itself is a subset
of the ASIC capability.
Cc: Alexander Deucher <Alexander.Deucher@amd.com>
Cc: John Clements <john.clements@amd.com>
Cc: Hawking Zhang <Hawking.Zhang@amd.com>
Signed-off-by: Luben Tuikov <luben.tuikov@amd.com>
Acked-by: Christian König <christian.koenig@amd.com>
Reviewed-by: John Clements <John.Clements@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
add hdp block ras error query and reset support in
amdgpu ras error count query and reset interface
Signed-off-by: Hawking Zhang <Hawking.Zhang@amd.com>
Reviewed-by: John Clements <John.Clements@amd.com>
Reviewed-by: Dennis Li <Dennis.Li@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
hdp v4_0 support ras features
Signed-off-by: Hawking Zhang <Hawking.Zhang@amd.com>
Reviewed-by: John Clements <John.Clements@amd.com>
Reviewed-by: Dennis Li <Dennis.Li@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Aldebaran has a hw fix so no longer requires the workaround.
Reviewed-by: Hawking Zhang <Hawking.Zhang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Use amdgpu_gmc_vram_pa and amdgpu_gmc_vram_cpu_pa
to simplify codes. No logic change.
Signed-off-by: Oak Zeng <Oak.Zeng@amd.com>
Signed-off-by: Harish Kasiviswanathan <Harish.Kasiviswanathan@amd.com>
Acked-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
mmhub ras is always owned by gpu driver. ras_funcs
initialization shall be done at ip level, instead of
putting it in common gmc interface file
Signed-off-by: Hawking Zhang <Hawking.Zhang@amd.com>
Reviewed-by: Lijo Lazar <lijo.lazar@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
mmhub ras is only avaiable in cerntain mmhub ip
generation.
Signed-off-by: Hawking Zhang <Hawking.Zhang@amd.com>
Reviewed-by: Dennis Li <Dennis.Li@amd.com>
Reviewed-by: John Clements <John.Clements@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
df/mca ras is not managed by gpu driver when gpu
is connected to cpu through xgmi. gpu driver should
register x86 mca notifier for umc ras error
notification
Signed-off-by: Hawking Zhang <Hawking.Zhang@amd.com>
Reviewed-by: Dennis Li <Dennis.Li@amd.com>
Reviewed-by: John Clements <John.Clements@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
umc ras is not managed by gpu driver when gpu is
connected to cpu through xgmi. split umc callbacks
into ras and non-ras ones so gpu driver only
initializes umc ras callbacks when it manages
umc ras.
Signed-off-by: Hawking Zhang <Hawking.Zhang@amd.com>
Reviewed-by: Dennis Li <Dennis.Li@amd.com>
Reviewed-by: John Clements <John.Clements@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
xgmi ras is not managed by gpu driver when gpu is
connected to cpu through xgmi. move all xgmi ras
functions to xgmi_ras_funcs so gpu driver only
initializes xgmi ras functions when it manages
xgmi ras.
Signed-off-by: Hawking Zhang <Hawking.Zhang@amd.com>
Reviewed-by: Dennis Li <Dennis.Li@amd.com>
Reviewed-by: John Clements <John.Clements@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Cleanup pdb0 bo before bo_fini gets called
Signed-off-by: Lijo Lazar <lijo.lazar@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Hawking Zhang <Hawking.Zhang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
The usage of in_interrupt() in gmc_v*_process_interrupt() is intended to
use a different code path if invoked from the interrupt handler vs
invoked from the workqueue.
The usage of in_interrupt() in drivers is phased out and Linus clearly
requested that code which changes behaviour depending on context should
either be separated or the context be conveyed in an argument passed by the
caller, which usually knows the context.
gmc_v*_process_interrupt() is invoked via the ->process() callback
from amdgpu_ih_process() which in turn is invoked either from
amdgpu_irq_handler() (the interrupt handler) or from
amdgpu_irq_handle_*() which is a workqueue.
amdgpu_irq::ih is always processed from the interrupt handler, the other
three struct amdgpu_ih_ring members are processed from a workqueue.
Replace the in_interrupt() check with a comparison against adev->irq.ih.
A similar check is already done to check if the ih pointer is from
ih_soft.
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
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>
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>
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>
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>
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>
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>
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>
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>
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>
Add more function pointers to amdgpu_mmhub_funcs. ASIC specific
implementation of most mmhub functions are called from a general
function pointer, instead of calling different function for
different ASIC.
V2: Split patch into upstreamable and aldebaran
Signed-off-by: Oak Zeng <Oak.Zeng@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Reviewed-by: Rajneesh Bhardwaj <rajneesh.bhardwaj@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
To support new cache coherence HW on A+A platform mainly in KFD.
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>
Aldebaran should be the same as Arcturus in the PTE SNOOPED bit handling.
Signed-off-by: Yong Zhao <Yong.Zhao@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
This applies to AMD Accelerated Processing Platforms that support host
gpu interconnect throguh a special link (xgmi). Aldebaran systems will
support this special feature for utilizing the benefits of host-gpu
cache coherence. This change outlines the basic framework for mapping
the GPU VRAM (HBM) to system address space making it accesible to the
host but managed by the amdgpu driver since this region is marked as
reserved memory in host address space by the underlying system firmware.
v2: switch to smuio callback function to check the type
of host-gpu interface (Hawking)
v3: use hub callbacks rather than direct function calls (Alex)
Reviewed-by: Oak Zeng <oak.zeng@amd.com>
Reviewed-by: Hawking Zhang <Hawking.Zhang@amd.com>
Signed-off-by: Rajneesh Bhardwaj <rajneesh.bhardwaj@amd.com>
Signed-off-by: Hawking Zhang <Hawking.Zhang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
This gives more information and improves productivity.
Signed-off-by: Yong Zhao <Yong.Zhao@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Add gfx memory controller support
Signed-off-by: Le Ma <le.ma@amd.com>
Reviewed-by: Hawking Zhang <Hawking.Zhang@amd.com>
Acked-by: Evan Quan <evan.quan@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
The hw interface changed on arcturus so the old numbering
scheme doesn't work.
Acked-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
