We seem to have a case where SDMA will sometimes miss a doorbell
if GFX is entering the powergating state when the doorbell comes in.
To workaround this, we can update the wptr via MMIO, however,
this is only safe because we disallow gfxoff in begin_ring() for
SDMA 5.2 and then allow it again in end_ring().
Enable this workaround while we are root causing the issue with
the HW team.
Bug: https://gitlab.freedesktop.org/drm/amd/-/issues/3440
Tested-by: Friedrich Vock <friedrich.vock@gmx.de>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
1. Use pa_pfn as the radix-tree key index to log
deferred error info.
2. Use local array to store a row of bad pages.
Signed-off-by: YiPeng Chai <YiPeng.Chai@amd.com>
Reviewed-by: Hawking Zhang <Hawking.Zhang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
For unified queue, DPG pause for encoding is done inside VCN firmware,
so there is no need to pause dpg based on ring type in kernel.
For VCN3 and below, pausing DPG for encoding in kernel is still needed.
v2: add more comments
v3: update commit message
Signed-off-by: Boyuan Zhang <boyuan.zhang@amd.com>
Acked-by: Alex Deucher <alexander.deucher@amd.com>
Reviewed-by: Ruijing Dong <ruijing.dong@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Increase the KMS minor version to indicate GFX12 DCC support since this
contains a major change in how DCC is managed across IPs like GFX, DCN
etc. This will be used mainly by userspace like Mesa to figure out
DCC support on GFX12 hardware.
v2: fix version number (Alex)
Signed-off-by: Aurabindo Pillai <aurabindo.pillai@amd.com>
Acked-by: Alex Deucher <alexander.deucher@amd.com>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
When the `power_saving_policy` property is set to bit mask
"Require color accuracy" ABM should be disabled immediately and
any requests by sysfs to update will return an -EBUSY error.
When the `power_saving_policy` property is set to bit mask
"Require low latency" PSR should be disabled.
When the property is restored to an empty bit mask ABM and PSR
can be enabled again.
Signed-off-by: Mario Limonciello <mario.limonciello@amd.com>
Reviewed-by: Leo Li <sunpeng.li@amd.com>
Signed-off-by: Hamza Mahfooz <hamza.mahfooz@amd.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20240703051722.328-3-mario.limonciello@amd.com
The problem case is as follows:
1. GPU A triggers a gpu ras reset, and GPU A drives
GPU B to also perform a gpu ras reset.
2. After gpu B ras reset started, gpu B queried a DE
data. Since the DE data was queried in the ras reset
thread instead of the page retirement thread, bad
page retirement work would not be triggered. Then
even if all gpu resets are completed, the bad pages
will be cached in RAM until GPU B's bad page retirement
work is triggered again and then saved to eeprom.
This patch can save the bad pages to eeprom in time after gpu
ras reset is completed.
v2:
1. Add the above description to code comments.
2. Reuse existing function.
Signed-off-by: YiPeng Chai <YiPeng.Chai@amd.com>
Reviewed-by: Tao Zhou <tao.zhou1@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Before uninstalling gpu driver, flush all cached ras
bad pages to eeprom.
v2:
Put the same code into a function and reuse the function.
Signed-off-by: YiPeng Chai <YiPeng.Chai@amd.com>
Reviewed-by: Tao Zhou <tao.zhou1@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
GFX ME right now is one but this could change in
future SOC's. Use no of ME for GFX as start point
for ME for compute for GFX12.
Signed-off-by: Sunil Khatri <sunil.khatri@amd.com>
Acked-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
GFX ME right now is one but this could change in
future SOC's. Use no of ME for GFX as start point
for ME for compute for GFX11.
Signed-off-by: Sunil Khatri <sunil.khatri@amd.com>
Acked-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
GFX ME right now is one but this could change in
future SOC's. Use no of ME for GFX as start point
for ME for compute for GFX10.
Signed-off-by: Sunil Khatri <sunil.khatri@amd.com>
Acked-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
For SOCs with GFX v9.4.3, a VF may have multiple compute partitions.
Fetch the partition information during init and initialize partition
nodes. There is no support to switch partition mode in VF mode, hence
disable the same.
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>
sdma has 2 instances in SRIOV cpx mode. Odd numbered VFs have
sdma0/sdma1 instances. Even numbered vfs have sdma2/sdma3. For
Even numbered vfs, the sdma2 & sdma3 (irq srouce id
CLIENTID_SDMA2 and CLIENTID_SDMA3) should map to irq seq 0 & 1.
Signed-off-by: Gavin Wan <Gavin.Wan@amd.com>
Reviewed-by: Hawking Zhang <Hawking.Zhang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
To address the problem with hitches moving when bulk move
sublists are lru-bumped, register the list cursors with the
ttm_lru_bulk_move structure when traversing its list, and
when lru-bumping the list, move the cursor hitch to the tail.
This also means it's mandatory for drivers to call
ttm_lru_bulk_move_init() and ttm_lru_bulk_move_fini() when
initializing and finalizing the bulk move structure, so add
those calls to the amdgpu- and xe driver.
Compared to v1 this is slightly more code but less fragile
and hopefully easier to understand.
Changes in previous series:
- Completely rework the functionality
- Avoid a NULL pointer dereference assigning manager->mem_type
- Remove some leftover code causing build problems
v2:
- For hitch bulk tail moves, store the mem_type in the cursor
instead of with the manager.
v3:
- Remove leftover mem_type member from change in v2.
v6:
- Add some lockdep asserts (Matthew Brost)
- Avoid NULL pointer dereference (Matthew Brost)
- No need to check bo->resource before dereferencing
bo->bulk_move (Matthew Brost)
Cc: Christian König <christian.koenig@amd.com>
Cc: Somalapuram Amaranath <Amaranath.Somalapuram@amd.com>
Cc: Matthew Brost <matthew.brost@intel.com>
Cc: <dri-devel@lists.freedesktop.org>
Signed-off-by: Thomas Hellström <thomas.hellstrom@linux.intel.com>
Reviewed-by: Matthew Brost <matthew.brost@intel.com>
Acked-by: Christian König <christian.koenig@amd.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20240705153206.68526-5-thomas.hellstrom@linux.intel.com
Signed-off-by: Christian König <christian.koenig@amd.com>
to avoid reading wrong WPTR from doorbell in sriov vf, set
CP_HQD_PQ_DOORBELL_CONTROL.DOORBELL_MODE to 1 to read WPTR from MQD.
Signed-off-by: Zhigang Luo <Zhigang.Luo@amd.com>
Acked-by: Lijo Lazar <lijo.lazar@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
v1:
- use unified event id to manage ras events
- add a new function amdgpu_ras_query_error_status_with_event() to accept
event type as parameter.
v2:
add a warn log to show the location of function failure
when calling amdgpu_ras_mark_event(). (Tao Zhou)
v3:
change RAS_EVENT_TYPE_ISR to RAS_EVENT_TYPE_FATAL.
v4:
rename amdgpu_ras_get_recovery_event() to
amdgpu_ras_get_fatal_error_event().
Signed-off-by: Yang Wang <kevinyang.wang@amd.com>
Reviewed-by: Tao Zhou <tao.zhou1@amd.com>
Reviewed-by: Hawking Zhang <Hawking.Zhang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
remove redundant semicolons in RAS_EVENT_LOG to avoid
code format check warning.
Fixes: b712d7c201 ("drm/amdgpu: fix compiler 'side-effect' check issue for RAS_EVENT_LOG()")
Signed-off-by: Yang Wang <kevinyang.wang@amd.com>
Reviewed-by: Tao Zhou <tao.zhou1@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
While moving buffer which has dcc tiling config, it is needed to restore
its original dcc tiling.
1. extend copy flag to cover tiling bits
2. add logic to restore original dcc tiling config
Signed-off-by: Frank Min <Frank.Min@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>
Enable redirection of irq for pagefaults for specific
clients to avoid overflow without dropping interrupts.
So here we redirect the interrupts to another IH ring
i.e ring1 where only these interrupts are processed.
Signed-off-by: Sunil Khatri <sunil.khatri@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
We need IH ring1 for handling the pagefault
interrupts which over flow in default
ring for specific usecases.
Enable ring1 allows software to redirect
high interrupts to ring1 from default IH
ring.
Signed-off-by: Sunil Khatri <sunil.khatri@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
[Why]
If VF request full GPU access and the request failed,
the VF driver can get stuck accessing registers for an extended period during
the unload of KMS.
[How]
Set no_hw_access flag when VF request for full GPU access fails
This prevents further hardware access attempts, avoiding the prolonged
stuck state.
Signed-off-by: Yifan Zha <Yifan.Zha@amd.com>
Acked-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
