upages is assigned under cpages = 0, so it isn't really used in this function.
Signed-off-by: James Zhu <James.Zhu@amd.com>
Reviewed-by: Philip.Yang<Philip.Yang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Since KFD proc content was moved to kernel debugfs, we can't destroy KFD
debugfs before kfd_process_destroy_wq. Move kfd_process_destroy_wq prior
to kfd_debugfs_fini to fix a kernel NULL pointer problem. It happens
when /sys/kernel/debug/kfd was already destroyed in kfd_debugfs_fini but
kfd_process_destroy_wq calls kfd_debugfs_remove_process. This line
debugfs_remove_recursive(entry->proc_dentry);
tries to remove /sys/kernel/debug/kfd/proc/<pid> while
/sys/kernel/debug/kfd is already gone. It hangs the kernel by kernel
NULL pointer.
Signed-off-by: Amber Lin <Amber.Lin@amd.com>
Reviewed-by: Eric Huang <jinhuieric.huang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
GPUs with multi-xcc have multiple MQDs per queue. This patch saves and
restores all the MQDs within the partition.
Signed-off-by: David Yat Sin <David.YatSin@amd.com>
Reviewed-by: Felix Kuehling <felix.kuehling@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
(cherry picked from commit a578f2a58c)
Cc: stable@vger.kernel.org
GPUs with multi-xcc have multiple MQDs per queue. This patch saves and
restores all the MQDs within the partition.
Signed-off-by: David Yat Sin <David.YatSin@amd.com>
Reviewed-by: Felix Kuehling <felix.kuehling@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Previously we tried to keep the HW specific PTE flags in each mapping,
but for CRIU that isn't sufficient any more since the original value is
needed for the checkpoint procedure.
So rework the whole handling, nuke the early mapping function, keep the
UAPI flags in each mapping instead of the HW flags and translate them to
the HW flags while filling in the PTEs.
Only tested on Navi 23 for now, so probably needs quite a bit of more
work.
v2: fix KFD and SVN handling
v3: one more SVN fix pointed out by Felix
v4: squash in gfx12 fix from David
Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Felix Kuehling <felix.kuehling@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
KFD has been confirmed that can run on LoongArch systems.
It's necessary to support CONFIG_HSA_AMD on LoongArch.
Signed-off-by: Han Gao <rabenda.cn@gmail.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>
Add a parameter to amdgpu_sdma_reset_engine() to let the
caller handle the kernel rings. This allows the kernel
rings to back up their unprocessed state if the reset comes in
via the drm scheduler rather than KFD.
Reviewed-by: Jesse Zhang <Jesse.Zhang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
If ring reset is disabled, skip resetting queues. Instead, fall back to
device based reset.
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>
If the process is exiting, the mmput inside mmu notifier callback from
compactd or fork or numa balancing could release the last reference
of mm struct to call exit_mmap and free_pgtable, this triggers deadlock
with below backtrace.
The deadlock will leak kfd process as mmu notifier release is not called
and cause VRAM leaking.
The fix is to take mm reference mmget_non_zero when adding prange to the
deferred list to pair with mmput in deferred list work.
If prange split and add into pchild list, the pchild work_item.mm is not
used, so remove the mm parameter from svm_range_unmap_split and
svm_range_add_child.
The backtrace of hung task:
INFO: task python:348105 blocked for more than 64512 seconds.
Call Trace:
__schedule+0x1c3/0x550
schedule+0x46/0xb0
rwsem_down_write_slowpath+0x24b/0x4c0
unlink_anon_vmas+0xb1/0x1c0
free_pgtables+0xa9/0x130
exit_mmap+0xbc/0x1a0
mmput+0x5a/0x140
svm_range_cpu_invalidate_pagetables+0x2b/0x40 [amdgpu]
mn_itree_invalidate+0x72/0xc0
__mmu_notifier_invalidate_range_start+0x48/0x60
try_to_unmap_one+0x10fa/0x1400
rmap_walk_anon+0x196/0x460
try_to_unmap+0xbb/0x210
migrate_page_unmap+0x54d/0x7e0
migrate_pages_batch+0x1c3/0xae0
migrate_pages_sync+0x98/0x240
migrate_pages+0x25c/0x520
compact_zone+0x29d/0x590
compact_zone_order+0xb6/0xf0
try_to_compact_pages+0xbe/0x220
__alloc_pages_direct_compact+0x96/0x1a0
__alloc_pages_slowpath+0x410/0x930
__alloc_pages_nodemask+0x3a9/0x3e0
do_huge_pmd_anonymous_page+0xd7/0x3e0
__handle_mm_fault+0x5e3/0x5f0
handle_mm_fault+0xf7/0x2e0
hmm_vma_fault.isra.0+0x4d/0xa0
walk_pmd_range.isra.0+0xa8/0x310
walk_pud_range+0x167/0x240
walk_pgd_range+0x55/0x100
__walk_page_range+0x87/0x90
walk_page_range+0xf6/0x160
hmm_range_fault+0x4f/0x90
amdgpu_hmm_range_get_pages+0x123/0x230 [amdgpu]
amdgpu_ttm_tt_get_user_pages+0xb1/0x150 [amdgpu]
init_user_pages+0xb1/0x2a0 [amdgpu]
amdgpu_amdkfd_gpuvm_alloc_memory_of_gpu+0x543/0x7d0 [amdgpu]
kfd_ioctl_alloc_memory_of_gpu+0x24c/0x4e0 [amdgpu]
kfd_ioctl+0x29d/0x500 [amdgpu]
Fixes: fa582c6f36 ("drm/amdkfd: Use mmget_not_zero in MMU notifier")
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>
Fix the following compile time warning when building with W=1:
warning: EXPORT_SYMBOL() is not used, but #include <linux/export.h> is present
Acked-by: Christian König <christian.koenig@amd.com>
Signed-off-by: André Almeida <andrealmeid@igalia.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Set memory mtype to UC host memory when ext-coherent
flag is set and memory is registered as a SVM allocation.
Reviewed-by: Amber Lin <Amber.Lin@amd.com>
Signed-off-by: David Yat Sin <David.YatSin@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
q->gws is not updated atomically with qpd->mapped_gws_queue. If a
runlist is created between pqm_set_gws and update_queue it will
contain a queue which uses GWS in a process with no GWS allocated.
This will result in a scheduler hang.
Use q->properties.is_gws which is changed while holding the DQM lock.
Signed-off-by: Jay Cornwall <jay.cornwall@amd.com>
Reviewed-by: Harish Kasiviswanathan <Harish.Kasiviswanathan@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
For the suspend and resume process, exclusive access is not required.
Therefore, it can be moved out of the full access section to reduce the
duration of exclusive access.
v3:
Move suspend processes before hardware fini.
Remove twice call for bare metal.
v4:
Refine code
Signed-off-by: Emily Deng <Emily.Deng@amd.com>
Acked-by: Lijo Lazar <lijo.lazar@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
The KFD currently bars a compute partition mode switch while a KFD
process exists.
Since cgroup excluded devices remain excluded for the lifetime of a KFD
process and user space is able to mode switch single devices, allow
users to mode switch a device with any running process that has been
cgroup excluded from this device.
Signed-off-by: Jonathan Kim <jonathan.kim@amd.com>
Reviewed-by: Harish Kasiviswanathan <harish.kasiviswanathan@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Relocate the per-SDMA queue reset capability check from
kfd_topology_set_capabilities() to node_show() to ensure we read the
latest value of sdma.supported_reset after all IP blocks are initialized.
Fixes: ceb7114c96 ("drm/amdkfd: flag per-sdma queue reset supported to user space")
Reviewed-by: Jonathan Kim <jonathan.kim@amd.com>
Signed-off-by: Jesse Zhang <Jesse.Zhang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
To avoid a cast when calling drm_dev_wedged_event(), replace pid and
task name inside of struct amdgpu_task_info with struct
drm_wedge_task_info.
Reviewed-by: Christian König <christian.koenig@amd.com>
Link: https://lore.kernel.org/r/20250617124949.2151549-6-andrealmeid@igalia.com
Signed-off-by: André Almeida <andrealmeid@igalia.com>
KFD has been confirmed that can run on RISCV systems. It's necessary to
support CONFIG_HSA_AMD on RISCV.
Signed-off-by: Xuemei Liu <liu.xuemei1@zte.com.cn>
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>
For simulation C models that don't run CP FW where adev->mes.sched_version
is not populated correctly. This causes NULL dereference in
amdgpu_amdkfd_free_gtt_mem(dev->adev, (void **)&pqn->q->wptr_bo_gart)
and warning on unpinned BO in amdgpu_bo_gpu_offset(q->properties.wptr_bo).
Compared with adding version check here and there,
always map wptr BO to GART simplifies things.
v2: Add NULL check in amdgpu_amdkfd_free_gtt_mem.(Philip)
Signed-off-by: Lang Yu <lang.yu@amd.com>
Reviewed-by: Felix Kuehling <Felix.Kuehling@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
This patch is to fix a kfd_prcess ref leak.
Signed-off-by: Yifan Zhang <yifan1.zhang@amd.com>
Reviewed-by: Philip Yang <Philip.Yang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
This patch removes the if/else statement in the
cik_event_interrupt_wq function because it is redundant
with both branches resulting in identical outcomes,
this improves code readibility.
Signed-off-by: Sunday Clement <Sunday.Clement@amd.com>
Reviewed-by: Harish Kasiviswanathan <Harish.Kasiviswanathan@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Static analysis shows that pointer "svms" cannot be NULL because it points
to the object "struct svm_range_list". Remove the extra NULL check. It is
meaningless and harms the readability of the code.
In the function svm_range_get_info() there is no possibility of failure.
Therefore, the caller of the function svm_range_get_info() does not need
a return value. Change the function svm_range_get_info() return type from
"int" to "void".
Since the function svm_range_get_info() has a return type of "void". The
caller of the function svm_range_get_info() does not need a return value.
Delete extra code.
Found by Linux Verification Center (linuxtesting.org) with SVACE.
Signed-off-by: Andrey Vatoropin <a.vatoropin@crpt.ru>
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>
If the MEC firmware supports chaining runlists of XNACK+/XNACK-
processes, set SQ_CONFIG1 chicken bit and SET_RESOURCES bit 28.
When the MEC/HWS supports it, KFD checks the XNACK+/XNACK- processes mix
happens or not. If it does, enter over-subscription.
Signed-off-by: Amber Lin <Amber.Lin@amd.com>
Reviewed-by: Philip Yang <Philip.Yang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Commit ded8b3c36f17 ("drm/amdgpu: properly handle GC vs MM in amdgpu_vmid_mgr_init()")
enables all 16 vmids for MMHUB on GC 10 and newer for KGD since
there are no KFD resources using MMHUB. With this change, KFD
starts seeing MMHUB vmids in it's range with no pasid set. As
such there is no need to warn, we can just ignore those interrupts.
Fixes: aded8b3c36 ("drm/amdgpu: properly handle GC vs MM in amdgpu_vmid_mgr_init()")
Reviewed-by: Felix Kuehling <felix.kuehling@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
If applications unmap the memory before destroying the userptr, it needs
trigger a segfault to notify user space to correct the free sequence in
VM debug mode.
v2: Send gpu access fault to user space
v3: Report gpu address to user space, remove unnecessary params
v4: update pr_err into one line, remove userptr log info
Signed-off-by: Shane Xiao <shane.xiao@amd.com>
Acked-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Felix Kuehling <Felix.Kuehling@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
SDMA doesn't support oversubsciption, it is the user matter to create
queues over HW limit, but not supposed to be a KFD error.
Signed-off-by: Eric Huang <jinhuieric.huang@amd.com>
Reviewed-by: Harish Kasiviswanathan <Harish.Kasiviswanathan@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
the entries will be appearing at
/sys/kernel/debug/kfd/proc/<pid>/pasid_<gpuid>.
Signed-off-by: Eric Huang <jinhuieric.huang@amd.com>
Reviewed-by: Harish Kasiviswanathan <Harish.Kasiviswanathan@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
When submitting MQD to CP, set SDMA_RLCx_IB_CNTL/SWITCH_INSIDE_IB bit so
it'll allow SDMA preemption if there is a massive command buffer of
long-running SDMA commands.
Signed-off-by: Amber Lin <Amber.Lin@amd.com>
Acked-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
rocm-smi with superuser permission doesn't show some
of smi events, i.e. page fault/migration, because the
condition of "(events & all)" is false. Superuser
should be able to detect all events, the condiiton of
"(events & all)" seems redundant, so removing it will
fix the issue.
Signed-off-by: Eric Huang <jinhuieric.huang@amd.com>
Reviewed-by: Kent Russell <kent.russell@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
The mistake will lead to NULL kernel oops, so fix it.
Fixes: 4172b556fd ("drm/amdkfd: add smi events for process start and end")
Signed-off-by: Eric Huang <jinhuieric.huang@amd.com>
Reviewed-by: Kent Russell <kent.russell@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
This patch adds recommended SDMA engines with limited XGMI SDMA engines.
It will help improve overall performance for device to device copies
with this optimization.
v2: Update the formatting issues and data type
Signed-off-by: Shane Xiao <shane.xiao@amd.com>
Suggested-by: Jonathan Kim <jonathan.kim@amd.com>
Reviewed-by: Jonathan Kim <jonathan.kim@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
rocm-smi will be able to show the events for KFD process
start/end, it is the implementation of this feature.
Signed-off-by: Eric Huang <jinhuieric.huang@amd.com>
Reviewed-by: Kent Russell <kent.russell@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Switch from a bool to an enum and allow more options
for enforce isolation. There are now 3 modes of operation:
- Disabled (0)
- Enabled (serialization and cleaner shader) (1)
- Enabled in legacy mode (no serialization or cleaner shader) (2)
This provides better flexibility for more use cases.
Acked-by: Srinivasan Shanmugam <srinivasan.shanmugam@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Remove workaround code for the early engineering
samples GC v9.4.3 SOCs with revID 0
Reviewed-by: Amber Lin <Amber.Lin@amd.com>
Signed-off-by: Apurv Mishra <Apurv.Mishra@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
To get the device context, replace pr_ with dev_ functions.
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>
ASICs post GFX 9 are being flagged as SDMA per queue reset supported
in the KGD but KFD and scheduler FW currently have no support.
Limit SDMA queue reset capabilities to GFX 9.
Fixes: ceb7114c96 ("drm/amdkfd: flag per-sdma queue reset supported to user space")
Signed-off-by: Jonathan Kim <jonathan.kim@amd.com>
Reviewed-by: David Belanger <david.belanger@amd.com>
Reviewed-by: Harish Kasiviswanathan <harish.kasiviswanathan@amd.com>
Reviewed-by: Jesse Zhang <jesse.zhang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Cc: stable@vger.kernel.org
Disable per queue reset for sriov.
Signed-off-by: Emily Deng <Emily.Deng@amd.com>
Reviewed-by: Jonathan Kim <jonathan.kim@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Move the kfd suspend/resume code into the caller. That
is where the KFD is likely to detect a reset so on the KFD
side there is no need to call them. Also add a mutex to
lock the actual reset sequence.
v2: make the locking per instance
Fixes: bac38ca8c4 ("drm/amdkfd: implement per queue sdma reset for gfx 9.4+")
Reviewed-by: Jesse Zhang <jesse.zhang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Clause instructions with precise memory enabled currently hang the
shader so set capabilities flag to disabled since it's unsafe to use
for debugging.
Signed-off-by: Jonathan Kim <jonathan.kim@amd.com>
Tested-by: Lancelot Six <lancelot.six@amd.com>
Reviewed-by: Harish Kasiviswanathan <harish.kasiviswanathan@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
For certain ASICs where dequeue_wait_count don't need to be initialized,
pm_config_dequeue_wait_counts_v9 return without filling in the packet
information. However, the calling function interprets this as a success
and sends the uninitialized packet to firmware causing hang.
Fix the above bug by not calling pm_config_dequeue_wait_counts_v9 for
ASICs that don't need the value to be initialized.
v2: Removed redudant code.
Tidy up code based on review comments
v3: Don't call pm_config_dequeue_wait_counts_v9 for certain ASICs
Fixes: ed962f8d06 ("drm/amdkfd: Add pm_config_dequeue_wait_counts API")
Signed-off-by: Harish Kasiviswanathan <Harish.Kasiviswanathan@amd.com>
Reviewed-by: Jonathan Kim <jonathan.kim@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Always use MTYPE_UC if UNCACHED flag is specified.
This makes kernarg region uncached and it restores
usermode cache disable debug flag functionality.
Do not set MTYPE_UC for COHERENT flag, on GFX12 coherence is handled by
shader code.
Signed-off-by: David Belanger <david.belanger@amd.com>
Reviewed-by: Felix Kuehling <felix.kuehling@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Correct F8_MODE setting for gfx950 that was removed
Fixes: 61972cd93a ("drm/amdkfd: Set per-process flags only once for gfx9/10/11/12")
Signed-off-by: Amber Lin <Amber.Lin@amd.com>
Reviewed-by: Harish Kasiviswanathan <Harish.Kasiviwanathan@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
In sriov multiple vf, Set CP_HQD_PQ_DOORBELL_CONTROL.DOORBELL_MODE to 1 to read WPTR from MQD.
Signed-off-by: Emily Deng <Emily.Deng@amd.com>
Acked-by: Lijo Lazar <lijo.lazar@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
VALU instructions with SGPR source need wait states to avoid hazard
with SALU using different SGPR.
v2: Eliminate some hazards to reduce code explosion
Signed-off-by: Jay Cornwall <jay.cornwall@amd.com>
Reviewed-by: Lancelot Six <lancelot.six@amd.com>
Acked-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Dequeue retry timeout controls the interval between checks for unmet
conditions. On MI series, reduce this from 0x40 to 0x1 (~ 1 uS). The
cost of additional bandwidth consumed by CP when polling memory
shouldn't be substantial.
Signed-off-by: Harish Kasiviswanathan <Harish.Kasiviswanathan@amd.com>
Reviewed-by: Jonathan Kim <jonathan.kim@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Issue:
In the scenario where svm_range_restore_pages is called, but
svm->checkpoint_ts has not been set and the retry fault has not been
drained, svm_range_unmap_from_cpu is triggered and calls svm_range_free.
Meanwhile, svm_range_restore_pages continues execution and reaches
svm_range_from_addr. This results in a "failed to find prange..." error,
causing the page recovery to fail.
How to fix:
Move the timestamp check code under the protection of svm->lock.
v2:
Make sure all right locks are released before go out.
v3:
Directly goto out_unlock_svms, and return -EAGAIN.
v4:
Refine code.
Signed-off-by: Emily Deng <Emily.Deng@amd.com>
Reviewed-by: Felix Kuehling <felix.kuehling@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
[Why]
If reset is detected and kfd need to evict working queues, HWS moving queue will be failed.
Then remaining queues are not evicted and in active state.
After reset done, kfd uses HWS to termination remaining activated queues but HWS is resetted.
So remove queue will be failed again.
[How]
Keep removing all queues even if HWS returns failed.
It will not affect cpsch as it checks reset_domain->sem.
v2: If any queue failed, evict queue returns error.
v3: Declare err inside the if-block.
Reviewed-by: Felix Kuehling <felix.kuehling@amd.com>
Signed-off-by: Yifan Zha <Yifan.Zha@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
These lines are indented one tab more than they should be. Delete
the stray tabs.
Signed-off-by: Dan Carpenter <dan.carpenter@linaro.org>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Update pm_update_grace_period() to more cleaner
pm_config_dequeue_wait_counts(). Previously, grace_period variable was
overloaded as a variable and a macro, making it inflexible to configure
additional dequeue wait times.
pm_config_dequeue_wait_counts() now takes in a cmd / variable. This
allows flexibility to update different dequeue wait times.
Signed-off-by: Harish Kasiviswanathan <Harish.Kasiviswanathan@amd.com>
Reviewed-by: Jonathan Kim <jonathan.kim@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Add support for more per-process flags starting with option to configure
MFMA precision for gfx 9.5
v2: Change flag name to KFD_PROC_FLAG_MFMA_HIGH_PRECISION
Remove unused else condition
v3: Bump the KFD API version
v4: Missed SH_MEM_CONFIG__PRECISION_MODE__SHIFT define. Added it.
Signed-off-by: Harish Kasiviswanathan <Harish.Kasiviswanathan@amd.com>
Reviewed-by: Felix Kuehling <felix.kuehling@amd.com>
Reviewed-by: Amber Lin <Amber.Lin@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Define set_cache_memory_policy() for these asics and move all static
changes from update_qpd() which is called each time a queue is created
to set_cache_memory_policy() which is called once during process
initialization
Signed-off-by: Harish Kasiviswanathan <Harish.Kasiviswanathan@amd.com>
Reviewed-by: Amber Lin <Amber.Lin@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Set per-process static sh_mem config only once during process
initialization. Move all static changes from update_qpd() which is
called each time a queue is created to set_cache_memory_policy() which
is called once during process initialization.
set_cache_memory_policy() is currently defined only for cik and vi
family. So this commit only focuses on these two. A separate commit will
address other asics.
Signed-off-by: Harish Kasiviswanathan <Harish.Kasiviswanathan@amd.com>
Reviewed-by: Amber Lin <Amber.Lin@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Default F8_MODE should be OCP format on gfx950.
Signed-off-by: Alex Sierra <alex.sierra@amd.com>
Reviewed-by: Harish Kasiviswanathan <Harish.Kasiviswanathan@amd.com>
Signed-off-by: Amber Lin <Amber.Lin@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
While the entry get in svm_range_unmap_from_cpu is the last entry, and
the entry is page fault, it also need to be dropped. So for equal case,
it also need to be dropped.
v2:
Only modify the svm_range_restore_pages.
Signed-off-by: Emily Deng <Emily.Deng@amd.com>
Reviewed-by: Xiaogang Chen<xiaogang.chen@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Remove unused declaration of gws_debug_workaround.
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>
Similar to compute queue reset, flag SDMA queue reset capabilities to
user space for safe testing.
Signed-off-by: Jonathan Kim <jonathan.kim@amd.com>
Reviewed-by: Harish Kasiviswanathan <harish.kasiviswanathan@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
To reset hung SDMA queues on GFX 9.4+ for the GFX9 family, a soft reset
must be issued through SMU. Since soft resets will reset an entire SDMA
engine, use a common KGD call to do the reset as the KGD will handle
avoiding a reset of in flight GFX and paging queues on that engine.
In addition, create a common call for all reset types to simplify
the handling of module parameter settings that block gpu resets.
Signed-off-by: Jonathan Kim <jonathan.kim@amd.com>
Reviewed-by: Harish Kasiviswanathan <harish.kasiviswanathan@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Through KFD IOCTL Fuzzing we encountered a NULL pointer derefrence
when calling kfd_queue_acquire_buffers.
Fixes: 629568d25f ("drm/amdkfd: Validate queue cwsr area and eop buffer size")
Signed-off-by: Andrew Martin <Andrew.Martin@amd.com>
Reviewed-by: Philip Yang <Philip.Yang@amd.com>
Signed-off-by: Andrew Martin <Andrew.Martin@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
pqm_get_kernel_queue() has been unused since 2022's
commit 5bdd3eb253 ("drm/amdkfd: Remove unused old debugger
implementation")
Remove it.
Signed-off-by: Dr. David Alan Gilbert <linux@treblig.org>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
debugfs hang_hws is used by GPU reset test with HWS, for MES this crash
the kernel with NULL pointer access because dqm->packet_mgr is not setup
for MES path.
Skip GPU with MES for now, MES hang_hws debugfs interface will be
supported later.
Signed-off-by: Philip Yang <Philip.Yang@amd.com>
Reviewed-by: Kent Russell <kent.russell@amd.com>
Reviewed-by: Felix Kuehling <felix.kuehling@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
If GPU in reset, destroy_queue return -EIO, pqm_destroy_queue should
delete the queue from process_queue_list and free the resource.
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>
If HW scheduler hangs and mode1 reset is used to recover GPU, KFD signal
user space to abort the processes. After process abort exit, user queues
still use the GPU to access system memory before h/w is reset while KFD
cleanup worker free system memory and free VRAM.
There is use-after-free race bug that KFD allocate and reuse the freed
system memory, and user queue write to the same system memory to corrupt
the data structure and cause driver crash.
To fix this race, KFD cleanup worker terminate user queues, then flush
reset_domain wq to wait for any GPU ongoing reset complete, and then
free outstanding BOs.
Signed-off-by: Philip Yang <Philip.Yang@amd.com>
Reviewed-by: Lijo Lazar <lijo.lazar@amd.com>
Reviewed-by: Felix Kuehling <felix.kuehling@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
If waiting for gpu reset done in KFD release_work, thers is WARNING:
possible circular locking dependency detected
#2 kfd_create_process
kfd_process_mutex
flush kfd release work
#1 kfd release work
wait for amdgpu reset work
#0 amdgpu_device_gpu_reset
kgd2kfd_pre_reset
kfd_process_mutex
Possible unsafe locking scenario:
CPU0 CPU1
---- ----
lock((work_completion)(&p->release_work));
lock((wq_completion)kfd_process_wq);
lock((work_completion)(&p->release_work));
lock((wq_completion)amdgpu-reset-dev);
To fix this, KFD create process move flush release work outside
kfd_process_mutex.
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>
With GPU reset-domain worker implemented, KFD hw_exception worker is not
needed any more, just call amdgpu_amdkfd_gpu_reset directly from
kfd_hws_hang.
Suggested-by: Felix Kuehling <felix.kuehling@amd.com>
Signed-off-by: Philip Yang <Philip.Yang@amd.com>
Reviewed-by: Lijo Lazar <lijo.lazar@amd.com>
Reviewed-by: Felix Kuehling <felix.kuehling@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Since kfd uses pasid values from graphic driver now do not need use kfd pasid
fucntions.
Signed-off-by: Xiaogang Chen <xiaogang.chen@amd.com>
Reviewed-by: Felix Kuehling <felix.kuehling@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
If queue size is less than minimum, clamp it to minimum to prevent
underflow when writing queue mqd.
Signed-off-by: David Yat Sin <David.YatSin@amd.com>
Reviewed-by: Jay Cornwall <jay.cornwall@amd.com>
Reviewed-by: Harish Kasiviswanathan <Harish.Kasiviswanathan@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Deprecate KFD XGMI peer info calls in favour of calling directly from
simplified XGMI peer info functions.
Signed-off-by: Jonathan Kim <jonathan.kim@amd.com>
Reviewed-by: Lijo Lazar <lijo.lazar@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Even though GWS no longer exists, to maintain runtime usage for
cooperative launch, SW set legacy GWS size.
Signed-off-by: Jonathan Kim <jonathan.kim@amd.com>
Acked-by: Mukul Joshi <mukul.joshi@amd.com>
Reviewed-by: Harish Kasiviswanathan <harish.kasiviswanathan@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
When userspace applications call AMDKFD_IOC_UPDATE_QUEUE. Preserve
bitfields that do not need to be modified as they contain flags to
track queue states that are used by CP FW.
Signed-off-by: David Yat Sin <David.YatSin@amd.com>
Reviewed-by: Jay Cornwall <jay.cornwall@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
In the kfd_process_device_init_vm function, a valid error code is now
returned when the associated Process Address Space ID (PASID) is not
present.
If the address space virtual memory (avm) does not have an associated
PASID, the function sets the ret variable to -EINVAL before proceeding
to the error handling section. This ensures that the calling function,
such as kfd_ioctl_acquire_vm, can appropriately handle the error,
thereby preventing any issues during virtual memory initialization.
Fixes the below:
drivers/gpu/drm/amd/amdgpu/../amdkfd/kfd_process.c:1694 kfd_process_device_init_vm()
warn: missing error code 'ret'
drivers/gpu/drm/amd/amdgpu/../amdkfd/kfd_process.c
1647 int kfd_process_device_init_vm(struct kfd_process_device *pdd,
1648 struct file *drm_file)
1649 {
...
1690
1691 if (unlikely(!avm->pasid)) {
1692 dev_warn(pdd->dev->adev->dev, "WARN: vm %p has no pasid associated",
1693 avm);
--> 1694 goto err_get_pasid;
ret = -EINVAL?
1695 }
Fixes: 8544374c0f ("drm/amdkfd: Have kfd driver use same PASID values from graphic driver")
Reported by: Dan Carpenter <dan.carpenter@linaro.org>
Cc: Xiaogang Chen <xiaogang.chen@amd.com>
Cc: Felix Kuehling <felix.kuehling@amd.com>
Cc: Christian König <christian.koenig@amd.com>
Cc: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Srinivasan Shanmugam <srinivasan.shanmugam@amd.com>
Reviewed-by: Felix Kuehling <felix.kuehling@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
To workaround queue full h/w issue on Gfx7/8, when application create
AQL queue, the ring buffer bo allocate size is queue_size/2 and
map queue_size ring buffer to GPU in 2 pieces using 2 attachments, each
attachment map size is queue_size/2, with same ring_bo backing memory.
For Gfx7/8, user queue buffer validation should use queue_size/2 to
verify ring_bo allocation and mapping size.
Fixes: 68e599db7a ("drm/amdkfd: Validate user queue buffers")
Suggested-by: Tomáš Trnka <trnka@scm.com>
Signed-off-by: Philip Yang <Philip.Yang@amd.com>
Acked-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Curret kfd does not allocate pasid values, instead uses pasid value for each
vm from graphic driver. So should not prevent graphic driver from releasing
pasid values since the values are allocated by graphic driver, not kfd driver
anymore. This patch does not stop graphic driver release pasid values.
Fixes: 8544374c0f ("drm/amdkfd: Have kfd driver use same PASID values from graphic driver")
Signed-off-by: Xiaogang Chen <xiaogang.chen@amd.com>
Reviewed-by: Felix Kuehling <felix.kuehling@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
In some ASICs L2 cache info may miss in kfd topology,
because the first bitmap may be empty, that means
the first cu may be inactive, so to find the first
active cu will solve the issue.
v2: Only find the first active cu in the first xcc
Signed-off-by: Eric Huang <jinhuieric.huang@amd.com>
Acked-by: Alex Deucher <alexander.deucher@amd.com>
Acked-by: Lijo Lazar <lijo.lazar@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
find_system_memory() pulls out two fields from an SMBIOS type 17
device and sets them on KFD devices. The data offsets are counted
to find interesting data.
Instead use a struct representation to access the members and pull
out the two specific fields.
No intended functional changes.
Link: https://www.dmtf.org/sites/default/files/standards/documents/DSP0134_3.8.0.pdf p99
Reviewed-by: Felix Kuehling <felix.kuehling@amd.com>
Link: https://lore.kernel.org/r/20250206214847.3334595-1-superm1@kernel.org
Signed-off-by: Mario Limonciello <mario.limonciello@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
On big and small APUs we send KFD VRAM allocations to GTT
since the carve out is either non-existent or relatively
small. However, if someone sets the carve out size to be
relatively large, we may end up using GTT rather than VRAM.
No change of logic with this patch, but it allows the
driver to determine which logic to use based on the
carve out size in the future.
Reviewed-by: Mario Limonciello <mario.limonciello@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
kfd_device_by_pci_dev(), kfd_get_pasid_limit() and kfd_set_pasid_limit()
have been unused since 2023's
commit c99a2e7ae2 ("drm/amdkfd: drop IOMMUv2 support")
Remove them.
Signed-off-by: Dr. David Alan Gilbert <linux@treblig.org>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Current kfd driver has its own PASID value for a kfd process and uses it to
locate vm at interrupt handler or mapping between kfd process and vm. That
design is not working when a physical gpu device has multiple spatial
partitions, ex: adev in CPX mode. This patch has kfd driver use same pasid
values that graphic driver generated which is per vm per pasid.
These pasid values are passed to fw/hardware. We do not need change interrupt
handler though more pasid values are used. Also, pasid values at log are
replaced by user process pid; pasid values are not exposed to user. Users see
their process pids that have meaning in user space.
Signed-off-by: Xiaogang Chen <xiaogang.chen@amd.com>
Reviewed-by: Felix Kuehling <felix.kuehling@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
This initializes SDMA IP version 6.1.3.
Signed-off-by: Tim Huang <tim.huang@amd.com>
Reviewed-by: Yifan Zhang <yifan1.zhang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
This initializes GC IP version 11.5.3.
Signed-off-by: Tim Huang <tim.huang@amd.com>
Reviewed-by: Yifan Zhang <yifan1.zhang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
It is possible for some waves in a workgroup to finish their save
sequence before the group leader has had time to capture the workgroup
barrier state. When this happens, having those waves exit do impact the
barrier state. As a consequence, the state captured by the group leader
is invalid, and is eventually incorrectly restored.
This patch proposes to have all waves in a workgroup wait for each other
at the end of their save sequence (just before calling s_endpgm_saved).
Signed-off-by: Lancelot SIX <lancelot.six@amd.com>
Reviewed-by: Jay Cornwall <jay.cornwall@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Cc: stable@vger.kernel.org # 6.12.x
The destructor of a gtt bo is declared as
void amdgpu_amdkfd_free_gtt_mem(struct amdgpu_device *adev, void **mem_obj);
Which takes void** as the second parameter.
GCC allows passing void* to the function because void* can be implicitly
casted to any other types, so it can pass compiling.
However, passing this void* parameter into the function's
execution process(which expects void** and dereferencing void**)
will result in errors.
Signed-off-by: Zhu Lingshan <lingshan.zhu@amd.com>
Reviewed-by: Felix Kuehling <felix.kuehling@amd.com>
Fixes: fb91065851 ("drm/amdkfd: Refactor queue wptr_bo GART mapping")
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
The following page fault was observed duringthe KFD process release.
In this particular error case, the HIP test (./MemcpyPerformance -h)
does not require the queue. As a result, the process_context_addr was
not assigned when the KFD process was released, ultimately leading to
this page fault during the execution of the function
kfd_process_dequeue_from_all_devices().
[345962.294891] amdgpu 0000:03:00.0: amdgpu: [gfxhub] page fault (src_id:0 ring:153 vmid:0 pasid:0)
[345962.295333] amdgpu 0000:03:00.0: amdgpu: in page starting at address 0x0000000000000000 from client 10
[345962.295775] amdgpu 0000:03:00.0: amdgpu: GCVM_L2_PROTECTION_FAULT_STATUS:0x00000B33
[345962.296097] amdgpu 0000:03:00.0: amdgpu: Faulty UTCL2 client ID: CPC (0x5)
[345962.296394] amdgpu 0000:03:00.0: amdgpu: MORE_FAULTS: 0x1
[345962.296633] amdgpu 0000:03:00.0: amdgpu: WALKER_ERROR: 0x1
[345962.296876] amdgpu 0000:03:00.0: amdgpu: PERMISSION_FAULTS: 0x3
[345962.297135] amdgpu 0000:03:00.0: amdgpu: MAPPING_ERROR: 0x1
[345962.297377] amdgpu 0000:03:00.0: amdgpu: RW: 0x0
[345962.297682] amdgpu 0000:03:00.0: amdgpu: [gfxhub] page fault (src_id:0 ring:169 vmid:0 pasid:0)
Signed-off-by: Prike Liang <Prike.Liang@amd.com>
Reviewed-by: Jonathan Kim <jonathan.kim@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Cc: stable@vger.kernel.org
The purpose of halt_if_hws_hang is to preserve GPU state for driver
debugging when queue preemption fails. Issuing per-queue reset may
kill wavefronts which caused the preemption failure.
Signed-off-by: Jay Cornwall <jay.cornwall@amd.com>
Reviewed-by: Jonathan Kim <Jonathan.Kim@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Cc: stable@vger.kernel.org # 6.12.x
If user shader issues S_SETVSKIP then this state will persist when
executing the trap handler, causing vector instructions to be
skipped.
VSKIP state is already saved/restored through the MODE register.
Signed-off-by: Jay Cornwall <jay.cornwall@amd.com>
Reviewed-by: Lancelot Six <lancelot.six@amd.com>
Suggested-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Source and binary have become mismatched during branch activity.
Signed-off-by: Jay Cornwall <jay.cornwall@amd.com>
Reviewed-by: Lancelot Six <lancelot.six@amd.com>
Suggested-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
For partial migrate from ram to vram, the migrate->cpages is not
equal to migrate->npages, should use migrate->npages to check all needed
migrate pages which could be copied or not.
And only need to set those pages could be migrated to migrate->dst[i], or
the migrate_vma_pages will migrate the wrong pages based on the migrate->dst[i].
v2:
Add mpages to break the loop earlier.
v3:
Uses MIGRATE_PFN_MIGRATE to identify whether page could be migrated.
v4:
Correct the error part.
Signed-off-by: Emily Deng <Emily.Deng@amd.com>
Reviewed-by: Philip Yang<Philip.Yang@amd.com>
Suggested-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
gfx12 derivatives will have substantially different trap handler
implementations from gfx10/gfx11. Add a separate source file for
gfx12+ and remove unneeded conditional code.
No functional change.
v2: Revert copyright date to 2018, minor comment fixes
Signed-off-by: Jay Cornwall <jay.cornwall@amd.com>
Reviewed-by: Lancelot Six <lancelot.six@amd.com>
Cc: Jonathan Kim <jonathan.kim@amd.com>
Acked-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
The header usr/linux/kfd_ioctl.h is a duplicate of uapi/linux/kfd_ioctl.h.
And it is actually not a file in the source code tree.
Ideally, the usr version should be updated whenever the source code is recompiled.
However, I have noticed a discrepancy between the two headers
even after rebuilding the kernel.
This commit modifies kfd_priv.h to always include the header from uapi to ensure
the latest changes are reflected. We should always include the source
code header other than the duplication.
Signed-off-by: Zhu Lingshan <lingshan.zhu@amd.com>
Reviewed-by: Felix Kuehling <felix.kuehling@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
If IH primary ring and KFD ih fifo overflows, we may miss CP, SDMA
interrupts and cause application soft hang. Show warning message with
ring name if overflow happens.
Add function to get ih ring name to avoid duplicating it. To keep
warning message consistent between GPU generations, change all
*_ih.c except ASICs older than Vega which has only one ih ring.
Signed-off-by: Philip Yang <Philip.Yang@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Felix Kuehling <felix.kuehling@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
If event slot is not signaled, kfd_signal_event_interrupt goes to slow
path to scan all event slots to find the signaled event, this is needed
for old ASICs that don't have the event ID or the event IDs are
incorrect in the IH payload.
There is case that GPU signal the same event twice, then driver process
the first event interrupt, set_event and event slot is auto-reset, then
for the second event interrupt, KFD goes to slow path as event is not
signaled, just drop the second event interrupt because the application
only need wakeup once.
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>
For CPX mode, each KFD node has interrupt worker to process ih_fifo to
send events to user space. Currently all interrupt workers of same adev
queue to same CPU, all workers execution are actually serialized and
this cause KFD ih_fifo overflow when CPU usage is high.
Use per-GPU unbounded highpri queue with number of workers equals to
number of partitions, let queue_work select the next CPU round robin
among the local CPUs of same NUMA.
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>
After GPU page fault, there are lots of page fault interrupts generated
at short period even with CAM filter enabled because the fault address
is different. Each page fault copy to KFD ih fifo to send event to user
space by KFD interrupt worker, this could cause KFD ih fifo overflow
while other processes generate events at same time.
KFD process is aborted after GPU page fault, we only need one GPU page
fault interrupt sent to KFD ih fifo to send memory exception event to
user space.
Incease KFD ih fifo size to 2 times of IH primary ring size, to handle
the burst events case.
This patch handle the gfx v9 path, cover retry on/off and CAM filter
on/off cases.
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>
To handle 40000 to 80000 interrupts per second running CPX mode with 4
streams/queues per KFD node, KFD interrupt handler becomes the
performance bottleneck.
Remove the kfifo_out memcpy overhead by accessing ih_fifo data in-place
and updating rptr with kfifo_skip_count.
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>
This patch checks and warns if pdd is NULL.
Signed-off-by: Andrew Martin <Andrew.Martin@amd.com>
Reviewed-by: Harish Kasiviswanathan <Harish.Kasiviswanathan@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
This a pointer that is being passed into other functions, so it is best to
initialize it to NULL prior.
Signed-off-by: Andrew Martin <Andrew.Martin@amd.com>
Reviewed-by: Harish Kasiviswanathan <Harish.Kasiviswanathan@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Update cwsr area size for gfx950 to fit the new user queue buffer validation.
The size of LDS calculation is referred from gfx950 thunk implementation.
Signed-off-by: Le Ma <le.ma@amd.com>
Acked-by: Hawking Zhang <Hawking.Zhang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
The gfx-9 trap handler is reading LDS allocation size in 256 bytes
granularity (from SQ_WAVE_LDS_ALLOC), but it using the assumption that
this value is always even (i.e. the LDS allocation is really done in
multiple of 512 bytes). This was true so far, but gfx-950 allocates LDS
in chunks of 1280 bytes, making this assumption invalid. This can cause
the trap handler to try to save / restore past the end of LDS, and past
the LDS allocated slot in the save are, overriding data from the
following wave.
This patch updates the trap handler to support LDS allocated in 1280
bytes blocks:
- During restore, copy from main memory directly to LDS in batch of 1280
bytes.
- During save, continue to use 512 bytes blocks (we only have 2 VGPRs we
can use to hold data), making sure to mask the upper half of the wave
when handling when the LDS size is not a multiple of 512 bytes.
Signed-off-by: Lancelot SIX <lancelot.six@amd.com>
Co-authored-by: Alex Sierra <alex.sierra@amd.com>
Reviewed-by: Jay Cornwall <jay.cornwall@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
In gfx950, the SQ_WAVE_LDS_ALLOC.LDS_SIZE field is extended to bits 12
to 22. The LDS_SIZE granularity remains unchanged (units of 64 dwords,
or 256 bytes). This patch adjusts the CWSR trap handler to read the
full extent of LDS_SIZE.
Signed-off-by: Lancelot SIX <lancelot.six@amd.com>
Reviewed-by: Jay Cornwall <jay.cornwall@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Instruction modifiers of the untyped vector memory buffer instructions
(MUBUF encoded) changed in gfx940. The slc, scc and glc modifiers have
been replaced with sc0, sc1 and nt.
The current CWSR trap handler is written using pre-gfx940 modifier
names, making the source incompatible with a strict gfx940 assembler.
This patch updates the cwsr_trap_handler_gfx9.s source file to be
compatible with all gfx9 variants of the ISA. The binary assembled code
is unchanged (so the behaviour is unchanged as well), only the source
representation is updated.
Signed-off-by: Lancelot SIX <lancelot.six@amd.com>
Reviewed-by: Jay Cornwall <jay.cornwall@amd.com>
Acked-by: Felix Kuehling <felix.kuehling@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Initial support for GC 9.5.0.
v2: squash in pqm_clean_queue_resource() fix from Lijo
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>
Update mtype flags to meet gfx 9.5.0 requirements for remote GPU
memory and system memory.
Signed-off-by: Alex Sierra <alex.sierra@amd.com>
Reviewed-by: Felix Kuehling <felix.kuehling@amd.com>
Reviewed-by: Harish Kasiviswanathan <harish.kasiviswanathan@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
GC 9.5.0 local memory MTYPE default should be set as RW.
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>
For better readability. Also leftover orphaned code.
Signed-off-by: Alex Sierra <alex.sierra@amd.com>
Reviewed-by: Felix Kuehling <felix.kuehling@amd.com>
Reviewed-by: Harish Kasiviswanathan <harish.kasiviswanathan@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
To have user better understand the causes triggering runlist oversubscription.
No function change.
Signed-off-by: Xiaogang Chen <xiaogang.chen@amd.com>
Reviewed-by: Harish Kasiviswanathan <harish.kasiviswanathan@amd.com>
Reviewed-by: Felix Kuehling <felix.kuehling@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Before scheduling a recovery due to scheduler/job hang, check if a RAS
error is detected. If so, choose RAS recovery to handle the situation. A
scheduler/job hang could be the side effect of a RAS error. In such
cases, it is required to go through the RAS error recovery process. A
RAS error recovery process in certains cases also could avoid a full
device device reset.
An error state is maintained in RAS context to detect the block
affected. Fatal Error state uses unused block id. Set the block id when
error is detected. If the interrupt handler detected a poison error,
it's not required to look for a fatal error. Skip fatal error checking
in such cases.
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>
When using MES creating a pdd will require talking to the GPU to
setup the relevant context. The code here forgot to wake up the GPU
in case it was in suspend, this causes KVM to EFAULT for passthrough
GPU for example. This issue can be masked if the GPU was woken up by
other things (e.g. opening the KMS node) first and have not yet gone to sleep.
v4: do the allocation of proc_ctx_bo in a lazy fashion
when the first queue is created in a process (Felix)
Signed-off-by: Jesse Zhang <jesse.zhang@amd.com>
Reviewed-by: Yunxiang Li <Yunxiang.Li@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Cc: stable@vger.kernel.org
This information is not available in ip discovery table.
Signed-off-by: Harish Kasiviswanathan <Harish.Kasiviswanathan@amd.com>
Reviewed-by: David Belanger <david.belanger@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Cc: stable@vger.kernel.org
This information is not available in ip discovery table.
Signed-off-by: Harish Kasiviswanathan <Harish.Kasiviswanathan@amd.com>
Reviewed-by: David Belanger <david.belanger@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Cc: stable@vger.kernel.org
In the function pqm_uninit there is a call-assignment of "pdd =
kfd_get_process_device_data" which could be null, and this value was
later dereferenced without checking.
Fixes: fb91065851 ("drm/amdkfd: Refactor queue wptr_bo GART mapping")
Signed-off-by: Andrew Martin <Andrew.Martin@amd.com>
Reviewed-by: Felix Kuehling <felix.kuehling@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Cc: stable@vger.kernel.org
Cacheline size is not available in IP discovery for gc943,gc944.
Signed-off-by: David Yat Sin <David.YatSin@amd.com>
Reviewed-by: Harish Kasiviswanathan <Harish.Kasiviswanathan@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Cc: stable@vger.kernel.org
Add MEC version from which alternate support for no PCIe atomics
is provided so that device is not skipped during KFD device init in
GFX1200/GFX1201.
Signed-off-by: Sreekant Somasekharan <sreekant.somasekharan@amd.com>
Reviewed-by: Harish Kasiviswanathan <Harish.Kasiviswanathan@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Cc: stable@vger.kernel.org # 6.11.x
Write pointer could be 32-bit or 64-bit. Use the correct size during
initialization.
Signed-off-by: Lijo Lazar <lijo.lazar@amd.com>
Acked-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Cc: stable@vger.kernel.org
In a consecutive packet submission, for example unmap and query status,
when CP is reading wptr caused by unmap packet doorbell ring, if in some
case CP operates slower (e.g. doorbell_mode=1) and wptr has been updated
to next packet (query status), but the query status packet content has
not been flushed to memory yet, it will cause CP fetched stalled data.
Adding mb to ensure ring buffer has been updated before updating wptr.
Also adding a mb to ensure wptr updated before doorbell ring.
Signed-off-by: Victor Zhao <Victor.Zhao@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Add back kfd queues in start scheduling that originally been
removed on stop scheduling.
Signed-off-by: Shaoyun Liu <shaoyun.liu@amd.com>
Reviewed-by: Felix Kuehling <felix.kuehling@amd.com>
Reviewed-by: Philip Yang <Philip.Yang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
kfd process kref count(process->ref) is initialized to 1 by kref_init. After
it is created not need to increase its kref. Instad add kfd process kref at kfd
process mmu notifier allocation since we already decrease the kref at
free_notifier of mmu_notifier_ops, so pair them.
When user process opens kfd node multiple times the kfd process kref is
increased each time to balance with kfd node close operation.
Signed-off-by: Xiaogang Chen <xiaogang.chen@amd.com>
Reviewed-by: Felix Kuehling <felix.kuehling@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
In kfd_procfs_show(), the sdma_activity_work_handler is a local variable
and the sdma_activity_work_handler.sdma_activity_work should initialize
with INIT_WORK_ONSTACK() instead of INIT_WORK().
Fixes: 32cb59f313 ("drm/amdkfd: Track SDMA utilization per process")
Signed-off-by: Yuan Can <yuancan@huawei.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>
GFX 12 does not require a page size cap for the trap handler because
it does not require a CWSR work around like GFX 11 did.
Signed-off-by: Jonathan Kim <jonathan.kim@amd.com>
Reviewed-by: David Belanger <david.belanger@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
The `kfd_get_cu_occupancy` function previously declared a large
`cu_occupancy` array as a local variable, which could lead to stack
overflows due to excessive stack usage. This commit replaces the static
array allocation with dynamic memory allocation using `kcalloc`,
thereby reducing the stack size.
This change avoids the risk of stack overflows in kernel space, in
scenarios where `AMDGPU_MAX_QUEUES` is large. The allocated memory is
freed using `kfree` before the function returns to prevent memory
leaks.
Fixes the below with gcc W=1:
drivers/gpu/drm/amd/amdgpu/../amdkfd/kfd_process.c: In function ‘kfd_get_cu_occupancy’:
drivers/gpu/drm/amd/amdgpu/../amdkfd/kfd_process.c:322:1: warning: the frame size of 1056 bytes is larger than 1024 bytes [-Wframe-larger-than=]
322 | }
| ^
Fixes: 6ae9e1aba9 ("drm/amdkfd: Update logic for CU occupancy calculations")
Cc: Harish Kasiviswanathan <Harish.Kasiviswanathan@amd.com>
Cc: Felix Kuehling <felix.kuehling@amd.com>
Cc: Christian König <christian.koenig@amd.com>
Cc: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Srinivasan Shanmugam <srinivasan.shanmugam@amd.com>
Suggested-by: Mukul Joshi <mukul.joshi@amd.com>
Reviewed-by: Mukul Joshi <mukul.joshi@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Add an interface to query whether KFD has any active queues.
v2: fix build issues
Acked-by: Srinivasan Shanmugam <srinivasan.shanmugam@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Flag KFD support for per-queue reset on GFX9 devices.
Signed-off-by: Jonathan Kim <jonathan.kim@amd.com>
Reviewed-by: Harish Kasiviswanathan <harish.kasiviswanathan@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Host drivers can create partial hives per guest by disabling xgmi sharing
between certain peers in the main hive.
Typically, these partial hives are fully connected per guest session.
In the event that the host makes a mistake by adding a non-shared node
to a guest session, have the KFD reflect sharing disabled by severing
the IO link.
Signed-off-by: Jonathan Kim <jonathan.kim@amd.com>
Tested-by: James Yao <yiqing.yao@amd.com>
Reviewed-by: Harish Kasiviswanathan <harish.kasiviswanathan@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
as the adding of mb() should be sufficient in function unmap_queues_cpsch,
remove the add of volatile type as recommended
Signed-off-by: Victor Zhao <Victor.Zhao@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
This reverts commit a3ab2d45b9.
The userspace side for this code is not ready yet so revert
for now.
Reviewed-by: Philip Yang <Philip.Yang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Cc: Philip Yang <Philip.Yang@amd.com>
make sure KFD_FENCE_INIT write to fence_addr before pm_send_query_status
called, to avoid qcm fence timeout caused by incorrect ordering.
Signed-off-by: Victor Zhao <Victor.Zhao@amd.com>
Reviewed-by: Philip Yang <Philip.Yang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Process device data pdd->vram_usage is read by rocm-smi via sysfs, this
is currently missing the svm_bo usage accounting, so "rocm-smi
--showpids" per process VRAM usage report is incorrect.
Add pdd->vram_usage accounting when svm_bo allocation and release,
change to atomic64_t type because it is updated outside process mutex
now.
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 new SMI event to report the dropped event count.
When the event kfifo is full, drop count is not zero, or no enough space
left to store the event message, increase drop count.
After reading event out from kfifo, if event was dropped, drop_count is
not zero, generate a dropped event record and reset drop count to zero.
Signed-off-by: Philip Yang <Philip.Yang@amd.com>
Reviewed-by: James Zhu <James.Zhu@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
get_wave_state is not defined for sdma queue, copy_context_work_handler
calls it for sdma queue will crash.
Signed-off-by: Philip Yang <Philip.Yang@amd.com>
Reviewed-by: Jonathan Kim <jonathan.kim@amd.com>
Tested-by: Jonathan Kim <jonathan.kim@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
SMI event fifo size 1KB was enough to report GPU vm fault or reset
event, but could drop the more frequent SVM migration events. Increase
kfifo size to 8KB to store about 100 migrate events, less chance to drop
the migrate events if lots of migration happened in the short period of
time. Add KFD prefix to the macro name.
Signed-off-by: Philip Yang <Philip.Yang@amd.com>
Reviewed-by: James Zhu <James.Zhu@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
If page migration failed, also output migrate end event to match with
migrate start event, with failure error_code added to the end of the
migrate message macro. This will not break uAPI because application uses
old message macro sscanf drop and ignore the error_code.
Output GPU page fault restore end event if migration failed.
Signed-off-by: Philip Yang <Philip.Yang@amd.com>
Reviewed-by: James Zhu <James.Zhu@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Only creating a new reference for each process instead of each VM.
Fixes: 9a1c1339ab ("drm/amdkfd: Run restore_workers on freezable WQs")
Suggested-by: Felix Kuehling <felix.kuehling@amd.com>
Signed-off-by: Lang Yu <lang.yu@amd.com>
Reviewed-by: Felix Kuehling <felix.kuehling@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
struct file *f is unused in queue creation, remove it.
Signed-off-by: Lang Yu <lang.yu@amd.com>
Reviewed-by: Felix Kuehling <felix.kuehling@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
program SDMAx_QUEUEx_SCHEDULE_CNTL for context switch due to
quantum in KFD for GFX12.
Signed-off-by: Sreekant Somasekharan <sreekant.somasekharan@amd.com>
Reviewed-by: Harish Kasiviswanathan <Harish.Kasiviswanathan@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Cc: stable@vger.kernel.org # 6.11.x
Make CU occupancy calculations work on GFX 9.4.3 by
updating the logic to handle multiple XCCs correctly.
Signed-off-by: Mukul Joshi <mukul.joshi@amd.com>
Reviewed-by: Harish Kasiviswanathan <Harish.Kasiviswanathan@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Currently, the code uses the IH_VMID_X_LUT register to map
a queue's vmid to the corresponding PASID. This logic is racy
since CP can update the VMID-PASID mapping anytime especially
when there are more processes than number of vmids. Update the
logic to calculate CU occupancy by matching doorbell offset of
the queue with valid wave counts against the process's queues.
Signed-off-by: Mukul Joshi <mukul.joshi@amd.com>
Reviewed-by: Harish Kasiviswanathan <Harish.Kasiviswanathan@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Variable hub_inst is unused.
Fixes: e28604d833 ("drm/amdkfd: Drop poison hanlding from gfx v10")
Signed-off-by: Jesse Zhang <jesse.zhang@amd.com>
Reviewed-by: Tim Huang <tim.huang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
We were removing the kernfs entry for queue info before checking if the
queue could be destroyed. If it failed to get destroyed (e.g. during
some GPU resets), then we would try to delete it later during pqm
teardown, but the file was already removed. This led to a kernel WARN
trying to remove size, gpuid and type. Move the remove to after the
destroy check.
Signed-off-by: Kent Russell <kent.russell@amd.com>
Reviewed-by: Jonathan Kim <jonathan.kim@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Instead of trying to use close_fd() on failure exits, just have
criu_get_prime_handle() store the file reference without inserting
it into descriptor table.
Then, once the callers are past the last failure exit, they can go
and either insert all those file references into the corresponding
slots of descriptor table, or drop all those file references and
free the unused descriptors.
Reviewed-by: Felix Kuehling <felix.kuehling@amd.com>
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
To avoid memory leaks, release q_extra_data when exiting the restore queue.
v2: Correct the proto (Alex)
Signed-off-by: Jesse Zhang <jesse.zhang@amd.com>
Reviewed-by: Tim Huang <tim.huang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Document how to use SMI system management interface to enable and
receive SVM events. Document SVM event triggers.
Define SVM events message string format macro that could be used by user
mode for sscanf to parse the event. Add it to uAPI header file to make
it obvious that is changing uAPI in future.
No functional changes.
Signed-off-by: Philip Yang <Philip.Yang@amd.com>
Reviewed-by: James Zhu <James.Zhu@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Driver mode-2 is only supported by relative new
smc firmware.
Signed-off-by: Hawking Zhang <Hawking.Zhang@amd.com>
Reviewed-by: Tao Zhou <tao.zhou1@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
If a queue is being destroyed but causes a HWS hang on removal, the KFD
may issue an unnecessary gpu reset if the destroyed queue can be fixed
by a queue reset.
This is because the queue has been removed from the KFD's queue list
prior to the preemption action on destroy so the reset call will fail to
match the HQD PQ reset information against the KFD's queue record to do
the actual reset.
To fix this, deactivate the queue prior to preemption since it's being
destroyed anyways and remove the queue from the KFD's queue list after
preemption.
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>
Enables users to update SVM's default granularity, used in
buffer migration and handling of recoverable page faults.
Param value is set in terms of log(numPages(buffer)),
e.g. 9 for a 2 MIB buffer
Signed-off-by: Ramesh Errabolu <Ramesh.Errabolu@amd.com>
Reviewed-by: Philip Yang <Philip.Yang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Populate cache line size info in topology based on information from IP
discovery table.
Signed-off-by: David Belanger <david.belanger@amd.com>
Reviewed-by: Sreekant Somasekharan <Sreekant.Somasekharan@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
ih1 is not initialized for APUs. Don't drain it or NULL pointer
error will be triggered.
Fixes: 6ef29715ac ("drm/amdkfd: Change kfd/svm page fault drain handling")
Signed-off-by: Yifan Zhang <yifan1.zhang@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
When app unmap vm ranges(munmap) kfd/svm starts drain pending page fault and
not handle any incoming pages fault of this process until a deferred work item
got executed by default system wq. The time period of "not handle page fault"
can be long and is unpredicable. That is advese to kfd performance on page
faults recovery.
This patch uses time stamp of incoming page fault to decide to drop or recover
page fault. When app unmap vm ranges kfd records each gpu device's ih ring
current time stamp. These time stamps are used at kfd page fault recovery
routine.
Any page fault happened on unmapped ranges after unmap events is application
bug that accesses vm range after unmap. It is not driver work to cover that.
By using time stamp of page fault do not need drain page faults at deferred
work. So, the time period that kfd does not handle page faults is reduced
and can be controlled.
Signed-off-by: Xiaogang.Chen <Xiaogang.Chen@amd.com>
Reviewed-by: Philip Yang <Philip.Yang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
This needs to be set to 1 to avoid a potential deadlock in
the GC 10.x and newer. On GC 9.x and older, this needs
to be set to 0. This can lead to hangs in some mixed
graphics and compute workloads.
Closes: https://gitlab.freedesktop.org/drm/amd/-/issues/3575
Reviewed-by: Hawking Zhang <Hawking.Zhang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Not supported.
Signed-off-by: Hawking Zhang <Hawking.Zhang@amd.com>
Reviewed-by: Tao Zhou <tao.zhou1@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Traditional utcl2 fault_status polling does not
work in SRIOV environment. The polling of fault
status register from guest side will be dropped
by hardware.
Driver should switch to check utcl2 interrupt
source id to identify utcl2 poison event. It is
set to 1 when poisoned data interrupts are
signaled.
v2: drop the unused local variable (Tao)
Signed-off-by: Hawking Zhang <Hawking.Zhang@amd.com>
Reviewed-by: Tao Zhou <tao.zhou1@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Based on the recommendation of MEC FW, update BadOpcode interrupt
handling by unmapping all queues, removing the queue that got the
interrupt from scheduling and remapping rest of the queues back when
using MES scheduler. This is done to prevent the case where unmapping
of the bad queue can fail thereby causing a GPU reset.
Signed-off-by: Mukul Joshi <mukul.joshi@amd.com>
Acked-by: Harish Kasiviswanathan <Harish.Kasiviswanathan@amd.com>
Acked-by: Alex Deucher <alexander.deucher@amd.com>
Reviewed-by: Felix Kuehling <felix.kuehling@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
MEC FW expects MES to unmap all queues when a VM fault is observed
on a queue and then resumed once the affected process is terminated.
Use the MES Suspend and Resume APIs to achieve this.
Signed-off-by: Mukul Joshi <mukul.joshi@amd.com>
Acked-by: Alex Deucher <alexander.deucher@amd.com>
Reviewed-by: Harish Kasiviswanathan <Harish.Kasiviswanathan@amd.com>
Reviewed-by: Felix Kuehling <felix.kuehling@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
When amdgpu enable enforce_isolation, KFD enables single-process mode in
HWS and sets exec_cleaner_shader bit in MAP_PROCESS.
Signed-off-by: Amber Lin <Amber.Lin@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Provide amdgpu_amdkfd_stop_sched() for amdgpu to stop KFD scheduling
compute work on HIQ. amdgpu_amdkfd_start_sched() resumes the scheduling.
When amdgpu_amdkfd_stop_sched is called, KFD will unmap queues from
runlist. If users send ioctls to KFD to create queues, they'll be added
but those queues won't be mapped to runlist (so not scheduled) until
amdgpu_amdkfd_start_sched is called.
v2: fix build (Alex)
Signed-off-by: Amber Lin <Amber.Lin@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
If there are multiple nodes per kfd device, add nodeid to location_id to
differentiate.
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>
Add helper function kfd_queue_unreference_buffers to reduce queue buffer
refcount, separate it from release queue buffers.
Because it is circular locking to hold dqm_lock to take vm lock,
kfd_ioctl_destroy_queue should take vm lock, unreference queue buffers
first, but not release queue buffers, to handle error in case failed to
hold vm lock. Then hold dqm_lock to remove queue from queue list and
then release queue buffers.
Restore process worker restore queue hold dqm_lock, will always find
the queue with valid queue buffers.
v2 (Felix):
- renamed kfd_queue_unreference_buffer(s) to kfd_queue_unref_bo_va(s)
- added two FIXME comments for follow up
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>
When users dynamically set the partition mode through sysfs writes,
this can lead to a double lock situation where the KFD is trying to take
the partition lock when updating the recommended SDMA engines.
Have the KFD reference its saved socket device number count instead.
Also ensure we have enough SDMA xGMI engines to report the recommended
engines in the first place.
Fixes: e06b71b231 ("drm/amdkfd: allow users to target recommended SDMA engines")
Signed-off-by: Jonathan Kim <jonathan.kim@amd.com>
Reviewed-by: Lijo Lazar <lijo.lazar@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
The number of watchpoints should be set and constrained per logical
partition device, not by the socket device.
Signed-off-by: Jonathan Kim <jonathan.kim@amd.com>
Reviewed-by: Harish Kasiviswanathan <harish.kasiviswanathan@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Support per-queue reset for GFX9. The recommendation is for the driver
to target reset the HW queue via a SPI MMIO register write.
Since this requires pipe and HW queue info and MEC FW is limited to
doorbell reports of hung queues after an unmap failure, scan the HW
queue slots defined by SET_RESOURCES first to identify the user queue
candidates to reset.
Only signal reset events to processes that have had a queue reset.
If queue reset fails, fall back to GPU reset.
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>
Fixes the below if kernel config not enable HMM support
>> drivers/gpu/drm/amd/amdgpu/../amdkfd/kfd_queue.c:107:26: error:
implicit declaration of function 'svm_range_from_addr'
>> drivers/gpu/drm/amd/amdgpu/../amdkfd/kfd_queue.c:107:24: error:
assignment to 'struct svm_range *' from 'int' makes pointer from integer
without a cast [-Wint-conversion]
>> drivers/gpu/drm/amd/amdgpu/../amdkfd/kfd_queue.c:111:28: error:
invalid use of undefined type 'struct svm_range'
Fixes: b049504e21 ("drm/amdkfd: Validate user queue svm memory residency")
Reported-by: kernel test robot <lkp@intel.com>
Closes: https://lore.kernel.org/oe-kbuild-all/202407252127.zvnxaKRA-lkp@intel.com/
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>
Certain GPUs have better copy performance over xGMI on specific
SDMA engines depending on the source and destination GPU.
Allow users to create SDMA queues on these recommended engines.
Close to 2x overall performance has been observed with this
optimization.
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>
When creating KFD user compute queue, check if queue eop buffer size,
cwsr area size, ctl stack size equal to the size of KFD node
properities.
Check the entire cwsr area which may split into multiple svm ranges
aligned to granularity boundary.
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 the queue eop buffer size, cwsr area size, ctl stack size
calculation from Thunk, store the value to KFD node properties.
Those will be used to validate queue eop buffer size, cwsr area size,
ctl stack size when creating KFD user compute queue.
Those will be exposed to user space via sysfs KFD node properties, to
remove the duplicate calculation code from Thunk.
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>
Ensure update queue new ring buffer is mapped on GPU with correct size.
Decrease queue old ring_bo queue_refcount and increase new ring_bo
queue_refcount.
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>
Queue CWSR area maybe registered to GPU as svm memory, create queue to
ensure svm mapped to GPU with KFD_IOCTL_SVM_FLAG_GPU_ALWAYS_MAPPED flag.
Add queue_refcount to struct svm_range, to track queue CWSR area usage.
Because unmap mmu notifier callback return value is ignored, if
application unmap the CWSR area while queue is active, pr_warn message
in dmesg log. To be safe, evict user queue.
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>
Add atomic queue_refcount to struct bo_va, return -EBUSY to fail unmap
BO from the GPU if the bo_va queue_refcount is not zero.
Create queue to increase the bo_va queue_refcount, destroy queue to
decrease the bo_va queue_refcount, to ensure the queue buffers mapped on
the GPU when queue is active.
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>
Find user queue rptr, ring buf, eop buffer and cwsr area BOs, and
check BOs are mapped on the GPU with correct size and take the BO
reference.
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>
Add helper function kfd_queue_acquire_buffers to get queue wptr_bo
reference from queue write_ptr if it is mapped to the KFD node with
expected size.
Add wptr_bo to structure queue_properties because structure queue is
allocated after queue buffers are validated, then we can remove wptr_bo
parameter from pqm_create_queue.
Rename structure queue wptr_bo_gart to hold wptr_bo reference for GART
mapping and umapping. Move MES wptr_bo_gart mapping to init_user_queue,
the same location with queue ctx_bo GART mapping.
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>
Pass pointer reference to amdgpu_bo_unref to clear the correct pointer,
otherwise amdgpu_bo_unref clear the local variable, the original pointer
not set to NULL, this could cause use-after-free bug.
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>
Change amdgpu_amdkfd_bo_mapped_to_dev to use drm_priv as parameter
instead of adev, to support spatial partition. This is only used by CRIU
checkpoint restore now. No functional change.
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>
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>
add amdgpu ras POSION_CONSUMPTION event id support.
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>
Enable KFD setting SDMA info for SDMA 6.1.2.
Signed-off-by: Tim Huang <Tim.Huang@amd.com>
Reviewed-by: Yifan Zhang <yifan1.zhang@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Convert some pr_* to some dev_* APIs to identify the device.
Signed-off-by: Lijo Lazar <lijo.lazar@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
We recently added locking to add_queue_mes() but this error path was
overlooked. Add an unlock to the error path.
Fixes: 1802b042a3 ("drm/amdgpu/kfd: remove is_hws_hang and is_resetting")
Signed-off-by: Dan Carpenter <dan.carpenter@linaro.org>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Per firmware's requirement, replace mode2 with mode1.
Signed-off-by: Tao Zhou <tao.zhou1@amd.com>
Reviewed-by: Hawking Zhang <Hawking.Zhang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
In commit fda812ebe3 ("drm/amdkfd: gfx12 context save/restore trap handler fixes")
the following fix was introduced but incorrectly restricted to gfx12.
The same issue and a corresponding fix apply to gfx10 and gfx11.
Do not overwrite TRAPSTS.{SAVECTX,HOST_TRAP} when restoring this
register. Both of these fields can assert while the wavefront is
running the trap handler.
Signed-off-by: Jay Cornwall <jay.cornwall@amd.com>
Reviewed-by: Lancelot Six <lancelot.six@amd.com>
Acked-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
We need to take the reset domain lock before talking to MES. While in
this case we can take the lock inside the mes helper. We can't do so for
most other mes helpers since they are used during reset. So for
consistency sake we add the lock here.
Signed-off-by: Yunxiang Li <Yunxiang.Li@amd.com>
Reviewed-by: Felix Kuehling <felix.kuehling@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
is_hws_hang and is_resetting serves pretty much the same purpose and
they all duplicates the work of the reset_domain lock, just check that
directly instead. This also eliminate a few bugs listed below and get
rid of dqm->ops.pre_reset.
kfd_hws_hang did not need to avoid scheduling another reset. If the
on-going reset decided to skip GPU reset we have a bad time, otherwise
the extra reset will get cancelled anyway.
remove_queue_mes forgot to check is_resetting flag compared to the
pre-MES path unmap_queue_cpsch, so it did not block hw access during
reset correctly.
Signed-off-by: Yunxiang Li <Yunxiang.Li@amd.com>
Reviewed-by: Felix Kuehling <felix.kuehling@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
idr_for_each_entry can ensure that mem is not empty during the loop.
So don't need check mem again.
Signed-off-by: Jesse Zhang <Jesse.Zhang@amd.com>
Reviewed-by: Felix Kuehling <felix.kuehling@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
reset cause is requested by customer as additional
info for gpu reset smi event.
v2: integerate reset sources suggested by Lijo Lazar
Signed-off-by: Eric Huang <jinhuieric.huang@amd.com>
Reviewed-by: Lijo Lazar <lijo.lazar@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
kfd_create_vcrat_image_gpu itself checks the avail_size at the start.
So the value of avail_size is at least VCRAT_SIZE_FOR_GPU(16384),
minus struct crat_header(40UL) and struct crat_subtype_compute(40UL) it cannot be less than 0.
Signed-off-by: Jesse Zhang <Jesse.Zhang@amd.com>
Reviewed-by: Felix Kuehling <felix.kuehling@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
The expression caps | HSA_CAP_TRAP_DEBUG_PRECISE_MEMORY_OPERATIONS_SUPPORTED
and caps | HSA_CAP_TRAP_DEBUG_PRECISE_ALU_OPERATIONS_SUPPORTED
are always 1/true regardless of the values of its operand.
Fixes: 9243240bed ("drm/amdkfd: enable single alu ops for gfx12")
Signed-off-by: Jesse Zhang <Jesse.Zhang@amd.com>
Suggested-by: Felix Kuehling <felix.kuehling@amd.com>
Reviewed-by: Felix Kuehling <felix.kuehling@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
To fix the warning about unused value,
remove the use_static and use the parameter is_static directly.
Signed-off-by: Jesse Zhang <Jesse.Zhang@amd.com>
Suggested-by: Felix Kuehling <felix.kuehling@amd.com>
Reviewed-by: Felix Kuehling <felix.kuehling@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
A wavefront may deallocate its VGPRs at the end of a program while
waiting for memory transactions to complete. If it subsequently
receives a context save exception it will be unable to save,
since this requires VGPRs. In this case the trap handler should
terminate the wavefront.
Fixes intermittent VM faults under context switching load.
V2: Use S_ENDPGM instead of S_ENDPGM_SAVED for performance counters
Signed-off-by: Jay Cornwall <jay.cornwall@amd.com>
Reviewed-by: Lancelot Six <lancelot.six@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
The varible uncached set false, the condition uncached cannot be true.
So remove the dead code, mapping flags will set the flag AMDGPU_VM_MTYPE_UC in else.
Signed-off-by: Jesse Zhang <Jesse.Zhang@amd.com>
Reviewed-by: Felix Kuehling <felix.kuehling@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Fix LDS size interpretation: 512 bytes (>= gfx12) vs 256 (< gfx12).
Ensure STATE_PRIV.BARRIER_COMPLETE cannot change after reading or
before writing. Other waves in the threadgroup may cause this field
to assert if they complete the barrier.
Do not overwrite EXCP_FLAG_PRIV.{SAVE_CONTEXT,HOST_TRAP} when
restoring this register. Both of these fields can assert while the
wavefront is running the trap handler.
Signed-off-by: Jay Cornwall <jay.cornwall@amd.com>
Reviewed-by: Lancelot Six <lancelot.six@amd.com>
Acked-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Newer assemblers reject S_WAITCNT. All instances of S_WAITCNT can be
replaced by S_WAITCNT 0 (< gfx12) or S_WAIT_IDLE (>= gfx12) since
there is no concurrency of different memory instruction classes.
Signed-off-by: Jay Cornwall <jay.cornwall@amd.com>
Reviewed-by: Lancelot Six <lancelot.six@amd.com>
Acked-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Source and binary have become mismatched during branch activity.
Signed-off-by: Jay Cornwall <jay.cornwall@amd.com>
Reviewed-by: Lancelot Six <lancelot.six@amd.com>
Acked-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
With commit 89773b8559
("drm/amdkfd: Let VRAM allocations go to GTT domain on small APUs")
big and small APU "VRAM" handling in KFD was unified. Since AMD_IS_APU
is set for both big and small APUs, we can simplify the checks in
the code.
v2: clean up a few more places (Lang)
Acked-by: Felix Kuehling <felix.kuehling@amd.com>
Reviewed-by: Lang Yu <Lang.Yu@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
This reverts commit 28ebbb4981.
Revert this commit as apparently the LLVM code to take advantage of
this never landed.
Reviewed-by: Feifei Xu <Feifei.Xu@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Cc: Feifei Xu <feifei.xu@amd.com>
It was an enablement vehicle for MES 11 and was never
productized. Remove it.
v2: drop additional checks in the GFX10 code.
v3: drop mes_api_def.h
Acked-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Harish Kasiviswanathan <Harish.Kasiviswanathan@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
GFX1201 was missed in the commit below. Adding it in.
Fixes: 628e1ace23 ("drm/amdkfd: mark GFX12 system and peer GPU memory mappings as MTYPE_NC")
Signed-off-by: Sreekant Somasekharan <sreekant.somasekharan@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
gpu_id needs to be unique for user space to identify GPUs via KFD
interface. In the current implementation there is a very small
probability of having non unique gpu_ids.
v2: Add check to confirm if gpu_id is unique. If not unique, find one
Changed commit header to reflect the above
v3: Use crc16 as suggested-by: Lijo Lazar <lijo.lazar@amd.com>
Ensure that gpu_id != 0
Signed-off-by: Harish Kasiviswanathan <Harish.Kasiviswanathan@amd.com>
Reviewed-by: Lijo Lazar <lijo.lazar@amd.com>
Reviewed-by: Felix Kuehling <felix.kuehling@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
No functional change. This will help in moving gpu_id creation to next
step while still being able to identify the correct GPU
Signed-off-by: Harish Kasiviswanathan <Harish.Kasiviswanathan@amd.com>
Reviewed-by: Felix Kuehling <felix.kuehling@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
We are incorrectly passing the first XCC's MQD when
updating CU masks for other XCCs in the partition. Fix
this by passing the MQD for the XCC currently being
updated with CU mask to update_cu_mask function.
Fixes: fc6efed2c7 ("drm/amdkfd: Update CU masking for GFX 9.4.3")
Signed-off-by: Mukul Joshi <mukul.joshi@amd.com>
Reviewed-by: Harish Kasiviswanathan <Harish.Kasiviswanathan@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Analysis of code by Coverity, a static code analyser, has identified
a resource leak in the symbol hmm_range. This leak occurs when one of
the prior steps before it is released encounters an error.
Signed-off-by: Ramesh Errabolu <Ramesh.Errabolu@amd.com>
Reviewed-by: Felix Kuehling <felix.kuehling@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
On system with khugepaged enabled and user cases with THP buffer, the
hmm_range_fault may takes > 15 seconds to return -EBUSY, the arbitrary
timeout value is not accurate, cause memory allocation failure.
Remove the arbitrary timeout value, return EAGAIN to application if
hmm_range_fault return EBUSY, then userspace libdrm and Thunk will call
ioctl again.
Change EAGAIN to debug message as this is not error.
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>
Currently oem_id is defined as uint8_t[6] and casted to uint64_t*
in some use case. This would lead code scanner to complain about
access beyond. Re-define it in union to enforce 8-byte size and
alignment to avoid potential issue.
Signed-off-by: Michael Chen <michael.chen@amd.com>
Reviewed-by: Felix Kuehling <felix.kuehling@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
We don't get the right offset in that case. The GPU has
an unused 4K area of the register BAR space into which you can
remap registers. We remap the HDP flush registers into this
space to allow userspace (CPU or GPU) to flush the HDP when it
updates VRAM. However, on systems with >4K pages, we end up
exposing PAGE_SIZE of MMIO space.
Fixes: d8e408a827 ("drm/amdkfd: Expose HDP registers to user space")
Reviewed-by: Felix Kuehling <felix.kuehling@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Cc: stable@vger.kernel.org
In interrupt context, write dbg_ev_file will be run by work queue. It
will cause write dbg_ev_file execution after debug_trap_disable, which
will cause NULL pointer access.
v2: cancel work "debug_event_workarea" before set dbg_ev_file as NULL.
Signed-off-by: Lin.Cao <lincao12@amd.com>
Reviewed-by: Jonathan Kim <jonathan.kim@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
This reverts commit c37ce764cd.
RCCL library is currently not treating spatial partitions differently,
hence this change is causing issues. Revert temporarily till RCCL
implementation is ready for spatial partitions.
Signed-off-by: Lijo Lazar <lijo.lazar@amd.com>
Reviewed-by: Jonathan Kim <jonathan.kim@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Small APUs(i.e., consumer, embedded products) usually have a small
carveout device memory which can't satisfy most compute workloads
memory allocation requirements.
We can't even run a Basic MNIST Example with a default 512MB carveout.
https://github.com/pytorch/examples/tree/main/mnist. Error Log:
"torch.cuda.OutOfMemoryError: HIP out of memory. Tried to allocate
84.00 MiB. GPU 0 has a total capacity of 512.00 MiB of which 0 bytes
is free. Of the allocated memory 103.83 MiB is allocated by PyTorch,
and 22.17 MiB is reserved by PyTorch but unallocated"
Though we can change BIOS settings to enlarge carveout size,
which is inflexible and may bring complaint. On the other hand,
the memory resource can't be effectively used between host and device.
The solution is MI300A approach, i.e., let VRAM allocations go to GTT.
Then device and host can flexibly and effectively share memory resource.
v2: Report local_mem_size_private as 0. (Felix)
Signed-off-by: Lang Yu <Lang.Yu@amd.com>
Reviewed-by: Felix Kuehling <felix.kuehling@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Due to a HW bug, the system memory mappings and peer GPU mappings
on GFX12 need to be marked as MTYPE_NC.
Cc: Joe Greathouse <joseph.greathouse@amd.com>
Cc: David Belanger <david.belanger@amd.com>
Signed-off-by: Rajneesh Bhardwaj <rajneesh.bhardwaj@amd.com>
Signed-off-by: Sreekant Somasekharan <sreekant.somasekharan@amd.com>
Reviewed-by: Harish Kasiviswanathan <Harish.Kasiviswanathan@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Add new GFX12 PTE flag AMDGPU_PTE_IS_PTE to svm_range_get_pte_flags
function. This resolves the issues related to SVM enablement in GFX12.
Signed-off-by: Sreekant Somasekharan <sreekant.somasekharan@amd.com>
Reviewed-by: Felix Kuehling <felix.kuehling@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Enable flag in KFD and set the atomic support bit in MQD.
Signed-off-by: David Belanger <david.belanger@amd.com>
Reviewed-by: Harish Kasiviswanathan <Harish.Kasiviswanathan@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
mqd_stride function in gfx v12 is not implemented, that
causes NULL ptr error. Add the generic func to fix it.
Signed-off-by: Eric Huang <jinhuieric.huang@amd.com>
Reviewed-by: Harish Kasiviswanathan <Harish.Kasiviswanathan@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
GFX12 debugging requires setting up precise ALU operation for catching
ALU exceptions.
Signed-off-by: Jonathan Kim <jonathan.kim@amd.com>
Tested-by: Lancelot Six <lancelot.six@amd.com>
Reviewed-by: Eric Huang <jinhuieric.huang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Similar to GFX11, always enable the setup of trap temporaries on GFX12.
Signed-off-by: Jonathan Kim <jonathan.kim@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Updated switch statement to use GFX12 trap handler.
Signed-off-by: David Belanger <david.belanger@amd.com>
Reviewed-by: Harish Kasiviswanathan <Harish.Kasiviswanathan@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
When trap_ctrl.trap_after_inst is set, it is possible for a wave to
enter the trap handler, after single-stepping an instruction and a
save_context is raised, with only save_context set in excp_flag_priv.
Because excp_flag_priv.trap_after_inst is not reliably set, we need to
use the missed single-step workaround for gfx12 as well.
Also add wave_start and wave_end as exceptions that should be handled
by the 2nd level trap handler.
Signed-off-by: Laurent Morichetti <laurent.morichetti@amd.com>
Tested-by: Lancelot Six <lancelot.six@amd.com>
Reviewed-by: Jonathan Kim <jonathan.kim@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Add support to save and restore the work group barrier state in gfx12
CWSR trap handler.
There is no support to directly restore the signal count of a barrier
state, so instead this patch repeatedly calls s_barrier_signal to
increment the signal count to the desired value.
In this patch, I have implemented the logic to restore the barrier at
the end of the block restoring the HWREGs. This process needs to be
done by exactly 1 wave per work group. To achieve this, the initial
value of s_restore_spi_init_hi (containing a FIRST_WAVE bit) needs to be
saved up until that point. An alternative could be restore the barrier
earlier in the process (around when LDS is restored, as the same wave
does both). Doing this would break the pattern that the restore
procedure follows the CWSR area layout.
Before restoring the barrier, this patch checks if the barrier was whose
state was saved has the "valid" bit set, even if I don't think this
barrier can be in an invalid state during context save. I expect this
test to always be true.
Signed-off-by: Lancelot SIX <lancelot.six@amd.com>
Reviewed-by: Jay Cornwall <jay.cornwall@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
- HWREG changes since gfx11
- Save/restore barrier state
- get_wave_size is now reserved by assembler
v2: rebase (Alex)
Signed-off-by: Jay Cornwall <jay.cornwall@amd.com>
Reviewed-by: Harish Kasiviswanathan <Harish.Kasiviswanathan@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
No functional change. Preparation for gfx12 support.
v2: drop unrelated change (Alex)
Signed-off-by: Jay Cornwall <jay.cornwall@amd.com>
Reviewed-by: Harish Kasiviswanathan <Harish.Kasiviswanathan@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Initial implementation, based on GFX11.
v2: Removed functions not needed by cp scheduler.
v3: Fixed typos.
v4: squash in warning fix (Alex)
Signed-off-by: David Belanger <david.belanger@amd.com>
Acked-by: Jonathan Kim <jonathan.kim@amd.com>
Reviewed-by: Harish Kasiviswanathan <Harish.Kasiviswanathan@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Initial implementation, based on GFX11.
v2: squash in include fix from David (Alex)
Signed-off-by: David Belanger <david.belanger@amd.com>
Reviewed-by: Harish Kasiviswanathan <Harish.Kasiviswanathan@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Initial implementation, based on GFX11.
v2: Removed dbg_wa code as not needed on GFX12.
v3: squash in SDMA queue fixes (Alex)
v4: rebase (Alex)
Signed-off-by: David Belanger <david.belanger@amd.com>
Reviewed-by: Harish Kasiviswanathan <Harish.Kasiviswanathan@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Added cases for GFX12 in switch statement, code relying on GFX11
implementation until GFX12 implementation is complete.
Signed-off-by: David Belanger <david.belanger@amd.com>
Reviewed-by: Harish Kasiviswanathan <Harish.Kasiviswanathan@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Added GFX12 support to a few switch statements.
Signed-off-by: David Belanger <david.belanger@amd.com>
Reviewed-by: Harish Kasiviswanathan <Harish.Kasiviswanathan@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Add gfx v9_4_4 ip block support
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>
Add sdma v4_4_5 ip block support
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>
There is a race condition when re-creating a kfd_process for a process.
This has been observed when a process under the debugger executes
exec(3). In this scenario:
- The process executes exec.
- This will eventually release the process's mm, which will cause the
kfd_process object associated with the process to be freed
(kfd_process_free_notifier decrements the reference count to the
kfd_process to 0). This causes kfd_process_ref_release to enqueue
kfd_process_wq_release to the kfd_process_wq.
- The debugger receives the PTRACE_EVENT_EXEC notification, and tries to
re-enable AMDGPU traps (KFD_IOC_DBG_TRAP_ENABLE).
- When handling this request, KFD tries to re-create a kfd_process.
This eventually calls kfd_create_process and kobject_init_and_add.
At this point the call to kobject_init_and_add can fail because the
old kfd_process.kobj has not been freed yet by kfd_process_wq_release.
This patch proposes to avoid this race by making sure to drain
kfd_process_wq before creating a new kfd_process object. This way, we
know that any cleanup task is done executing when we reach
kobject_init_and_add.
Signed-off-by: Lancelot SIX <lancelot.six@amd.com>
Reviewed-by: Felix Kuehling <felix.kuehling@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Queue buffer, though it is in system memory, has to be created using the
correct amdgpu device. Enforce this as the BO needs to mapped to the
GART for MES Hardware scheduler to access it.
Signed-off-by: Harish Kasiviswanathan <Harish.Kasiviswanathan@amd.com>
Reviewed-by: Felix Kuehling <felix.kuehling@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Replace tmz flag into buffer flag to make it easier to understand
and extend
Signed-off-by: Likun Gao <Likun.Gao@amd.com>
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>
Do VRAM accounting when doing migrations to vram to make sure
there is enough available VRAM and migrating to VRAM doesn't evict
other possible non-unified memory BOs. If migrating to VRAM fails,
driver can fall back to using system memory seamlessly.
Signed-off-by: Mukul Joshi <mukul.joshi@amd.com>
Reviewed-by: Felix Kuehling <felix.kuehling@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Handle the case that the restore worker was already scheduled by another
eviction while the restore was in progress.
Fixes: 9a1c1339ab ("drm/amdkfd: Run restore_workers on freezable WQs")
Signed-off-by: Felix Kuehling <felix.kuehling@amd.com>
Reviewed-by: Philip Yang <Philip.Yang@amd.com>
Tested-by: Yunxiang Li <Yunxiang.Li@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
It's not really an error since the devices don't support
the necessary hardware functionality.
Closes: https://gitlab.freedesktop.org/drm/amd/-/issues/3331
Tested-by: Paul Menzel <pmenzel@molgen.mpg.de>
Reviewed-by: Paul Menzel <pmenzel@molgen.mpg.de>
Reviewed-by: Kent Russell <kent.russell@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
-----BEGIN PGP SIGNATURE-----
iQFSBAABCAA8FiEEq68RxlopcLEwq+PEeb4+QwBBGIYFAmYlapoeHHRvcnZhbGRz
QGxpbnV4LWZvdW5kYXRpb24ub3JnAAoJEHm+PkMAQRiGJLgH/RFrsXefpxAWACJv
wRmIx/YFjI3Z4nypBZPXysX/JOuf/cf4bR22jMHC1vhQDtxVgB7eUpEYx0C0Y5RD
ll7cunAZCpP+PB725/bOWbaF94eigl8xC4RrKPon4tG1CS9NAvCJFDyNgWpqh48s
sSKKae68a16zMtIX5Za8nvmsBQOeD67ZFBgI4m7wIDDhTXY3XlG84r8Rz/ZwcdE/
rExFkDMUqjpoJVAMBtETQNk4pvKZVMrW2B0f/xZSR+IKXw5l2FO0raXkIG/6TERs
CxJvCuUjKIosqFLIK2O/cU9G+5V2axZ/WD8YRwDr2vlnnOPht+dFQzyPcuSqezbG
ShIhQo8=
=GI5F
-----END PGP SIGNATURE-----
Backmerge tag 'v6.9-rc5' into drm-next
Linux 6.9-rc5
I've had a persistent msm failure on clang, and the fix is in fixes
so just pull it back to fix that.
Signed-off-by: Dave Airlie <airlied@redhat.com>
Handle case that dma_fence_get_rcu_safe returns NULL.
If restore work is already scheduled, only update its timer. The same
work item cannot be queued twice, so undo the extra queue eviction.
Fixes: 9a1c1339ab ("drm/amdkfd: Run restore_workers on freezable WQs")
Signed-off-by: Felix Kuehling <felix.kuehling@amd.com>
Reviewed-by: Philip Yang <Philip.Yang@amd.com>
Tested-by: Gang BA <Gang.Ba@amd.com>
Reviewed-by: Gang BA <Gang.Ba@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
mode-2 reset is the only reliable method that can get
GC/SDMA back when poison is consumed. mmhub requires
mode-1 reset.
Signed-off-by: Hawking Zhang <Hawking.Zhang@amd.com>
Reviewed-by: Tao Zhou <tao.zhou1@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Fix memory leak due to a leaked mmget reference on an error handling
code path that is triggered when attempting to create KFD processes
while a GPU reset is in progress.
Fixes: 0ab2d7532b ("drm/amdkfd: prepare per-process debug enable and disable")
CC: Xiaogang Chen <xiaogang.chen@amd.com>
Signed-off-by: Felix Kuehling <felix.kuehling@amd.com>
Tested-by: Harish Kasiviswanthan <Harish.Kasiviswanthan@amd.com>
Reviewed-by: Mukul Joshi <mukul.joshi@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Cc: stable@vger.kernel.org
Fix memory leak due to a leaked mmget reference on an error handling
code path that is triggered when attempting to create KFD processes
while a GPU reset is in progress.
Fixes: 0ab2d7532b ("drm/amdkfd: prepare per-process debug enable and disable")
CC: Xiaogang Chen <xiaogang.chen@amd.com>
Signed-off-by: Felix Kuehling <felix.kuehling@amd.com>
Tested-by: Harish Kasiviswanthan <Harish.Kasiviswanthan@amd.com>
Reviewed-by: Mukul Joshi <mukul.joshi@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
If there are more than one device doing reset in parallel, the first
device will call kfd_suspend_all_processes() to evict all processes
on all devices, this call takes time to finish. other device will
start reset and recover without waiting. if the process has not been
evicted before doing recover, it will be restored, then caused page
fault.
Signed-off-by: Zhigang Luo <Zhigang.Luo@amd.com>
Reviewed-by: Felix Kuehling <felix.kuehling@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Currently, with F32 HWS GPU reset is only when unmap queue fails.
However, if compute queue doesn't repond to preemption request in time
unmap will return without any error. In this case, only preemption error
is logged and Reset is not triggered. Call GPU reset in this case also.
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Harish Kasiviswanathan <Harish.Kasiviswanathan@amd.com>
Reviewed-by: Mukul Joshi <mukul.joshi@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Cc: stable@vger.kernel.org
If there are more than one device doing reset in parallel, the first
device will call kfd_suspend_all_processes() to evict all processes
on all devices, this call takes time to finish. other device will
start reset and recover without waiting. if the process has not been
evicted before doing recover, it will be restored, then caused page
fault.
Signed-off-by: Zhigang Luo <Zhigang.Luo@amd.com>
Reviewed-by: Felix Kuehling <felix.kuehling@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Due to a CP interrupt bug, bad packet garbage exception codes are raised.
Do a range check so that the debugger and runtime do not receive garbage
codes.
Update the user api to guard exception code type checking as well.
Signed-off-by: Jonathan Kim <jonathan.kim@amd.com>
Tested-by: Jesse Zhang <jesse.zhang@amd.com>
Reviewed-by: Felix Kuehling <felix.kuehling@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
TLB flush after unmap accidentially was removed on
gfx9.4.2. It is to add it back.
Signed-off-by: Eric Huang <jinhuieric.huang@amd.com>
Reviewed-by: Harish Kasiviswanathan <Harish.Kasiviswanathan@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Cc: stable@vger.kernel.org
Check cgroup permissions when returning DMA-buf info and
based on cgroup info return the GPU id of the GPU that have
access to the BO.
Signed-off-by: Mukul Joshi <mukul.joshi@amd.com>
Reviewed-by: Felix Kuehling <felix.kuehling@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Currently, with F32 HWS GPU reset is only when unmap queue fails.
However, if compute queue doesn't repond to preemption request in time
unmap will return without any error. In this case, only preemption error
is logged and Reset is not triggered. Call GPU reset in this case also.
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Harish Kasiviswanathan <Harish.Kasiviswanathan@amd.com>
Reviewed-by: Mukul Joshi <mukul.joshi@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Destroy the high priority workqueue that handles interrupts
during KFD node cleanup.
Signed-off-by: Mukul Joshi <mukul.joshi@amd.com>
Reviewed-by: Felix Kuehling <felix.kuehling@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Due to a CP interrupt bug, bad packet garbage exception codes are raised.
Do a range check so that the debugger and runtime do not receive garbage
codes.
Update the user api to guard exception code type checking as well.
Signed-off-by: Jonathan Kim <jonathan.kim@amd.com>
Tested-by: Jesse Zhang <jesse.zhang@amd.com>
Reviewed-by: Felix Kuehling <felix.kuehling@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
TLB flush after unmap accidentially was removed on
gfx9.4.2. It is to add it back.
Signed-off-by: Eric Huang <jinhuieric.huang@amd.com>
Reviewed-by: Harish Kasiviswanathan <Harish.Kasiviswanathan@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
