The KFD VMID assignment was hard-coded in a few places. Consolidate that in
a single variable adev->vm_manager.first_kfd_vmid. The value is still
assigned in gmc-ip-version-specific code.
Signed-off-by: Felix Kuehling <Felix.Kuehling@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
For HMM support we need the ability to invalidate PTEs from
a MM callback where we can't lock the root PD.
Add a new flag to better support this instead of assuming
that all invalidation updates are unlocked.
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>
To avoid confusion with direct ring submissions rename bottom
of pipe VM table changes to immediate updates.
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>
Define the TMZ (encryption) bit in the page table entry (PTE) for
Raven and newer asics.
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Reviewed-by: Huang Rui <ray.huang@amd.com>
Vega20 arbitrates pstate at hive level and not device level. Last peer to
remote buffer unmap could drop P-State while another process is still
remote buffer mapped.
With this fix, P-States still needs to be disabled for now as SMU bug
was discovered on synchronous P2P transfers. This should be fixed in the
next FW update.
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>
Job fence on page table should be a shared one, so add it to the root
page talbe bo resv.
last_delayed field is not needed anymore. so remove it.
Cc: Christian König <christian.koenig@amd.com>
Cc: Alex Deucher <alexander.deucher@amd.com>
Cc: Felix Kuehling <Felix.Kuehling@amd.com>
Suggested-by: Christian König <christian.koenig@amd.com>
Signed-off-by: xinhui pan <xinhui.pan@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
If provided we only sync to the BOs reservation
object and no longer to the root PD.
v2: update comment, cleanup amdgpu_bo_sync_wait_resv
v3: use correct reservation object while clearing
v4: fix typo in amdgpu_bo_sync_wait_resv
Signed-off-by: Christian König <christian.koenig@amd.com>
Tested-by: Tom St Denis <tom.stdenis@amd.com>
Reviewed-by: Felix Kuehling <Felix.Kuehling@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
[Why]
Avoid reclaim filesystem while eviction lock is held called from
MMU notifier.
[How]
Setting PF_MEMALLOC_NOFS flags while eviction mutex is locked.
Using memalloc_nofs_save / memalloc_nofs_restore API.
Signed-off-by: Alex Sierra <alex.sierra@amd.com>
Reviewed-by: Felix Kuehling <Felix.Kuehling@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
drm_sched_entity_init() takes drm gpu scheduler list instead of
drm_sched_rq list. This makes conversion of drm_sched_rq list
to drm gpu scheduler list unnecessary
Signed-off-by: Nirmoy Das <nirmoy.das@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
This allows to invalidate VM entries without taking the reservation lock.
v3: use -EBUSY
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>
Don't add the VM update fences to the resv object and remove
the handling to stop implicitely syncing to them.
Ongoing updates prevent page tables from being evicted and we manually
block for all updates to complete before releasing PDs and PTS.
This way we can do updates even without the resv obj locked.
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>
When a page tables needs to be evicted the VM code should
decide if that is possible or not.
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>
Flag added to indicate if the amdgpu vm context is used for compute or
graphics.
Signed-off-by: Alex Sierra <alex.sierra@amd.com>
Reviewed-by: Felix Kuehling <Felix.Kuehling@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Next step towards HMM support. For now just silence the retry fault and
optionally redirect the request to the dummy page.
v2: make sure the VM is not destroyed while we handle the fault.
v3: fix VM destroy check, cleanup comments
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>
For handling PDE updates directly in the fault handler.
v2: fix typo in comment
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>
This allows us to update page tables directly while in a page fault.
v2: use direct/delayed entities and still wait for moves
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>
For page fault handling we need to use a direct update which can't be
blocked by ongoing user CS.
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>
This hopefully helps reduce the contention for page tables.
v2: adjust maximum reported VRAM size as well
Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Chunming Zhou <david1.zhou@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
navi14 share same PTE format with navi10.
Reviewed-by: Hawking Zhang <Hawking.Zhang@amd.com>
Signed-off-by: tiancyin <tianci.yin@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
v2: set mmhub num under CHIP_ARCTURUS switch case and add one more mmhub id_mgr
Signed-off-by: Le Ma <le.ma@amd.com>
Reviewed-by: Hawking Zhang <Hawking.Zhang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
The number of GFXHUB/MMHUB may be expanded in later ASICs.
Signed-off-by: Le Ma <le.ma@amd.com>
Reviewed-by: Hawking Zhang <Hawking.Zhang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Due to GCR change from navi10, the PTE encoding of PRT
needs change VSCTL = 01111 (was 0XX1X).
Signed-off-by: Jack Xiao <Jack.Xiao@amd.com>
Reviewed-by: Hawking Zhang <Hawking.Zhang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
To differentiate the mtypes across asics.
Signed-off-by: Hawking Zhang <Hawking.Zhang@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>
The MTYPE filed moves from bits 58:57 to 50:48 for NV10
And the size of MTYPE field is now 3bits
Signed-off-by: Hawking Zhang <Hawking.Zhang@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>
We are going to need that for recoverable page faults.
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>
It is redundant with the job pointer.
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>
Clean that up further and also fix another case where the BO
wasn't kmapped for CPU based updates.
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>
Driver vote low to high pstate switch whenever there is an outstanding
XGMI mapping request. Driver vote high to low pstate when all the
outstanding XGMI mapping is terminated.
Signed-off-by: shaoyunl <shaoyun.liu@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
And remove the existing code when it is unused.
Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Chunming Zhou <david1.zhou@amd.com>
Reviewed-by: Felix Kuehling <Felix.Kuehling@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Separate out all functions for SDMA and CPU based page table
updates into separate backends.
This way we can keep most of the complexity of those from the
core VM code.
Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Chunming Zhou <david1.zhou@amd.com>
Reviewed-by: Felix Kuehling <Felix.Kuehling@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Move the update parameter into the VM header and rename them.
Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Chunming Zhou <david1.zhou@amd.com>
Reviewed-by: Felix Kuehling <Felix.Kuehling@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Not needed any more.
Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Chunming Zhou <david1.zhou@amd.com>
Reviewed-by: Felix Kuehling <Felix.Kuehling@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Make sure that not only the entities are flush, but that
we also wait for the HW to finish all processing.
Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Chunming Zhou <david1.zhou@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Remove the chash implementation for now since it isn't used any more.
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>
Not needed any more since we now free PDs/PTs on demand.
Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Felix Kuehling <Felix.Kuehling@amd.com>
Acked-by: Huang Rui <ray.huang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Let's start to allocate VM PDs/PTs on demand instead of pre-allocating
them during mapping.
Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Felix Kuehling <Felix.Kuehling@amd.com>
Acked-by: Huang Rui <ray.huang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
if lru is changed, we cannot do bulk moving.
v2:
root bo isn't in bulk moving, skip its change.
Signed-off-by: Chunming Zhou <david1.zhou@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
printk_ratelimit() is much better suited to limit the number of reported
VM faults.
Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Felix Kuehling <Felix.Kuehling@amd.com>
Acked-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Instead of the double linked list. Gets the size of amdgpu_vm_pt down to
64 bytes again.
We could even reduce it down to 32 bytes, but that would require some
rather extreme hacks.
Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Chunming Zhou <david1.zhou@amd.com>
Acked-by: Felix Kuehling <Felix.Kuehling@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
In stead of share one fault hash table per device, make it
per vm. This can avoid inter-process lock issue when fault
hash table is full.
Change-Id: I5d1281b7c41eddc8e26113e010516557588d3708
Signed-off-by: Oak Zeng <Oak.Zeng@amd.com>
Suggested-by: Christian Konig <Christian.Koenig@amd.com>
Suggested-by: Felix Kuehling <Felix.Kuehling@amd.com>
Reviewed-by: Christian Konig <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Correct sign extend the GMC addresses to 48bit.
v2: sign extending turned out easier than thought.
v3: clean up the defines and move them into amdgpu_gmc.h as well
Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Junwei Zhang <Jerry.Zhang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Allows us to avoid taking the spinlock in more places.
Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Junwei Zhang <Jerry.Zhang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
For compute vm acquired from amdgpu, vm.pasid is managed
by kfd. Decouple pasid from such vm on process destroy
to avoid duplicate pasid release.
Signed-off-by: Oak Zeng <Oak.Zeng@amd.com>
Reviewed-by: Felix Kuehling <Felix.Kuehling@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
To make a amdgpu vm to a compute vm, the old pasid will be freed and
replaced with a pasid managed by kfd. Kfd can't reuse original pasid
allocated by amdgpu because kfd uses different pasid policy with amdgpu.
For example, all graphic devices share one same pasid in a process.
v2: rebase (Alex)
Signed-off-by: Oak Zeng <Oak.Zeng@amd.com>
Reviewed-by: Felix Kuehling <Felix.Kuehling@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Just another leftover from radeon.
Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Reviewed-by: Junwei Zhang <Jerry.Zhang@amd.com>
Acked-by: Huang Rui <ray.huang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Set the VM size based on system memory size between the ASIC-specific
limits given by min_vm_size and max_bits. GFXv9 GPUs will keep their
default VM size of 256TB (48 bit). Only older GPUs will adjust VM size
depending on system memory size.
This makes more VM space available for ROCm applications on GFXv8 GPUs
that want to map all available VRAM and system memory in their SVM
address space.
v2:
* Clarify comment
* Round up memory size before >> 30
* Round up automatic vm_size to power of two
Signed-off-by: Felix Kuehling <Felix.Kuehling@amd.com>
Acked-by: Junwei Zhang <Jerry.Zhang@amd.com>
Reviewed-by: Huang Rui <ray.huang@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
I continue to work for bulk moving that based on the proposal by Christian.
Background:
amdgpu driver will move all PD/PT and PerVM BOs into idle list. Then move all of
them on the end of LRU list one by one. Thus, that cause so many BOs moved to
the end of the LRU, and impact performance seriously.
Then Christian provided a workaround to not move PD/PT BOs on LRU with below
patch:
Commit 0bbf32026cf5ba41e9922b30e26e1bed1ecd38ae ("drm/amdgpu: band aid
validating VM PTs")
However, the final solution should bulk move all PD/PT and PerVM BOs on the LRU
instead of one by one.
Whenever amdgpu_vm_validate_pt_bos() is called and we have BOs which need to be
validated we move all BOs together to the end of the LRU without dropping the
lock for the LRU.
While doing so we note the beginning and end of this block in the LRU list.
Now when amdgpu_vm_validate_pt_bos() is called and we don't have anything to do,
we don't move every BO one by one, but instead cut the LRU list into pieces so
that we bulk move everything to the end in just one operation.
Test data:
+--------------+-----------------+-----------+---------------------------------------+
| |The Talos |Clpeak(OCL)|BusSpeedReadback(OCL) |
| |Principle(Vulkan)| | |
+------------------------------------------------------------------------------------+
| | | |0.319 ms(1k) 0.314 ms(2K) 0.308 ms(4K) |
| Original | 147.7 FPS | 76.86 us |0.307 ms(8K) 0.310 ms(16K) |
+------------------------------------------------------------------------------------+
| Orignial + WA| | |0.254 ms(1K) 0.241 ms(2K) |
|(don't move | 162.1 FPS | 42.15 us |0.230 ms(4K) 0.223 ms(8K) 0.204 ms(16K)|
|PT BOs on LRU)| | | |
+------------------------------------------------------------------------------------+
| Bulk move | 163.1 FPS | 40.52 us |0.244 ms(1K) 0.252 ms(2K) 0.213 ms(4K) |
| | | |0.214 ms(8K) 0.225 ms(16K) |
+--------------+-----------------+-----------+---------------------------------------+
After test them with above three benchmarks include vulkan and opencl. We can
see the visible improvement than original, and even better than original with
workaround.
v2: move all BOs include idle, relocated, and moved list to the end of LRU and
put them together.
v3: remove unused parameter and use list_for_each_entry instead of the one with
save entry.
v4: move the amdgpu_vm_move_to_lru_tail after command submission, at that time,
all bo will be back on idle list.
v5: remove amdgpu_vm_move_to_lru_tail_by_list(), use bulk_moveable instread of
validated, and move ttm_bo_bulk_move_lru_tail() also into
amdgpu_vm_move_to_lru_tail().
v6: clean up and fix return value.
Signed-off-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Huang Rui <ray.huang@amd.com>
Tested-by: Mike Lothian <mike@fireburn.co.uk>
Tested-by: Dieter Nützel <Dieter@nuetzel-hh.de>
Acked-by: Chunming Zhou <david1.zhou@amd.com>
Reviewed-by: Junwei Zhang <Jerry.Zhang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Instead of the fixed round robin use let the scheduler balance the load
of page table updates.
Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Chunming Zhou <david1.zhou@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Demangle amdgpu.h.
Signed-off-by: Huang Rui <ray.huang@amd.com>
Acked-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
This allows us to trace all VM ranges which should be valid inside a CS.
v2: dump mappings without BO as well
Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Chunming Zhou <david1.zhou@amd.com>
Reviewed-and-tested-by: Andrey Grodzovsky <andrey.grodzovsky@amd.com> (v1)
Reviewed-by: Huang Rui <ray.huang@amd.com> (v1)
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Felix Kuehling