We can get that just by casting tv.bo.
v2: squash in kfd fix (Alex)
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>
cs dependencies handling doesn't need in vm resv
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>
Slowly leaking memory one page at a time :)
Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Andrey Grodzovsky <andrey.grodzovsky@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>
Avoid unlocking a lock we never locked.
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>
First step to fix the LRU corruption, we accidentially tried to move things
on the LRU after dropping the lock.
Signed-off-by: Christian König <christian.koenig@amd.com>
Tested-by: Michel Dänzer <michel.daenzer@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
We can't hold the mn_lock while allocating memory.
Signed-off-by: Christian König <christian.koenig@amd.com>
Acked-by: Chunming Zhou <david1.zhou@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
No more waiting for a fence done here.
Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Chunming Zhou <david1.zhou@amd.com>
Reviewed-by: Junwei Zhang <Jerry.Zhang@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>
Add a helper to get the root PD address and remove the workarounds from
the GMC9 code for that.
Signed-off-by: Christian König <christian.koenig@amd.com>
Acked-by: Felix Kuehling <Felix.Kuehling@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Add a helper function for getting the root PD addr and cleanup join the
two VM related functions and cleanup the function name.
No functional change.
Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Huang Rui <ray.huang@amd.com>
Reviewed-by: Junwei Zhang <Jerry.Zhang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Further demangle ring from entity handling.
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>
Not needed any more since that is now done by the scheduler.
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>
At this point the command submission can still be interrupted.
Signed-off-by: Christian König <christian.koenig@amd.com>
Acked-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Instead of having extra handling just create an empty bo_list when no
handle is provided.
Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Chunming Zhou <david1.zhou@amd.com>
Reviewed-by: Huang Rui <ray.huang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Add helpers to iterate over all entries in a bo_list.
Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Chunming Zhou <david1.zhou@amd.com>
Acked-by: Huang Rui <ray.huang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
The bo_list handle is allocated by OP_CREATE, so in OP_UPDATE here we just
re-create the bo_list object and replace the handle. This way we don't
need locking to protect the bo_list because it's always re-created when
changed.
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>
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>
Return -EINVAL when both the BOs as well as a list handle is provided in
the IOCTL.
Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Chunming Zhou <david1.zhou@amd.com>
Reviewed-by: Huang Rui <ray.huang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Otherwise we silently don't use a BO list when the handle is invalid.
Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Chunming Zhou <david1.zhou@amd.com>
Reviewed-by: Huang Rui <ray.huang@amd.com>
Reviewed-by: Michel Dänzer <michel.daenzer@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
We are going to need that for the second UVD instance on Vega20.
v2: rename to patch_cs_in_place
Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-and-tested-by: James Zhu <James.Zhu@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
The scheduler of the entity is decided by the run queue on which
it is queued. This patch avoids us the effort required to maintain
a sync between rq and sched field when we start shifting entites
among different rqs.
Signed-off-by: Nayan Deshmukh <nayan26deshmukh@gmail.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Eric Anholt <eric@anholt.net>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
entity has a scheduler field and we don't need the sched argument
in any of the functions where entity is provided.
Signed-off-by: Nayan Deshmukh <nayan26deshmukh@gmail.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Eric Anholt <eric@anholt.net>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Just rename functions, no functional change.
Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Pushing a job can change the ring assignment of an entity.
v2: squash in:
"drm/amdgpu: fix job priority handling" (Christian)
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>
We can easily get that from the scheduler.
Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Junwei Zhang <Jerry.Zhang@amd.com>
Acked-by: Chunming Zhou <david1.zhou@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Can be obtained directly from the fence as well.
Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Junwei Zhang <Jerry.Zhang@amd.com>
Acked-by: Chunming Zhou <david1.zhou@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
This change is to support MESA performace optimization.
Modify CS IOCTL to allow its input as command buffer and an array of
buffer handles to create a temporay bo list and then destroy it
when IOCTL completes.
This saves on calling for BO_LIST create and destry IOCTLs in MESA
and by this improves performance.
v2: Avoid inserting the temp list into idr struct.
v3:
Remove idr alloation from amdgpu_bo_list_create.
Remove useless argument from amdgpu_cs_parser_fini
Minor cosmetic stuff.
v4: Revert amdgpu_bo_list_destroy back to static
Signed-off-by: Andrey Grodzovsky <andrey.grodzovsky@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Chunming Zhou <david1.zhou@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Concurrent execution of the non-atomic arithmetic could result in
completely bogus values.
v2:
* Rebased on v2 of the previous patch
Cc: stable@vger.kernel.org
Bugzilla: https://bugs.freedesktop.org/106872
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Michel Dänzer <michel.daenzer@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Extract and present the reposnsible process and thread when
VM_FAULT happens.
v2: Use getter and setter functions.
Signed-off-by: Andrey Grodzovsky <andrey.grodzovsky@amd.com>
Acked-by: Jim Qu <Jim.Qu@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Without this, there could not be enough slots, which could trigger the
BUG_ON in reservation_object_add_shared_fence.
v2:
* Jump to the error label instead of returning directly (Jerry Zhang)
v3:
* Reserve slots for command submission after VM updates (Christian König)
Cc: stable@vger.kernel.org
Bugzilla: https://bugs.freedesktop.org/106418
Reported-by: mikhail.v.gavrilov@gmail.com
Signed-off-by: Michel Dänzer <michel.daenzer@amd.com>
Signed-off-by: Junwei Zhang <Jerry.Zhang@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Move all instnaces of this check into a function in amdgpu_gmc.h
Rename the original function to a more proper name.
v2:
Add more places to cleanup.
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Andrey Grodzovsky <andrey.grodzovsky@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
As now enabled per vm bo feature, the user mode driver won't supply the
bo_list generally, for this case, the gdb_base, gds_size, gws_base, gws_size and
oa_base, oa_size won't be set.
v2: fix warning (Chunming)
Signed-off-by: Emily Deng <Emily.Deng@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Main changes for 4.18. I'd like to do a separate pull for vega20 later
this week or next. Highlights:
- Reserve pre-OS scanout buffer during init for seemless transition from
console to driver
- VEGAM support
- Improved GPU scheduler documentation
- Initial gfxoff support for raven
- SR-IOV fixes
- Default to non-AGP on PowerPC for radeon
- Fine grained clock voltage control for vega10
- Power profiles for vega10
- Further clean up of powerplay/driver interface
- Underlay fixes
- Display link bw updates
- Gamma fixes
- Scatter/Gather display support on CZ/ST
- Misc bug fixes and clean ups
[airlied: fixup v3d vs scheduler API change]
Link: https://patchwork.freedesktop.org/patch/msgid/20180515185450.1113-1-alexander.deucher@amd.com
Signed-off-by: Dave Airlie <airlied@redhat.com>
This reverts commit 6f51d28bfe8e1a676de5cd877639245bed3cc818.
Makes fallback handling to complicated. This is just a feature for the
GEM interface and shouldn't leak into the core BO create function.
Signed-off-by: Christian König <christian.koenig@amd.com>
Acked-by: Chunming Zhou <david1.zhou@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
The detection if a BO was placed in CPU visible VRAM was incorrect.
Fix it and merge it with the correct detection in amdgpu_ttm.c
Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Michel Dänzer <michel.daenzer@amd.com>
Reviewed-by: Chunming Zhou <david1.zhou@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Instead of the global (inaccurate) counter.
Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Michel Dänzer <michel.daenzer@amd.com>
Reviewed-by: Chunming Zhou <david1.zhou@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
user cases:
1. KFD wraps amdgpu_bo_create, they have no fallback case which is different
with amdgpu_gem_object_create.
since upstream branch has no amdgpu_amdkfd_gpuvm.c, which need KFD
guys add this flag to __alloc_memory_of_gpu:
+ flags |= AMDGPU_GEM_CREATE_NO_FALLBACK;
2. UMD can specify this flag for their allocation as well if they like.
v2: squash in merge conflict fix (Chunming)
Signed-off-by: Chunming Zhou <david1.zhou@amd.com>
Acked-by: Felix Kuehling <Felix.Kuehling@amd.com>
Cc: felix.kuehling@amd.com
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
This commit adds the notion of MMU notifier types GFX and HSA. GFX
continues to work like MMU notifiers did before. HSA adds support for
KFD userptr BOs. The implementation of KFD userptr eviction is a stub
for now.
Signed-off-by: Felix Kuehling <Felix.Kuehling@amd.com>
Reviewed-by: Oded Gabbay <oded.gabbay@gmail.com>
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>
for saving memory and more bit flag can be used in future
Signed-off-by: Roger He <Hongbo.He@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
If these bos are evicted and are in the validated list
things blow up, so do not put them in there. Notably,
that tries to add the bo to the LRU twice, which results
in a BUG_ON in ttm_bo.c.
While for the bo_list an alternative would be to not allow
always valid bos in there, that does not work for the user
fence.
v2: Fixed whitespace issue pointed out by checkpatch.pl
Signed-off-by: Bas Nieuwenhuizen <basni@chromium.org>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Cc: stable@vger.kernel.org
And rename it to amdgpu_gmc as well.
Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Samuel Li <Samuel.Li@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Necessary for the next patch.
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>
enable eviction of other per VM BOs during allocation and allows
reaping of deleted BOs during CS.
Reviewed-by: Chunming Zhou <david1.zhou@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Roger He <Hongbo.He@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
This moves and renames the AMDGPU scheduler to a common location in DRM
in order to facilitate re-use by other drivers. This is mostly a straight
forward rename with no code changes.
One notable exception is the function to_drm_sched_fence(), which is no
longer a inline header function to avoid the need to export the
drm_sched_fence_ops_scheduled and drm_sched_fence_ops_finished structures.
Reviewed-by: Chunming Zhou <david1.zhou@amd.com>
Tested-by: Dieter Nützel <Dieter@nuetzel-hh.de>
Acked-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Lucas Stach <l.stach@pengutronix.de>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Instead mark fence as explicit in it's amdgpu_sync_entry.
v2:
Fix use after free bug and add new parameter description.
Signed-off-by: Andrey Grodzovsky <andrey.grodzovsky@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Similar to the CPU address space the VA on Vega10 has a hole in it.
v2: use dev_dbg instead of dev_err
v3: add some more comments to explain how the hw works
Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
CC: stable@vger.kernel.org
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Instead of the global statistics use the per context bytes moved counter.
v2: rebased
Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Michel Dänzer <michel.daenzer@amd.com>
Reviewed-by: Chunming Zhou <david1.zhou@amd.com>
Tested-by: Dieter Nützel <Dieter@nuetzel-hh.de>
Tested-by: Michel Dänzer <michel.daenzer@amd.com>
Acked-by: Felix Kuehling <Felix.Kuehling@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Give moving a BO into place an operation context to work with.
v2: rebased
Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Michel Dänzer <michel.daenzer@amd.com>
Reviewed-by: Chunming Zhou <david1.zhou@amd.com>
Tested-by: Dieter Nützel <Dieter@nuetzel-hh.de>
Tested-by: Michel Dänzer <michel.daenzer@amd.com>
Acked-by: Felix Kuehling <Felix.Kuehling@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Christian König