[PATCH 00/25] Add KFD GPUVM support for dGPUs

[Christian König]

Am 27.01.2018 um 02:09 schrieb Felix Kuehling:
> I split this into an AMDGPU and AMDKFD part. The bigger patches that
> add lots of new code are not cherry-picked and squashed. Instead I
> copied, reorganized and cleaned up the code by hand and then split it
> into some semblance of a sensible history. I acknowledged major
> contributors with signed-off-by lines but didn't list everyone who
> ever touched that code (would probably be most of the team).
>
> I pushed an updated Thunk (rebased on ROCm 1.7) that works with this
> KFD update. Most testing was done on Fiji with KFDTest (Yong started
> working on open-sourcing it). I was also able to run the OpenCL
> version of SHOC, though most sub-tests still fail.
>
> KFDTest can manage VRAM and system memory, submit shader dispatches,
> receive events. I haven't tested multi-GPU yet, but in theory that
> should also work, with system memory buffers shared between multiple
> GPUs.
>
> The big missing piece at this point is support for userptr memory
> (user-allocated memory mapped for GPU access). That's giong to be my
> next patch series that should enable a much wider range of real-world
> applications.
>
> AMDGPU:
> Patches 1-5 are minor cleanups and fixes

Patches #1-#5 are Reviewed-by: Christian König <christian.koenig at amd.com>.

> Patches 6-10 add and implement KFD->KGD interfaces for GPUVM
>
> AMDKFD:
> Patches 11-13 are minor cleanups and fixes
> Patches 14-25 add all the GPUVM memory management functionality
>
> Felix Kuehling (22):
>    drm/amdgpu: remove useless BUG_ONs
>    drm/amdgpu: Fix header file dependencies
>    drm/amdgpu: Fix wrong mask in get_atc_vmid_pasid_mapping_pasid
>    drm/amdgpu: Remove unused kfd2kgd interface
>    drm/amdgpu: Add KFD eviction fence
>    drm/amdgpu: Update kgd2kfd_shared_resources for dGPU support
>    drm/amdgpu: add amdgpu_sync_clone
>    drm/amdgpu: Add GPUVM memory management functions for KFD
>    drm/amdgpu: Add submit IB function for KFD
>    drm/amdkfd: Add missing #ifdef CONFIG_AMD_IOMMU_V2 guard
>    drm/amdkfd: Use per-device sched_policy
>    drm/amdkfd: Add GPUVM virtual address space to PDD
>    drm/amdkfd: Implement KFD process eviction/restore
>    uapi: Fix type used in ioctl parameter structures
>    drm/amdkfd: Remove limit on number of GPUs
>    drm/amdkfd: Aperture setup for dGPUs
>    drm/amdkfd: Add per-process IDR for buffer handles
>    drm/amdkfd: Allocate CWSR trap handler memory for dGPUs
>    drm/amdkfd: Add TC flush on VMID deallocation for Hawaii
>    drm/amdkfd: Add ioctls for GPUVM memory management
>    drm/amdkfd: Kmap event page for dGPUs
>    drm/amdkfd: Add module option for testing large-BAR functionality
>
> Harish Kasiviswanathan (1):
>    drm/amdkfd: Remove unaligned memory access
>
> Oak Zeng (1):
>    drm/amdkfd: Populate DRM render device minor
>
> Yong Zhao (1):
>    drm/amdgpu: Replace kgd_mem with amdgpu_bo for kernel pinned gtt mem
>
>   drivers/gpu/drm/amd/amdgpu/Makefile                |    2 +
>   drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c         |  127 +-
>   drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.h         |  115 +-
>   drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_fence.c   |  196 +++
>   drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v7.c  |   80 +-
>   drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v8.c  |   82 +-
>   drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c   | 1500 ++++++++++++++++++++
>   drivers/gpu/drm/amd/amdgpu/amdgpu_object.c         |    4 +
>   drivers/gpu/drm/amd/amdgpu/amdgpu_object.h         |    2 +
>   drivers/gpu/drm/amd/amdgpu/amdgpu_ring.h           |    6 +-
>   drivers/gpu/drm/amd/amdgpu/amdgpu_sync.c           |   53 +-
>   drivers/gpu/drm/amd/amdgpu/amdgpu_sync.h           |    1 +
>   drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c            |   25 +
>   drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h             |    1 +
>   drivers/gpu/drm/amd/amdkfd/kfd_chardev.c           |  484 +++++++
>   drivers/gpu/drm/amd/amdkfd/kfd_crat.c              |    3 +
>   drivers/gpu/drm/amd/amdkfd/kfd_device.c            |   65 +-
>   .../gpu/drm/amd/amdkfd/kfd_device_queue_manager.c  |  290 +++-
>   .../gpu/drm/amd/amdkfd/kfd_device_queue_manager.h  |    9 +
>   drivers/gpu/drm/amd/amdkfd/kfd_events.c            |   31 +-
>   drivers/gpu/drm/amd/amdkfd/kfd_flat_memory.c       |   59 +-
>   drivers/gpu/drm/amd/amdkfd/kfd_module.c            |    7 +
>   drivers/gpu/drm/amd/amdkfd/kfd_packet_manager.c    |   37 +
>   drivers/gpu/drm/amd/amdkfd/kfd_priv.h              |   79 +-
>   drivers/gpu/drm/amd/amdkfd/kfd_process.c           |  490 ++++++-
>   drivers/gpu/drm/amd/amdkfd/kfd_topology.c          |    4 +
>   drivers/gpu/drm/amd/amdkfd/kfd_topology.h          |    1 +
>   drivers/gpu/drm/amd/include/kgd_kfd_interface.h    |  101 +-
>   include/uapi/linux/kfd_ioctl.h                     |   87 +-
>   29 files changed, 3811 insertions(+), 130 deletions(-)
>   create mode 100644 drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_fence.c
>   create mode 100644 drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c
>