linux/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c, branch v5.0

drm/amdkfd: Add support for doorbell BOs

2018-12-07T23:14:00Z

This allows user mode to map doorbell pages into GPUVM address space. That way GPUs can submit to user mode queues (self-dispatch). Acked-by: Alex Deucher Signed-off-by: Felix Kuehling Signed-off-by: Alex Deucher

drm/amdkfd: Add DMABuf import functionality

2018-12-07T23:13:54Z

This is used for interoperability between ROCm compute and graphics APIs. It allows importing graphics driver BOs into the ROCm SVM address space for zero-copy GPU access. The API is split into two steps (query and import) to allow user mode to manage the virtual address space allocation for the imported buffer. Acked-by: Alex Deucher Signed-off-by: Felix Kuehling Signed-off-by: Alex Deucher

drm/amdgpu: Add KFD VRAM limit checking

2018-12-07T23:13:15Z

We don't want KFD processes evicting each other over VRAM usage. Therefore prevent overcommitting VRAM among KFD applications with a per-GPU limit. Also leave enough room for page tables on top of the application memory usage. Acked-by: Alex Deucher Signed-off-by: Felix Kuehling Reviewed-by: Eric Huang Signed-off-by: Alex Deucher

drm/ttm: allow reserving more than one shared slot v3

2018-12-07T22:53:44Z

Let's support simultaneous submissions to multiple engines. v2: rename the field to num_shared and fix up all users v3: rebased Signed-off-by: Christian König Reviewed-by: Michel Dänzer Reviewed-by: Junwei Zhang Reviewed-by: Huang Rui Signed-off-by: Alex Deucher

drm/amdkfd: change system memory overcommit limit

2018-11-20T19:01:54Z

It is to improve system limit by: 1. replacing userptrlimit with a total memory limit that conunts TTM memory usage and userptr usage. 2. counting acc size for all BOs. Signed-off-by: Eric Huang Reviewed-by: Felix Kuehling Signed-off-by: Felix Kuehling Signed-off-by: Alex Deucher

drm/amdgpu: Fix KFD doorbell SG BO mapping

2018-11-20T19:01:12Z

This change prepares for adding SG BOs that will be used for mapping doorbells into GPUVM address space. This type of BO would be mistaken for an invalid userptr BO. Improve that check to test that it's actually a userptr BO so that SG BOs that are still in the CPU domain can be validated and mapped correctly. Signed-off-by: Felix Kuehling Signed-off-by: Alex Deucher

drm/amdkfd: Fix and simplify sync object handling for KFD

2018-11-20T19:01:05Z

The adev parameter in amdgpu_sync_fence and amdgpu_sync_resv is only needed for updating sync->last_vm_update. This breaks if different adevs are passed to calls for the same sync object. Always pass NULL for calls from KFD because sync objects used for KFD don't belong to any particular device, and KFD doesn't need the sync->last_vm_update fence. This fixes kernel log warnings on multi-GPU systems after recent changes in amdgpu_amdkfd_gpuvm_restore_process_bos. Signed-off-by: Felix Kuehling Reviewed-by: Christian König Signed-off-by: Alex Deucher

drm/amdgpu: KFD Restore process: Optimize waiting

2018-11-20T19:00:58Z

Instead of waiting for each KFD BO after validation just wait for the last BO moving fence. Signed-off-by: Harish Kasiviswanathan Reviewed-by: Felix Kuehling Signed-off-by: Felix Kuehling Acked-by: Alex Deucher Signed-off-by: Alex Deucher

drm/amdgpu: Remove explicit wait after VM validate

2018-11-20T19:00:37Z

PD or PT might have to be moved during validation and this move has to be completed before updating it. If page table updates are done using SDMA then this serializing is done by SDMA command submission. And if PD/PT updates are done by CPU, then explicit waiting for PD/PT updates are done in amdgpu VM amdgpu_vm_wait_pd function. Sync to PD BO moving fence to handle corner case where none of the PTs are updated but PD is evicted. Signed-off-by: Harish Kasiviswanathan Reviewed-by: Felix Kuehling Signed-off-by: Felix Kuehling Acked-by: Alex Deucher Signed-off-by: Alex Deucher

drm/amdkfd: Added Vega12 and Polaris12 for KFD.

2018-11-19T21:38:13Z

Add Vega12 and Polaris12 device info and device IDs to KFD. Signed-off-by: Gang Ba Reviewed-by: Felix Kuehling Signed-off-by: Felix Kuehling Reviewed-by: Alex Deucher Signed-off-by: Alex Deucher