Commit c76f0b2c authored by Dave Airlie's avatar Dave Airlie
Browse files

Merge tag 'drm-amdkfd-next-2018-05-14' of git://people.freedesktop.org/~gabbayo/linux into drm-next 



This is amdkfd pull for 4.18. The major new features are:

- Add support for GFXv9 dGPUs (VEGA)
- Add support for userptr memory mapping

In addition, there are a couple of small fixes and improvements, such as:
- Fix lock handling
- Fix rollback packet in kernel kfd_queue
- Optimize kfd signal handling
- Fix CP hang in APU

Signed-off-by: default avatarDave Airlie <airlied@redhat.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20180514070126.GA1827@odedg-x270
parents 444ac87b af47b390

MAINTAINERS

+2 −0
Original line number Diff line number Diff line
@@ -767,12 +767,14 @@ F: drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c 
F:	drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.h

F:	drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v7.c

F:	drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v8.c

F:	drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v9.c

F:	drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_fence.c

F:	drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c

F:	drivers/gpu/drm/amd/amdkfd/

F:	drivers/gpu/drm/amd/include/cik_structs.h

F:	drivers/gpu/drm/amd/include/kgd_kfd_interface.h

F:	drivers/gpu/drm/amd/include/vi_structs.h

F:	drivers/gpu/drm/amd/include/v9_structs.h

F:	include/uapi/linux/kfd_ioctl.h



AMD SEATTLE DEVICE TREE SUPPORT

drivers/gpu/drm/amd/amdgpu/Makefile

+2 −1
Original line number Diff line number Diff line
@@ -130,7 +130,8 @@ amdgpu-y += \ 
	 amdgpu_amdkfd.o \

	 amdgpu_amdkfd_fence.o \

	 amdgpu_amdkfd_gpuvm.o \

	 amdgpu_amdkfd_gfx_v8.o

	 amdgpu_amdkfd_gfx_v8.o \

	 amdgpu_amdkfd_gfx_v9.o



# add cgs

amdgpu-y += amdgpu_cgs.o

drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c

+26 −0
Original line number Diff line number Diff line
@@ -92,6 +92,10 @@ void amdgpu_amdkfd_device_probe(struct amdgpu_device *adev) 
	case CHIP_POLARIS11:

		kfd2kgd = amdgpu_amdkfd_gfx_8_0_get_functions();

		break;

	case CHIP_VEGA10:

	case CHIP_RAVEN:

		kfd2kgd = amdgpu_amdkfd_gfx_9_0_get_functions();

		break;

	default:

		dev_dbg(adev->dev, "kfd not supported on this ASIC\n");

		return;
@@ -175,6 +179,28 @@ void amdgpu_amdkfd_device_init(struct amdgpu_device *adev) 
				&gpu_resources.doorbell_physical_address,

				&gpu_resources.doorbell_aperture_size,

				&gpu_resources.doorbell_start_offset);

		if (adev->asic_type >= CHIP_VEGA10) {

			/* On SOC15 the BIF is involved in routing

			 * doorbells using the low 12 bits of the

			 * address. Communicate the assignments to

			 * KFD. KFD uses two doorbell pages per

			 * process in case of 64-bit doorbells so we

			 * can use each doorbell assignment twice.

			 */

			gpu_resources.sdma_doorbell[0][0] =

				AMDGPU_DOORBELL64_sDMA_ENGINE0;

			gpu_resources.sdma_doorbell[0][1] =

				AMDGPU_DOORBELL64_sDMA_ENGINE0 + 0x200;

			gpu_resources.sdma_doorbell[1][0] =

				AMDGPU_DOORBELL64_sDMA_ENGINE1;

			gpu_resources.sdma_doorbell[1][1] =

				AMDGPU_DOORBELL64_sDMA_ENGINE1 + 0x200;

			/* Doorbells 0x0f0-0ff and 0x2f0-2ff are reserved for

			 * SDMA, IH and VCN. So don't use them for the CP.

			 */

			gpu_resources.reserved_doorbell_mask = 0x1f0;

			gpu_resources.reserved_doorbell_val  = 0x0f0;

		}



		kgd2kfd->device_init(adev->kfd, &gpu_resources);

	}

drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.h

+13 −0
Original line number Diff line number Diff line
@@ -28,6 +28,7 @@ 
#include <linux/types.h>

#include <linux/mm.h>

#include <linux/mmu_context.h>

#include <linux/workqueue.h>

#include <kgd_kfd_interface.h>

#include <drm/ttm/ttm_execbuf_util.h>

#include "amdgpu_sync.h"
@@ -59,7 +60,9 @@ struct kgd_mem { 


	uint32_t mapping_flags;



	atomic_t invalid;

	struct amdkfd_process_info *process_info;

	struct page **user_pages;



	struct amdgpu_sync sync;
@@ -84,6 +87,9 @@ struct amdkfd_process_info { 
	struct list_head vm_list_head;

	/* List head for all KFD BOs that belong to a KFD process. */

	struct list_head kfd_bo_list;

	/* List of userptr BOs that are valid or invalid */

	struct list_head userptr_valid_list;

	struct list_head userptr_inval_list;

	/* Lock to protect kfd_bo_list */

	struct mutex lock;
@@ -91,6 +97,11 @@ struct amdkfd_process_info { 
	unsigned int n_vms;

	/* Eviction Fence */

	struct amdgpu_amdkfd_fence *eviction_fence;



	/* MMU-notifier related fields */

	atomic_t evicted_bos;

	struct delayed_work restore_userptr_work;

	struct pid *pid;

};



int amdgpu_amdkfd_init(void);
@@ -104,12 +115,14 @@ void amdgpu_amdkfd_device_probe(struct amdgpu_device *adev); 
void amdgpu_amdkfd_device_init(struct amdgpu_device *adev);

void amdgpu_amdkfd_device_fini(struct amdgpu_device *adev);



int amdgpu_amdkfd_evict_userptr(struct kgd_mem *mem, struct mm_struct *mm);

int amdgpu_amdkfd_submit_ib(struct kgd_dev *kgd, enum kgd_engine_type engine,

				uint32_t vmid, uint64_t gpu_addr,

				uint32_t *ib_cmd, uint32_t ib_len);



struct kfd2kgd_calls *amdgpu_amdkfd_gfx_7_get_functions(void);

struct kfd2kgd_calls *amdgpu_amdkfd_gfx_8_0_get_functions(void);

struct kfd2kgd_calls *amdgpu_amdkfd_gfx_9_0_get_functions(void);



bool amdgpu_amdkfd_is_kfd_vmid(struct amdgpu_device *adev, u32 vmid);

drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v7.c

+0 −10
Original line number Diff line number Diff line
@@ -98,8 +98,6 @@ static void kgd_program_sh_mem_settings(struct kgd_dev *kgd, uint32_t vmid, 
static int kgd_set_pasid_vmid_mapping(struct kgd_dev *kgd, unsigned int pasid,

					unsigned int vmid);



static int kgd_init_pipeline(struct kgd_dev *kgd, uint32_t pipe_id,

				uint32_t hpd_size, uint64_t hpd_gpu_addr);

static int kgd_init_interrupts(struct kgd_dev *kgd, uint32_t pipe_id);

static int kgd_hqd_load(struct kgd_dev *kgd, void *mqd, uint32_t pipe_id,

			uint32_t queue_id, uint32_t __user *wptr,
@@ -183,7 +181,6 @@ static const struct kfd2kgd_calls kfd2kgd = { 
	.free_pasid = amdgpu_pasid_free,

	.program_sh_mem_settings = kgd_program_sh_mem_settings,

	.set_pasid_vmid_mapping = kgd_set_pasid_vmid_mapping,

	.init_pipeline = kgd_init_pipeline,

	.init_interrupts = kgd_init_interrupts,

	.hqd_load = kgd_hqd_load,

	.hqd_sdma_load = kgd_hqd_sdma_load,
@@ -309,13 +306,6 @@ static int kgd_set_pasid_vmid_mapping(struct kgd_dev *kgd, unsigned int pasid, 
	return 0;

}



static int kgd_init_pipeline(struct kgd_dev *kgd, uint32_t pipe_id,

				uint32_t hpd_size, uint64_t hpd_gpu_addr)

{

	/* amdgpu owns the per-pipe state */

	return 0;

}



static int kgd_init_interrupts(struct kgd_dev *kgd, uint32_t pipe_id)

{

	struct amdgpu_device *adev = get_amdgpu_device(kgd);

CRA Git | Maintained and supported by SUSTech CRA and CCSE