[linux-2.6-block.git] / drivers / gpu / drm / amd / amdgpu / amdgpu_amdkfd.h

/*
 * Copyright 2014 Advanced Micro Devices, Inc.
 *
 * Permission is hereby granted, free of charge, to any person obtaining a
 * copy of this software and associated documentation files (the "Software"),
 * to deal in the Software without restriction, including without limitation
 * the rights to use, copy, modify, merge, publish, distribute, sublicense,
 * and/or sell copies of the Software, and to permit persons to whom the
 * Software is furnished to do so, subject to the following conditions:
 *
 * The above copyright notice and this permission notice shall be included in
 * all copies or substantial portions of the Software.
 *
 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
 * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
 * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
 * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
 * OTHER DEALINGS IN THE SOFTWARE.
 */

/* amdgpu_amdkfd.h defines the private interface between amdgpu and amdkfd. */

#ifndef AMDGPU_AMDKFD_H_INCLUDED
#define AMDGPU_AMDKFD_H_INCLUDED

#include <linux/list.h>
#include <linux/types.h>
#include <linux/mm.h>
#include <linux/kthread.h>
#include <linux/workqueue.h>
#include <linux/mmu_notifier.h>
#include <linux/memremap.h>
#include <kgd_kfd_interface.h>
#include "amdgpu_sync.h"
#include "amdgpu_vm.h"
#include "amdgpu_xcp.h"

extern uint64_t amdgpu_amdkfd_total_mem_size;

enum TLB_FLUSH_TYPE {
	TLB_FLUSH_LEGACY = 0,
	TLB_FLUSH_LIGHTWEIGHT,
	TLB_FLUSH_HEAVYWEIGHT
};

struct amdgpu_device;

enum kfd_mem_attachment_type {
	KFD_MEM_ATT_SHARED,	/* Share kgd_mem->bo or another attachment's */
	KFD_MEM_ATT_USERPTR,	/* SG bo to DMA map pages from a userptr bo */
	KFD_MEM_ATT_DMABUF,	/* DMAbuf to DMA map TTM BOs */
	KFD_MEM_ATT_SG		/* Tag to DMA map SG BOs */
};

struct kfd_mem_attachment {
	struct list_head list;
	enum kfd_mem_attachment_type type;
	bool is_mapped;
	struct amdgpu_bo_va *bo_va;
	struct amdgpu_device *adev;
	uint64_t va;
	uint64_t pte_flags;
};

struct kgd_mem {
	struct mutex lock;
	struct amdgpu_bo *bo;
	struct dma_buf *dmabuf;
	struct hmm_range *range;
	struct list_head attachments;
	/* protected by amdkfd_process_info.lock */
	struct list_head validate_list;
	uint32_t domain;
	unsigned int mapped_to_gpu_memory;
	uint64_t va;

	uint32_t alloc_flags;

	uint32_t invalid;
	struct amdkfd_process_info *process_info;

	struct amdgpu_sync sync;

	bool aql_queue;
	bool is_imported;
};

/* KFD Memory Eviction */
struct amdgpu_amdkfd_fence {
	struct dma_fence base;
	struct mm_struct *mm;
	spinlock_t lock;
	char timeline_name[TASK_COMM_LEN];
	struct svm_range_bo *svm_bo;
};

struct amdgpu_kfd_dev {
	struct kfd_dev *dev;
	int64_t vram_used[MAX_XCP];
	uint64_t vram_used_aligned[MAX_XCP];
	bool init_complete;
	struct work_struct reset_work;

	/* HMM page migration MEMORY_DEVICE_PRIVATE mapping */
	struct dev_pagemap pgmap;
};

enum kgd_engine_type {
	KGD_ENGINE_PFP = 1,
	KGD_ENGINE_ME,
	KGD_ENGINE_CE,
	KGD_ENGINE_MEC1,
	KGD_ENGINE_MEC2,
	KGD_ENGINE_RLC,
	KGD_ENGINE_SDMA1,
	KGD_ENGINE_SDMA2,
	KGD_ENGINE_MAX
};


struct amdkfd_process_info {
	/* List head of all VMs that belong to a KFD process */
	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;

	/* Number of VMs */
	unsigned int n_vms;
	/* Eviction Fence */
	struct amdgpu_amdkfd_fence *eviction_fence;

	/* MMU-notifier related fields */
	struct mutex notifier_lock;
	uint32_t evicted_bos;
	struct delayed_work restore_userptr_work;
	struct pid *pid;
	bool block_mmu_notifications;
};

int amdgpu_amdkfd_init(void);
void amdgpu_amdkfd_fini(void);

void amdgpu_amdkfd_suspend(struct amdgpu_device *adev, bool run_pm);
int amdgpu_amdkfd_resume(struct amdgpu_device *adev, bool run_pm);
void amdgpu_amdkfd_interrupt(struct amdgpu_device *adev,
			const void *ih_ring_entry);
void amdgpu_amdkfd_device_probe(struct amdgpu_device *adev);
void amdgpu_amdkfd_device_init(struct amdgpu_device *adev);
void amdgpu_amdkfd_device_fini_sw(struct amdgpu_device *adev);
int amdgpu_amdkfd_check_and_lock_kfd(struct amdgpu_device *adev);
void amdgpu_amdkfd_unlock_kfd(struct amdgpu_device *adev);
int amdgpu_amdkfd_submit_ib(struct amdgpu_device *adev,
				enum kgd_engine_type engine,
				uint32_t vmid, uint64_t gpu_addr,
				uint32_t *ib_cmd, uint32_t ib_len);
void amdgpu_amdkfd_set_compute_idle(struct amdgpu_device *adev, bool idle);
bool amdgpu_amdkfd_have_atomics_support(struct amdgpu_device *adev);
int amdgpu_amdkfd_flush_gpu_tlb_vmid(struct amdgpu_device *adev,
				uint16_t vmid);
int amdgpu_amdkfd_flush_gpu_tlb_pasid(struct amdgpu_device *adev,
				uint16_t pasid, enum TLB_FLUSH_TYPE flush_type,
				uint32_t inst);

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

int amdgpu_amdkfd_pre_reset(struct amdgpu_device *adev);

int amdgpu_amdkfd_post_reset(struct amdgpu_device *adev);

void amdgpu_amdkfd_gpu_reset(struct amdgpu_device *adev);

int amdgpu_queue_mask_bit_to_set_resource_bit(struct amdgpu_device *adev,
					int queue_bit);

struct amdgpu_amdkfd_fence *amdgpu_amdkfd_fence_create(u64 context,
				struct mm_struct *mm,
				struct svm_range_bo *svm_bo);
#if defined(CONFIG_DEBUG_FS)
int kfd_debugfs_kfd_mem_limits(struct seq_file *m, void *data);
#endif
#if IS_ENABLED(CONFIG_HSA_AMD)
bool amdkfd_fence_check_mm(struct dma_fence *f, struct mm_struct *mm);
struct amdgpu_amdkfd_fence *to_amdgpu_amdkfd_fence(struct dma_fence *f);
int amdgpu_amdkfd_remove_fence_on_pt_pd_bos(struct amdgpu_bo *bo);
int amdgpu_amdkfd_evict_userptr(struct mmu_interval_notifier *mni,
				unsigned long cur_seq, struct kgd_mem *mem);
#else
static inline
bool amdkfd_fence_check_mm(struct dma_fence *f, struct mm_struct *mm)
{
	return false;
}

static inline
struct amdgpu_amdkfd_fence *to_amdgpu_amdkfd_fence(struct dma_fence *f)
{
	return NULL;
}

static inline
int amdgpu_amdkfd_remove_fence_on_pt_pd_bos(struct amdgpu_bo *bo)
{
	return 0;
}

static inline
int amdgpu_amdkfd_evict_userptr(struct mmu_interval_notifier *mni,
				unsigned long cur_seq, struct kgd_mem *mem)
{
	return 0;
}
#endif
/* Shared API */
int amdgpu_amdkfd_alloc_gtt_mem(struct amdgpu_device *adev, size_t size,
				void **mem_obj, uint64_t *gpu_addr,
				void **cpu_ptr, bool mqd_gfx9);
void amdgpu_amdkfd_free_gtt_mem(struct amdgpu_device *adev, void *mem_obj);
int amdgpu_amdkfd_alloc_gws(struct amdgpu_device *adev, size_t size,
				void **mem_obj);
void amdgpu_amdkfd_free_gws(struct amdgpu_device *adev, void *mem_obj);
int amdgpu_amdkfd_add_gws_to_process(void *info, void *gws, struct kgd_mem **mem);
int amdgpu_amdkfd_remove_gws_from_process(void *info, void *mem);
uint32_t amdgpu_amdkfd_get_fw_version(struct amdgpu_device *adev,
				      enum kgd_engine_type type);
void amdgpu_amdkfd_get_local_mem_info(struct amdgpu_device *adev,
				      struct kfd_local_mem_info *mem_info,
				      struct amdgpu_xcp *xcp);
uint64_t amdgpu_amdkfd_get_gpu_clock_counter(struct amdgpu_device *adev);

uint32_t amdgpu_amdkfd_get_max_engine_clock_in_mhz(struct amdgpu_device *adev);
void amdgpu_amdkfd_get_cu_info(struct amdgpu_device *adev,
			       struct kfd_cu_info *cu_info);
int amdgpu_amdkfd_get_dmabuf_info(struct amdgpu_device *adev, int dma_buf_fd,
				  struct amdgpu_device **dmabuf_adev,
				  uint64_t *bo_size, void *metadata_buffer,
				  size_t buffer_size, uint32_t *metadata_size,
				  uint32_t *flags, int8_t *xcp_id);
uint8_t amdgpu_amdkfd_get_xgmi_hops_count(struct amdgpu_device *dst,
					  struct amdgpu_device *src);
int amdgpu_amdkfd_get_xgmi_bandwidth_mbytes(struct amdgpu_device *dst,
					    struct amdgpu_device *src,
					    bool is_min);
int amdgpu_amdkfd_get_pcie_bandwidth_mbytes(struct amdgpu_device *adev, bool is_min);
int amdgpu_amdkfd_send_close_event_drain_irq(struct amdgpu_device *adev,
					uint32_t *payload);
int amdgpu_amdkfd_unmap_hiq(struct amdgpu_device *adev, u32 doorbell_off,
				u32 inst);

/* Read user wptr from a specified user address space with page fault
 * disabled. The memory must be pinned and mapped to the hardware when
 * this is called in hqd_load functions, so it should never fault in
 * the first place. This resolves a circular lock dependency involving
 * four locks, including the DQM lock and mmap_lock.
 */
#define read_user_wptr(mmptr, wptr, dst)				\
	({								\
		bool valid = false;					\
		if ((mmptr) && (wptr)) {				\
			pagefault_disable();				\
			if ((mmptr) == current->mm) {			\
				valid = !get_user((dst), (wptr));	\
			} else if (current->flags & PF_KTHREAD) {	\
				kthread_use_mm(mmptr);			\
				valid = !get_user((dst), (wptr));	\
				kthread_unuse_mm(mmptr);		\
			}						\
			pagefault_enable();				\
		}							\
		valid;							\
	})

/* GPUVM API */
#define drm_priv_to_vm(drm_priv)					\
	(&((struct amdgpu_fpriv *)					\
		((struct drm_file *)(drm_priv))->driver_priv)->vm)

int amdgpu_amdkfd_gpuvm_set_vm_pasid(struct amdgpu_device *adev,
				     struct amdgpu_vm *avm, u32 pasid);
int amdgpu_amdkfd_gpuvm_acquire_process_vm(struct amdgpu_device *adev,
					struct amdgpu_vm *avm,
					void **process_info,
					struct dma_fence **ef);
void amdgpu_amdkfd_gpuvm_release_process_vm(struct amdgpu_device *adev,
					void *drm_priv);
uint64_t amdgpu_amdkfd_gpuvm_get_process_page_dir(void *drm_priv);
size_t amdgpu_amdkfd_get_available_memory(struct amdgpu_device *adev,
					uint8_t xcp_id);
int amdgpu_amdkfd_gpuvm_alloc_memory_of_gpu(
		struct amdgpu_device *adev, uint64_t va, uint64_t size,
		void *drm_priv, struct kgd_mem **mem,
		uint64_t *offset, uint32_t flags, bool criu_resume);
int amdgpu_amdkfd_gpuvm_free_memory_of_gpu(
		struct amdgpu_device *adev, struct kgd_mem *mem, void *drm_priv,
		uint64_t *size);
int amdgpu_amdkfd_gpuvm_map_memory_to_gpu(struct amdgpu_device *adev,
					  struct kgd_mem *mem, void *drm_priv);
int amdgpu_amdkfd_gpuvm_unmap_memory_from_gpu(
		struct amdgpu_device *adev, struct kgd_mem *mem, void *drm_priv);
int amdgpu_amdkfd_gpuvm_sync_memory(
		struct amdgpu_device *adev, struct kgd_mem *mem, bool intr);
int amdgpu_amdkfd_gpuvm_map_gtt_bo_to_kernel(struct kgd_mem *mem,
					     void **kptr, uint64_t *size);
void amdgpu_amdkfd_gpuvm_unmap_gtt_bo_from_kernel(struct kgd_mem *mem);

int amdgpu_amdkfd_map_gtt_bo_to_gart(struct amdgpu_device *adev, struct amdgpu_bo *bo);

int amdgpu_amdkfd_gpuvm_restore_process_bos(void *process_info,
					    struct dma_fence **ef);
int amdgpu_amdkfd_gpuvm_get_vm_fault_info(struct amdgpu_device *adev,
					      struct kfd_vm_fault_info *info);
int amdgpu_amdkfd_gpuvm_import_dmabuf(struct amdgpu_device *adev,
				      struct dma_buf *dmabuf,
				      uint64_t va, void *drm_priv,
				      struct kgd_mem **mem, uint64_t *size,
				      uint64_t *mmap_offset);
int amdgpu_amdkfd_gpuvm_export_dmabuf(struct kgd_mem *mem,
				      struct dma_buf **dmabuf);
void amdgpu_amdkfd_debug_mem_fence(struct amdgpu_device *adev);
int amdgpu_amdkfd_get_tile_config(struct amdgpu_device *adev,
				struct tile_config *config);
void amdgpu_amdkfd_ras_poison_consumption_handler(struct amdgpu_device *adev,
				bool reset);
bool amdgpu_amdkfd_bo_mapped_to_dev(struct amdgpu_device *adev, struct kgd_mem *mem);
void amdgpu_amdkfd_block_mmu_notifications(void *p);
int amdgpu_amdkfd_criu_resume(void *p);
bool amdgpu_amdkfd_ras_query_utcl2_poison_status(struct amdgpu_device *adev);
int amdgpu_amdkfd_reserve_mem_limit(struct amdgpu_device *adev,
		uint64_t size, u32 alloc_flag, int8_t xcp_id);
void amdgpu_amdkfd_unreserve_mem_limit(struct amdgpu_device *adev,
		uint64_t size, u32 alloc_flag, int8_t xcp_id);

u64 amdgpu_amdkfd_xcp_memory_size(struct amdgpu_device *adev, int xcp_id);

#define KFD_XCP_MEM_ID(adev, xcp_id) \
		((adev)->xcp_mgr && (xcp_id) >= 0 ?\
		(adev)->xcp_mgr->xcp[(xcp_id)].mem_id : -1)

#define KFD_XCP_MEMORY_SIZE(adev, xcp_id) amdgpu_amdkfd_xcp_memory_size((adev), (xcp_id))


#if IS_ENABLED(CONFIG_HSA_AMD)
void amdgpu_amdkfd_gpuvm_init_mem_limits(void);
void amdgpu_amdkfd_gpuvm_destroy_cb(struct amdgpu_device *adev,
				struct amdgpu_vm *vm);

/**
 * @amdgpu_amdkfd_release_notify() - Notify KFD when GEM object is released
 *
 * Allows KFD to release its resources associated with the GEM object.
 */
void amdgpu_amdkfd_release_notify(struct amdgpu_bo *bo);
void amdgpu_amdkfd_reserve_system_mem(uint64_t size);
#else
static inline
void amdgpu_amdkfd_gpuvm_init_mem_limits(void)
{
}

static inline
void amdgpu_amdkfd_gpuvm_destroy_cb(struct amdgpu_device *adev,
					struct amdgpu_vm *vm)
{
}

static inline
void amdgpu_amdkfd_release_notify(struct amdgpu_bo *bo)
{
}
#endif

#if IS_ENABLED(CONFIG_HSA_AMD_SVM)
int kgd2kfd_init_zone_device(struct amdgpu_device *adev);
#else
static inline
int kgd2kfd_init_zone_device(struct amdgpu_device *adev)
{
	return 0;
}
#endif

/* KGD2KFD callbacks */
int kgd2kfd_quiesce_mm(struct mm_struct *mm, uint32_t trigger);
int kgd2kfd_resume_mm(struct mm_struct *mm);
int kgd2kfd_schedule_evict_and_restore_process(struct mm_struct *mm,
						struct dma_fence *fence);
#if IS_ENABLED(CONFIG_HSA_AMD)
int kgd2kfd_init(void);
void kgd2kfd_exit(void);
struct kfd_dev *kgd2kfd_probe(struct amdgpu_device *adev, bool vf);
bool kgd2kfd_device_init(struct kfd_dev *kfd,
			 const struct kgd2kfd_shared_resources *gpu_resources);
void kgd2kfd_device_exit(struct kfd_dev *kfd);
void kgd2kfd_suspend(struct kfd_dev *kfd, bool run_pm);
int kgd2kfd_resume(struct kfd_dev *kfd, bool run_pm);
int kgd2kfd_pre_reset(struct kfd_dev *kfd);
int kgd2kfd_post_reset(struct kfd_dev *kfd);
void kgd2kfd_interrupt(struct kfd_dev *kfd, const void *ih_ring_entry);
void kgd2kfd_set_sram_ecc_flag(struct kfd_dev *kfd);
void kgd2kfd_smi_event_throttle(struct kfd_dev *kfd, uint64_t throttle_bitmask);
int kgd2kfd_check_and_lock_kfd(void);
void kgd2kfd_unlock_kfd(void);
#else
static inline int kgd2kfd_init(void)
{
	return -ENOENT;
}

static inline void kgd2kfd_exit(void)
{
}

static inline
struct kfd_dev *kgd2kfd_probe(struct amdgpu_device *adev, bool vf)
{
	return NULL;
}

static inline
bool kgd2kfd_device_init(struct kfd_dev *kfd,
				const struct kgd2kfd_shared_resources *gpu_resources)
{
	return false;
}

static inline void kgd2kfd_device_exit(struct kfd_dev *kfd)
{
}

static inline void kgd2kfd_suspend(struct kfd_dev *kfd, bool run_pm)
{
}

static inline int kgd2kfd_resume(struct kfd_dev *kfd, bool run_pm)
{
	return 0;
}

static inline int kgd2kfd_pre_reset(struct kfd_dev *kfd)
{
	return 0;
}

static inline int kgd2kfd_post_reset(struct kfd_dev *kfd)
{
	return 0;
}

static inline
void kgd2kfd_interrupt(struct kfd_dev *kfd, const void *ih_ring_entry)
{
}

static inline
void kgd2kfd_set_sram_ecc_flag(struct kfd_dev *kfd)
{
}

static inline
void kgd2kfd_smi_event_throttle(struct kfd_dev *kfd, uint64_t throttle_bitmask)
{
}

static inline int kgd2kfd_check_and_lock_kfd(void)
{
	return 0;
}

static inline void kgd2kfd_unlock_kfd(void)
{
}
#endif
#endif /* AMDGPU_AMDKFD_H_INCLUDED */
Commit	Line	Data
130e0371 OG	1	/*
	2	* Copyright 2014 Advanced Micro Devices, Inc.
	3	*
	4	* Permission is hereby granted, free of charge, to any person obtaining a
	5	* copy of this software and associated documentation files (the "Software"),
	6	* to deal in the Software without restriction, including without limitation
	7	* the rights to use, copy, modify, merge, publish, distribute, sublicense,
	8	* and/or sell copies of the Software, and to permit persons to whom the
	9	* Software is furnished to do so, subject to the following conditions:
	10	*
	11	* The above copyright notice and this permission notice shall be included in
	12	* all copies or substantial portions of the Software.
	13	*
	14	* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
	15	* IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
	16	* FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
	17	* THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
	18	* OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
	19	* ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
	20	* OTHER DEALINGS IN THE SOFTWARE.
	21	*/
	22
	23	/* amdgpu_amdkfd.h defines the private interface between amdgpu and amdkfd. */
	24
	25	#ifndef AMDGPU_AMDKFD_H_INCLUDED
	26	#define AMDGPU_AMDKFD_H_INCLUDED
	27
8abc1eb2	28	#include <linux/list.h>
130e0371	29	#include <linux/types.h>
7420f482	30	#include <linux/mm.h>
9bf5b9eb	31	#include <linux/kthread.h>
5ae0283e	32	#include <linux/workqueue.h>
f95f51a4	33	#include <linux/mmu_notifier.h>
610dab11	34	#include <linux/memremap.h>
130e0371	35	#include <kgd_kfd_interface.h>
a46a2cd1 FK	36	#include "amdgpu_sync.h"
a46a2cd1 FK	37	#include "amdgpu_vm.h"
1c77527a	38	#include "amdgpu_xcp.h"
130e0371	39
611736d8	40	extern uint64_t amdgpu_amdkfd_total_mem_size;
d8d019cc	41
765385ec PY	42	enum TLB_FLUSH_TYPE {
	43	TLB_FLUSH_LEGACY = 0,
	44	TLB_FLUSH_LIGHTWEIGHT,
	45	TLB_FLUSH_HEAVYWEIGHT
	46	};
	47
130e0371 OG	48	struct amdgpu_device;
130e0371 OG	49
264fb4d3 FK	50	enum kfd_mem_attachment_type {
	51	KFD_MEM_ATT_SHARED, /* Share kgd_mem->bo or another attachment's */
	52	KFD_MEM_ATT_USERPTR, /* SG bo to DMA map pages from a userptr bo */
5ac3c3e4	53	KFD_MEM_ATT_DMABUF, /* DMAbuf to DMA map TTM BOs */
08a2fd23	54	KFD_MEM_ATT_SG /* Tag to DMA map SG BOs */
264fb4d3 FK	55	};
264fb4d3 FK	56
c780b2ee FK	57	struct kfd_mem_attachment {
c780b2ee FK	58	struct list_head list;
264fb4d3	59	enum kfd_mem_attachment_type type;
a46a2cd1	60	bool is_mapped;
a46a2cd1	61	struct amdgpu_bo_va *bo_va;
c780b2ee	62	struct amdgpu_device *adev;
a46a2cd1 FK	63	uint64_t va;
	64	uint64_t pte_flags;
	65	};
	66
130e0371	67	struct kgd_mem {
a46a2cd1	68	struct mutex lock;
130e0371	69	struct amdgpu_bo *bo;
5ac3c3e4	70	struct dma_buf *dmabuf;
f95f51a4	71	struct hmm_range *range;
c780b2ee	72	struct list_head attachments;
a46a2cd1	73	/* protected by amdkfd_process_info.lock */
8abc1eb2	74	struct list_head validate_list;
a46a2cd1 FK	75	uint32_t domain;
	76	unsigned int mapped_to_gpu_memory;
	77	uint64_t va;
	78
d0ba51b1	79	uint32_t alloc_flags;
a46a2cd1	80
f95f51a4	81	uint32_t invalid;
a46a2cd1 FK	82	struct amdkfd_process_info *process_info;
	83
	84	struct amdgpu_sync sync;
	85
	86	bool aql_queue;
d4566dee	87	bool is_imported;
130e0371 OG	88	};
130e0371 OG	89
d8d019cc FK	90	/* KFD Memory Eviction */
	91	struct amdgpu_amdkfd_fence {
	92	struct dma_fence base;
	93	struct mm_struct *mm;
	94	spinlock_t lock;
	95	char timeline_name[TASK_COMM_LEN];
eb2cec55	96	struct svm_range_bo *svm_bo;
d8d019cc FK	97	};
d8d019cc FK	98
611736d8 FK	99	struct amdgpu_kfd_dev {
611736d8 FK	100	struct kfd_dev *dev;
1c77527a MJ	101	int64_t vram_used[MAX_XCP];
1c77527a MJ	102	uint64_t vram_used_aligned[MAX_XCP];
8e2712e7	103	bool init_complete;
b5fd0cf3	104	struct work_struct reset_work;
610dab11 PY	105
	106	/* HMM page migration MEMORY_DEVICE_PRIVATE mapping */
	107	struct dev_pagemap pgmap;
611736d8 FK	108	};
611736d8 FK	109
0da8b10e AL	110	enum kgd_engine_type {
	111	KGD_ENGINE_PFP = 1,
	112	KGD_ENGINE_ME,
	113	KGD_ENGINE_CE,
	114	KGD_ENGINE_MEC1,
	115	KGD_ENGINE_MEC2,
	116	KGD_ENGINE_RLC,
	117	KGD_ENGINE_SDMA1,
	118	KGD_ENGINE_SDMA2,
	119	KGD_ENGINE_MAX
	120	};
	121
d8d019cc	122
a46a2cd1 FK	123	struct amdkfd_process_info {
	124	/* List head of all VMs that belong to a KFD process */
	125	struct list_head vm_list_head;
	126	/* List head for all KFD BOs that belong to a KFD process. */
	127	struct list_head kfd_bo_list;
5ae0283e FK	128	/* List of userptr BOs that are valid or invalid */
	129	struct list_head userptr_valid_list;
	130	struct list_head userptr_inval_list;
a46a2cd1 FK	131	/* Lock to protect kfd_bo_list */
	132	struct mutex lock;
	133
	134	/* Number of VMs */
	135	unsigned int n_vms;
	136	/* Eviction Fence */
	137	struct amdgpu_amdkfd_fence *eviction_fence;
5ae0283e FK	138
5ae0283e FK	139	/* MMU-notifier related fields */
f95f51a4 FK	140	struct mutex notifier_lock;
f95f51a4 FK	141	uint32_t evicted_bos;
5ae0283e FK	142	struct delayed_work restore_userptr_work;
5ae0283e FK	143	struct pid *pid;
011bbb03	144	bool block_mmu_notifications;
a46a2cd1 FK	145	};
a46a2cd1 FK	146
efb1c658	147	int amdgpu_amdkfd_init(void);
130e0371 OG	148	void amdgpu_amdkfd_fini(void);
130e0371 OG	149
9593f4d6 RB	150	void amdgpu_amdkfd_suspend(struct amdgpu_device *adev, bool run_pm);
9593f4d6 RB	151	int amdgpu_amdkfd_resume(struct amdgpu_device *adev, bool run_pm);
dc102c43	152	void amdgpu_amdkfd_interrupt(struct amdgpu_device *adev,
130e0371	153	const void *ih_ring_entry);
dc102c43 AR	154	void amdgpu_amdkfd_device_probe(struct amdgpu_device *adev);
dc102c43 AR	155	void amdgpu_amdkfd_device_init(struct amdgpu_device *adev);
e9669fb7	156	void amdgpu_amdkfd_device_fini_sw(struct amdgpu_device *adev);
0c7315e7 MJ	157	int amdgpu_amdkfd_check_and_lock_kfd(struct amdgpu_device *adev);
0c7315e7 MJ	158	void amdgpu_amdkfd_unlock_kfd(struct amdgpu_device *adev);
6bfc7c7e GS	159	int amdgpu_amdkfd_submit_ib(struct amdgpu_device *adev,
6bfc7c7e GS	160	enum kgd_engine_type engine,
4c660c8f FK	161	uint32_t vmid, uint64_t gpu_addr,
4c660c8f FK	162	uint32_t *ib_cmd, uint32_t ib_len);
6bfc7c7e GS	163	void amdgpu_amdkfd_set_compute_idle(struct amdgpu_device *adev, bool idle);
	164	bool amdgpu_amdkfd_have_atomics_support(struct amdgpu_device *adev);
	165	int amdgpu_amdkfd_flush_gpu_tlb_vmid(struct amdgpu_device *adev,
	166	uint16_t vmid);
	167	int amdgpu_amdkfd_flush_gpu_tlb_pasid(struct amdgpu_device *adev,
f87f6864 MJ	168	uint16_t pasid, enum TLB_FLUSH_TYPE flush_type,
f87f6864 MJ	169	uint32_t inst);
4c660c8f	170
155494db FK	171	bool amdgpu_amdkfd_is_kfd_vmid(struct amdgpu_device *adev, u32 vmid);
155494db FK	172
5c6dd71e SL	173	int amdgpu_amdkfd_pre_reset(struct amdgpu_device *adev);
	174
	175	int amdgpu_amdkfd_post_reset(struct amdgpu_device *adev);
	176
6bfc7c7e	177	void amdgpu_amdkfd_gpu_reset(struct amdgpu_device *adev);
24da5a9c	178
d09f85d5 YZ	179	int amdgpu_queue_mask_bit_to_set_resource_bit(struct amdgpu_device *adev,
	180	int queue_bit);
	181
cd63989e	182	struct amdgpu_amdkfd_fence *amdgpu_amdkfd_fence_create(u64 context,
eb2cec55 AS	183	struct mm_struct *mm,
eb2cec55 AS	184	struct svm_range_bo *svm_bo);
3d2af401 AS	185	#if defined(CONFIG_DEBUG_FS)
	186	int kfd_debugfs_kfd_mem_limits(struct seq_file m, void data);
	187	#endif
cd63989e LY	188	#if IS_ENABLED(CONFIG_HSA_AMD)
	189	bool amdkfd_fence_check_mm(struct dma_fence f, struct mm_struct mm);
	190	struct amdgpu_amdkfd_fence to_amdgpu_amdkfd_fence(struct dma_fence f);
	191	int amdgpu_amdkfd_remove_fence_on_pt_pd_bos(struct amdgpu_bo *bo);
f95f51a4 FK	192	int amdgpu_amdkfd_evict_userptr(struct mmu_interval_notifier *mni,
f95f51a4 FK	193	unsigned long cur_seq, struct kgd_mem *mem);
cd63989e LY	194	#else
	195	static inline
	196	bool amdkfd_fence_check_mm(struct dma_fence f, struct mm_struct mm)
	197	{
	198	return false;
	199	}
	200
	201	static inline
	202	struct amdgpu_amdkfd_fence to_amdgpu_amdkfd_fence(struct dma_fence f)
	203	{
	204	return NULL;
	205	}
	206
	207	static inline
	208	int amdgpu_amdkfd_remove_fence_on_pt_pd_bos(struct amdgpu_bo *bo)
	209	{
	210	return 0;
	211	}
	212
	213	static inline
f95f51a4 FK	214	int amdgpu_amdkfd_evict_userptr(struct mmu_interval_notifier *mni,
f95f51a4 FK	215	unsigned long cur_seq, struct kgd_mem *mem)
cd63989e LY	216	{
	217	return 0;
	218	}
	219	#endif
130e0371	220	/* Shared API */
6bfc7c7e	221	int amdgpu_amdkfd_alloc_gtt_mem(struct amdgpu_device *adev, size_t size,
7cd52c91 AL	222	void *mem_obj, uint64_t gpu_addr,
7cd52c91 AL	223	void **cpu_ptr, bool mqd_gfx9);
6bfc7c7e GS	224	void amdgpu_amdkfd_free_gtt_mem(struct amdgpu_device adev, void mem_obj);
	225	int amdgpu_amdkfd_alloc_gws(struct amdgpu_device *adev, size_t size,
	226	void **mem_obj);
	227	void amdgpu_amdkfd_free_gws(struct amdgpu_device adev, void mem_obj);
71efab6a OZ	228	int amdgpu_amdkfd_add_gws_to_process(void info, void gws, struct kgd_mem **mem);
71efab6a OZ	229	int amdgpu_amdkfd_remove_gws_from_process(void info, void mem);
574c4183	230	uint32_t amdgpu_amdkfd_get_fw_version(struct amdgpu_device *adev,
0da8b10e	231	enum kgd_engine_type type);
574c4183	232	void amdgpu_amdkfd_get_local_mem_info(struct amdgpu_device *adev,
315e29ec	233	struct kfd_local_mem_info *mem_info,
9a3ce1a7	234	struct amdgpu_xcp *xcp);
574c4183	235	uint64_t amdgpu_amdkfd_get_gpu_clock_counter(struct amdgpu_device *adev);
7cd52c91	236
574c4183 GS	237	uint32_t amdgpu_amdkfd_get_max_engine_clock_in_mhz(struct amdgpu_device *adev);
	238	void amdgpu_amdkfd_get_cu_info(struct amdgpu_device *adev,
	239	struct kfd_cu_info *cu_info);
	240	int amdgpu_amdkfd_get_dmabuf_info(struct amdgpu_device *adev, int dma_buf_fd,
	241	struct amdgpu_device **dmabuf_adev,
1dde0ea9 FK	242	uint64_t bo_size, void metadata_buffer,
1dde0ea9 FK	243	size_t buffer_size, uint32_t *metadata_size,
2fa9ff25	244	uint32_t flags, int8_t xcp_id);
574c4183 GS	245	uint8_t amdgpu_amdkfd_get_xgmi_hops_count(struct amdgpu_device *dst,
	246	struct amdgpu_device *src);
	247	int amdgpu_amdkfd_get_xgmi_bandwidth_mbytes(struct amdgpu_device *dst,
	248	struct amdgpu_device *src,
	249	bool is_min);
	250	int amdgpu_amdkfd_get_pcie_bandwidth_mbytes(struct amdgpu_device *adev, bool is_min);
12fb1ad7 JK	251	int amdgpu_amdkfd_send_close_event_drain_irq(struct amdgpu_device *adev,
12fb1ad7 JK	252	uint32_t *payload);
9041b53a MJ	253	int amdgpu_amdkfd_unmap_hiq(struct amdgpu_device *adev, u32 doorbell_off,
9041b53a MJ	254	u32 inst);
130e0371	255
cd05c865 FK	256	/* Read user wptr from a specified user address space with page fault
	257	* disabled. The memory must be pinned and mapped to the hardware when
	258	* this is called in hqd_load functions, so it should never fault in
	259	* the first place. This resolves a circular lock dependency involving
c1e8d7c6	260	* four locks, including the DQM lock and mmap_lock.
cd05c865	261	*/
70539bd7 FK	262	#define read_user_wptr(mmptr, wptr, dst) \
	263	({ \
	264	bool valid = false; \
	265	if ((mmptr) && (wptr)) { \
cd05c865	266	pagefault_disable(); \
70539bd7 FK	267	if ((mmptr) == current->mm) { \
70539bd7 FK	268	valid = !get_user((dst), (wptr)); \
8449d150	269	} else if (current->flags & PF_KTHREAD) { \
f5678e7f	270	kthread_use_mm(mmptr); \
70539bd7	271	valid = !get_user((dst), (wptr)); \
f5678e7f	272	kthread_unuse_mm(mmptr); \
70539bd7	273	} \
cd05c865	274	pagefault_enable(); \
70539bd7 FK	275	} \
	276	valid; \
	277	})
	278
a46a2cd1	279	/* GPUVM API */
f80fe9d3 FK	280	#define drm_priv_to_vm(drm_priv) \
	281	(&((struct amdgpu_fpriv *) \
	282	((struct drm_file *)(drm_priv))->driver_priv)->vm)
	283
41d82649	284	int amdgpu_amdkfd_gpuvm_set_vm_pasid(struct amdgpu_device *adev,
23b02b0e	285	struct amdgpu_vm *avm, u32 pasid);
dff63da9	286	int amdgpu_amdkfd_gpuvm_acquire_process_vm(struct amdgpu_device *adev,
23b02b0e	287	struct amdgpu_vm *avm,
b40a6ab2	288	void **process_info,
fcdfa432	289	struct dma_fence **ef);
dff63da9 GS	290	void amdgpu_amdkfd_gpuvm_release_process_vm(struct amdgpu_device *adev,
dff63da9 GS	291	void *drm_priv);
b40a6ab2	292	uint64_t amdgpu_amdkfd_gpuvm_get_process_page_dir(void *drm_priv);
1c77527a MJ	293	size_t amdgpu_amdkfd_get_available_memory(struct amdgpu_device *adev,
1c77527a MJ	294	uint8_t xcp_id);
a46a2cd1	295	int amdgpu_amdkfd_gpuvm_alloc_memory_of_gpu(
dff63da9	296	struct amdgpu_device *adev, uint64_t va, uint64_t size,
b40a6ab2	297	void drm_priv, struct kgd_mem *mem,
011bbb03	298	uint64_t *offset, uint32_t flags, bool criu_resume);
a46a2cd1	299	int amdgpu_amdkfd_gpuvm_free_memory_of_gpu(
dff63da9	300	struct amdgpu_device adev, struct kgd_mem mem, void *drm_priv,
d4ec4bdc	301	uint64_t *size);
4d30a83c CK	302	int amdgpu_amdkfd_gpuvm_map_memory_to_gpu(struct amdgpu_device *adev,
4d30a83c CK	303	struct kgd_mem mem, void drm_priv);
a46a2cd1	304	int amdgpu_amdkfd_gpuvm_unmap_memory_from_gpu(
dff63da9	305	struct amdgpu_device adev, struct kgd_mem mem, void *drm_priv);
a46a2cd1	306	int amdgpu_amdkfd_gpuvm_sync_memory(
dff63da9	307	struct amdgpu_device adev, struct kgd_mem mem, bool intr);
4e2d1044 FK	308	int amdgpu_amdkfd_gpuvm_map_gtt_bo_to_kernel(struct kgd_mem *mem,
	309	void *kptr, uint64_t size);
	310	void amdgpu_amdkfd_gpuvm_unmap_gtt_bo_from_kernel(struct kgd_mem *mem);
68df0f19	311
e77a541f GS	312	int amdgpu_amdkfd_map_gtt_bo_to_gart(struct amdgpu_device adev, struct amdgpu_bo bo);
e77a541f GS	313
a46a2cd1 FK	314	int amdgpu_amdkfd_gpuvm_restore_process_bos(void *process_info,
a46a2cd1 FK	315	struct dma_fence **ef);
dff63da9	316	int amdgpu_amdkfd_gpuvm_get_vm_fault_info(struct amdgpu_device *adev,
b97dfa27	317	struct kfd_vm_fault_info *info);
dff63da9	318	int amdgpu_amdkfd_gpuvm_import_dmabuf(struct amdgpu_device *adev,
1dde0ea9	319	struct dma_buf *dmabuf,
b40a6ab2	320	uint64_t va, void *drm_priv,
1dde0ea9 FK	321	struct kgd_mem *mem, uint64_t size,
1dde0ea9 FK	322	uint64_t *mmap_offset);
fd234e75 FK	323	int amdgpu_amdkfd_gpuvm_export_dmabuf(struct kgd_mem *mem,
fd234e75 FK	324	struct dma_buf **dmabuf);
a70a93fa	325	void amdgpu_amdkfd_debug_mem_fence(struct amdgpu_device *adev);
dff63da9	326	int amdgpu_amdkfd_get_tile_config(struct amdgpu_device *adev,
fd7d08ba	327	struct tile_config *config);
b6485bed TZ	328	void amdgpu_amdkfd_ras_poison_consumption_handler(struct amdgpu_device *adev,
b6485bed TZ	329	bool reset);
5ccbb057	330	bool amdgpu_amdkfd_bo_mapped_to_dev(struct amdgpu_device adev, struct kgd_mem mem);
011bbb03 RB	331	void amdgpu_amdkfd_block_mmu_notifications(void *p);
011bbb03 RB	332	int amdgpu_amdkfd_criu_resume(void *p);
6475ae2b	333	bool amdgpu_amdkfd_ras_query_utcl2_poison_status(struct amdgpu_device *adev);
f9af3c16	334	int amdgpu_amdkfd_reserve_mem_limit(struct amdgpu_device *adev,
1c77527a	335	uint64_t size, u32 alloc_flag, int8_t xcp_id);
f9af3c16	336	void amdgpu_amdkfd_unreserve_mem_limit(struct amdgpu_device *adev,
1c77527a	337	uint64_t size, u32 alloc_flag, int8_t xcp_id);
011bbb03	338
45b3a914 AD	339	u64 amdgpu_amdkfd_xcp_memory_size(struct amdgpu_device *adev, int xcp_id);
45b3a914 AD	340
3ebfd221 PY	341	#define KFD_XCP_MEM_ID(adev, xcp_id) \
	342	((adev)->xcp_mgr && (xcp_id) >= 0 ?\
	343	(adev)->xcp_mgr->xcp[(xcp_id)].mem_id : -1)
	344
45b3a914 AD	345	#define KFD_XCP_MEMORY_SIZE(adev, xcp_id) amdgpu_amdkfd_xcp_memory_size((adev), (xcp_id))
45b3a914 AD	346
4c6ce75f	347
cd63989e LY	348	#if IS_ENABLED(CONFIG_HSA_AMD)
	349	void amdgpu_amdkfd_gpuvm_init_mem_limits(void);
	350	void amdgpu_amdkfd_gpuvm_destroy_cb(struct amdgpu_device *adev,
	351	struct amdgpu_vm *vm);
f441dd33 RE	352
	353	/**
	354	* @amdgpu_amdkfd_release_notify() - Notify KFD when GEM object is released
	355	*
	356	* Allows KFD to release its resources associated with the GEM object.
	357	*/
5702d052	358	void amdgpu_amdkfd_release_notify(struct amdgpu_bo *bo);
c46ebb6a	359	void amdgpu_amdkfd_reserve_system_mem(uint64_t size);
cd63989e LY	360	#else
	361	static inline
	362	void amdgpu_amdkfd_gpuvm_init_mem_limits(void)
	363	{
	364	}
fd7d08ba	365
cd63989e LY	366	static inline
	367	void amdgpu_amdkfd_gpuvm_destroy_cb(struct amdgpu_device *adev,
	368	struct amdgpu_vm *vm)
	369	{
	370	}
	371
	372	static inline
5702d052	373	void amdgpu_amdkfd_release_notify(struct amdgpu_bo *bo)
cd63989e LY	374	{
	375	}
	376	#endif
84b4dd3f PY	377
	378	#if IS_ENABLED(CONFIG_HSA_AMD_SVM)
	379	int kgd2kfd_init_zone_device(struct amdgpu_device *adev);
	380	#else
	381	static inline
	382	int kgd2kfd_init_zone_device(struct amdgpu_device *adev)
	383	{
	384	return 0;
	385	}
	386	#endif
	387
2d3d25b6	388	/* KGD2KFD callbacks */
c7f21978	389	int kgd2kfd_quiesce_mm(struct mm_struct *mm, uint32_t trigger);
cd63989e LY	390	int kgd2kfd_resume_mm(struct mm_struct *mm);
	391	int kgd2kfd_schedule_evict_and_restore_process(struct mm_struct *mm,
	392	struct dma_fence *fence);
	393	#if IS_ENABLED(CONFIG_HSA_AMD)
308176d6	394	int kgd2kfd_init(void);
2d3d25b6	395	void kgd2kfd_exit(void);
b5d1d755	396	struct kfd_dev kgd2kfd_probe(struct amdgpu_device adev, bool vf);
2d3d25b6 AL	397	bool kgd2kfd_device_init(struct kfd_dev *kfd,
	398	const struct kgd2kfd_shared_resources *gpu_resources);
	399	void kgd2kfd_device_exit(struct kfd_dev *kfd);
9593f4d6 RB	400	void kgd2kfd_suspend(struct kfd_dev *kfd, bool run_pm);
9593f4d6 RB	401	int kgd2kfd_resume(struct kfd_dev *kfd, bool run_pm);
2d3d25b6 AL	402	int kgd2kfd_pre_reset(struct kfd_dev *kfd);
	403	int kgd2kfd_post_reset(struct kfd_dev *kfd);
	404	void kgd2kfd_interrupt(struct kfd_dev kfd, const void ih_ring_entry);
9b54d201	405	void kgd2kfd_set_sram_ecc_flag(struct kfd_dev *kfd);
410e302e	406	void kgd2kfd_smi_event_throttle(struct kfd_dev *kfd, uint64_t throttle_bitmask);
0c7315e7 MJ	407	int kgd2kfd_check_and_lock_kfd(void);
0c7315e7 MJ	408	void kgd2kfd_unlock_kfd(void);
cd63989e LY	409	#else
	410	static inline int kgd2kfd_init(void)
	411	{
	412	return -ENOENT;
	413	}
2d3d25b6	414
cd63989e LY	415	static inline void kgd2kfd_exit(void)
	416	{
	417	}
	418
	419	static inline
b5d1d755	420	struct kfd_dev kgd2kfd_probe(struct amdgpu_device adev, bool vf)
cd63989e LY	421	{
	422	return NULL;
	423	}
	424
	425	static inline
d69a3b76	426	bool kgd2kfd_device_init(struct kfd_dev *kfd,
cd63989e LY	427	const struct kgd2kfd_shared_resources *gpu_resources)
	428	{
	429	return false;
	430	}
	431
	432	static inline void kgd2kfd_device_exit(struct kfd_dev *kfd)
	433	{
	434	}
	435
	436	static inline void kgd2kfd_suspend(struct kfd_dev *kfd, bool run_pm)
	437	{
	438	}
	439
	440	static inline int kgd2kfd_resume(struct kfd_dev *kfd, bool run_pm)
	441	{
	442	return 0;
	443	}
	444
	445	static inline int kgd2kfd_pre_reset(struct kfd_dev *kfd)
	446	{
	447	return 0;
	448	}
	449
	450	static inline int kgd2kfd_post_reset(struct kfd_dev *kfd)
	451	{
	452	return 0;
	453	}
	454
	455	static inline
	456	void kgd2kfd_interrupt(struct kfd_dev kfd, const void ih_ring_entry)
	457	{
	458	}
	459
	460	static inline
	461	void kgd2kfd_set_sram_ecc_flag(struct kfd_dev *kfd)
	462	{
	463	}
	464
	465	static inline
410e302e	466	void kgd2kfd_smi_event_throttle(struct kfd_dev *kfd, uint64_t throttle_bitmask)
cd63989e LY	467	{
cd63989e LY	468	}
0c7315e7 MJ	469
	470	static inline int kgd2kfd_check_and_lock_kfd(void)
	471	{
	472	return 0;
	473	}
	474
	475	static inline void kgd2kfd_unlock_kfd(void)
	476	{
	477	}
cd63989e	478	#endif
130e0371	479	#endif /* AMDGPU_AMDKFD_H_INCLUDED */