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

/*
 * Copyright 2016 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.
 *
 * Authors: Christian König
 */
#ifndef __AMDGPU_VM_H__
#define __AMDGPU_VM_H__

#include <linux/rbtree.h>

#include "gpu_scheduler.h"
#include "amdgpu_sync.h"
#include "amdgpu_ring.h"

struct amdgpu_bo_va;
struct amdgpu_job;
struct amdgpu_bo_list_entry;

/*
 * GPUVM handling
 */

/* maximum number of VMIDs */
#define AMDGPU_NUM_VM	16

/* Maximum number of PTEs the hardware can write with one command */
#define AMDGPU_VM_MAX_UPDATE_SIZE	0x3FFFF

/* number of entries in page table */
#define AMDGPU_VM_PTE_COUNT(adev) (1 << (adev)->vm_manager.block_size)

/* PTBs (Page Table Blocks) need to be aligned to 32K */
#define AMDGPU_VM_PTB_ALIGN_SIZE   32768

/* LOG2 number of continuous pages for the fragment field */
#define AMDGPU_LOG2_PAGES_PER_FRAG 4

#define AMDGPU_PTE_VALID	(1ULL << 0)
#define AMDGPU_PTE_SYSTEM	(1ULL << 1)
#define AMDGPU_PTE_SNOOPED	(1ULL << 2)

/* VI only */
#define AMDGPU_PTE_EXECUTABLE	(1ULL << 4)

#define AMDGPU_PTE_READABLE	(1ULL << 5)
#define AMDGPU_PTE_WRITEABLE	(1ULL << 6)

#define AMDGPU_PTE_FRAG(x)	((x & 0x1fULL) << 7)

/* TILED for VEGA10, reserved for older ASICs  */
#define AMDGPU_PTE_PRT		(1ULL << 51)

/* VEGA10 only */
#define AMDGPU_PTE_MTYPE(a)    ((uint64_t)a << 57)
#define AMDGPU_PTE_MTYPE_MASK	AMDGPU_PTE_MTYPE(3ULL)

/* How to programm VM fault handling */
#define AMDGPU_VM_FAULT_STOP_NEVER	0
#define AMDGPU_VM_FAULT_STOP_FIRST	1
#define AMDGPU_VM_FAULT_STOP_ALWAYS	2

/* max number of VMHUB */
#define AMDGPU_MAX_VMHUBS			2
#define AMDGPU_GFXHUB				0
#define AMDGPU_MMHUB				1

/* hardcode that limit for now */
#define AMDGPU_VA_RESERVED_SIZE			(8 << 20)
/* max vmids dedicated for process */
#define AMDGPU_VM_MAX_RESERVED_VMID	1

struct amdgpu_vm_pt {
	struct amdgpu_bo	*bo;
	uint64_t		addr;

	/* array of page tables, one for each directory entry */
	struct amdgpu_vm_pt	*entries;
	unsigned		last_entry_used;
};

struct amdgpu_vm {
	/* tree of virtual addresses mapped */
	struct rb_root		va;

	/* protecting invalidated */
	spinlock_t		status_lock;

	/* BOs moved, but not yet updated in the PT */
	struct list_head	invalidated;

	/* BOs cleared in the PT because of a move */
	struct list_head	cleared;

	/* BO mappings freed, but not yet updated in the PT */
	struct list_head	freed;

	/* contains the page directory */
	struct amdgpu_vm_pt     root;
	struct dma_fence	*last_dir_update;
	uint64_t		last_eviction_counter;

	/* protecting freed */
	spinlock_t		freed_lock;

	/* Scheduler entity for page table updates */
	struct amd_sched_entity	entity;

	/* client id */
	u64                     client_id;
	/* dedicated to vm */
	struct amdgpu_vm_id	*reserved_vmid[AMDGPU_MAX_VMHUBS];
	/* each VM will map on CSA */
	struct amdgpu_bo_va *csa_bo_va;
};

struct amdgpu_vm_id {
	struct list_head	list;
	struct amdgpu_sync	active;
	struct dma_fence		*last_flush;
	atomic64_t		owner;

	uint64_t		pd_gpu_addr;
	/* last flushed PD/PT update */
	struct dma_fence		*flushed_updates;

	uint32_t                current_gpu_reset_count;

	uint32_t		gds_base;
	uint32_t		gds_size;
	uint32_t		gws_base;
	uint32_t		gws_size;
	uint32_t		oa_base;
	uint32_t		oa_size;
};

struct amdgpu_vm_id_manager {
	struct mutex		lock;
	unsigned		num_ids;
	struct list_head	ids_lru;
	struct amdgpu_vm_id	ids[AMDGPU_NUM_VM];
	atomic_t		reserved_vmid_num;
};

struct amdgpu_vm_manager {
	/* Handling of VMIDs */
	struct amdgpu_vm_id_manager		id_mgr[AMDGPU_MAX_VMHUBS];

	/* Handling of VM fences */
	u64					fence_context;
	unsigned				seqno[AMDGPU_MAX_RINGS];

	uint64_t				max_pfn;
	uint32_t				num_level;
	uint64_t				vm_size;
	uint32_t				block_size;
	/* vram base address for page table entry  */
	u64					vram_base_offset;
	/* vm pte handling */
	const struct amdgpu_vm_pte_funcs        *vm_pte_funcs;
	struct amdgpu_ring                      *vm_pte_rings[AMDGPU_MAX_RINGS];
	unsigned				vm_pte_num_rings;
	atomic_t				vm_pte_next_ring;
	/* client id counter */
	atomic64_t				client_counter;

	/* partial resident texture handling */
	spinlock_t				prt_lock;
	atomic_t				num_prt_users;
};

void amdgpu_vm_manager_init(struct amdgpu_device *adev);
void amdgpu_vm_manager_fini(struct amdgpu_device *adev);
int amdgpu_vm_init(struct amdgpu_device *adev, struct amdgpu_vm *vm);
void amdgpu_vm_fini(struct amdgpu_device *adev, struct amdgpu_vm *vm);
void amdgpu_vm_get_pd_bo(struct amdgpu_vm *vm,
			 struct list_head *validated,
			 struct amdgpu_bo_list_entry *entry);
int amdgpu_vm_validate_pt_bos(struct amdgpu_device *adev, struct amdgpu_vm *vm,
			      int (*callback)(void *p, struct amdgpu_bo *bo),
			      void *param);
void amdgpu_vm_move_pt_bos_in_lru(struct amdgpu_device *adev,
				  struct amdgpu_vm *vm);
int amdgpu_vm_alloc_pts(struct amdgpu_device *adev,
			struct amdgpu_vm *vm,
			uint64_t saddr, uint64_t size);
int amdgpu_vm_grab_id(struct amdgpu_vm *vm, struct amdgpu_ring *ring,
		      struct amdgpu_sync *sync, struct dma_fence *fence,
		      struct amdgpu_job *job);
int amdgpu_vm_flush(struct amdgpu_ring *ring, struct amdgpu_job *job);
void amdgpu_vm_reset_id(struct amdgpu_device *adev, unsigned vmhub,
			unsigned vmid);
void amdgpu_vm_reset_all_ids(struct amdgpu_device *adev);
int amdgpu_vm_update_directories(struct amdgpu_device *adev,
				 struct amdgpu_vm *vm);
int amdgpu_vm_clear_freed(struct amdgpu_device *adev,
			  struct amdgpu_vm *vm,
			  struct dma_fence **fence);
int amdgpu_vm_clear_invalids(struct amdgpu_device *adev, struct amdgpu_vm *vm,
			     struct amdgpu_sync *sync);
int amdgpu_vm_bo_update(struct amdgpu_device *adev,
			struct amdgpu_bo_va *bo_va,
			bool clear);
void amdgpu_vm_bo_invalidate(struct amdgpu_device *adev,
			     struct amdgpu_bo *bo);
struct amdgpu_bo_va *amdgpu_vm_bo_find(struct amdgpu_vm *vm,
				       struct amdgpu_bo *bo);
struct amdgpu_bo_va *amdgpu_vm_bo_add(struct amdgpu_device *adev,
				      struct amdgpu_vm *vm,
				      struct amdgpu_bo *bo);
int amdgpu_vm_bo_map(struct amdgpu_device *adev,
		     struct amdgpu_bo_va *bo_va,
		     uint64_t addr, uint64_t offset,
		     uint64_t size, uint64_t flags);
int amdgpu_vm_bo_replace_map(struct amdgpu_device *adev,
			     struct amdgpu_bo_va *bo_va,
			     uint64_t addr, uint64_t offset,
			     uint64_t size, uint64_t flags);
int amdgpu_vm_bo_unmap(struct amdgpu_device *adev,
		       struct amdgpu_bo_va *bo_va,
		       uint64_t addr);
int amdgpu_vm_bo_clear_mappings(struct amdgpu_device *adev,
				struct amdgpu_vm *vm,
				uint64_t saddr, uint64_t size);
void amdgpu_vm_bo_rmv(struct amdgpu_device *adev,
		      struct amdgpu_bo_va *bo_va);
void amdgpu_vm_adjust_size(struct amdgpu_device *adev, uint64_t vm_size);
int amdgpu_vm_ioctl(struct drm_device *dev, void *data, struct drm_file *filp);
bool amdgpu_vm_need_pipeline_sync(struct amdgpu_ring *ring,
				  struct amdgpu_job *job);
void amdgpu_vm_check_compute_bug(struct amdgpu_device *adev);

#endif
Commit	Line	Data
073440d2 CK	1	/*
	2	* Copyright 2016 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	* Authors: Christian König
	23	*/
	24	#ifndef __AMDGPU_VM_H__
	25	#define __AMDGPU_VM_H__
	26
	27	#include <linux/rbtree.h>
	28
	29	#include "gpu_scheduler.h"
	30	#include "amdgpu_sync.h"
	31	#include "amdgpu_ring.h"
	32
	33	struct amdgpu_bo_va;
	34	struct amdgpu_job;
	35	struct amdgpu_bo_list_entry;
	36
	37	/*
	38	* GPUVM handling
	39	*/
	40
	41	/* maximum number of VMIDs */
	42	#define AMDGPU_NUM_VM 16
	43
	44	/* Maximum number of PTEs the hardware can write with one command */
	45	#define AMDGPU_VM_MAX_UPDATE_SIZE 0x3FFFF
	46
	47	/* number of entries in page table */
36b32a68	48	#define AMDGPU_VM_PTE_COUNT(adev) (1 << (adev)->vm_manager.block_size)
073440d2 CK	49
	50	/* PTBs (Page Table Blocks) need to be aligned to 32K */
	51	#define AMDGPU_VM_PTB_ALIGN_SIZE 32768
	52
	53	/* LOG2 number of continuous pages for the fragment field */
	54	#define AMDGPU_LOG2_PAGES_PER_FRAG 4
	55
35ba15f0 CK	56	#define AMDGPU_PTE_VALID (1ULL << 0)
	57	#define AMDGPU_PTE_SYSTEM (1ULL << 1)
	58	#define AMDGPU_PTE_SNOOPED (1ULL << 2)
073440d2 CK	59
073440d2 CK	60	/* VI only */
35ba15f0	61	#define AMDGPU_PTE_EXECUTABLE (1ULL << 4)
073440d2	62
35ba15f0 CK	63	#define AMDGPU_PTE_READABLE (1ULL << 5)
35ba15f0 CK	64	#define AMDGPU_PTE_WRITEABLE (1ULL << 6)
073440d2	65
982a1348	66	#define AMDGPU_PTE_FRAG(x) ((x & 0x1fULL) << 7)
073440d2	67
d0766e98 ZJ	68	/* TILED for VEGA10, reserved for older ASICs */
d0766e98 ZJ	69	#define AMDGPU_PTE_PRT (1ULL << 51)
284710fa	70
ca02061c AD	71	/* VEGA10 only */
	72	#define AMDGPU_PTE_MTYPE(a) ((uint64_t)a << 57)
	73	#define AMDGPU_PTE_MTYPE_MASK AMDGPU_PTE_MTYPE(3ULL)
	74
073440d2 CK	75	/* How to programm VM fault handling */
	76	#define AMDGPU_VM_FAULT_STOP_NEVER 0
	77	#define AMDGPU_VM_FAULT_STOP_FIRST 1
	78	#define AMDGPU_VM_FAULT_STOP_ALWAYS 2
	79
eb60ef2b CK	80	/* max number of VMHUB */
	81	#define AMDGPU_MAX_VMHUBS 2
	82	#define AMDGPU_GFXHUB 0
	83	#define AMDGPU_MMHUB 1
	84
	85	/* hardcode that limit for now */
	86	#define AMDGPU_VA_RESERVED_SIZE (8 << 20)
c3505770 CZ	87	/* max vmids dedicated for process */
c3505770 CZ	88	#define AMDGPU_VM_MAX_RESERVED_VMID 1
eb60ef2b	89
073440d2 CK	90	struct amdgpu_vm_pt {
	91	struct amdgpu_bo *bo;
	92	uint64_t addr;
67003a15 CK	93
	94	/* array of page tables, one for each directory entry */
	95	struct amdgpu_vm_pt *entries;
	96	unsigned last_entry_used;
073440d2 CK	97	};
	98
	99	struct amdgpu_vm {
	100	/* tree of virtual addresses mapped */
	101	struct rb_root va;
	102
	103	/* protecting invalidated */
	104	spinlock_t status_lock;
	105
	106	/* BOs moved, but not yet updated in the PT */
	107	struct list_head invalidated;
	108
	109	/* BOs cleared in the PT because of a move */
	110	struct list_head cleared;
	111
	112	/* BO mappings freed, but not yet updated in the PT */
	113	struct list_head freed;
	114
	115	/* contains the page directory */
67003a15	116	struct amdgpu_vm_pt root;
a24960f3	117	struct dma_fence *last_dir_update;
073440d2 CK	118	uint64_t last_eviction_counter;
073440d2 CK	119
073440d2 CK	120	/* protecting freed */
	121	spinlock_t freed_lock;
	122
	123	/* Scheduler entity for page table updates */
	124	struct amd_sched_entity entity;
	125
	126	/* client id */
	127	u64 client_id;
36bbf3bf CZ	128	/* dedicated to vm */
36bbf3bf CZ	129	struct amdgpu_vm_id *reserved_vmid[AMDGPU_MAX_VMHUBS];
bd7de27d ML	130	/* each VM will map on CSA */
bd7de27d ML	131	struct amdgpu_bo_va *csa_bo_va;
073440d2 CK	132	};
	133
	134	struct amdgpu_vm_id {
	135	struct list_head list;
073440d2	136	struct amdgpu_sync active;
220196b3	137	struct dma_fence *last_flush;
073440d2 CK	138	atomic64_t owner;
	139
	140	uint64_t pd_gpu_addr;
	141	/* last flushed PD/PT update */
220196b3	142	struct dma_fence *flushed_updates;
073440d2 CK	143
	144	uint32_t current_gpu_reset_count;
	145
	146	uint32_t gds_base;
	147	uint32_t gds_size;
	148	uint32_t gws_base;
	149	uint32_t gws_size;
	150	uint32_t oa_base;
	151	uint32_t oa_size;
	152	};
	153
7645670d CK	154	struct amdgpu_vm_id_manager {
	155	struct mutex lock;
	156	unsigned num_ids;
	157	struct list_head ids_lru;
	158	struct amdgpu_vm_id ids[AMDGPU_NUM_VM];
c3505770	159	atomic_t reserved_vmid_num;
7645670d CK	160	};
7645670d CK	161
073440d2 CK	162	struct amdgpu_vm_manager {
073440d2 CK	163	/* Handling of VMIDs */
7645670d	164	struct amdgpu_vm_id_manager id_mgr[AMDGPU_MAX_VMHUBS];
073440d2 CK	165
	166	/* Handling of VM fences */
	167	u64 fence_context;
	168	unsigned seqno[AMDGPU_MAX_RINGS];
	169
22770e5a	170	uint64_t max_pfn;
8437a097	171	uint32_t num_level;
36b32a68 ZJ	172	uint64_t vm_size;
36b32a68 ZJ	173	uint32_t block_size;
073440d2 CK	174	/* vram base address for page table entry */
073440d2 CK	175	u64 vram_base_offset;
073440d2 CK	176	/* vm pte handling */
	177	const struct amdgpu_vm_pte_funcs *vm_pte_funcs;
	178	struct amdgpu_ring *vm_pte_rings[AMDGPU_MAX_RINGS];
	179	unsigned vm_pte_num_rings;
	180	atomic_t vm_pte_next_ring;
	181	/* client id counter */
	182	atomic64_t client_counter;
284710fa CK	183
	184	/* partial resident texture handling */
	185	spinlock_t prt_lock;
451bc8eb	186	atomic_t num_prt_users;
073440d2 CK	187	};
	188
	189	void amdgpu_vm_manager_init(struct amdgpu_device *adev);
	190	void amdgpu_vm_manager_fini(struct amdgpu_device *adev);
	191	int amdgpu_vm_init(struct amdgpu_device adev, struct amdgpu_vm vm);
	192	void amdgpu_vm_fini(struct amdgpu_device adev, struct amdgpu_vm vm);
	193	void amdgpu_vm_get_pd_bo(struct amdgpu_vm *vm,
	194	struct list_head *validated,
	195	struct amdgpu_bo_list_entry *entry);
	196	int amdgpu_vm_validate_pt_bos(struct amdgpu_device adev, struct amdgpu_vm vm,
	197	int (callback)(void p, struct amdgpu_bo *bo),
	198	void *param);
	199	void amdgpu_vm_move_pt_bos_in_lru(struct amdgpu_device *adev,
	200	struct amdgpu_vm *vm);
663e4577 CK	201	int amdgpu_vm_alloc_pts(struct amdgpu_device *adev,
	202	struct amdgpu_vm *vm,
	203	uint64_t saddr, uint64_t size);
073440d2	204	int amdgpu_vm_grab_id(struct amdgpu_vm vm, struct amdgpu_ring ring,
220196b3	205	struct amdgpu_sync sync, struct dma_fence fence,
073440d2 CK	206	struct amdgpu_job *job);
073440d2 CK	207	int amdgpu_vm_flush(struct amdgpu_ring ring, struct amdgpu_job job);
7645670d CK	208	void amdgpu_vm_reset_id(struct amdgpu_device *adev, unsigned vmhub,
7645670d CK	209	unsigned vmid);
32601d48	210	void amdgpu_vm_reset_all_ids(struct amdgpu_device *adev);
194d2161 CK	211	int amdgpu_vm_update_directories(struct amdgpu_device *adev,
194d2161 CK	212	struct amdgpu_vm *vm);
073440d2	213	int amdgpu_vm_clear_freed(struct amdgpu_device *adev,
f3467818 NH	214	struct amdgpu_vm *vm,
f3467818 NH	215	struct dma_fence **fence);
073440d2 CK	216	int amdgpu_vm_clear_invalids(struct amdgpu_device adev, struct amdgpu_vm vm,
	217	struct amdgpu_sync *sync);
	218	int amdgpu_vm_bo_update(struct amdgpu_device *adev,
	219	struct amdgpu_bo_va *bo_va,
	220	bool clear);
	221	void amdgpu_vm_bo_invalidate(struct amdgpu_device *adev,
	222	struct amdgpu_bo *bo);
	223	struct amdgpu_bo_va amdgpu_vm_bo_find(struct amdgpu_vm vm,
	224	struct amdgpu_bo *bo);
	225	struct amdgpu_bo_va amdgpu_vm_bo_add(struct amdgpu_device adev,
	226	struct amdgpu_vm *vm,
	227	struct amdgpu_bo *bo);
	228	int amdgpu_vm_bo_map(struct amdgpu_device *adev,
	229	struct amdgpu_bo_va *bo_va,
	230	uint64_t addr, uint64_t offset,
268c3001	231	uint64_t size, uint64_t flags);
80f95c57 CK	232	int amdgpu_vm_bo_replace_map(struct amdgpu_device *adev,
	233	struct amdgpu_bo_va *bo_va,
	234	uint64_t addr, uint64_t offset,
	235	uint64_t size, uint64_t flags);
073440d2 CK	236	int amdgpu_vm_bo_unmap(struct amdgpu_device *adev,
	237	struct amdgpu_bo_va *bo_va,
	238	uint64_t addr);
dc54d3d1 CK	239	int amdgpu_vm_bo_clear_mappings(struct amdgpu_device *adev,
	240	struct amdgpu_vm *vm,
	241	uint64_t saddr, uint64_t size);
073440d2 CK	242	void amdgpu_vm_bo_rmv(struct amdgpu_device *adev,
073440d2 CK	243	struct amdgpu_bo_va *bo_va);
bab4fee7	244	void amdgpu_vm_adjust_size(struct amdgpu_device *adev, uint64_t vm_size);
cfbcacf4	245	int amdgpu_vm_ioctl(struct drm_device dev, void data, struct drm_file *filp);
b9bf33d5 CZ	246	bool amdgpu_vm_need_pipeline_sync(struct amdgpu_ring *ring,
b9bf33d5 CZ	247	struct amdgpu_job *job);
e59c0205	248	void amdgpu_vm_check_compute_bug(struct amdgpu_device *adev);
073440d2 CK	249
073440d2 CK	250	#endif