2 * Copyright 2016 Advanced Micro Devices, Inc.
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:
11 * The above copyright notice and this permission notice shall be included in
12 * all copies or substantial portions of the Software.
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.
22 * Authors: Christian König
24 #ifndef __AMDGPU_VM_H__
25 #define __AMDGPU_VM_H__
27 #include <linux/rbtree.h>
29 #include "gpu_scheduler.h"
30 #include "amdgpu_sync.h"
31 #include "amdgpu_ring.h"
35 struct amdgpu_bo_list_entry;
41 /* maximum number of VMIDs */
42 #define AMDGPU_NUM_VM 16
44 /* Maximum number of PTEs the hardware can write with one command */
45 #define AMDGPU_VM_MAX_UPDATE_SIZE 0x3FFFF
47 /* number of entries in page table */
48 #define AMDGPU_VM_PTE_COUNT (1 << amdgpu_vm_block_size)
50 /* PTBs (Page Table Blocks) need to be aligned to 32K */
51 #define AMDGPU_VM_PTB_ALIGN_SIZE 32768
53 /* LOG2 number of continuous pages for the fragment field */
54 #define AMDGPU_LOG2_PAGES_PER_FRAG 4
56 #define AMDGPU_PTE_VALID (1ULL << 0)
57 #define AMDGPU_PTE_SYSTEM (1ULL << 1)
58 #define AMDGPU_PTE_SNOOPED (1ULL << 2)
61 #define AMDGPU_PTE_EXECUTABLE (1ULL << 4)
63 #define AMDGPU_PTE_READABLE (1ULL << 5)
64 #define AMDGPU_PTE_WRITEABLE (1ULL << 6)
66 #define AMDGPU_PTE_FRAG(x) ((x & 0x1fULL) << 7)
68 #define AMDGPU_PTE_PRT (1ULL << 63)
70 /* How to programm VM fault handling */
71 #define AMDGPU_VM_FAULT_STOP_NEVER 0
72 #define AMDGPU_VM_FAULT_STOP_FIRST 1
73 #define AMDGPU_VM_FAULT_STOP_ALWAYS 2
81 /* tree of virtual addresses mapped */
84 /* protecting invalidated */
85 spinlock_t status_lock;
87 /* BOs moved, but not yet updated in the PT */
88 struct list_head invalidated;
90 /* BOs cleared in the PT because of a move */
91 struct list_head cleared;
93 /* BO mappings freed, but not yet updated in the PT */
94 struct list_head freed;
96 /* contains the page directory */
97 struct amdgpu_bo *page_directory;
98 unsigned max_pde_used;
99 struct dma_fence *page_directory_fence;
100 uint64_t last_eviction_counter;
102 /* array of page tables, one for each page directory entry */
103 struct amdgpu_vm_pt *page_tables;
105 /* for id and flush management per ring */
106 struct amdgpu_vm_id *ids[AMDGPU_MAX_RINGS];
108 /* protecting freed */
109 spinlock_t freed_lock;
111 /* Scheduler entity for page table updates */
112 struct amd_sched_entity entity;
116 /* each VM will map on CSA */
117 struct amdgpu_bo_va *csa_bo_va;
120 struct amdgpu_vm_id {
121 struct list_head list;
122 struct dma_fence *first;
123 struct amdgpu_sync active;
124 struct dma_fence *last_flush;
127 uint64_t pd_gpu_addr;
128 /* last flushed PD/PT update */
129 struct dma_fence *flushed_updates;
131 uint32_t current_gpu_reset_count;
141 struct amdgpu_vm_manager {
142 /* Handling of VMIDs */
145 struct list_head ids_lru;
146 struct amdgpu_vm_id ids[AMDGPU_NUM_VM];
148 /* Handling of VM fences */
150 unsigned seqno[AMDGPU_MAX_RINGS];
153 /* vram base address for page table entry */
154 u64 vram_base_offset;
157 /* vm pte handling */
158 const struct amdgpu_vm_pte_funcs *vm_pte_funcs;
159 struct amdgpu_ring *vm_pte_rings[AMDGPU_MAX_RINGS];
160 unsigned vm_pte_num_rings;
161 atomic_t vm_pte_next_ring;
162 /* client id counter */
163 atomic64_t client_counter;
165 /* partial resident texture handling */
167 atomic_t num_prt_users;
170 void amdgpu_vm_manager_init(struct amdgpu_device *adev);
171 void amdgpu_vm_manager_fini(struct amdgpu_device *adev);
172 int amdgpu_vm_init(struct amdgpu_device *adev, struct amdgpu_vm *vm);
173 void amdgpu_vm_fini(struct amdgpu_device *adev, struct amdgpu_vm *vm);
174 void amdgpu_vm_get_pd_bo(struct amdgpu_vm *vm,
175 struct list_head *validated,
176 struct amdgpu_bo_list_entry *entry);
177 int amdgpu_vm_validate_pt_bos(struct amdgpu_device *adev, struct amdgpu_vm *vm,
178 int (*callback)(void *p, struct amdgpu_bo *bo),
180 void amdgpu_vm_move_pt_bos_in_lru(struct amdgpu_device *adev,
181 struct amdgpu_vm *vm);
182 int amdgpu_vm_alloc_pts(struct amdgpu_device *adev,
183 struct amdgpu_vm *vm,
184 uint64_t saddr, uint64_t size);
185 int amdgpu_vm_grab_id(struct amdgpu_vm *vm, struct amdgpu_ring *ring,
186 struct amdgpu_sync *sync, struct dma_fence *fence,
187 struct amdgpu_job *job);
188 int amdgpu_vm_flush(struct amdgpu_ring *ring, struct amdgpu_job *job);
189 void amdgpu_vm_reset_id(struct amdgpu_device *adev, unsigned vm_id);
190 int amdgpu_vm_update_page_directory(struct amdgpu_device *adev,
191 struct amdgpu_vm *vm);
192 int amdgpu_vm_clear_freed(struct amdgpu_device *adev,
193 struct amdgpu_vm *vm);
194 int amdgpu_vm_clear_invalids(struct amdgpu_device *adev, struct amdgpu_vm *vm,
195 struct amdgpu_sync *sync);
196 int amdgpu_vm_bo_update(struct amdgpu_device *adev,
197 struct amdgpu_bo_va *bo_va,
199 void amdgpu_vm_bo_invalidate(struct amdgpu_device *adev,
200 struct amdgpu_bo *bo);
201 struct amdgpu_bo_va *amdgpu_vm_bo_find(struct amdgpu_vm *vm,
202 struct amdgpu_bo *bo);
203 struct amdgpu_bo_va *amdgpu_vm_bo_add(struct amdgpu_device *adev,
204 struct amdgpu_vm *vm,
205 struct amdgpu_bo *bo);
206 int amdgpu_vm_bo_map(struct amdgpu_device *adev,
207 struct amdgpu_bo_va *bo_va,
208 uint64_t addr, uint64_t offset,
209 uint64_t size, uint64_t flags);
210 int amdgpu_vm_bo_replace_map(struct amdgpu_device *adev,
211 struct amdgpu_bo_va *bo_va,
212 uint64_t addr, uint64_t offset,
213 uint64_t size, uint64_t flags);
214 int amdgpu_vm_bo_unmap(struct amdgpu_device *adev,
215 struct amdgpu_bo_va *bo_va,
217 int amdgpu_vm_bo_clear_mappings(struct amdgpu_device *adev,
218 struct amdgpu_vm *vm,
219 uint64_t saddr, uint64_t size);
220 void amdgpu_vm_bo_rmv(struct amdgpu_device *adev,
221 struct amdgpu_bo_va *bo_va);