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 */
23
24 #ifndef __AMDGPU_TTM_H__
25 #define __AMDGPU_TTM_H__
26
27 #include <linux/dma-direction.h>
28 #include <drm/gpu_scheduler.h>
29 #include "amdgpu_vram_mgr.h"
30 #include "amdgpu.h"
31
32 #define AMDGPU_PL_GDS (TTM_PL_PRIV + 0)
33 #define AMDGPU_PL_GWS (TTM_PL_PRIV + 1)
34 #define AMDGPU_PL_OA (TTM_PL_PRIV + 2)
35 #define AMDGPU_PL_PREEMPT (TTM_PL_PRIV + 3)
36 #define AMDGPU_PL_DOORBELL (TTM_PL_PRIV + 4)
37
38 #define AMDGPU_GTT_MAX_TRANSFER_SIZE 512
39 #define AMDGPU_GTT_NUM_TRANSFER_WINDOWS 2
40
41 extern const struct attribute_group amdgpu_vram_mgr_attr_group;
42 extern const struct attribute_group amdgpu_gtt_mgr_attr_group;
43
44 struct hmm_range;
45
46 struct amdgpu_gtt_mgr {
47 struct ttm_resource_manager manager;
48 struct drm_mm mm;
49 spinlock_t lock;
50 };
51
52 struct amdgpu_mman {
53 struct ttm_device bdev;
54 struct ttm_pool *ttm_pools;
55 bool initialized;
56 void __iomem *aper_base_kaddr;
57
58 /* buffer handling */
59 const struct amdgpu_buffer_funcs *buffer_funcs;
60 struct amdgpu_ring *buffer_funcs_ring;
61 bool buffer_funcs_enabled;
62
63 struct mutex gtt_window_lock;
64 /* High priority scheduler entity for buffer moves */
65 struct drm_sched_entity high_pr;
66 /* Low priority scheduler entity for VRAM clearing */
67 struct drm_sched_entity low_pr;
68
69 struct amdgpu_vram_mgr vram_mgr;
70 struct amdgpu_gtt_mgr gtt_mgr;
71 struct ttm_resource_manager preempt_mgr;
72
73 uint64_t stolen_vga_size;
74 struct amdgpu_bo *stolen_vga_memory;
75 uint64_t stolen_extended_size;
76 struct amdgpu_bo *stolen_extended_memory;
77 bool keep_stolen_vga_memory;
78
79 struct amdgpu_bo *stolen_reserved_memory;
80 uint64_t stolen_reserved_offset;
81 uint64_t stolen_reserved_size;
82
83 /* discovery */
84 uint8_t *discovery_bin;
85 uint32_t discovery_tmr_size;
86 /* fw reserved memory */
87 struct amdgpu_bo *fw_reserved_memory;
88
89 /* firmware VRAM reservation */
90 u64 fw_vram_usage_start_offset;
91 u64 fw_vram_usage_size;
92 struct amdgpu_bo *fw_vram_usage_reserved_bo;
93 void *fw_vram_usage_va;
94
95 /* driver VRAM reservation */
96 u64 drv_vram_usage_start_offset;
97 u64 drv_vram_usage_size;
98 struct amdgpu_bo *drv_vram_usage_reserved_bo;
99 void *drv_vram_usage_va;
100
101 /* PAGE_SIZE'd BO for process memory r/w over SDMA. */
102 struct amdgpu_bo *sdma_access_bo;
103 void *sdma_access_ptr;
104 };
105
106 struct amdgpu_copy_mem {
107 struct ttm_buffer_object *bo;
108 struct ttm_resource *mem;
109 unsigned long offset;
110 };
111
112 #define AMDGPU_COPY_FLAGS_TMZ (1 << 0)
113 #define AMDGPU_COPY_FLAGS_READ_DECOMPRESSED (1 << 1)
114 #define AMDGPU_COPY_FLAGS_WRITE_COMPRESSED (1 << 2)
115 #define AMDGPU_COPY_FLAGS_MAX_COMPRESSED_SHIFT 3
116 #define AMDGPU_COPY_FLAGS_MAX_COMPRESSED_MASK 0x03
117 #define AMDGPU_COPY_FLAGS_NUMBER_TYPE_SHIFT 5
118 #define AMDGPU_COPY_FLAGS_NUMBER_TYPE_MASK 0x07
119 #define AMDGPU_COPY_FLAGS_DATA_FORMAT_SHIFT 8
120 #define AMDGPU_COPY_FLAGS_DATA_FORMAT_MASK 0x3f
121 #define AMDGPU_COPY_FLAGS_WRITE_COMPRESS_DISABLE_SHIFT 14
122 #define AMDGPU_COPY_FLAGS_WRITE_COMPRESS_DISABLE_MASK 0x1
123
124 #define AMDGPU_COPY_FLAGS_SET(field, value) \
125 (((__u32)(value) & AMDGPU_COPY_FLAGS_##field##_MASK) << AMDGPU_COPY_FLAGS_##field##_SHIFT)
126 #define AMDGPU_COPY_FLAGS_GET(value, field) \
127 (((__u32)(value) >> AMDGPU_COPY_FLAGS_##field##_SHIFT) & AMDGPU_COPY_FLAGS_##field##_MASK)
128
129 int amdgpu_gtt_mgr_init(struct amdgpu_device *adev, uint64_t gtt_size);
130 void amdgpu_gtt_mgr_fini(struct amdgpu_device *adev);
131 int amdgpu_preempt_mgr_init(struct amdgpu_device *adev);
132 void amdgpu_preempt_mgr_fini(struct amdgpu_device *adev);
133 int amdgpu_vram_mgr_init(struct amdgpu_device *adev);
134 void amdgpu_vram_mgr_fini(struct amdgpu_device *adev);
135
136 bool amdgpu_gtt_mgr_has_gart_addr(struct ttm_resource *mem);
137 void amdgpu_gtt_mgr_recover(struct amdgpu_gtt_mgr *mgr);
138
139 uint64_t amdgpu_preempt_mgr_usage(struct ttm_resource_manager *man);
140
141 u64 amdgpu_vram_mgr_bo_visible_size(struct amdgpu_bo *bo);
142 int amdgpu_vram_mgr_alloc_sgt(struct amdgpu_device *adev,
143 struct ttm_resource *mem,
144 u64 offset, u64 size,
145 struct device *dev,
146 enum dma_data_direction dir,
147 struct sg_table **sgt);
148 void amdgpu_vram_mgr_free_sgt(struct device *dev,
149 enum dma_data_direction dir,
150 struct sg_table *sgt);
151 uint64_t amdgpu_vram_mgr_vis_usage(struct amdgpu_vram_mgr *mgr);
152 int amdgpu_vram_mgr_reserve_range(struct amdgpu_vram_mgr *mgr,
153 uint64_t start, uint64_t size);
154 int amdgpu_vram_mgr_query_page_status(struct amdgpu_vram_mgr *mgr,
155 uint64_t start);
156 void amdgpu_vram_mgr_clear_reset_blocks(struct amdgpu_device *adev);
157
158 bool amdgpu_res_cpu_visible(struct amdgpu_device *adev,
159 struct ttm_resource *res);
160
161 int amdgpu_ttm_init(struct amdgpu_device *adev);
162 void amdgpu_ttm_fini(struct amdgpu_device *adev);
163 void amdgpu_ttm_set_buffer_funcs_status(struct amdgpu_device *adev,
164 bool enable);
165 int amdgpu_copy_buffer(struct amdgpu_ring *ring, uint64_t src_offset,
166 uint64_t dst_offset, uint32_t byte_count,
167 struct dma_resv *resv,
168 struct dma_fence **fence, bool direct_submit,
169 bool vm_needs_flush, uint32_t copy_flags);
170 int amdgpu_ttm_copy_mem_to_mem(struct amdgpu_device *adev,
171 const struct amdgpu_copy_mem *src,
172 const struct amdgpu_copy_mem *dst,
173 uint64_t size, bool tmz,
174 struct dma_resv *resv,
175 struct dma_fence **f);
176 int amdgpu_ttm_clear_buffer(struct amdgpu_bo *bo,
177 struct dma_resv *resv,
178 struct dma_fence **fence);
179 int amdgpu_fill_buffer(struct amdgpu_bo *bo,
180 uint32_t src_data,
181 struct dma_resv *resv,
182 struct dma_fence **fence,
183 bool delayed);
184
185 int amdgpu_ttm_alloc_gart(struct ttm_buffer_object *bo);
186 void amdgpu_ttm_recover_gart(struct ttm_buffer_object *tbo);
187 uint64_t amdgpu_ttm_domain_start(struct amdgpu_device *adev, uint32_t type);
188
189 #if IS_ENABLED(CONFIG_DRM_AMDGPU_USERPTR)
190 int amdgpu_ttm_tt_get_user_pages(struct amdgpu_bo *bo, struct page **pages,
191 struct hmm_range **range);
192 void amdgpu_ttm_tt_discard_user_pages(struct ttm_tt *ttm,
193 struct hmm_range *range);
194 bool amdgpu_ttm_tt_get_user_pages_done(struct ttm_tt *ttm,
195 struct hmm_range *range);
196 #else
amdgpu_ttm_tt_get_user_pages(struct amdgpu_bo * bo,struct page ** pages,struct hmm_range ** range)197 static inline int amdgpu_ttm_tt_get_user_pages(struct amdgpu_bo *bo,
198 struct page **pages,
199 struct hmm_range **range)
200 {
201 return -EPERM;
202 }
amdgpu_ttm_tt_discard_user_pages(struct ttm_tt * ttm,struct hmm_range * range)203 static inline void amdgpu_ttm_tt_discard_user_pages(struct ttm_tt *ttm,
204 struct hmm_range *range)
205 {
206 }
amdgpu_ttm_tt_get_user_pages_done(struct ttm_tt * ttm,struct hmm_range * range)207 static inline bool amdgpu_ttm_tt_get_user_pages_done(struct ttm_tt *ttm,
208 struct hmm_range *range)
209 {
210 return false;
211 }
212 #endif
213
214 void amdgpu_ttm_tt_set_user_pages(struct ttm_tt *ttm, struct page **pages);
215 int amdgpu_ttm_tt_get_userptr(const struct ttm_buffer_object *tbo,
216 uint64_t *user_addr);
217 int amdgpu_ttm_tt_set_userptr(struct ttm_buffer_object *bo,
218 uint64_t addr, uint32_t flags);
219 bool amdgpu_ttm_tt_has_userptr(struct ttm_tt *ttm);
220 struct mm_struct *amdgpu_ttm_tt_get_usermm(struct ttm_tt *ttm);
221 bool amdgpu_ttm_tt_affect_userptr(struct ttm_tt *ttm, unsigned long start,
222 unsigned long end, unsigned long *userptr);
223 bool amdgpu_ttm_tt_userptr_invalidated(struct ttm_tt *ttm,
224 int *last_invalidated);
225 bool amdgpu_ttm_tt_is_userptr(struct ttm_tt *ttm);
226 bool amdgpu_ttm_tt_is_readonly(struct ttm_tt *ttm);
227 uint64_t amdgpu_ttm_tt_pde_flags(struct ttm_tt *ttm, struct ttm_resource *mem);
228 uint64_t amdgpu_ttm_tt_pte_flags(struct amdgpu_device *adev, struct ttm_tt *ttm,
229 struct ttm_resource *mem);
230 int amdgpu_ttm_evict_resources(struct amdgpu_device *adev, int mem_type);
231
232 void amdgpu_ttm_debugfs_init(struct amdgpu_device *adev);
233
234 #endif
235