| 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 <drm/ttm/ttm_placement.h> |
| 30 | #include "amdgpu_vram_mgr.h" |
| 31 | #include "amdgpu_hmm.h" |
| 32 | |
| 33 | #define AMDGPU_PL_GDS (TTM_PL_PRIV + 0) |
| 34 | #define AMDGPU_PL_GWS (TTM_PL_PRIV + 1) |
| 35 | #define AMDGPU_PL_OA (TTM_PL_PRIV + 2) |
| 36 | #define AMDGPU_PL_PREEMPT (TTM_PL_PRIV + 3) |
| 37 | #define AMDGPU_PL_DOORBELL (TTM_PL_PRIV + 4) |
| 38 | #define AMDGPU_PL_MMIO_REMAP (TTM_PL_PRIV + 5) |
| 39 | #define __AMDGPU_PL_NUM (TTM_PL_PRIV + 6) |
| 40 | |
| 41 | #define AMDGPU_GTT_MAX_TRANSFER_SIZE 512 |
| 42 | #define AMDGPU_GTT_NUM_TRANSFER_WINDOWS 2 |
| 43 | |
| 44 | extern const struct attribute_group amdgpu_vram_mgr_attr_group; |
| 45 | extern const struct attribute_group amdgpu_gtt_mgr_attr_group; |
| 46 | |
| 47 | struct hmm_range; |
| 48 | |
| 49 | struct amdgpu_gtt_mgr { |
| 50 | struct ttm_resource_manager manager; |
| 51 | struct drm_mm mm; |
| 52 | spinlock_t lock; |
| 53 | }; |
| 54 | |
| 55 | struct amdgpu_mman { |
| 56 | struct ttm_device bdev; |
| 57 | struct ttm_pool *ttm_pools; |
| 58 | bool initialized; |
| 59 | void __iomem *aper_base_kaddr; |
| 60 | |
| 61 | /* buffer handling */ |
| 62 | const struct amdgpu_buffer_funcs *buffer_funcs; |
| 63 | struct amdgpu_ring *buffer_funcs_ring; |
| 64 | bool buffer_funcs_enabled; |
| 65 | |
| 66 | struct mutex gtt_window_lock; |
| 67 | /* High priority scheduler entity for buffer moves */ |
| 68 | struct drm_sched_entity high_pr; |
| 69 | /* Low priority scheduler entity for VRAM clearing */ |
| 70 | struct drm_sched_entity low_pr; |
| 71 | |
| 72 | struct amdgpu_vram_mgr vram_mgr; |
| 73 | struct amdgpu_gtt_mgr gtt_mgr; |
| 74 | struct ttm_resource_manager preempt_mgr; |
| 75 | |
| 76 | uint64_t stolen_vga_size; |
| 77 | struct amdgpu_bo *stolen_vga_memory; |
| 78 | uint64_t stolen_extended_size; |
| 79 | struct amdgpu_bo *stolen_extended_memory; |
| 80 | bool keep_stolen_vga_memory; |
| 81 | |
| 82 | struct amdgpu_bo *stolen_reserved_memory; |
| 83 | uint64_t stolen_reserved_offset; |
| 84 | uint64_t stolen_reserved_size; |
| 85 | |
| 86 | /* fw reserved memory */ |
| 87 | struct amdgpu_bo *fw_reserved_memory; |
| 88 | struct amdgpu_bo *fw_reserved_memory_extend; |
| 89 | |
| 90 | /* firmware VRAM reservation */ |
| 91 | u64 fw_vram_usage_start_offset; |
| 92 | u64 fw_vram_usage_size; |
| 93 | struct amdgpu_bo *fw_vram_usage_reserved_bo; |
| 94 | void *fw_vram_usage_va; |
| 95 | |
| 96 | /* driver VRAM reservation */ |
| 97 | u64 drv_vram_usage_start_offset; |
| 98 | u64 drv_vram_usage_size; |
| 99 | struct amdgpu_bo *drv_vram_usage_reserved_bo; |
| 100 | void *drv_vram_usage_va; |
| 101 | |
| 102 | /* PAGE_SIZE'd BO for process memory r/w over SDMA. */ |
| 103 | struct amdgpu_bo *sdma_access_bo; |
| 104 | void *sdma_access_ptr; |
| 105 | }; |
| 106 | |
| 107 | struct amdgpu_copy_mem { |
| 108 | struct ttm_buffer_object *bo; |
| 109 | struct ttm_resource *mem; |
| 110 | unsigned long offset; |
| 111 | }; |
| 112 | |
| 113 | #define AMDGPU_COPY_FLAGS_TMZ (1 << 0) |
| 114 | #define AMDGPU_COPY_FLAGS_READ_DECOMPRESSED (1 << 1) |
| 115 | #define AMDGPU_COPY_FLAGS_WRITE_COMPRESSED (1 << 2) |
| 116 | #define AMDGPU_COPY_FLAGS_MAX_COMPRESSED_SHIFT 3 |
| 117 | #define AMDGPU_COPY_FLAGS_MAX_COMPRESSED_MASK 0x03 |
| 118 | #define AMDGPU_COPY_FLAGS_NUMBER_TYPE_SHIFT 5 |
| 119 | #define AMDGPU_COPY_FLAGS_NUMBER_TYPE_MASK 0x07 |
| 120 | #define AMDGPU_COPY_FLAGS_DATA_FORMAT_SHIFT 8 |
| 121 | #define AMDGPU_COPY_FLAGS_DATA_FORMAT_MASK 0x3f |
| 122 | #define AMDGPU_COPY_FLAGS_WRITE_COMPRESS_DISABLE_SHIFT 14 |
| 123 | #define AMDGPU_COPY_FLAGS_WRITE_COMPRESS_DISABLE_MASK 0x1 |
| 124 | |
| 125 | #define AMDGPU_COPY_FLAGS_SET(field, value) \ |
| 126 | (((__u32)(value) & AMDGPU_COPY_FLAGS_##field##_MASK) << AMDGPU_COPY_FLAGS_##field##_SHIFT) |
| 127 | #define AMDGPU_COPY_FLAGS_GET(value, field) \ |
| 128 | (((__u32)(value) >> AMDGPU_COPY_FLAGS_##field##_SHIFT) & AMDGPU_COPY_FLAGS_##field##_MASK) |
| 129 | |
| 130 | int amdgpu_gtt_mgr_init(struct amdgpu_device *adev, uint64_t gtt_size); |
| 131 | void amdgpu_gtt_mgr_fini(struct amdgpu_device *adev); |
| 132 | int amdgpu_preempt_mgr_init(struct amdgpu_device *adev); |
| 133 | void amdgpu_preempt_mgr_fini(struct amdgpu_device *adev); |
| 134 | int amdgpu_vram_mgr_init(struct amdgpu_device *adev); |
| 135 | void amdgpu_vram_mgr_fini(struct amdgpu_device *adev); |
| 136 | |
| 137 | bool amdgpu_gtt_mgr_has_gart_addr(struct ttm_resource *mem); |
| 138 | void amdgpu_gtt_mgr_recover(struct amdgpu_gtt_mgr *mgr); |
| 139 | |
| 140 | uint64_t amdgpu_preempt_mgr_usage(struct ttm_resource_manager *man); |
| 141 | |
| 142 | u64 amdgpu_vram_mgr_bo_visible_size(struct amdgpu_bo *bo); |
| 143 | int amdgpu_vram_mgr_alloc_sgt(struct amdgpu_device *adev, |
| 144 | struct ttm_resource *mem, |
| 145 | u64 offset, u64 size, |
| 146 | struct device *dev, |
| 147 | enum dma_data_direction dir, |
| 148 | struct sg_table **sgt); |
| 149 | void amdgpu_vram_mgr_free_sgt(struct device *dev, |
| 150 | enum dma_data_direction dir, |
| 151 | struct sg_table *sgt); |
| 152 | uint64_t amdgpu_vram_mgr_vis_usage(struct amdgpu_vram_mgr *mgr); |
| 153 | int amdgpu_vram_mgr_reserve_range(struct amdgpu_vram_mgr *mgr, |
| 154 | uint64_t start, uint64_t size); |
| 155 | int amdgpu_vram_mgr_query_page_status(struct amdgpu_vram_mgr *mgr, |
| 156 | uint64_t start); |
| 157 | void amdgpu_vram_mgr_clear_reset_blocks(struct amdgpu_device *adev); |
| 158 | |
| 159 | bool amdgpu_res_cpu_visible(struct amdgpu_device *adev, |
| 160 | struct ttm_resource *res); |
| 161 | |
| 162 | int amdgpu_ttm_init(struct amdgpu_device *adev); |
| 163 | void amdgpu_ttm_fini(struct amdgpu_device *adev); |
| 164 | void amdgpu_ttm_set_buffer_funcs_status(struct amdgpu_device *adev, |
| 165 | bool enable); |
| 166 | int amdgpu_copy_buffer(struct amdgpu_ring *ring, uint64_t src_offset, |
| 167 | uint64_t dst_offset, uint32_t byte_count, |
| 168 | struct dma_resv *resv, |
| 169 | struct dma_fence **fence, bool direct_submit, |
| 170 | bool vm_needs_flush, uint32_t copy_flags); |
| 171 | int amdgpu_ttm_clear_buffer(struct amdgpu_bo *bo, |
| 172 | struct dma_resv *resv, |
| 173 | struct dma_fence **fence); |
| 174 | int amdgpu_fill_buffer(struct amdgpu_bo *bo, |
| 175 | uint32_t src_data, |
| 176 | struct dma_resv *resv, |
| 177 | struct dma_fence **fence, |
| 178 | bool delayed, |
| 179 | u64 k_job_id); |
| 180 | |
| 181 | int amdgpu_ttm_alloc_gart(struct ttm_buffer_object *bo); |
| 182 | void amdgpu_ttm_recover_gart(struct ttm_buffer_object *tbo); |
| 183 | uint64_t amdgpu_ttm_domain_start(struct amdgpu_device *adev, uint32_t type); |
| 184 | |
| 185 | #if IS_ENABLED(CONFIG_DRM_AMDGPU_USERPTR) |
| 186 | int amdgpu_ttm_tt_get_user_pages(struct amdgpu_bo *bo, |
| 187 | struct amdgpu_hmm_range *range); |
| 188 | #else |
| 189 | static inline int amdgpu_ttm_tt_get_user_pages(struct amdgpu_bo *bo, |
| 190 | struct amdgpu_hmm_range *range) |
| 191 | { |
| 192 | return -EPERM; |
| 193 | } |
| 194 | #endif |
| 195 | |
| 196 | void amdgpu_ttm_tt_set_user_pages(struct ttm_tt *ttm, struct amdgpu_hmm_range *range); |
| 197 | int amdgpu_ttm_tt_get_userptr(const struct ttm_buffer_object *tbo, |
| 198 | uint64_t *user_addr); |
| 199 | int amdgpu_ttm_tt_set_userptr(struct ttm_buffer_object *bo, |
| 200 | uint64_t addr, uint32_t flags); |
| 201 | bool amdgpu_ttm_tt_has_userptr(struct ttm_tt *ttm); |
| 202 | struct mm_struct *amdgpu_ttm_tt_get_usermm(struct ttm_tt *ttm); |
| 203 | bool amdgpu_ttm_tt_affect_userptr(struct ttm_tt *ttm, unsigned long start, |
| 204 | unsigned long end, unsigned long *userptr); |
| 205 | bool amdgpu_ttm_tt_userptr_invalidated(struct ttm_tt *ttm, |
| 206 | int *last_invalidated); |
| 207 | bool amdgpu_ttm_tt_is_userptr(struct ttm_tt *ttm); |
| 208 | bool amdgpu_ttm_tt_is_readonly(struct ttm_tt *ttm); |
| 209 | uint64_t amdgpu_ttm_tt_pde_flags(struct ttm_tt *ttm, struct ttm_resource *mem); |
| 210 | uint64_t amdgpu_ttm_tt_pte_flags(struct amdgpu_device *adev, struct ttm_tt *ttm, |
| 211 | struct ttm_resource *mem); |
| 212 | int amdgpu_ttm_evict_resources(struct amdgpu_device *adev, int mem_type); |
| 213 | |
| 214 | void amdgpu_ttm_debugfs_init(struct amdgpu_device *adev); |
| 215 | |
| 216 | #endif |
| 217 | |