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_VCN_H__ 25 #define __AMDGPU_VCN_H__ 26 27 #include "amdgpu_ras.h" 28 29 #define AMDGPU_VCN_STACK_SIZE (128*1024) 30 #define AMDGPU_VCN_CONTEXT_SIZE (512*1024) 31 32 #define AMDGPU_VCN_FIRMWARE_OFFSET 256 33 #define AMDGPU_VCN_MAX_ENC_RINGS 3 34 35 #define AMDGPU_MAX_VCN_INSTANCES 2 36 #define AMDGPU_MAX_VCN_ENC_RINGS AMDGPU_VCN_MAX_ENC_RINGS * AMDGPU_MAX_VCN_INSTANCES 37 38 #define AMDGPU_VCN_HARVEST_VCN0 (1 << 0) 39 #define AMDGPU_VCN_HARVEST_VCN1 (1 << 1) 40 41 #define VCN_DEC_KMD_CMD 0x80000000 42 #define VCN_DEC_CMD_FENCE 0x00000000 43 #define VCN_DEC_CMD_TRAP 0x00000001 44 #define VCN_DEC_CMD_WRITE_REG 0x00000004 45 #define VCN_DEC_CMD_REG_READ_COND_WAIT 0x00000006 46 #define VCN_DEC_CMD_PACKET_START 0x0000000a 47 #define VCN_DEC_CMD_PACKET_END 0x0000000b 48 49 #define VCN_DEC_SW_CMD_NO_OP 0x00000000 50 #define VCN_DEC_SW_CMD_END 0x00000001 51 #define VCN_DEC_SW_CMD_IB 0x00000002 52 #define VCN_DEC_SW_CMD_FENCE 0x00000003 53 #define VCN_DEC_SW_CMD_TRAP 0x00000004 54 #define VCN_DEC_SW_CMD_IB_AUTO 0x00000005 55 #define VCN_DEC_SW_CMD_SEMAPHORE 0x00000006 56 #define VCN_DEC_SW_CMD_PREEMPT_FENCE 0x00000009 57 #define VCN_DEC_SW_CMD_REG_WRITE 0x0000000b 58 #define VCN_DEC_SW_CMD_REG_WAIT 0x0000000c 59 60 #define VCN_ENC_CMD_NO_OP 0x00000000 61 #define VCN_ENC_CMD_END 0x00000001 62 #define VCN_ENC_CMD_IB 0x00000002 63 #define VCN_ENC_CMD_FENCE 0x00000003 64 #define VCN_ENC_CMD_TRAP 0x00000004 65 #define VCN_ENC_CMD_REG_WRITE 0x0000000b 66 #define VCN_ENC_CMD_REG_WAIT 0x0000000c 67 68 #define VCN_VID_SOC_ADDRESS_2_0 0x1fa00 69 #define VCN1_VID_SOC_ADDRESS_3_0 0x48200 70 #define VCN_AON_SOC_ADDRESS_2_0 0x1f800 71 #define VCN1_AON_SOC_ADDRESS_3_0 0x48000 72 #define VCN_VID_IP_ADDRESS_2_0 0x0 73 #define VCN_AON_IP_ADDRESS_2_0 0x30000 74 75 #define mmUVD_RBC_XX_IB_REG_CHECK 0x026b 76 #define mmUVD_RBC_XX_IB_REG_CHECK_BASE_IDX 1 77 #define mmUVD_REG_XX_MASK 0x026c 78 #define mmUVD_REG_XX_MASK_BASE_IDX 1 79 80 /* 1 second timeout */ 81 #define VCN_IDLE_TIMEOUT msecs_to_jiffies(1000) 82 83 #define RREG32_SOC15_DPG_MODE_1_0(ip, inst_idx, reg, mask, sram_sel) \ 84 ({ WREG32_SOC15(ip, inst_idx, mmUVD_DPG_LMA_MASK, mask); \ 85 WREG32_SOC15(ip, inst_idx, mmUVD_DPG_LMA_CTL, \ 86 UVD_DPG_LMA_CTL__MASK_EN_MASK | \ 87 ((adev->reg_offset[ip##_HWIP][inst_idx][reg##_BASE_IDX] + reg) \ 88 << UVD_DPG_LMA_CTL__READ_WRITE_ADDR__SHIFT) | \ 89 (sram_sel << UVD_DPG_LMA_CTL__SRAM_SEL__SHIFT)); \ 90 RREG32_SOC15(ip, inst_idx, mmUVD_DPG_LMA_DATA); \ 91 }) 92 93 #define WREG32_SOC15_DPG_MODE_1_0(ip, inst_idx, reg, value, mask, sram_sel) \ 94 do { \ 95 WREG32_SOC15(ip, inst_idx, mmUVD_DPG_LMA_DATA, value); \ 96 WREG32_SOC15(ip, inst_idx, mmUVD_DPG_LMA_MASK, mask); \ 97 WREG32_SOC15(ip, inst_idx, mmUVD_DPG_LMA_CTL, \ 98 UVD_DPG_LMA_CTL__READ_WRITE_MASK | \ 99 ((adev->reg_offset[ip##_HWIP][inst_idx][reg##_BASE_IDX] + reg) \ 100 << UVD_DPG_LMA_CTL__READ_WRITE_ADDR__SHIFT) | \ 101 (sram_sel << UVD_DPG_LMA_CTL__SRAM_SEL__SHIFT)); \ 102 } while (0) 103 104 #define SOC15_DPG_MODE_OFFSET(ip, inst_idx, reg) \ 105 ({ \ 106 uint32_t internal_reg_offset, addr; \ 107 bool video_range, video1_range, aon_range, aon1_range; \ 108 \ 109 addr = (adev->reg_offset[ip##_HWIP][inst_idx][reg##_BASE_IDX] + reg); \ 110 addr <<= 2; \ 111 video_range = ((((0xFFFFF & addr) >= (VCN_VID_SOC_ADDRESS_2_0)) && \ 112 ((0xFFFFF & addr) < ((VCN_VID_SOC_ADDRESS_2_0 + 0x2600))))); \ 113 video1_range = ((((0xFFFFF & addr) >= (VCN1_VID_SOC_ADDRESS_3_0)) && \ 114 ((0xFFFFF & addr) < ((VCN1_VID_SOC_ADDRESS_3_0 + 0x2600))))); \ 115 aon_range = ((((0xFFFFF & addr) >= (VCN_AON_SOC_ADDRESS_2_0)) && \ 116 ((0xFFFFF & addr) < ((VCN_AON_SOC_ADDRESS_2_0 + 0x600))))); \ 117 aon1_range = ((((0xFFFFF & addr) >= (VCN1_AON_SOC_ADDRESS_3_0)) && \ 118 ((0xFFFFF & addr) < ((VCN1_AON_SOC_ADDRESS_3_0 + 0x600))))); \ 119 if (video_range) \ 120 internal_reg_offset = ((0xFFFFF & addr) - (VCN_VID_SOC_ADDRESS_2_0) + \ 121 (VCN_VID_IP_ADDRESS_2_0)); \ 122 else if (aon_range) \ 123 internal_reg_offset = ((0xFFFFF & addr) - (VCN_AON_SOC_ADDRESS_2_0) + \ 124 (VCN_AON_IP_ADDRESS_2_0)); \ 125 else if (video1_range) \ 126 internal_reg_offset = ((0xFFFFF & addr) - (VCN1_VID_SOC_ADDRESS_3_0) + \ 127 (VCN_VID_IP_ADDRESS_2_0)); \ 128 else if (aon1_range) \ 129 internal_reg_offset = ((0xFFFFF & addr) - (VCN1_AON_SOC_ADDRESS_3_0) + \ 130 (VCN_AON_IP_ADDRESS_2_0)); \ 131 else \ 132 internal_reg_offset = (0xFFFFF & addr); \ 133 \ 134 internal_reg_offset >>= 2; \ 135 }) 136 137 #define RREG32_SOC15_DPG_MODE(inst_idx, offset, mask_en) \ 138 ({ \ 139 WREG32_SOC15(VCN, inst_idx, mmUVD_DPG_LMA_CTL, \ 140 (0x0 << UVD_DPG_LMA_CTL__READ_WRITE__SHIFT | \ 141 mask_en << UVD_DPG_LMA_CTL__MASK_EN__SHIFT | \ 142 offset << UVD_DPG_LMA_CTL__READ_WRITE_ADDR__SHIFT)); \ 143 RREG32_SOC15(VCN, inst_idx, mmUVD_DPG_LMA_DATA); \ 144 }) 145 146 #define WREG32_SOC15_DPG_MODE(inst_idx, offset, value, mask_en, indirect) \ 147 do { \ 148 if (!indirect) { \ 149 WREG32_SOC15(VCN, inst_idx, mmUVD_DPG_LMA_DATA, value); \ 150 WREG32_SOC15(VCN, inst_idx, mmUVD_DPG_LMA_CTL, \ 151 (0x1 << UVD_DPG_LMA_CTL__READ_WRITE__SHIFT | \ 152 mask_en << UVD_DPG_LMA_CTL__MASK_EN__SHIFT | \ 153 offset << UVD_DPG_LMA_CTL__READ_WRITE_ADDR__SHIFT)); \ 154 } else { \ 155 *adev->vcn.inst[inst_idx].dpg_sram_curr_addr++ = offset; \ 156 *adev->vcn.inst[inst_idx].dpg_sram_curr_addr++ = value; \ 157 } \ 158 } while (0) 159 160 #define AMDGPU_VCN_FW_SHARED_FLAG_0_RB (1 << 6) 161 #define AMDGPU_VCN_MULTI_QUEUE_FLAG (1 << 8) 162 #define AMDGPU_VCN_SW_RING_FLAG (1 << 9) 163 #define AMDGPU_VCN_FW_LOGGING_FLAG (1 << 10) 164 #define AMDGPU_VCN_SMU_VERSION_INFO_FLAG (1 << 11) 165 166 #define AMDGPU_VCN_IB_FLAG_DECODE_BUFFER 0x00000001 167 #define AMDGPU_VCN_CMD_FLAG_MSG_BUFFER 0x00000001 168 169 #define VCN_CODEC_DISABLE_MASK_AV1 (1 << 0) 170 #define VCN_CODEC_DISABLE_MASK_VP9 (1 << 1) 171 #define VCN_CODEC_DISABLE_MASK_HEVC (1 << 2) 172 #define VCN_CODEC_DISABLE_MASK_H264 (1 << 3) 173 174 enum fw_queue_mode { 175 FW_QUEUE_RING_RESET = 1, 176 FW_QUEUE_DPG_HOLD_OFF = 2, 177 }; 178 179 enum engine_status_constants { 180 UVD_PGFSM_STATUS__UVDM_UVDU_PWR_ON = 0x2AAAA0, 181 UVD_PGFSM_STATUS__UVDM_UVDU_PWR_ON_2_0 = 0xAAAA0, 182 UVD_PGFSM_STATUS__UVDM_UVDU_UVDLM_PWR_ON_3_0 = 0x2A2A8AA0, 183 UVD_PGFSM_CONFIG__UVDM_UVDU_PWR_ON = 0x00000002, 184 UVD_STATUS__UVD_BUSY = 0x00000004, 185 GB_ADDR_CONFIG_DEFAULT = 0x26010011, 186 UVD_STATUS__IDLE = 0x2, 187 UVD_STATUS__BUSY = 0x5, 188 UVD_POWER_STATUS__UVD_POWER_STATUS_TILES_OFF = 0x1, 189 UVD_STATUS__RBC_BUSY = 0x1, 190 UVD_PGFSM_STATUS_UVDJ_PWR_ON = 0, 191 }; 192 193 enum internal_dpg_state { 194 VCN_DPG_STATE__UNPAUSE = 0, 195 VCN_DPG_STATE__PAUSE, 196 }; 197 198 struct dpg_pause_state { 199 enum internal_dpg_state fw_based; 200 enum internal_dpg_state jpeg; 201 }; 202 203 struct amdgpu_vcn_reg{ 204 unsigned data0; 205 unsigned data1; 206 unsigned cmd; 207 unsigned nop; 208 unsigned context_id; 209 unsigned ib_vmid; 210 unsigned ib_bar_low; 211 unsigned ib_bar_high; 212 unsigned ib_size; 213 unsigned gp_scratch8; 214 unsigned scratch9; 215 }; 216 217 struct amdgpu_vcn_fw_shared { 218 void *cpu_addr; 219 uint64_t gpu_addr; 220 uint32_t mem_size; 221 uint32_t log_offset; 222 }; 223 224 struct amdgpu_vcn_inst { 225 struct amdgpu_bo *vcpu_bo; 226 void *cpu_addr; 227 uint64_t gpu_addr; 228 void *saved_bo; 229 struct amdgpu_ring ring_dec; 230 struct amdgpu_ring ring_enc[AMDGPU_VCN_MAX_ENC_RINGS]; 231 atomic_t sched_score; 232 struct amdgpu_irq_src irq; 233 struct amdgpu_vcn_reg external; 234 struct amdgpu_bo *dpg_sram_bo; 235 struct dpg_pause_state pause_state; 236 void *dpg_sram_cpu_addr; 237 uint64_t dpg_sram_gpu_addr; 238 uint32_t *dpg_sram_curr_addr; 239 atomic_t dpg_enc_submission_cnt; 240 struct amdgpu_vcn_fw_shared fw_shared; 241 }; 242 243 struct amdgpu_vcn_ras { 244 struct amdgpu_ras_block_object ras_block; 245 }; 246 247 struct amdgpu_vcn { 248 unsigned fw_version; 249 struct delayed_work idle_work; 250 const struct firmware *fw; /* VCN firmware */ 251 unsigned num_enc_rings; 252 enum amd_powergating_state cur_state; 253 bool indirect_sram; 254 255 uint8_t num_vcn_inst; 256 struct amdgpu_vcn_inst inst[AMDGPU_MAX_VCN_INSTANCES]; 257 uint8_t vcn_config[AMDGPU_MAX_VCN_INSTANCES]; 258 uint32_t vcn_codec_disable_mask[AMDGPU_MAX_VCN_INSTANCES]; 259 struct amdgpu_vcn_reg internal; 260 struct mutex vcn_pg_lock; 261 struct mutex vcn1_jpeg1_workaround; 262 atomic_t total_submission_cnt; 263 264 unsigned harvest_config; 265 int (*pause_dpg_mode)(struct amdgpu_device *adev, 266 int inst_idx, struct dpg_pause_state *new_state); 267 268 struct ras_common_if *ras_if; 269 struct amdgpu_vcn_ras *ras; 270 }; 271 272 struct amdgpu_fw_shared_rb_ptrs_struct { 273 /* to WA DPG R/W ptr issues.*/ 274 uint32_t rptr; 275 uint32_t wptr; 276 }; 277 278 struct amdgpu_fw_shared_multi_queue { 279 uint8_t decode_queue_mode; 280 uint8_t encode_generalpurpose_queue_mode; 281 uint8_t encode_lowlatency_queue_mode; 282 uint8_t encode_realtime_queue_mode; 283 uint8_t padding[4]; 284 }; 285 286 struct amdgpu_fw_shared_sw_ring { 287 uint8_t is_enabled; 288 uint8_t padding[3]; 289 }; 290 291 struct amdgpu_fw_shared_fw_logging { 292 uint8_t is_enabled; 293 uint32_t addr_lo; 294 uint32_t addr_hi; 295 uint32_t size; 296 }; 297 298 struct amdgpu_fw_shared_smu_interface_info { 299 uint8_t smu_interface_type; 300 uint8_t padding[3]; 301 }; 302 303 struct amdgpu_fw_shared { 304 uint32_t present_flag_0; 305 uint8_t pad[44]; 306 struct amdgpu_fw_shared_rb_ptrs_struct rb; 307 uint8_t pad1[1]; 308 struct amdgpu_fw_shared_multi_queue multi_queue; 309 struct amdgpu_fw_shared_sw_ring sw_ring; 310 struct amdgpu_fw_shared_fw_logging fw_log; 311 struct amdgpu_fw_shared_smu_interface_info smu_interface_info; 312 }; 313 314 struct amdgpu_vcn_fwlog { 315 uint32_t rptr; 316 uint32_t wptr; 317 uint32_t buffer_size; 318 uint32_t header_size; 319 uint8_t wrapped; 320 }; 321 322 struct amdgpu_vcn_decode_buffer { 323 uint32_t valid_buf_flag; 324 uint32_t msg_buffer_address_hi; 325 uint32_t msg_buffer_address_lo; 326 uint32_t pad[30]; 327 }; 328 329 #define VCN_BLOCK_ENCODE_DISABLE_MASK 0x80 330 #define VCN_BLOCK_DECODE_DISABLE_MASK 0x40 331 #define VCN_BLOCK_QUEUE_DISABLE_MASK 0xC0 332 333 enum vcn_ring_type { 334 VCN_ENCODE_RING, 335 VCN_DECODE_RING, 336 VCN_UNIFIED_RING, 337 }; 338 339 int amdgpu_vcn_sw_init(struct amdgpu_device *adev); 340 int amdgpu_vcn_sw_fini(struct amdgpu_device *adev); 341 int amdgpu_vcn_suspend(struct amdgpu_device *adev); 342 int amdgpu_vcn_resume(struct amdgpu_device *adev); 343 void amdgpu_vcn_ring_begin_use(struct amdgpu_ring *ring); 344 void amdgpu_vcn_ring_end_use(struct amdgpu_ring *ring); 345 346 bool amdgpu_vcn_is_disabled_vcn(struct amdgpu_device *adev, 347 enum vcn_ring_type type, uint32_t vcn_instance); 348 349 int amdgpu_vcn_dec_ring_test_ring(struct amdgpu_ring *ring); 350 int amdgpu_vcn_dec_ring_test_ib(struct amdgpu_ring *ring, long timeout); 351 int amdgpu_vcn_dec_sw_ring_test_ring(struct amdgpu_ring *ring); 352 int amdgpu_vcn_dec_sw_ring_test_ib(struct amdgpu_ring *ring, long timeout); 353 354 int amdgpu_vcn_enc_ring_test_ring(struct amdgpu_ring *ring); 355 int amdgpu_vcn_enc_ring_test_ib(struct amdgpu_ring *ring, long timeout); 356 357 enum amdgpu_ring_priority_level amdgpu_vcn_get_enc_ring_prio(int ring); 358 359 void amdgpu_vcn_setup_ucode(struct amdgpu_device *adev); 360 361 void amdgpu_vcn_fwlog_init(struct amdgpu_vcn_inst *vcn); 362 void amdgpu_debugfs_vcn_fwlog_init(struct amdgpu_device *adev, 363 uint8_t i, struct amdgpu_vcn_inst *vcn); 364 #endif 365