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 4 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_AON_SOC_ADDRESS_2_0 0x1f800 69 #define VCN1_AON_SOC_ADDRESS_3_0 0x48000 70 #define VCN_VID_IP_ADDRESS_2_0 0x0 71 #define VCN_AON_IP_ADDRESS_2_0 0x30000 72 73 #define mmUVD_RBC_XX_IB_REG_CHECK 0x026b 74 #define mmUVD_RBC_XX_IB_REG_CHECK_BASE_IDX 1 75 #define mmUVD_REG_XX_MASK 0x026c 76 #define mmUVD_REG_XX_MASK_BASE_IDX 1 77 78 /* 1 second timeout */ 79 #define VCN_IDLE_TIMEOUT msecs_to_jiffies(1000) 80 81 #define RREG32_SOC15_DPG_MODE_1_0(ip, inst_idx, reg, mask, sram_sel) \ 82 ({ WREG32_SOC15(ip, inst_idx, mmUVD_DPG_LMA_MASK, mask); \ 83 WREG32_SOC15(ip, inst_idx, mmUVD_DPG_LMA_CTL, \ 84 UVD_DPG_LMA_CTL__MASK_EN_MASK | \ 85 ((adev->reg_offset[ip##_HWIP][inst_idx][reg##_BASE_IDX] + reg) \ 86 << UVD_DPG_LMA_CTL__READ_WRITE_ADDR__SHIFT) | \ 87 (sram_sel << UVD_DPG_LMA_CTL__SRAM_SEL__SHIFT)); \ 88 RREG32_SOC15(ip, inst_idx, mmUVD_DPG_LMA_DATA); \ 89 }) 90 91 #define WREG32_SOC15_DPG_MODE_1_0(ip, inst_idx, reg, value, mask, sram_sel) \ 92 do { \ 93 WREG32_SOC15(ip, inst_idx, mmUVD_DPG_LMA_DATA, value); \ 94 WREG32_SOC15(ip, inst_idx, mmUVD_DPG_LMA_MASK, mask); \ 95 WREG32_SOC15(ip, inst_idx, mmUVD_DPG_LMA_CTL, \ 96 UVD_DPG_LMA_CTL__READ_WRITE_MASK | \ 97 ((adev->reg_offset[ip##_HWIP][inst_idx][reg##_BASE_IDX] + reg) \ 98 << UVD_DPG_LMA_CTL__READ_WRITE_ADDR__SHIFT) | \ 99 (sram_sel << UVD_DPG_LMA_CTL__SRAM_SEL__SHIFT)); \ 100 } while (0) 101 102 #define SOC15_DPG_MODE_OFFSET(ip, inst_idx, reg) \ 103 ({ \ 104 uint32_t internal_reg_offset, addr; \ 105 bool video_range, video1_range, aon_range, aon1_range; \ 106 \ 107 addr = (adev->reg_offset[ip##_HWIP][inst_idx][reg##_BASE_IDX] + reg); \ 108 addr <<= 2; \ 109 video_range = ((((0xFFFFF & addr) >= (VCN_VID_SOC_ADDRESS_2_0)) && \ 110 ((0xFFFFF & addr) < ((VCN_VID_SOC_ADDRESS_2_0 + 0x2600))))); \ 111 video1_range = ((((0xFFFFF & addr) >= (VCN1_VID_SOC_ADDRESS_3_0)) && \ 112 ((0xFFFFF & addr) < ((VCN1_VID_SOC_ADDRESS_3_0 + 0x2600))))); \ 113 aon_range = ((((0xFFFFF & addr) >= (VCN_AON_SOC_ADDRESS_2_0)) && \ 114 ((0xFFFFF & addr) < ((VCN_AON_SOC_ADDRESS_2_0 + 0x600))))); \ 115 aon1_range = ((((0xFFFFF & addr) >= (VCN1_AON_SOC_ADDRESS_3_0)) && \ 116 ((0xFFFFF & addr) < ((VCN1_AON_SOC_ADDRESS_3_0 + 0x600))))); \ 117 if (video_range) \ 118 internal_reg_offset = ((0xFFFFF & addr) - (VCN_VID_SOC_ADDRESS_2_0) + \ 119 (VCN_VID_IP_ADDRESS_2_0)); \ 120 else if (aon_range) \ 121 internal_reg_offset = ((0xFFFFF & addr) - (VCN_AON_SOC_ADDRESS_2_0) + \ 122 (VCN_AON_IP_ADDRESS_2_0)); \ 123 else if (video1_range) \ 124 internal_reg_offset = ((0xFFFFF & addr) - (VCN1_VID_SOC_ADDRESS_3_0) + \ 125 (VCN_VID_IP_ADDRESS_2_0)); \ 126 else if (aon1_range) \ 127 internal_reg_offset = ((0xFFFFF & addr) - (VCN1_AON_SOC_ADDRESS_3_0) + \ 128 (VCN_AON_IP_ADDRESS_2_0)); \ 129 else \ 130 internal_reg_offset = (0xFFFFF & addr); \ 131 \ 132 internal_reg_offset >>= 2; \ 133 }) 134 135 #define RREG32_SOC15_DPG_MODE(inst_idx, offset, mask_en) \ 136 ({ \ 137 WREG32_SOC15(VCN, inst_idx, mmUVD_DPG_LMA_CTL, \ 138 (0x0 << UVD_DPG_LMA_CTL__READ_WRITE__SHIFT | \ 139 mask_en << UVD_DPG_LMA_CTL__MASK_EN__SHIFT | \ 140 offset << UVD_DPG_LMA_CTL__READ_WRITE_ADDR__SHIFT)); \ 141 RREG32_SOC15(VCN, inst_idx, mmUVD_DPG_LMA_DATA); \ 142 }) 143 144 #define WREG32_SOC15_DPG_MODE(inst_idx, offset, value, mask_en, indirect) \ 145 do { \ 146 if (!indirect) { \ 147 WREG32_SOC15(VCN, GET_INST(VCN, inst_idx), \ 148 mmUVD_DPG_LMA_DATA, value); \ 149 WREG32_SOC15( \ 150 VCN, GET_INST(VCN, inst_idx), \ 151 mmUVD_DPG_LMA_CTL, \ 152 (0x1 << UVD_DPG_LMA_CTL__READ_WRITE__SHIFT | \ 153 mask_en << UVD_DPG_LMA_CTL__MASK_EN__SHIFT | \ 154 offset << UVD_DPG_LMA_CTL__READ_WRITE_ADDR__SHIFT)); \ 155 } else { \ 156 *adev->vcn.inst[inst_idx].dpg_sram_curr_addr++ = \ 157 offset; \ 158 *adev->vcn.inst[inst_idx].dpg_sram_curr_addr++ = \ 159 value; \ 160 } \ 161 } while (0) 162 163 #define AMDGPU_FW_SHARED_FLAG_0_UNIFIED_QUEUE (1 << 2) 164 #define AMDGPU_FW_SHARED_FLAG_0_DRM_KEY_INJECT (1 << 4) 165 #define AMDGPU_VCN_FW_SHARED_FLAG_0_RB (1 << 6) 166 #define AMDGPU_VCN_MULTI_QUEUE_FLAG (1 << 8) 167 #define AMDGPU_VCN_SW_RING_FLAG (1 << 9) 168 #define AMDGPU_VCN_FW_LOGGING_FLAG (1 << 10) 169 #define AMDGPU_VCN_SMU_VERSION_INFO_FLAG (1 << 11) 170 #define AMDGPU_VCN_SMU_DPM_INTERFACE_FLAG (1 << 11) 171 #define AMDGPU_VCN_VF_RB_SETUP_FLAG (1 << 14) 172 #define AMDGPU_VCN_VF_RB_DECOUPLE_FLAG (1 << 15) 173 174 #define MAX_NUM_VCN_RB_SETUP 4 175 176 #define AMDGPU_VCN_IB_FLAG_DECODE_BUFFER 0x00000001 177 #define AMDGPU_VCN_CMD_FLAG_MSG_BUFFER 0x00000001 178 179 #define VCN_CODEC_DISABLE_MASK_AV1 (1 << 0) 180 #define VCN_CODEC_DISABLE_MASK_VP9 (1 << 1) 181 #define VCN_CODEC_DISABLE_MASK_HEVC (1 << 2) 182 #define VCN_CODEC_DISABLE_MASK_H264 (1 << 3) 183 184 #define AMDGPU_VCN_SMU_DPM_INTERFACE_DGPU (0) 185 #define AMDGPU_VCN_SMU_DPM_INTERFACE_APU (1) 186 187 #define AMDGPU_DRM_KEY_INJECT_WORKAROUND_VCNFW_ASD_HANDSHAKING 2 188 189 enum fw_queue_mode { 190 FW_QUEUE_RING_RESET = 1, 191 FW_QUEUE_DPG_HOLD_OFF = 2, 192 }; 193 194 enum engine_status_constants { 195 UVD_PGFSM_STATUS__UVDM_UVDU_PWR_ON = 0x2AAAA0, 196 UVD_PGFSM_STATUS__UVDM_UVDU_PWR_ON_2_0 = 0xAAAA0, 197 UVD_PGFSM_STATUS__UVDM_UVDU_UVDLM_PWR_ON_3_0 = 0x2A2A8AA0, 198 UVD_PGFSM_CONFIG__UVDM_UVDU_PWR_ON = 0x00000002, 199 UVD_STATUS__UVD_BUSY = 0x00000004, 200 GB_ADDR_CONFIG_DEFAULT = 0x26010011, 201 UVD_STATUS__IDLE = 0x2, 202 UVD_STATUS__BUSY = 0x5, 203 UVD_POWER_STATUS__UVD_POWER_STATUS_TILES_OFF = 0x1, 204 UVD_STATUS__RBC_BUSY = 0x1, 205 UVD_PGFSM_STATUS_UVDJ_PWR_ON = 0, 206 }; 207 208 enum internal_dpg_state { 209 VCN_DPG_STATE__UNPAUSE = 0, 210 VCN_DPG_STATE__PAUSE, 211 }; 212 213 struct dpg_pause_state { 214 enum internal_dpg_state fw_based; 215 enum internal_dpg_state jpeg; 216 }; 217 218 struct amdgpu_vcn_reg{ 219 unsigned data0; 220 unsigned data1; 221 unsigned cmd; 222 unsigned nop; 223 unsigned context_id; 224 unsigned ib_vmid; 225 unsigned ib_bar_low; 226 unsigned ib_bar_high; 227 unsigned ib_size; 228 unsigned gp_scratch8; 229 unsigned scratch9; 230 }; 231 232 struct amdgpu_vcn_fw_shared { 233 void *cpu_addr; 234 uint64_t gpu_addr; 235 uint32_t mem_size; 236 uint32_t log_offset; 237 }; 238 239 struct amdgpu_vcn_inst { 240 struct amdgpu_bo *vcpu_bo; 241 void *cpu_addr; 242 uint64_t gpu_addr; 243 void *saved_bo; 244 struct amdgpu_ring ring_dec; 245 struct amdgpu_ring ring_enc[AMDGPU_VCN_MAX_ENC_RINGS]; 246 atomic_t sched_score; 247 struct amdgpu_irq_src irq; 248 struct amdgpu_irq_src ras_poison_irq; 249 struct amdgpu_vcn_reg external; 250 struct amdgpu_bo *dpg_sram_bo; 251 struct dpg_pause_state pause_state; 252 void *dpg_sram_cpu_addr; 253 uint64_t dpg_sram_gpu_addr; 254 uint32_t *dpg_sram_curr_addr; 255 atomic_t dpg_enc_submission_cnt; 256 struct amdgpu_vcn_fw_shared fw_shared; 257 uint8_t aid_id; 258 }; 259 260 struct amdgpu_vcn_ras { 261 struct amdgpu_ras_block_object ras_block; 262 }; 263 264 struct amdgpu_vcn { 265 unsigned fw_version; 266 struct delayed_work idle_work; 267 const struct firmware *fw; /* VCN firmware */ 268 unsigned num_enc_rings; 269 enum amd_powergating_state cur_state; 270 bool indirect_sram; 271 272 uint8_t num_vcn_inst; 273 struct amdgpu_vcn_inst inst[AMDGPU_MAX_VCN_INSTANCES]; 274 uint8_t vcn_config[AMDGPU_MAX_VCN_INSTANCES]; 275 uint32_t vcn_codec_disable_mask[AMDGPU_MAX_VCN_INSTANCES]; 276 struct amdgpu_vcn_reg internal; 277 struct mutex vcn_pg_lock; 278 struct mutex vcn1_jpeg1_workaround; 279 atomic_t total_submission_cnt; 280 281 unsigned harvest_config; 282 int (*pause_dpg_mode)(struct amdgpu_device *adev, 283 int inst_idx, struct dpg_pause_state *new_state); 284 285 struct ras_common_if *ras_if; 286 struct amdgpu_vcn_ras *ras; 287 288 uint16_t inst_mask; 289 uint8_t num_inst_per_aid; 290 }; 291 292 struct amdgpu_fw_shared_rb_ptrs_struct { 293 /* to WA DPG R/W ptr issues.*/ 294 uint32_t rptr; 295 uint32_t wptr; 296 }; 297 298 struct amdgpu_fw_shared_multi_queue { 299 uint8_t decode_queue_mode; 300 uint8_t encode_generalpurpose_queue_mode; 301 uint8_t encode_lowlatency_queue_mode; 302 uint8_t encode_realtime_queue_mode; 303 uint8_t padding[4]; 304 }; 305 306 struct amdgpu_fw_shared_sw_ring { 307 uint8_t is_enabled; 308 uint8_t padding[3]; 309 }; 310 311 struct amdgpu_fw_shared_unified_queue_struct { 312 uint8_t is_enabled; 313 uint8_t queue_mode; 314 uint8_t queue_status; 315 uint8_t padding[5]; 316 }; 317 318 struct amdgpu_fw_shared_fw_logging { 319 uint8_t is_enabled; 320 uint32_t addr_lo; 321 uint32_t addr_hi; 322 uint32_t size; 323 }; 324 325 struct amdgpu_fw_shared_smu_interface_info { 326 uint8_t smu_interface_type; 327 uint8_t padding[3]; 328 }; 329 330 struct amdgpu_fw_shared { 331 uint32_t present_flag_0; 332 uint8_t pad[44]; 333 struct amdgpu_fw_shared_rb_ptrs_struct rb; 334 uint8_t pad1[1]; 335 struct amdgpu_fw_shared_multi_queue multi_queue; 336 struct amdgpu_fw_shared_sw_ring sw_ring; 337 struct amdgpu_fw_shared_fw_logging fw_log; 338 struct amdgpu_fw_shared_smu_interface_info smu_interface_info; 339 }; 340 341 struct amdgpu_vcn_rb_setup_info { 342 uint32_t rb_addr_lo; 343 uint32_t rb_addr_hi; 344 uint32_t rb_size; 345 }; 346 347 struct amdgpu_fw_shared_rb_setup { 348 uint32_t is_rb_enabled_flags; 349 350 union { 351 struct { 352 uint32_t rb_addr_lo; 353 uint32_t rb_addr_hi; 354 uint32_t rb_size; 355 uint32_t rb4_addr_lo; 356 uint32_t rb4_addr_hi; 357 uint32_t rb4_size; 358 uint32_t reserved[6]; 359 }; 360 361 struct { 362 struct amdgpu_vcn_rb_setup_info rb_info[MAX_NUM_VCN_RB_SETUP]; 363 }; 364 }; 365 }; 366 367 struct amdgpu_fw_shared_drm_key_wa { 368 uint8_t method; 369 uint8_t reserved[3]; 370 }; 371 372 struct amdgpu_fw_shared_queue_decouple { 373 uint8_t is_enabled; 374 uint8_t reserved[7]; 375 }; 376 377 struct amdgpu_vcn4_fw_shared { 378 uint32_t present_flag_0; 379 uint8_t pad[12]; 380 struct amdgpu_fw_shared_unified_queue_struct sq; 381 uint8_t pad1[8]; 382 struct amdgpu_fw_shared_fw_logging fw_log; 383 uint8_t pad2[20]; 384 struct amdgpu_fw_shared_rb_setup rb_setup; 385 struct amdgpu_fw_shared_smu_interface_info smu_dpm_interface; 386 struct amdgpu_fw_shared_drm_key_wa drm_key_wa; 387 uint8_t pad3[9]; 388 struct amdgpu_fw_shared_queue_decouple decouple; 389 }; 390 391 struct amdgpu_vcn_fwlog { 392 uint32_t rptr; 393 uint32_t wptr; 394 uint32_t buffer_size; 395 uint32_t header_size; 396 uint8_t wrapped; 397 }; 398 399 struct amdgpu_vcn_decode_buffer { 400 uint32_t valid_buf_flag; 401 uint32_t msg_buffer_address_hi; 402 uint32_t msg_buffer_address_lo; 403 uint32_t pad[30]; 404 }; 405 406 struct amdgpu_vcn_rb_metadata { 407 uint32_t size; 408 uint32_t present_flag_0; 409 410 uint8_t version; 411 uint8_t ring_id; 412 uint8_t pad[26]; 413 }; 414 415 #define VCN_BLOCK_ENCODE_DISABLE_MASK 0x80 416 #define VCN_BLOCK_DECODE_DISABLE_MASK 0x40 417 #define VCN_BLOCK_QUEUE_DISABLE_MASK 0xC0 418 419 enum vcn_ring_type { 420 VCN_ENCODE_RING, 421 VCN_DECODE_RING, 422 VCN_UNIFIED_RING, 423 }; 424 425 int amdgpu_vcn_early_init(struct amdgpu_device *adev); 426 int amdgpu_vcn_sw_init(struct amdgpu_device *adev); 427 int amdgpu_vcn_sw_fini(struct amdgpu_device *adev); 428 int amdgpu_vcn_suspend(struct amdgpu_device *adev); 429 int amdgpu_vcn_resume(struct amdgpu_device *adev); 430 void amdgpu_vcn_ring_begin_use(struct amdgpu_ring *ring); 431 void amdgpu_vcn_ring_end_use(struct amdgpu_ring *ring); 432 433 bool amdgpu_vcn_is_disabled_vcn(struct amdgpu_device *adev, 434 enum vcn_ring_type type, uint32_t vcn_instance); 435 436 int amdgpu_vcn_dec_ring_test_ring(struct amdgpu_ring *ring); 437 int amdgpu_vcn_dec_ring_test_ib(struct amdgpu_ring *ring, long timeout); 438 int amdgpu_vcn_dec_sw_ring_test_ring(struct amdgpu_ring *ring); 439 int amdgpu_vcn_dec_sw_ring_test_ib(struct amdgpu_ring *ring, long timeout); 440 int amdgpu_vcn_unified_ring_test_ib(struct amdgpu_ring *ring, long timeout); 441 442 int amdgpu_vcn_enc_ring_test_ring(struct amdgpu_ring *ring); 443 int amdgpu_vcn_enc_ring_test_ib(struct amdgpu_ring *ring, long timeout); 444 445 enum amdgpu_ring_priority_level amdgpu_vcn_get_enc_ring_prio(int ring); 446 447 void amdgpu_vcn_setup_ucode(struct amdgpu_device *adev); 448 449 void amdgpu_vcn_fwlog_init(struct amdgpu_vcn_inst *vcn); 450 void amdgpu_debugfs_vcn_fwlog_init(struct amdgpu_device *adev, 451 uint8_t i, struct amdgpu_vcn_inst *vcn); 452 453 int amdgpu_vcn_process_poison_irq(struct amdgpu_device *adev, 454 struct amdgpu_irq_src *source, 455 struct amdgpu_iv_entry *entry); 456 int amdgpu_vcn_ras_late_init(struct amdgpu_device *adev, 457 struct ras_common_if *ras_block); 458 int amdgpu_vcn_ras_sw_init(struct amdgpu_device *adev); 459 460 int amdgpu_vcn_psp_update_sram(struct amdgpu_device *adev, int inst_idx, 461 enum AMDGPU_UCODE_ID ucode_id); 462 463 #endif 464