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 #include <linux/firmware.h> 25 #include <drm/drmP.h> 26 #include "amdgpu.h" 27 #include "amdgpu_vcn.h" 28 #include "soc15.h" 29 #include "soc15d.h" 30 #include "soc15_common.h" 31 32 #include "vcn/vcn_1_0_offset.h" 33 #include "vcn/vcn_1_0_sh_mask.h" 34 #include "hdp/hdp_4_0_offset.h" 35 #include "mmhub/mmhub_9_1_offset.h" 36 #include "mmhub/mmhub_9_1_sh_mask.h" 37 38 #include "ivsrcid/vcn/irqsrcs_vcn_1_0.h" 39 40 #define mmUVD_RBC_XX_IB_REG_CHECK 0x05ab 41 #define mmUVD_RBC_XX_IB_REG_CHECK_BASE_IDX 1 42 #define mmUVD_REG_XX_MASK 0x05ac 43 #define mmUVD_REG_XX_MASK_BASE_IDX 1 44 45 static int vcn_v1_0_stop(struct amdgpu_device *adev); 46 static void vcn_v1_0_set_dec_ring_funcs(struct amdgpu_device *adev); 47 static void vcn_v1_0_set_enc_ring_funcs(struct amdgpu_device *adev); 48 static void vcn_v1_0_set_jpeg_ring_funcs(struct amdgpu_device *adev); 49 static void vcn_v1_0_set_irq_funcs(struct amdgpu_device *adev); 50 static void vcn_v1_0_jpeg_ring_set_patch_ring(struct amdgpu_ring *ring, uint32_t ptr); 51 static int vcn_v1_0_set_powergating_state(void *handle, enum amd_powergating_state state); 52 53 /** 54 * vcn_v1_0_early_init - set function pointers 55 * 56 * @handle: amdgpu_device pointer 57 * 58 * Set ring and irq function pointers 59 */ 60 static int vcn_v1_0_early_init(void *handle) 61 { 62 struct amdgpu_device *adev = (struct amdgpu_device *)handle; 63 64 adev->vcn.num_enc_rings = 2; 65 66 vcn_v1_0_set_dec_ring_funcs(adev); 67 vcn_v1_0_set_enc_ring_funcs(adev); 68 vcn_v1_0_set_jpeg_ring_funcs(adev); 69 vcn_v1_0_set_irq_funcs(adev); 70 71 return 0; 72 } 73 74 /** 75 * vcn_v1_0_sw_init - sw init for VCN block 76 * 77 * @handle: amdgpu_device pointer 78 * 79 * Load firmware and sw initialization 80 */ 81 static int vcn_v1_0_sw_init(void *handle) 82 { 83 struct amdgpu_ring *ring; 84 int i, r; 85 struct amdgpu_device *adev = (struct amdgpu_device *)handle; 86 87 /* VCN DEC TRAP */ 88 r = amdgpu_irq_add_id(adev, SOC15_IH_CLIENTID_VCN, VCN_1_0__SRCID__UVD_SYSTEM_MESSAGE_INTERRUPT, &adev->vcn.irq); 89 if (r) 90 return r; 91 92 /* VCN ENC TRAP */ 93 for (i = 0; i < adev->vcn.num_enc_rings; ++i) { 94 r = amdgpu_irq_add_id(adev, SOC15_IH_CLIENTID_VCN, i + VCN_1_0__SRCID__UVD_ENC_GENERAL_PURPOSE, 95 &adev->vcn.irq); 96 if (r) 97 return r; 98 } 99 100 /* VCN JPEG TRAP */ 101 r = amdgpu_irq_add_id(adev, SOC15_IH_CLIENTID_VCN, 126, &adev->vcn.irq); 102 if (r) 103 return r; 104 105 r = amdgpu_vcn_sw_init(adev); 106 if (r) 107 return r; 108 109 if (adev->firmware.load_type == AMDGPU_FW_LOAD_PSP) { 110 const struct common_firmware_header *hdr; 111 hdr = (const struct common_firmware_header *)adev->vcn.fw->data; 112 adev->firmware.ucode[AMDGPU_UCODE_ID_VCN].ucode_id = AMDGPU_UCODE_ID_VCN; 113 adev->firmware.ucode[AMDGPU_UCODE_ID_VCN].fw = adev->vcn.fw; 114 adev->firmware.fw_size += 115 ALIGN(le32_to_cpu(hdr->ucode_size_bytes), PAGE_SIZE); 116 DRM_INFO("PSP loading VCN firmware\n"); 117 } 118 119 r = amdgpu_vcn_resume(adev); 120 if (r) 121 return r; 122 123 ring = &adev->vcn.ring_dec; 124 sprintf(ring->name, "vcn_dec"); 125 r = amdgpu_ring_init(adev, ring, 512, &adev->vcn.irq, 0); 126 if (r) 127 return r; 128 129 for (i = 0; i < adev->vcn.num_enc_rings; ++i) { 130 ring = &adev->vcn.ring_enc[i]; 131 sprintf(ring->name, "vcn_enc%d", i); 132 r = amdgpu_ring_init(adev, ring, 512, &adev->vcn.irq, 0); 133 if (r) 134 return r; 135 } 136 137 ring = &adev->vcn.ring_jpeg; 138 sprintf(ring->name, "vcn_jpeg"); 139 r = amdgpu_ring_init(adev, ring, 512, &adev->vcn.irq, 0); 140 if (r) 141 return r; 142 143 return r; 144 } 145 146 /** 147 * vcn_v1_0_sw_fini - sw fini for VCN block 148 * 149 * @handle: amdgpu_device pointer 150 * 151 * VCN suspend and free up sw allocation 152 */ 153 static int vcn_v1_0_sw_fini(void *handle) 154 { 155 int r; 156 struct amdgpu_device *adev = (struct amdgpu_device *)handle; 157 158 r = amdgpu_vcn_suspend(adev); 159 if (r) 160 return r; 161 162 r = amdgpu_vcn_sw_fini(adev); 163 164 return r; 165 } 166 167 /** 168 * vcn_v1_0_hw_init - start and test VCN block 169 * 170 * @handle: amdgpu_device pointer 171 * 172 * Initialize the hardware, boot up the VCPU and do some testing 173 */ 174 static int vcn_v1_0_hw_init(void *handle) 175 { 176 struct amdgpu_device *adev = (struct amdgpu_device *)handle; 177 struct amdgpu_ring *ring = &adev->vcn.ring_dec; 178 int i, r; 179 180 r = amdgpu_ring_test_helper(ring); 181 if (r) 182 goto done; 183 184 for (i = 0; i < adev->vcn.num_enc_rings; ++i) { 185 ring = &adev->vcn.ring_enc[i]; 186 ring->sched.ready = true; 187 r = amdgpu_ring_test_helper(ring); 188 if (r) 189 goto done; 190 } 191 192 ring = &adev->vcn.ring_jpeg; 193 r = amdgpu_ring_test_helper(ring); 194 if (r) 195 goto done; 196 197 done: 198 if (!r) 199 DRM_INFO("VCN decode and encode initialized successfully(under %s).\n", 200 (adev->pg_flags & AMD_PG_SUPPORT_VCN_DPG)?"DPG Mode":"SPG Mode"); 201 202 return r; 203 } 204 205 /** 206 * vcn_v1_0_hw_fini - stop the hardware block 207 * 208 * @handle: amdgpu_device pointer 209 * 210 * Stop the VCN block, mark ring as not ready any more 211 */ 212 static int vcn_v1_0_hw_fini(void *handle) 213 { 214 struct amdgpu_device *adev = (struct amdgpu_device *)handle; 215 struct amdgpu_ring *ring = &adev->vcn.ring_dec; 216 217 if ((adev->pg_flags & AMD_PG_SUPPORT_VCN_DPG) || 218 RREG32_SOC15(VCN, 0, mmUVD_STATUS)) 219 vcn_v1_0_set_powergating_state(adev, AMD_PG_STATE_GATE); 220 221 ring->sched.ready = false; 222 223 return 0; 224 } 225 226 /** 227 * vcn_v1_0_suspend - suspend VCN block 228 * 229 * @handle: amdgpu_device pointer 230 * 231 * HW fini and suspend VCN block 232 */ 233 static int vcn_v1_0_suspend(void *handle) 234 { 235 int r; 236 struct amdgpu_device *adev = (struct amdgpu_device *)handle; 237 238 r = vcn_v1_0_hw_fini(adev); 239 if (r) 240 return r; 241 242 r = amdgpu_vcn_suspend(adev); 243 244 return r; 245 } 246 247 /** 248 * vcn_v1_0_resume - resume VCN block 249 * 250 * @handle: amdgpu_device pointer 251 * 252 * Resume firmware and hw init VCN block 253 */ 254 static int vcn_v1_0_resume(void *handle) 255 { 256 int r; 257 struct amdgpu_device *adev = (struct amdgpu_device *)handle; 258 259 r = amdgpu_vcn_resume(adev); 260 if (r) 261 return r; 262 263 r = vcn_v1_0_hw_init(adev); 264 265 return r; 266 } 267 268 /** 269 * vcn_v1_0_mc_resume_spg_mode - memory controller programming 270 * 271 * @adev: amdgpu_device pointer 272 * 273 * Let the VCN memory controller know it's offsets 274 */ 275 static void vcn_v1_0_mc_resume_spg_mode(struct amdgpu_device *adev) 276 { 277 uint32_t size = AMDGPU_GPU_PAGE_ALIGN(adev->vcn.fw->size + 4); 278 uint32_t offset; 279 280 /* cache window 0: fw */ 281 if (adev->firmware.load_type == AMDGPU_FW_LOAD_PSP) { 282 WREG32_SOC15(UVD, 0, mmUVD_LMI_VCPU_CACHE_64BIT_BAR_LOW, 283 (adev->firmware.ucode[AMDGPU_UCODE_ID_VCN].tmr_mc_addr_lo)); 284 WREG32_SOC15(UVD, 0, mmUVD_LMI_VCPU_CACHE_64BIT_BAR_HIGH, 285 (adev->firmware.ucode[AMDGPU_UCODE_ID_VCN].tmr_mc_addr_hi)); 286 WREG32_SOC15(UVD, 0, mmUVD_VCPU_CACHE_OFFSET0, 0); 287 offset = 0; 288 } else { 289 WREG32_SOC15(UVD, 0, mmUVD_LMI_VCPU_CACHE_64BIT_BAR_LOW, 290 lower_32_bits(adev->vcn.gpu_addr)); 291 WREG32_SOC15(UVD, 0, mmUVD_LMI_VCPU_CACHE_64BIT_BAR_HIGH, 292 upper_32_bits(adev->vcn.gpu_addr)); 293 offset = size; 294 WREG32_SOC15(UVD, 0, mmUVD_VCPU_CACHE_OFFSET0, 295 AMDGPU_UVD_FIRMWARE_OFFSET >> 3); 296 } 297 298 WREG32_SOC15(UVD, 0, mmUVD_VCPU_CACHE_SIZE0, size); 299 300 /* cache window 1: stack */ 301 WREG32_SOC15(UVD, 0, mmUVD_LMI_VCPU_CACHE1_64BIT_BAR_LOW, 302 lower_32_bits(adev->vcn.gpu_addr + offset)); 303 WREG32_SOC15(UVD, 0, mmUVD_LMI_VCPU_CACHE1_64BIT_BAR_HIGH, 304 upper_32_bits(adev->vcn.gpu_addr + offset)); 305 WREG32_SOC15(UVD, 0, mmUVD_VCPU_CACHE_OFFSET1, 0); 306 WREG32_SOC15(UVD, 0, mmUVD_VCPU_CACHE_SIZE1, AMDGPU_VCN_STACK_SIZE); 307 308 /* cache window 2: context */ 309 WREG32_SOC15(UVD, 0, mmUVD_LMI_VCPU_CACHE2_64BIT_BAR_LOW, 310 lower_32_bits(adev->vcn.gpu_addr + offset + AMDGPU_VCN_STACK_SIZE)); 311 WREG32_SOC15(UVD, 0, mmUVD_LMI_VCPU_CACHE2_64BIT_BAR_HIGH, 312 upper_32_bits(adev->vcn.gpu_addr + offset + AMDGPU_VCN_STACK_SIZE)); 313 WREG32_SOC15(UVD, 0, mmUVD_VCPU_CACHE_OFFSET2, 0); 314 WREG32_SOC15(UVD, 0, mmUVD_VCPU_CACHE_SIZE2, AMDGPU_VCN_CONTEXT_SIZE); 315 316 WREG32_SOC15(UVD, 0, mmUVD_UDEC_ADDR_CONFIG, 317 adev->gfx.config.gb_addr_config); 318 WREG32_SOC15(UVD, 0, mmUVD_UDEC_DB_ADDR_CONFIG, 319 adev->gfx.config.gb_addr_config); 320 WREG32_SOC15(UVD, 0, mmUVD_UDEC_DBW_ADDR_CONFIG, 321 adev->gfx.config.gb_addr_config); 322 WREG32_SOC15(UVD, 0, mmUVD_UDEC_DBW_UV_ADDR_CONFIG, 323 adev->gfx.config.gb_addr_config); 324 WREG32_SOC15(UVD, 0, mmUVD_MIF_CURR_ADDR_CONFIG, 325 adev->gfx.config.gb_addr_config); 326 WREG32_SOC15(UVD, 0, mmUVD_MIF_CURR_UV_ADDR_CONFIG, 327 adev->gfx.config.gb_addr_config); 328 WREG32_SOC15(UVD, 0, mmUVD_MIF_RECON1_ADDR_CONFIG, 329 adev->gfx.config.gb_addr_config); 330 WREG32_SOC15(UVD, 0, mmUVD_MIF_RECON1_UV_ADDR_CONFIG, 331 adev->gfx.config.gb_addr_config); 332 WREG32_SOC15(UVD, 0, mmUVD_MIF_REF_ADDR_CONFIG, 333 adev->gfx.config.gb_addr_config); 334 WREG32_SOC15(UVD, 0, mmUVD_MIF_REF_UV_ADDR_CONFIG, 335 adev->gfx.config.gb_addr_config); 336 WREG32_SOC15(UVD, 0, mmUVD_JPEG_ADDR_CONFIG, 337 adev->gfx.config.gb_addr_config); 338 WREG32_SOC15(UVD, 0, mmUVD_JPEG_UV_ADDR_CONFIG, 339 adev->gfx.config.gb_addr_config); 340 } 341 342 static void vcn_v1_0_mc_resume_dpg_mode(struct amdgpu_device *adev) 343 { 344 uint32_t size = AMDGPU_GPU_PAGE_ALIGN(adev->vcn.fw->size + 4); 345 uint32_t offset; 346 347 /* cache window 0: fw */ 348 if (adev->firmware.load_type == AMDGPU_FW_LOAD_PSP) { 349 WREG32_SOC15_DPG_MODE(UVD, 0, mmUVD_LMI_VCPU_CACHE_64BIT_BAR_LOW, 350 (adev->firmware.ucode[AMDGPU_UCODE_ID_VCN].tmr_mc_addr_lo), 351 0xFFFFFFFF, 0); 352 WREG32_SOC15_DPG_MODE(UVD, 0, mmUVD_LMI_VCPU_CACHE_64BIT_BAR_HIGH, 353 (adev->firmware.ucode[AMDGPU_UCODE_ID_VCN].tmr_mc_addr_hi), 354 0xFFFFFFFF, 0); 355 WREG32_SOC15_DPG_MODE(UVD, 0, mmUVD_VCPU_CACHE_OFFSET0, 0, 356 0xFFFFFFFF, 0); 357 offset = 0; 358 } else { 359 WREG32_SOC15_DPG_MODE(UVD, 0, mmUVD_LMI_VCPU_CACHE_64BIT_BAR_LOW, 360 lower_32_bits(adev->vcn.gpu_addr), 0xFFFFFFFF, 0); 361 WREG32_SOC15_DPG_MODE(UVD, 0, mmUVD_LMI_VCPU_CACHE_64BIT_BAR_HIGH, 362 upper_32_bits(adev->vcn.gpu_addr), 0xFFFFFFFF, 0); 363 offset = size; 364 WREG32_SOC15_DPG_MODE(UVD, 0, mmUVD_VCPU_CACHE_OFFSET0, 365 AMDGPU_UVD_FIRMWARE_OFFSET >> 3, 0xFFFFFFFF, 0); 366 } 367 368 WREG32_SOC15_DPG_MODE(UVD, 0, mmUVD_VCPU_CACHE_SIZE0, size, 0xFFFFFFFF, 0); 369 370 /* cache window 1: stack */ 371 WREG32_SOC15_DPG_MODE(UVD, 0, mmUVD_LMI_VCPU_CACHE1_64BIT_BAR_LOW, 372 lower_32_bits(adev->vcn.gpu_addr + offset), 0xFFFFFFFF, 0); 373 WREG32_SOC15_DPG_MODE(UVD, 0, mmUVD_LMI_VCPU_CACHE1_64BIT_BAR_HIGH, 374 upper_32_bits(adev->vcn.gpu_addr + offset), 0xFFFFFFFF, 0); 375 WREG32_SOC15_DPG_MODE(UVD, 0, mmUVD_VCPU_CACHE_OFFSET1, 0, 376 0xFFFFFFFF, 0); 377 WREG32_SOC15_DPG_MODE(UVD, 0, mmUVD_VCPU_CACHE_SIZE1, AMDGPU_VCN_STACK_SIZE, 378 0xFFFFFFFF, 0); 379 380 /* cache window 2: context */ 381 WREG32_SOC15_DPG_MODE(UVD, 0, mmUVD_LMI_VCPU_CACHE2_64BIT_BAR_LOW, 382 lower_32_bits(adev->vcn.gpu_addr + offset + AMDGPU_VCN_STACK_SIZE), 383 0xFFFFFFFF, 0); 384 WREG32_SOC15_DPG_MODE(UVD, 0, mmUVD_LMI_VCPU_CACHE2_64BIT_BAR_HIGH, 385 upper_32_bits(adev->vcn.gpu_addr + offset + AMDGPU_VCN_STACK_SIZE), 386 0xFFFFFFFF, 0); 387 WREG32_SOC15_DPG_MODE(UVD, 0, mmUVD_VCPU_CACHE_OFFSET2, 0, 0xFFFFFFFF, 0); 388 WREG32_SOC15_DPG_MODE(UVD, 0, mmUVD_VCPU_CACHE_SIZE2, AMDGPU_VCN_CONTEXT_SIZE, 389 0xFFFFFFFF, 0); 390 391 /* VCN global tiling registers */ 392 WREG32_SOC15_DPG_MODE(UVD, 0, mmUVD_UDEC_ADDR_CONFIG, 393 adev->gfx.config.gb_addr_config, 0xFFFFFFFF, 0); 394 WREG32_SOC15_DPG_MODE(UVD, 0, mmUVD_UDEC_DB_ADDR_CONFIG, 395 adev->gfx.config.gb_addr_config, 0xFFFFFFFF, 0); 396 WREG32_SOC15_DPG_MODE(UVD, 0, mmUVD_UDEC_DBW_ADDR_CONFIG, 397 adev->gfx.config.gb_addr_config, 0xFFFFFFFF, 0); 398 WREG32_SOC15_DPG_MODE(UVD, 0, mmUVD_UDEC_DBW_UV_ADDR_CONFIG, 399 adev->gfx.config.gb_addr_config, 0xFFFFFFFF, 0); 400 WREG32_SOC15_DPG_MODE(UVD, 0, mmUVD_MIF_CURR_ADDR_CONFIG, 401 adev->gfx.config.gb_addr_config, 0xFFFFFFFF, 0); 402 WREG32_SOC15_DPG_MODE(UVD, 0, mmUVD_MIF_CURR_UV_ADDR_CONFIG, 403 adev->gfx.config.gb_addr_config, 0xFFFFFFFF, 0); 404 WREG32_SOC15_DPG_MODE(UVD, 0, mmUVD_MIF_RECON1_ADDR_CONFIG, 405 adev->gfx.config.gb_addr_config, 0xFFFFFFFF, 0); 406 WREG32_SOC15_DPG_MODE(UVD, 0, mmUVD_MIF_RECON1_UV_ADDR_CONFIG, 407 adev->gfx.config.gb_addr_config, 0xFFFFFFFF, 0); 408 WREG32_SOC15_DPG_MODE(UVD, 0, mmUVD_MIF_REF_ADDR_CONFIG, 409 adev->gfx.config.gb_addr_config, 0xFFFFFFFF, 0); 410 WREG32_SOC15_DPG_MODE(UVD, 0, mmUVD_MIF_REF_UV_ADDR_CONFIG, 411 adev->gfx.config.gb_addr_config, 0xFFFFFFFF, 0); 412 } 413 414 /** 415 * vcn_v1_0_disable_clock_gating - disable VCN clock gating 416 * 417 * @adev: amdgpu_device pointer 418 * @sw: enable SW clock gating 419 * 420 * Disable clock gating for VCN block 421 */ 422 static void vcn_v1_0_disable_clock_gating(struct amdgpu_device *adev) 423 { 424 uint32_t data; 425 426 /* JPEG disable CGC */ 427 data = RREG32_SOC15(VCN, 0, mmJPEG_CGC_CTRL); 428 429 if (adev->cg_flags & AMD_CG_SUPPORT_VCN_MGCG) 430 data |= 1 << JPEG_CGC_CTRL__DYN_CLOCK_MODE__SHIFT; 431 else 432 data &= ~JPEG_CGC_CTRL__DYN_CLOCK_MODE_MASK; 433 434 data |= 1 << JPEG_CGC_CTRL__CLK_GATE_DLY_TIMER__SHIFT; 435 data |= 4 << JPEG_CGC_CTRL__CLK_OFF_DELAY__SHIFT; 436 WREG32_SOC15(VCN, 0, mmJPEG_CGC_CTRL, data); 437 438 data = RREG32_SOC15(VCN, 0, mmJPEG_CGC_GATE); 439 data &= ~(JPEG_CGC_GATE__JPEG_MASK | JPEG_CGC_GATE__JPEG2_MASK); 440 WREG32_SOC15(VCN, 0, mmJPEG_CGC_GATE, data); 441 442 /* UVD disable CGC */ 443 data = RREG32_SOC15(VCN, 0, mmUVD_CGC_CTRL); 444 if (adev->cg_flags & AMD_CG_SUPPORT_VCN_MGCG) 445 data |= 1 << UVD_CGC_CTRL__DYN_CLOCK_MODE__SHIFT; 446 else 447 data &= ~ UVD_CGC_CTRL__DYN_CLOCK_MODE_MASK; 448 449 data |= 1 << UVD_CGC_CTRL__CLK_GATE_DLY_TIMER__SHIFT; 450 data |= 4 << UVD_CGC_CTRL__CLK_OFF_DELAY__SHIFT; 451 WREG32_SOC15(VCN, 0, mmUVD_CGC_CTRL, data); 452 453 data = RREG32_SOC15(VCN, 0, mmUVD_CGC_GATE); 454 data &= ~(UVD_CGC_GATE__SYS_MASK 455 | UVD_CGC_GATE__UDEC_MASK 456 | UVD_CGC_GATE__MPEG2_MASK 457 | UVD_CGC_GATE__REGS_MASK 458 | UVD_CGC_GATE__RBC_MASK 459 | UVD_CGC_GATE__LMI_MC_MASK 460 | UVD_CGC_GATE__LMI_UMC_MASK 461 | UVD_CGC_GATE__IDCT_MASK 462 | UVD_CGC_GATE__MPRD_MASK 463 | UVD_CGC_GATE__MPC_MASK 464 | UVD_CGC_GATE__LBSI_MASK 465 | UVD_CGC_GATE__LRBBM_MASK 466 | UVD_CGC_GATE__UDEC_RE_MASK 467 | UVD_CGC_GATE__UDEC_CM_MASK 468 | UVD_CGC_GATE__UDEC_IT_MASK 469 | UVD_CGC_GATE__UDEC_DB_MASK 470 | UVD_CGC_GATE__UDEC_MP_MASK 471 | UVD_CGC_GATE__WCB_MASK 472 | UVD_CGC_GATE__VCPU_MASK 473 | UVD_CGC_GATE__SCPU_MASK); 474 WREG32_SOC15(VCN, 0, mmUVD_CGC_GATE, data); 475 476 data = RREG32_SOC15(VCN, 0, mmUVD_CGC_CTRL); 477 data &= ~(UVD_CGC_CTRL__UDEC_RE_MODE_MASK 478 | UVD_CGC_CTRL__UDEC_CM_MODE_MASK 479 | UVD_CGC_CTRL__UDEC_IT_MODE_MASK 480 | UVD_CGC_CTRL__UDEC_DB_MODE_MASK 481 | UVD_CGC_CTRL__UDEC_MP_MODE_MASK 482 | UVD_CGC_CTRL__SYS_MODE_MASK 483 | UVD_CGC_CTRL__UDEC_MODE_MASK 484 | UVD_CGC_CTRL__MPEG2_MODE_MASK 485 | UVD_CGC_CTRL__REGS_MODE_MASK 486 | UVD_CGC_CTRL__RBC_MODE_MASK 487 | UVD_CGC_CTRL__LMI_MC_MODE_MASK 488 | UVD_CGC_CTRL__LMI_UMC_MODE_MASK 489 | UVD_CGC_CTRL__IDCT_MODE_MASK 490 | UVD_CGC_CTRL__MPRD_MODE_MASK 491 | UVD_CGC_CTRL__MPC_MODE_MASK 492 | UVD_CGC_CTRL__LBSI_MODE_MASK 493 | UVD_CGC_CTRL__LRBBM_MODE_MASK 494 | UVD_CGC_CTRL__WCB_MODE_MASK 495 | UVD_CGC_CTRL__VCPU_MODE_MASK 496 | UVD_CGC_CTRL__SCPU_MODE_MASK); 497 WREG32_SOC15(VCN, 0, mmUVD_CGC_CTRL, data); 498 499 /* turn on */ 500 data = RREG32_SOC15(VCN, 0, mmUVD_SUVD_CGC_GATE); 501 data |= (UVD_SUVD_CGC_GATE__SRE_MASK 502 | UVD_SUVD_CGC_GATE__SIT_MASK 503 | UVD_SUVD_CGC_GATE__SMP_MASK 504 | UVD_SUVD_CGC_GATE__SCM_MASK 505 | UVD_SUVD_CGC_GATE__SDB_MASK 506 | UVD_SUVD_CGC_GATE__SRE_H264_MASK 507 | UVD_SUVD_CGC_GATE__SRE_HEVC_MASK 508 | UVD_SUVD_CGC_GATE__SIT_H264_MASK 509 | UVD_SUVD_CGC_GATE__SIT_HEVC_MASK 510 | UVD_SUVD_CGC_GATE__SCM_H264_MASK 511 | UVD_SUVD_CGC_GATE__SCM_HEVC_MASK 512 | UVD_SUVD_CGC_GATE__SDB_H264_MASK 513 | UVD_SUVD_CGC_GATE__SDB_HEVC_MASK 514 | UVD_SUVD_CGC_GATE__SCLR_MASK 515 | UVD_SUVD_CGC_GATE__UVD_SC_MASK 516 | UVD_SUVD_CGC_GATE__ENT_MASK 517 | UVD_SUVD_CGC_GATE__SIT_HEVC_DEC_MASK 518 | UVD_SUVD_CGC_GATE__SIT_HEVC_ENC_MASK 519 | UVD_SUVD_CGC_GATE__SITE_MASK 520 | UVD_SUVD_CGC_GATE__SRE_VP9_MASK 521 | UVD_SUVD_CGC_GATE__SCM_VP9_MASK 522 | UVD_SUVD_CGC_GATE__SIT_VP9_DEC_MASK 523 | UVD_SUVD_CGC_GATE__SDB_VP9_MASK 524 | UVD_SUVD_CGC_GATE__IME_HEVC_MASK); 525 WREG32_SOC15(VCN, 0, mmUVD_SUVD_CGC_GATE, data); 526 527 data = RREG32_SOC15(VCN, 0, mmUVD_SUVD_CGC_CTRL); 528 data &= ~(UVD_SUVD_CGC_CTRL__SRE_MODE_MASK 529 | UVD_SUVD_CGC_CTRL__SIT_MODE_MASK 530 | UVD_SUVD_CGC_CTRL__SMP_MODE_MASK 531 | UVD_SUVD_CGC_CTRL__SCM_MODE_MASK 532 | UVD_SUVD_CGC_CTRL__SDB_MODE_MASK 533 | UVD_SUVD_CGC_CTRL__SCLR_MODE_MASK 534 | UVD_SUVD_CGC_CTRL__UVD_SC_MODE_MASK 535 | UVD_SUVD_CGC_CTRL__ENT_MODE_MASK 536 | UVD_SUVD_CGC_CTRL__IME_MODE_MASK 537 | UVD_SUVD_CGC_CTRL__SITE_MODE_MASK); 538 WREG32_SOC15(VCN, 0, mmUVD_SUVD_CGC_CTRL, data); 539 } 540 541 /** 542 * vcn_v1_0_enable_clock_gating - enable VCN clock gating 543 * 544 * @adev: amdgpu_device pointer 545 * @sw: enable SW clock gating 546 * 547 * Enable clock gating for VCN block 548 */ 549 static void vcn_v1_0_enable_clock_gating(struct amdgpu_device *adev) 550 { 551 uint32_t data = 0; 552 553 /* enable JPEG CGC */ 554 data = RREG32_SOC15(VCN, 0, mmJPEG_CGC_CTRL); 555 if (adev->cg_flags & AMD_CG_SUPPORT_VCN_MGCG) 556 data |= 1 << JPEG_CGC_CTRL__DYN_CLOCK_MODE__SHIFT; 557 else 558 data |= 0 << JPEG_CGC_CTRL__DYN_CLOCK_MODE__SHIFT; 559 data |= 1 << JPEG_CGC_CTRL__CLK_GATE_DLY_TIMER__SHIFT; 560 data |= 4 << JPEG_CGC_CTRL__CLK_OFF_DELAY__SHIFT; 561 WREG32_SOC15(VCN, 0, mmJPEG_CGC_CTRL, data); 562 563 data = RREG32_SOC15(VCN, 0, mmJPEG_CGC_GATE); 564 data |= (JPEG_CGC_GATE__JPEG_MASK | JPEG_CGC_GATE__JPEG2_MASK); 565 WREG32_SOC15(VCN, 0, mmJPEG_CGC_GATE, data); 566 567 /* enable UVD CGC */ 568 data = RREG32_SOC15(VCN, 0, mmUVD_CGC_CTRL); 569 if (adev->cg_flags & AMD_CG_SUPPORT_VCN_MGCG) 570 data |= 1 << UVD_CGC_CTRL__DYN_CLOCK_MODE__SHIFT; 571 else 572 data |= 0 << UVD_CGC_CTRL__DYN_CLOCK_MODE__SHIFT; 573 data |= 1 << UVD_CGC_CTRL__CLK_GATE_DLY_TIMER__SHIFT; 574 data |= 4 << UVD_CGC_CTRL__CLK_OFF_DELAY__SHIFT; 575 WREG32_SOC15(VCN, 0, mmUVD_CGC_CTRL, data); 576 577 data = RREG32_SOC15(VCN, 0, mmUVD_CGC_CTRL); 578 data |= (UVD_CGC_CTRL__UDEC_RE_MODE_MASK 579 | UVD_CGC_CTRL__UDEC_CM_MODE_MASK 580 | UVD_CGC_CTRL__UDEC_IT_MODE_MASK 581 | UVD_CGC_CTRL__UDEC_DB_MODE_MASK 582 | UVD_CGC_CTRL__UDEC_MP_MODE_MASK 583 | UVD_CGC_CTRL__SYS_MODE_MASK 584 | UVD_CGC_CTRL__UDEC_MODE_MASK 585 | UVD_CGC_CTRL__MPEG2_MODE_MASK 586 | UVD_CGC_CTRL__REGS_MODE_MASK 587 | UVD_CGC_CTRL__RBC_MODE_MASK 588 | UVD_CGC_CTRL__LMI_MC_MODE_MASK 589 | UVD_CGC_CTRL__LMI_UMC_MODE_MASK 590 | UVD_CGC_CTRL__IDCT_MODE_MASK 591 | UVD_CGC_CTRL__MPRD_MODE_MASK 592 | UVD_CGC_CTRL__MPC_MODE_MASK 593 | UVD_CGC_CTRL__LBSI_MODE_MASK 594 | UVD_CGC_CTRL__LRBBM_MODE_MASK 595 | UVD_CGC_CTRL__WCB_MODE_MASK 596 | UVD_CGC_CTRL__VCPU_MODE_MASK 597 | UVD_CGC_CTRL__SCPU_MODE_MASK); 598 WREG32_SOC15(VCN, 0, mmUVD_CGC_CTRL, data); 599 600 data = RREG32_SOC15(VCN, 0, mmUVD_SUVD_CGC_CTRL); 601 data |= (UVD_SUVD_CGC_CTRL__SRE_MODE_MASK 602 | UVD_SUVD_CGC_CTRL__SIT_MODE_MASK 603 | UVD_SUVD_CGC_CTRL__SMP_MODE_MASK 604 | UVD_SUVD_CGC_CTRL__SCM_MODE_MASK 605 | UVD_SUVD_CGC_CTRL__SDB_MODE_MASK 606 | UVD_SUVD_CGC_CTRL__SCLR_MODE_MASK 607 | UVD_SUVD_CGC_CTRL__UVD_SC_MODE_MASK 608 | UVD_SUVD_CGC_CTRL__ENT_MODE_MASK 609 | UVD_SUVD_CGC_CTRL__IME_MODE_MASK 610 | UVD_SUVD_CGC_CTRL__SITE_MODE_MASK); 611 WREG32_SOC15(VCN, 0, mmUVD_SUVD_CGC_CTRL, data); 612 } 613 614 static void vcn_v1_0_clock_gating_dpg_mode(struct amdgpu_device *adev, uint8_t sram_sel) 615 { 616 uint32_t reg_data = 0; 617 618 /* disable JPEG CGC */ 619 if (adev->cg_flags & AMD_CG_SUPPORT_VCN_MGCG) 620 reg_data = 1 << JPEG_CGC_CTRL__DYN_CLOCK_MODE__SHIFT; 621 else 622 reg_data = 0 << JPEG_CGC_CTRL__DYN_CLOCK_MODE__SHIFT; 623 reg_data |= 1 << JPEG_CGC_CTRL__CLK_GATE_DLY_TIMER__SHIFT; 624 reg_data |= 4 << JPEG_CGC_CTRL__CLK_OFF_DELAY__SHIFT; 625 WREG32_SOC15_DPG_MODE(UVD, 0, mmJPEG_CGC_CTRL, reg_data, 0xFFFFFFFF, sram_sel); 626 627 WREG32_SOC15_DPG_MODE(UVD, 0, mmJPEG_CGC_GATE, 0, 0xFFFFFFFF, sram_sel); 628 629 /* enable sw clock gating control */ 630 if (adev->cg_flags & AMD_CG_SUPPORT_VCN_MGCG) 631 reg_data = 1 << UVD_CGC_CTRL__DYN_CLOCK_MODE__SHIFT; 632 else 633 reg_data = 0 << UVD_CGC_CTRL__DYN_CLOCK_MODE__SHIFT; 634 reg_data |= 1 << UVD_CGC_CTRL__CLK_GATE_DLY_TIMER__SHIFT; 635 reg_data |= 4 << UVD_CGC_CTRL__CLK_OFF_DELAY__SHIFT; 636 reg_data &= ~(UVD_CGC_CTRL__UDEC_RE_MODE_MASK | 637 UVD_CGC_CTRL__UDEC_CM_MODE_MASK | 638 UVD_CGC_CTRL__UDEC_IT_MODE_MASK | 639 UVD_CGC_CTRL__UDEC_DB_MODE_MASK | 640 UVD_CGC_CTRL__UDEC_MP_MODE_MASK | 641 UVD_CGC_CTRL__SYS_MODE_MASK | 642 UVD_CGC_CTRL__UDEC_MODE_MASK | 643 UVD_CGC_CTRL__MPEG2_MODE_MASK | 644 UVD_CGC_CTRL__REGS_MODE_MASK | 645 UVD_CGC_CTRL__RBC_MODE_MASK | 646 UVD_CGC_CTRL__LMI_MC_MODE_MASK | 647 UVD_CGC_CTRL__LMI_UMC_MODE_MASK | 648 UVD_CGC_CTRL__IDCT_MODE_MASK | 649 UVD_CGC_CTRL__MPRD_MODE_MASK | 650 UVD_CGC_CTRL__MPC_MODE_MASK | 651 UVD_CGC_CTRL__LBSI_MODE_MASK | 652 UVD_CGC_CTRL__LRBBM_MODE_MASK | 653 UVD_CGC_CTRL__WCB_MODE_MASK | 654 UVD_CGC_CTRL__VCPU_MODE_MASK | 655 UVD_CGC_CTRL__SCPU_MODE_MASK); 656 WREG32_SOC15_DPG_MODE(UVD, 0, mmUVD_CGC_CTRL, reg_data, 0xFFFFFFFF, sram_sel); 657 658 /* turn off clock gating */ 659 WREG32_SOC15_DPG_MODE(UVD, 0, mmUVD_CGC_GATE, 0, 0xFFFFFFFF, sram_sel); 660 661 /* turn on SUVD clock gating */ 662 WREG32_SOC15_DPG_MODE(UVD, 0, mmUVD_SUVD_CGC_GATE, 1, 0xFFFFFFFF, sram_sel); 663 664 /* turn on sw mode in UVD_SUVD_CGC_CTRL */ 665 WREG32_SOC15_DPG_MODE(UVD, 0, mmUVD_SUVD_CGC_CTRL, 0, 0xFFFFFFFF, sram_sel); 666 } 667 668 static void vcn_1_0_disable_static_power_gating(struct amdgpu_device *adev) 669 { 670 uint32_t data = 0; 671 int ret; 672 673 if (adev->pg_flags & AMD_PG_SUPPORT_VCN) { 674 data = (1 << UVD_PGFSM_CONFIG__UVDM_PWR_CONFIG__SHIFT 675 | 1 << UVD_PGFSM_CONFIG__UVDU_PWR_CONFIG__SHIFT 676 | 2 << UVD_PGFSM_CONFIG__UVDF_PWR_CONFIG__SHIFT 677 | 2 << UVD_PGFSM_CONFIG__UVDC_PWR_CONFIG__SHIFT 678 | 2 << UVD_PGFSM_CONFIG__UVDB_PWR_CONFIG__SHIFT 679 | 2 << UVD_PGFSM_CONFIG__UVDIL_PWR_CONFIG__SHIFT 680 | 2 << UVD_PGFSM_CONFIG__UVDIR_PWR_CONFIG__SHIFT 681 | 2 << UVD_PGFSM_CONFIG__UVDTD_PWR_CONFIG__SHIFT 682 | 2 << UVD_PGFSM_CONFIG__UVDTE_PWR_CONFIG__SHIFT 683 | 2 << UVD_PGFSM_CONFIG__UVDE_PWR_CONFIG__SHIFT 684 | 2 << UVD_PGFSM_CONFIG__UVDW_PWR_CONFIG__SHIFT); 685 686 WREG32_SOC15(VCN, 0, mmUVD_PGFSM_CONFIG, data); 687 SOC15_WAIT_ON_RREG(VCN, 0, mmUVD_PGFSM_STATUS, UVD_PGFSM_STATUS__UVDM_UVDU_PWR_ON, 0xFFFFFF, ret); 688 } else { 689 data = (1 << UVD_PGFSM_CONFIG__UVDM_PWR_CONFIG__SHIFT 690 | 1 << UVD_PGFSM_CONFIG__UVDU_PWR_CONFIG__SHIFT 691 | 1 << UVD_PGFSM_CONFIG__UVDF_PWR_CONFIG__SHIFT 692 | 1 << UVD_PGFSM_CONFIG__UVDC_PWR_CONFIG__SHIFT 693 | 1 << UVD_PGFSM_CONFIG__UVDB_PWR_CONFIG__SHIFT 694 | 1 << UVD_PGFSM_CONFIG__UVDIL_PWR_CONFIG__SHIFT 695 | 1 << UVD_PGFSM_CONFIG__UVDIR_PWR_CONFIG__SHIFT 696 | 1 << UVD_PGFSM_CONFIG__UVDTD_PWR_CONFIG__SHIFT 697 | 1 << UVD_PGFSM_CONFIG__UVDTE_PWR_CONFIG__SHIFT 698 | 1 << UVD_PGFSM_CONFIG__UVDE_PWR_CONFIG__SHIFT 699 | 1 << UVD_PGFSM_CONFIG__UVDW_PWR_CONFIG__SHIFT); 700 WREG32_SOC15(VCN, 0, mmUVD_PGFSM_CONFIG, data); 701 SOC15_WAIT_ON_RREG(VCN, 0, mmUVD_PGFSM_STATUS, 0, 0xFFFFFFFF, ret); 702 } 703 704 /* polling UVD_PGFSM_STATUS to confirm UVDM_PWR_STATUS , UVDU_PWR_STATUS are 0 (power on) */ 705 706 data = RREG32_SOC15(VCN, 0, mmUVD_POWER_STATUS); 707 data &= ~0x103; 708 if (adev->pg_flags & AMD_PG_SUPPORT_VCN) 709 data |= UVD_PGFSM_CONFIG__UVDM_UVDU_PWR_ON | UVD_POWER_STATUS__UVD_PG_EN_MASK; 710 711 WREG32_SOC15(VCN, 0, mmUVD_POWER_STATUS, data); 712 } 713 714 static void vcn_1_0_enable_static_power_gating(struct amdgpu_device *adev) 715 { 716 uint32_t data = 0; 717 int ret; 718 719 if (adev->pg_flags & AMD_PG_SUPPORT_VCN) { 720 /* Before power off, this indicator has to be turned on */ 721 data = RREG32_SOC15(VCN, 0, mmUVD_POWER_STATUS); 722 data &= ~UVD_POWER_STATUS__UVD_POWER_STATUS_MASK; 723 data |= UVD_POWER_STATUS__UVD_POWER_STATUS_TILES_OFF; 724 WREG32_SOC15(VCN, 0, mmUVD_POWER_STATUS, data); 725 726 727 data = (2 << UVD_PGFSM_CONFIG__UVDM_PWR_CONFIG__SHIFT 728 | 2 << UVD_PGFSM_CONFIG__UVDU_PWR_CONFIG__SHIFT 729 | 2 << UVD_PGFSM_CONFIG__UVDF_PWR_CONFIG__SHIFT 730 | 2 << UVD_PGFSM_CONFIG__UVDC_PWR_CONFIG__SHIFT 731 | 2 << UVD_PGFSM_CONFIG__UVDB_PWR_CONFIG__SHIFT 732 | 2 << UVD_PGFSM_CONFIG__UVDIL_PWR_CONFIG__SHIFT 733 | 2 << UVD_PGFSM_CONFIG__UVDIR_PWR_CONFIG__SHIFT 734 | 2 << UVD_PGFSM_CONFIG__UVDTD_PWR_CONFIG__SHIFT 735 | 2 << UVD_PGFSM_CONFIG__UVDTE_PWR_CONFIG__SHIFT 736 | 2 << UVD_PGFSM_CONFIG__UVDE_PWR_CONFIG__SHIFT 737 | 2 << UVD_PGFSM_CONFIG__UVDW_PWR_CONFIG__SHIFT); 738 739 WREG32_SOC15(VCN, 0, mmUVD_PGFSM_CONFIG, data); 740 741 data = (2 << UVD_PGFSM_STATUS__UVDM_PWR_STATUS__SHIFT 742 | 2 << UVD_PGFSM_STATUS__UVDU_PWR_STATUS__SHIFT 743 | 2 << UVD_PGFSM_STATUS__UVDF_PWR_STATUS__SHIFT 744 | 2 << UVD_PGFSM_STATUS__UVDC_PWR_STATUS__SHIFT 745 | 2 << UVD_PGFSM_STATUS__UVDB_PWR_STATUS__SHIFT 746 | 2 << UVD_PGFSM_STATUS__UVDIL_PWR_STATUS__SHIFT 747 | 2 << UVD_PGFSM_STATUS__UVDIR_PWR_STATUS__SHIFT 748 | 2 << UVD_PGFSM_STATUS__UVDTD_PWR_STATUS__SHIFT 749 | 2 << UVD_PGFSM_STATUS__UVDTE_PWR_STATUS__SHIFT 750 | 2 << UVD_PGFSM_STATUS__UVDE_PWR_STATUS__SHIFT 751 | 2 << UVD_PGFSM_STATUS__UVDW_PWR_STATUS__SHIFT); 752 SOC15_WAIT_ON_RREG(VCN, 0, mmUVD_PGFSM_STATUS, data, 0xFFFFFFFF, ret); 753 } 754 } 755 756 /** 757 * vcn_v1_0_start - start VCN block 758 * 759 * @adev: amdgpu_device pointer 760 * 761 * Setup and start the VCN block 762 */ 763 static int vcn_v1_0_start_spg_mode(struct amdgpu_device *adev) 764 { 765 struct amdgpu_ring *ring = &adev->vcn.ring_dec; 766 uint32_t rb_bufsz, tmp; 767 uint32_t lmi_swap_cntl; 768 int i, j, r; 769 770 /* disable byte swapping */ 771 lmi_swap_cntl = 0; 772 773 vcn_1_0_disable_static_power_gating(adev); 774 775 tmp = RREG32_SOC15(UVD, 0, mmUVD_STATUS) | UVD_STATUS__UVD_BUSY; 776 WREG32_SOC15(UVD, 0, mmUVD_STATUS, tmp); 777 778 /* disable clock gating */ 779 vcn_v1_0_disable_clock_gating(adev); 780 781 /* disable interupt */ 782 WREG32_P(SOC15_REG_OFFSET(UVD, 0, mmUVD_MASTINT_EN), 0, 783 ~UVD_MASTINT_EN__VCPU_EN_MASK); 784 785 /* initialize VCN memory controller */ 786 tmp = RREG32_SOC15(UVD, 0, mmUVD_LMI_CTRL); 787 WREG32_SOC15(UVD, 0, mmUVD_LMI_CTRL, tmp | 788 UVD_LMI_CTRL__WRITE_CLEAN_TIMER_EN_MASK | 789 UVD_LMI_CTRL__MASK_MC_URGENT_MASK | 790 UVD_LMI_CTRL__DATA_COHERENCY_EN_MASK | 791 UVD_LMI_CTRL__VCPU_DATA_COHERENCY_EN_MASK); 792 793 #ifdef __BIG_ENDIAN 794 /* swap (8 in 32) RB and IB */ 795 lmi_swap_cntl = 0xa; 796 #endif 797 WREG32_SOC15(UVD, 0, mmUVD_LMI_SWAP_CNTL, lmi_swap_cntl); 798 799 tmp = RREG32_SOC15(UVD, 0, mmUVD_MPC_CNTL); 800 tmp &= ~UVD_MPC_CNTL__REPLACEMENT_MODE_MASK; 801 tmp |= 0x2 << UVD_MPC_CNTL__REPLACEMENT_MODE__SHIFT; 802 WREG32_SOC15(UVD, 0, mmUVD_MPC_CNTL, tmp); 803 804 WREG32_SOC15(UVD, 0, mmUVD_MPC_SET_MUXA0, 805 ((0x1 << UVD_MPC_SET_MUXA0__VARA_1__SHIFT) | 806 (0x2 << UVD_MPC_SET_MUXA0__VARA_2__SHIFT) | 807 (0x3 << UVD_MPC_SET_MUXA0__VARA_3__SHIFT) | 808 (0x4 << UVD_MPC_SET_MUXA0__VARA_4__SHIFT))); 809 810 WREG32_SOC15(UVD, 0, mmUVD_MPC_SET_MUXB0, 811 ((0x1 << UVD_MPC_SET_MUXB0__VARB_1__SHIFT) | 812 (0x2 << UVD_MPC_SET_MUXB0__VARB_2__SHIFT) | 813 (0x3 << UVD_MPC_SET_MUXB0__VARB_3__SHIFT) | 814 (0x4 << UVD_MPC_SET_MUXB0__VARB_4__SHIFT))); 815 816 WREG32_SOC15(UVD, 0, mmUVD_MPC_SET_MUX, 817 ((0x0 << UVD_MPC_SET_MUX__SET_0__SHIFT) | 818 (0x1 << UVD_MPC_SET_MUX__SET_1__SHIFT) | 819 (0x2 << UVD_MPC_SET_MUX__SET_2__SHIFT))); 820 821 vcn_v1_0_mc_resume_spg_mode(adev); 822 823 WREG32_SOC15(UVD, 0, mmUVD_REG_XX_MASK, 0x10); 824 WREG32_SOC15(UVD, 0, mmUVD_RBC_XX_IB_REG_CHECK, 825 RREG32_SOC15(UVD, 0, mmUVD_RBC_XX_IB_REG_CHECK) | 0x3); 826 827 /* enable VCPU clock */ 828 WREG32_SOC15(UVD, 0, mmUVD_VCPU_CNTL, UVD_VCPU_CNTL__CLK_EN_MASK); 829 830 /* boot up the VCPU */ 831 WREG32_P(SOC15_REG_OFFSET(UVD, 0, mmUVD_SOFT_RESET), 0, 832 ~UVD_SOFT_RESET__VCPU_SOFT_RESET_MASK); 833 834 /* enable UMC */ 835 WREG32_P(SOC15_REG_OFFSET(UVD, 0, mmUVD_LMI_CTRL2), 0, 836 ~UVD_LMI_CTRL2__STALL_ARB_UMC_MASK); 837 838 tmp = RREG32_SOC15(UVD, 0, mmUVD_SOFT_RESET); 839 tmp &= ~UVD_SOFT_RESET__LMI_SOFT_RESET_MASK; 840 tmp &= ~UVD_SOFT_RESET__LMI_UMC_SOFT_RESET_MASK; 841 WREG32_SOC15(UVD, 0, mmUVD_SOFT_RESET, tmp); 842 843 for (i = 0; i < 10; ++i) { 844 uint32_t status; 845 846 for (j = 0; j < 100; ++j) { 847 status = RREG32_SOC15(UVD, 0, mmUVD_STATUS); 848 if (status & UVD_STATUS__IDLE) 849 break; 850 mdelay(10); 851 } 852 r = 0; 853 if (status & UVD_STATUS__IDLE) 854 break; 855 856 DRM_ERROR("VCN decode not responding, trying to reset the VCPU!!!\n"); 857 WREG32_P(SOC15_REG_OFFSET(UVD, 0, mmUVD_SOFT_RESET), 858 UVD_SOFT_RESET__VCPU_SOFT_RESET_MASK, 859 ~UVD_SOFT_RESET__VCPU_SOFT_RESET_MASK); 860 mdelay(10); 861 WREG32_P(SOC15_REG_OFFSET(UVD, 0, mmUVD_SOFT_RESET), 0, 862 ~UVD_SOFT_RESET__VCPU_SOFT_RESET_MASK); 863 mdelay(10); 864 r = -1; 865 } 866 867 if (r) { 868 DRM_ERROR("VCN decode not responding, giving up!!!\n"); 869 return r; 870 } 871 /* enable master interrupt */ 872 WREG32_P(SOC15_REG_OFFSET(UVD, 0, mmUVD_MASTINT_EN), 873 UVD_MASTINT_EN__VCPU_EN_MASK, ~UVD_MASTINT_EN__VCPU_EN_MASK); 874 875 /* enable system interrupt for JRBC, TODO: move to set interrupt*/ 876 WREG32_P(SOC15_REG_OFFSET(UVD, 0, mmUVD_SYS_INT_EN), 877 UVD_SYS_INT_EN__UVD_JRBC_EN_MASK, 878 ~UVD_SYS_INT_EN__UVD_JRBC_EN_MASK); 879 880 /* clear the busy bit of UVD_STATUS */ 881 tmp = RREG32_SOC15(UVD, 0, mmUVD_STATUS) & ~UVD_STATUS__UVD_BUSY; 882 WREG32_SOC15(UVD, 0, mmUVD_STATUS, tmp); 883 884 /* force RBC into idle state */ 885 rb_bufsz = order_base_2(ring->ring_size); 886 tmp = REG_SET_FIELD(0, UVD_RBC_RB_CNTL, RB_BUFSZ, rb_bufsz); 887 tmp = REG_SET_FIELD(tmp, UVD_RBC_RB_CNTL, RB_BLKSZ, 1); 888 tmp = REG_SET_FIELD(tmp, UVD_RBC_RB_CNTL, RB_NO_FETCH, 1); 889 tmp = REG_SET_FIELD(tmp, UVD_RBC_RB_CNTL, RB_NO_UPDATE, 1); 890 tmp = REG_SET_FIELD(tmp, UVD_RBC_RB_CNTL, RB_RPTR_WR_EN, 1); 891 WREG32_SOC15(UVD, 0, mmUVD_RBC_RB_CNTL, tmp); 892 893 /* set the write pointer delay */ 894 WREG32_SOC15(UVD, 0, mmUVD_RBC_RB_WPTR_CNTL, 0); 895 896 /* set the wb address */ 897 WREG32_SOC15(UVD, 0, mmUVD_RBC_RB_RPTR_ADDR, 898 (upper_32_bits(ring->gpu_addr) >> 2)); 899 900 /* programm the RB_BASE for ring buffer */ 901 WREG32_SOC15(UVD, 0, mmUVD_LMI_RBC_RB_64BIT_BAR_LOW, 902 lower_32_bits(ring->gpu_addr)); 903 WREG32_SOC15(UVD, 0, mmUVD_LMI_RBC_RB_64BIT_BAR_HIGH, 904 upper_32_bits(ring->gpu_addr)); 905 906 /* Initialize the ring buffer's read and write pointers */ 907 WREG32_SOC15(UVD, 0, mmUVD_RBC_RB_RPTR, 0); 908 909 WREG32_SOC15(UVD, 0, mmUVD_SCRATCH2, 0); 910 911 ring->wptr = RREG32_SOC15(UVD, 0, mmUVD_RBC_RB_RPTR); 912 WREG32_SOC15(UVD, 0, mmUVD_RBC_RB_WPTR, 913 lower_32_bits(ring->wptr)); 914 915 WREG32_P(SOC15_REG_OFFSET(UVD, 0, mmUVD_RBC_RB_CNTL), 0, 916 ~UVD_RBC_RB_CNTL__RB_NO_FETCH_MASK); 917 918 ring = &adev->vcn.ring_enc[0]; 919 WREG32_SOC15(UVD, 0, mmUVD_RB_RPTR, lower_32_bits(ring->wptr)); 920 WREG32_SOC15(UVD, 0, mmUVD_RB_WPTR, lower_32_bits(ring->wptr)); 921 WREG32_SOC15(UVD, 0, mmUVD_RB_BASE_LO, ring->gpu_addr); 922 WREG32_SOC15(UVD, 0, mmUVD_RB_BASE_HI, upper_32_bits(ring->gpu_addr)); 923 WREG32_SOC15(UVD, 0, mmUVD_RB_SIZE, ring->ring_size / 4); 924 925 ring = &adev->vcn.ring_enc[1]; 926 WREG32_SOC15(UVD, 0, mmUVD_RB_RPTR2, lower_32_bits(ring->wptr)); 927 WREG32_SOC15(UVD, 0, mmUVD_RB_WPTR2, lower_32_bits(ring->wptr)); 928 WREG32_SOC15(UVD, 0, mmUVD_RB_BASE_LO2, ring->gpu_addr); 929 WREG32_SOC15(UVD, 0, mmUVD_RB_BASE_HI2, upper_32_bits(ring->gpu_addr)); 930 WREG32_SOC15(UVD, 0, mmUVD_RB_SIZE2, ring->ring_size / 4); 931 932 ring = &adev->vcn.ring_jpeg; 933 WREG32_SOC15(UVD, 0, mmUVD_LMI_JRBC_RB_VMID, 0); 934 WREG32_SOC15(UVD, 0, mmUVD_JRBC_RB_CNTL, UVD_JRBC_RB_CNTL__RB_NO_FETCH_MASK | 935 UVD_JRBC_RB_CNTL__RB_RPTR_WR_EN_MASK); 936 WREG32_SOC15(UVD, 0, mmUVD_LMI_JRBC_RB_64BIT_BAR_LOW, lower_32_bits(ring->gpu_addr)); 937 WREG32_SOC15(UVD, 0, mmUVD_LMI_JRBC_RB_64BIT_BAR_HIGH, upper_32_bits(ring->gpu_addr)); 938 WREG32_SOC15(UVD, 0, mmUVD_JRBC_RB_RPTR, 0); 939 WREG32_SOC15(UVD, 0, mmUVD_JRBC_RB_WPTR, 0); 940 WREG32_SOC15(UVD, 0, mmUVD_JRBC_RB_CNTL, UVD_JRBC_RB_CNTL__RB_RPTR_WR_EN_MASK); 941 942 /* initialize wptr */ 943 ring->wptr = RREG32_SOC15(UVD, 0, mmUVD_JRBC_RB_WPTR); 944 945 /* copy patch commands to the jpeg ring */ 946 vcn_v1_0_jpeg_ring_set_patch_ring(ring, 947 (ring->wptr + ring->max_dw * amdgpu_sched_hw_submission)); 948 949 return 0; 950 } 951 952 static int vcn_v1_0_start_dpg_mode(struct amdgpu_device *adev) 953 { 954 struct amdgpu_ring *ring = &adev->vcn.ring_dec; 955 uint32_t rb_bufsz, tmp; 956 uint32_t lmi_swap_cntl; 957 958 /* disable byte swapping */ 959 lmi_swap_cntl = 0; 960 961 vcn_1_0_enable_static_power_gating(adev); 962 963 /* enable dynamic power gating mode */ 964 tmp = RREG32_SOC15(UVD, 0, mmUVD_POWER_STATUS); 965 tmp |= UVD_POWER_STATUS__UVD_PG_MODE_MASK; 966 tmp |= UVD_POWER_STATUS__UVD_PG_EN_MASK; 967 WREG32_SOC15(UVD, 0, mmUVD_POWER_STATUS, tmp); 968 969 /* enable clock gating */ 970 vcn_v1_0_clock_gating_dpg_mode(adev, 0); 971 972 /* enable VCPU clock */ 973 tmp = (0xFF << UVD_VCPU_CNTL__PRB_TIMEOUT_VAL__SHIFT); 974 tmp |= UVD_VCPU_CNTL__CLK_EN_MASK; 975 tmp |= UVD_VCPU_CNTL__MIF_WR_LOW_THRESHOLD_BP_MASK; 976 WREG32_SOC15_DPG_MODE(UVD, 0, mmUVD_VCPU_CNTL, tmp, 0xFFFFFFFF, 0); 977 978 /* disable interupt */ 979 WREG32_SOC15_DPG_MODE(UVD, 0, mmUVD_MASTINT_EN, 980 0, UVD_MASTINT_EN__VCPU_EN_MASK, 0); 981 982 /* initialize VCN memory controller */ 983 WREG32_SOC15_DPG_MODE(UVD, 0, mmUVD_LMI_CTRL, 984 (8 << UVD_LMI_CTRL__WRITE_CLEAN_TIMER__SHIFT) | 985 UVD_LMI_CTRL__WRITE_CLEAN_TIMER_EN_MASK | 986 UVD_LMI_CTRL__DATA_COHERENCY_EN_MASK | 987 UVD_LMI_CTRL__VCPU_DATA_COHERENCY_EN_MASK | 988 UVD_LMI_CTRL__REQ_MODE_MASK | 989 UVD_LMI_CTRL__CRC_RESET_MASK | 990 UVD_LMI_CTRL__MASK_MC_URGENT_MASK | 991 0x00100000L, 0xFFFFFFFF, 0); 992 993 #ifdef __BIG_ENDIAN 994 /* swap (8 in 32) RB and IB */ 995 lmi_swap_cntl = 0xa; 996 #endif 997 WREG32_SOC15_DPG_MODE(UVD, 0, mmUVD_LMI_SWAP_CNTL, lmi_swap_cntl, 0xFFFFFFFF, 0); 998 999 WREG32_SOC15_DPG_MODE(UVD, 0, mmUVD_MPC_CNTL, 1000 0x2 << UVD_MPC_CNTL__REPLACEMENT_MODE__SHIFT, 0xFFFFFFFF, 0); 1001 1002 WREG32_SOC15_DPG_MODE(UVD, 0, mmUVD_MPC_SET_MUXA0, 1003 ((0x1 << UVD_MPC_SET_MUXA0__VARA_1__SHIFT) | 1004 (0x2 << UVD_MPC_SET_MUXA0__VARA_2__SHIFT) | 1005 (0x3 << UVD_MPC_SET_MUXA0__VARA_3__SHIFT) | 1006 (0x4 << UVD_MPC_SET_MUXA0__VARA_4__SHIFT)), 0xFFFFFFFF, 0); 1007 1008 WREG32_SOC15_DPG_MODE(UVD, 0, mmUVD_MPC_SET_MUXB0, 1009 ((0x1 << UVD_MPC_SET_MUXB0__VARB_1__SHIFT) | 1010 (0x2 << UVD_MPC_SET_MUXB0__VARB_2__SHIFT) | 1011 (0x3 << UVD_MPC_SET_MUXB0__VARB_3__SHIFT) | 1012 (0x4 << UVD_MPC_SET_MUXB0__VARB_4__SHIFT)), 0xFFFFFFFF, 0); 1013 1014 WREG32_SOC15_DPG_MODE(UVD, 0, mmUVD_MPC_SET_MUX, 1015 ((0x0 << UVD_MPC_SET_MUX__SET_0__SHIFT) | 1016 (0x1 << UVD_MPC_SET_MUX__SET_1__SHIFT) | 1017 (0x2 << UVD_MPC_SET_MUX__SET_2__SHIFT)), 0xFFFFFFFF, 0); 1018 1019 vcn_v1_0_mc_resume_dpg_mode(adev); 1020 1021 WREG32_SOC15_DPG_MODE(UVD, 0, mmUVD_REG_XX_MASK, 0x10, 0xFFFFFFFF, 0); 1022 WREG32_SOC15_DPG_MODE(UVD, 0, mmUVD_RBC_XX_IB_REG_CHECK, 0x3, 0xFFFFFFFF, 0); 1023 1024 /* boot up the VCPU */ 1025 WREG32_SOC15_DPG_MODE(UVD, 0, mmUVD_SOFT_RESET, 0, 0xFFFFFFFF, 0); 1026 1027 /* enable UMC */ 1028 WREG32_SOC15_DPG_MODE(UVD, 0, mmUVD_LMI_CTRL2, 1029 0x1F << UVD_LMI_CTRL2__RE_OFLD_MIF_WR_REQ_NUM__SHIFT, 1030 0xFFFFFFFF, 0); 1031 1032 /* enable master interrupt */ 1033 WREG32_SOC15_DPG_MODE(UVD, 0, mmUVD_MASTINT_EN, 1034 UVD_MASTINT_EN__VCPU_EN_MASK, UVD_MASTINT_EN__VCPU_EN_MASK, 0); 1035 1036 vcn_v1_0_clock_gating_dpg_mode(adev, 1); 1037 /* setup mmUVD_LMI_CTRL */ 1038 WREG32_SOC15_DPG_MODE(UVD, 0, mmUVD_LMI_CTRL, 1039 (8 << UVD_LMI_CTRL__WRITE_CLEAN_TIMER__SHIFT) | 1040 UVD_LMI_CTRL__WRITE_CLEAN_TIMER_EN_MASK | 1041 UVD_LMI_CTRL__DATA_COHERENCY_EN_MASK | 1042 UVD_LMI_CTRL__VCPU_DATA_COHERENCY_EN_MASK | 1043 UVD_LMI_CTRL__REQ_MODE_MASK | 1044 UVD_LMI_CTRL__CRC_RESET_MASK | 1045 UVD_LMI_CTRL__MASK_MC_URGENT_MASK | 1046 0x00100000L, 0xFFFFFFFF, 1); 1047 1048 tmp = adev->gfx.config.gb_addr_config; 1049 /* setup VCN global tiling registers */ 1050 WREG32_SOC15_DPG_MODE(UVD, 0, mmUVD_JPEG_ADDR_CONFIG, tmp, 0xFFFFFFFF, 1); 1051 WREG32_SOC15_DPG_MODE(UVD, 0, mmUVD_JPEG_UV_ADDR_CONFIG, tmp, 0xFFFFFFFF, 1); 1052 1053 /* enable System Interrupt for JRBC */ 1054 WREG32_SOC15_DPG_MODE(UVD, 0, mmUVD_SYS_INT_EN, 1055 UVD_SYS_INT_EN__UVD_JRBC_EN_MASK, 0xFFFFFFFF, 1); 1056 1057 /* force RBC into idle state */ 1058 rb_bufsz = order_base_2(ring->ring_size); 1059 tmp = REG_SET_FIELD(0, UVD_RBC_RB_CNTL, RB_BUFSZ, rb_bufsz); 1060 tmp = REG_SET_FIELD(tmp, UVD_RBC_RB_CNTL, RB_BLKSZ, 1); 1061 tmp = REG_SET_FIELD(tmp, UVD_RBC_RB_CNTL, RB_NO_FETCH, 1); 1062 tmp = REG_SET_FIELD(tmp, UVD_RBC_RB_CNTL, RB_NO_UPDATE, 1); 1063 tmp = REG_SET_FIELD(tmp, UVD_RBC_RB_CNTL, RB_RPTR_WR_EN, 1); 1064 WREG32_SOC15(UVD, 0, mmUVD_RBC_RB_CNTL, tmp); 1065 1066 /* set the write pointer delay */ 1067 WREG32_SOC15(UVD, 0, mmUVD_RBC_RB_WPTR_CNTL, 0); 1068 1069 /* set the wb address */ 1070 WREG32_SOC15(UVD, 0, mmUVD_RBC_RB_RPTR_ADDR, 1071 (upper_32_bits(ring->gpu_addr) >> 2)); 1072 1073 /* programm the RB_BASE for ring buffer */ 1074 WREG32_SOC15(UVD, 0, mmUVD_LMI_RBC_RB_64BIT_BAR_LOW, 1075 lower_32_bits(ring->gpu_addr)); 1076 WREG32_SOC15(UVD, 0, mmUVD_LMI_RBC_RB_64BIT_BAR_HIGH, 1077 upper_32_bits(ring->gpu_addr)); 1078 1079 /* Initialize the ring buffer's read and write pointers */ 1080 WREG32_SOC15(UVD, 0, mmUVD_RBC_RB_RPTR, 0); 1081 1082 WREG32_SOC15(UVD, 0, mmUVD_SCRATCH2, 0); 1083 1084 ring->wptr = RREG32_SOC15(UVD, 0, mmUVD_RBC_RB_RPTR); 1085 WREG32_SOC15(UVD, 0, mmUVD_RBC_RB_WPTR, 1086 lower_32_bits(ring->wptr)); 1087 1088 WREG32_P(SOC15_REG_OFFSET(UVD, 0, mmUVD_RBC_RB_CNTL), 0, 1089 ~UVD_RBC_RB_CNTL__RB_NO_FETCH_MASK); 1090 1091 /* initialize JPEG wptr */ 1092 ring = &adev->vcn.ring_jpeg; 1093 ring->wptr = RREG32_SOC15(UVD, 0, mmUVD_JRBC_RB_WPTR); 1094 1095 /* copy patch commands to the jpeg ring */ 1096 vcn_v1_0_jpeg_ring_set_patch_ring(ring, 1097 (ring->wptr + ring->max_dw * amdgpu_sched_hw_submission)); 1098 1099 return 0; 1100 } 1101 1102 static int vcn_v1_0_start(struct amdgpu_device *adev) 1103 { 1104 int r; 1105 1106 if (adev->pg_flags & AMD_PG_SUPPORT_VCN_DPG) 1107 r = vcn_v1_0_start_dpg_mode(adev); 1108 else 1109 r = vcn_v1_0_start_spg_mode(adev); 1110 return r; 1111 } 1112 1113 /** 1114 * vcn_v1_0_stop - stop VCN block 1115 * 1116 * @adev: amdgpu_device pointer 1117 * 1118 * stop the VCN block 1119 */ 1120 static int vcn_v1_0_stop_spg_mode(struct amdgpu_device *adev) 1121 { 1122 int ret_code, tmp; 1123 1124 SOC15_WAIT_ON_RREG(UVD, 0, mmUVD_STATUS, UVD_STATUS__IDLE, 0x7, ret_code); 1125 1126 tmp = UVD_LMI_STATUS__VCPU_LMI_WRITE_CLEAN_MASK | 1127 UVD_LMI_STATUS__READ_CLEAN_MASK | 1128 UVD_LMI_STATUS__WRITE_CLEAN_MASK | 1129 UVD_LMI_STATUS__WRITE_CLEAN_RAW_MASK; 1130 SOC15_WAIT_ON_RREG(UVD, 0, mmUVD_LMI_STATUS, tmp, tmp, ret_code); 1131 1132 /* put VCPU into reset */ 1133 WREG32_P(SOC15_REG_OFFSET(UVD, 0, mmUVD_SOFT_RESET), 1134 UVD_SOFT_RESET__VCPU_SOFT_RESET_MASK, 1135 ~UVD_SOFT_RESET__VCPU_SOFT_RESET_MASK); 1136 1137 tmp = UVD_LMI_STATUS__UMC_READ_CLEAN_RAW_MASK | 1138 UVD_LMI_STATUS__UMC_WRITE_CLEAN_RAW_MASK; 1139 SOC15_WAIT_ON_RREG(UVD, 0, mmUVD_LMI_STATUS, tmp, tmp, ret_code); 1140 1141 /* disable VCPU clock */ 1142 WREG32_P(SOC15_REG_OFFSET(UVD, 0, mmUVD_VCPU_CNTL), 0, 1143 ~UVD_VCPU_CNTL__CLK_EN_MASK); 1144 1145 /* reset LMI UMC/LMI */ 1146 WREG32_P(SOC15_REG_OFFSET(UVD, 0, mmUVD_SOFT_RESET), 1147 UVD_SOFT_RESET__LMI_UMC_SOFT_RESET_MASK, 1148 ~UVD_SOFT_RESET__LMI_UMC_SOFT_RESET_MASK); 1149 1150 WREG32_P(SOC15_REG_OFFSET(UVD, 0, mmUVD_SOFT_RESET), 1151 UVD_SOFT_RESET__LMI_SOFT_RESET_MASK, 1152 ~UVD_SOFT_RESET__LMI_SOFT_RESET_MASK); 1153 1154 WREG32_SOC15(UVD, 0, mmUVD_STATUS, 0); 1155 1156 vcn_v1_0_enable_clock_gating(adev); 1157 vcn_1_0_enable_static_power_gating(adev); 1158 return 0; 1159 } 1160 1161 static int vcn_v1_0_stop_dpg_mode(struct amdgpu_device *adev) 1162 { 1163 int ret_code = 0; 1164 uint32_t tmp; 1165 1166 /* Wait for power status to be UVD_POWER_STATUS__UVD_POWER_STATUS_TILES_OFF */ 1167 SOC15_WAIT_ON_RREG(UVD, 0, mmUVD_POWER_STATUS, 1168 UVD_POWER_STATUS__UVD_POWER_STATUS_TILES_OFF, 1169 UVD_POWER_STATUS__UVD_POWER_STATUS_MASK, ret_code); 1170 1171 /* wait for read ptr to be equal to write ptr */ 1172 tmp = RREG32_SOC15(UVD, 0, mmUVD_RB_WPTR); 1173 SOC15_WAIT_ON_RREG(UVD, 0, mmUVD_RB_RPTR, tmp, 0xFFFFFFFF, ret_code); 1174 1175 tmp = RREG32_SOC15(UVD, 0, mmUVD_RB_WPTR2); 1176 SOC15_WAIT_ON_RREG(UVD, 0, mmUVD_RB_RPTR2, tmp, 0xFFFFFFFF, ret_code); 1177 1178 tmp = RREG32_SOC15(UVD, 0, mmUVD_JRBC_RB_WPTR); 1179 SOC15_WAIT_ON_RREG(UVD, 0, mmUVD_JRBC_RB_RPTR, tmp, 0xFFFFFFFF, ret_code); 1180 1181 tmp = RREG32_SOC15(UVD, 0, mmUVD_RBC_RB_WPTR) & 0x7FFFFFFF; 1182 SOC15_WAIT_ON_RREG(UVD, 0, mmUVD_RBC_RB_RPTR, tmp, 0xFFFFFFFF, ret_code); 1183 1184 SOC15_WAIT_ON_RREG(UVD, 0, mmUVD_POWER_STATUS, 1185 UVD_POWER_STATUS__UVD_POWER_STATUS_TILES_OFF, 1186 UVD_POWER_STATUS__UVD_POWER_STATUS_MASK, ret_code); 1187 1188 /* disable dynamic power gating mode */ 1189 WREG32_P(SOC15_REG_OFFSET(UVD, 0, mmUVD_POWER_STATUS), 0, 1190 ~UVD_POWER_STATUS__UVD_PG_MODE_MASK); 1191 1192 return 0; 1193 } 1194 1195 static int vcn_v1_0_stop(struct amdgpu_device *adev) 1196 { 1197 int r; 1198 1199 if (adev->pg_flags & AMD_PG_SUPPORT_VCN_DPG) 1200 r = vcn_v1_0_stop_dpg_mode(adev); 1201 else 1202 r = vcn_v1_0_stop_spg_mode(adev); 1203 1204 return r; 1205 } 1206 1207 static bool vcn_v1_0_is_idle(void *handle) 1208 { 1209 struct amdgpu_device *adev = (struct amdgpu_device *)handle; 1210 1211 return (RREG32_SOC15(VCN, 0, mmUVD_STATUS) == UVD_STATUS__IDLE); 1212 } 1213 1214 static int vcn_v1_0_wait_for_idle(void *handle) 1215 { 1216 struct amdgpu_device *adev = (struct amdgpu_device *)handle; 1217 int ret = 0; 1218 1219 SOC15_WAIT_ON_RREG(VCN, 0, mmUVD_STATUS, UVD_STATUS__IDLE, 1220 UVD_STATUS__IDLE, ret); 1221 1222 return ret; 1223 } 1224 1225 static int vcn_v1_0_set_clockgating_state(void *handle, 1226 enum amd_clockgating_state state) 1227 { 1228 struct amdgpu_device *adev = (struct amdgpu_device *)handle; 1229 bool enable = (state == AMD_CG_STATE_GATE) ? true : false; 1230 1231 if (enable) { 1232 /* wait for STATUS to clear */ 1233 if (vcn_v1_0_is_idle(handle)) 1234 return -EBUSY; 1235 vcn_v1_0_enable_clock_gating(adev); 1236 } else { 1237 /* disable HW gating and enable Sw gating */ 1238 vcn_v1_0_disable_clock_gating(adev); 1239 } 1240 return 0; 1241 } 1242 1243 /** 1244 * vcn_v1_0_dec_ring_get_rptr - get read pointer 1245 * 1246 * @ring: amdgpu_ring pointer 1247 * 1248 * Returns the current hardware read pointer 1249 */ 1250 static uint64_t vcn_v1_0_dec_ring_get_rptr(struct amdgpu_ring *ring) 1251 { 1252 struct amdgpu_device *adev = ring->adev; 1253 1254 return RREG32_SOC15(UVD, 0, mmUVD_RBC_RB_RPTR); 1255 } 1256 1257 /** 1258 * vcn_v1_0_dec_ring_get_wptr - get write pointer 1259 * 1260 * @ring: amdgpu_ring pointer 1261 * 1262 * Returns the current hardware write pointer 1263 */ 1264 static uint64_t vcn_v1_0_dec_ring_get_wptr(struct amdgpu_ring *ring) 1265 { 1266 struct amdgpu_device *adev = ring->adev; 1267 1268 return RREG32_SOC15(UVD, 0, mmUVD_RBC_RB_WPTR); 1269 } 1270 1271 /** 1272 * vcn_v1_0_dec_ring_set_wptr - set write pointer 1273 * 1274 * @ring: amdgpu_ring pointer 1275 * 1276 * Commits the write pointer to the hardware 1277 */ 1278 static void vcn_v1_0_dec_ring_set_wptr(struct amdgpu_ring *ring) 1279 { 1280 struct amdgpu_device *adev = ring->adev; 1281 1282 if (adev->pg_flags & AMD_PG_SUPPORT_VCN_DPG) 1283 WREG32_SOC15(UVD, 0, mmUVD_SCRATCH2, 1284 lower_32_bits(ring->wptr) | 0x80000000); 1285 1286 WREG32_SOC15(UVD, 0, mmUVD_RBC_RB_WPTR, lower_32_bits(ring->wptr)); 1287 } 1288 1289 /** 1290 * vcn_v1_0_dec_ring_insert_start - insert a start command 1291 * 1292 * @ring: amdgpu_ring pointer 1293 * 1294 * Write a start command to the ring. 1295 */ 1296 static void vcn_v1_0_dec_ring_insert_start(struct amdgpu_ring *ring) 1297 { 1298 struct amdgpu_device *adev = ring->adev; 1299 1300 amdgpu_ring_write(ring, 1301 PACKET0(SOC15_REG_OFFSET(UVD, 0, mmUVD_GPCOM_VCPU_DATA0), 0)); 1302 amdgpu_ring_write(ring, 0); 1303 amdgpu_ring_write(ring, 1304 PACKET0(SOC15_REG_OFFSET(UVD, 0, mmUVD_GPCOM_VCPU_CMD), 0)); 1305 amdgpu_ring_write(ring, VCN_DEC_CMD_PACKET_START << 1); 1306 } 1307 1308 /** 1309 * vcn_v1_0_dec_ring_insert_end - insert a end command 1310 * 1311 * @ring: amdgpu_ring pointer 1312 * 1313 * Write a end command to the ring. 1314 */ 1315 static void vcn_v1_0_dec_ring_insert_end(struct amdgpu_ring *ring) 1316 { 1317 struct amdgpu_device *adev = ring->adev; 1318 1319 amdgpu_ring_write(ring, 1320 PACKET0(SOC15_REG_OFFSET(UVD, 0, mmUVD_GPCOM_VCPU_CMD), 0)); 1321 amdgpu_ring_write(ring, VCN_DEC_CMD_PACKET_END << 1); 1322 } 1323 1324 /** 1325 * vcn_v1_0_dec_ring_emit_fence - emit an fence & trap command 1326 * 1327 * @ring: amdgpu_ring pointer 1328 * @fence: fence to emit 1329 * 1330 * Write a fence and a trap command to the ring. 1331 */ 1332 static void vcn_v1_0_dec_ring_emit_fence(struct amdgpu_ring *ring, u64 addr, u64 seq, 1333 unsigned flags) 1334 { 1335 struct amdgpu_device *adev = ring->adev; 1336 1337 WARN_ON(flags & AMDGPU_FENCE_FLAG_64BIT); 1338 1339 amdgpu_ring_write(ring, 1340 PACKET0(SOC15_REG_OFFSET(UVD, 0, mmUVD_CONTEXT_ID), 0)); 1341 amdgpu_ring_write(ring, seq); 1342 amdgpu_ring_write(ring, 1343 PACKET0(SOC15_REG_OFFSET(UVD, 0, mmUVD_GPCOM_VCPU_DATA0), 0)); 1344 amdgpu_ring_write(ring, addr & 0xffffffff); 1345 amdgpu_ring_write(ring, 1346 PACKET0(SOC15_REG_OFFSET(UVD, 0, mmUVD_GPCOM_VCPU_DATA1), 0)); 1347 amdgpu_ring_write(ring, upper_32_bits(addr) & 0xff); 1348 amdgpu_ring_write(ring, 1349 PACKET0(SOC15_REG_OFFSET(UVD, 0, mmUVD_GPCOM_VCPU_CMD), 0)); 1350 amdgpu_ring_write(ring, VCN_DEC_CMD_FENCE << 1); 1351 1352 amdgpu_ring_write(ring, 1353 PACKET0(SOC15_REG_OFFSET(UVD, 0, mmUVD_GPCOM_VCPU_DATA0), 0)); 1354 amdgpu_ring_write(ring, 0); 1355 amdgpu_ring_write(ring, 1356 PACKET0(SOC15_REG_OFFSET(UVD, 0, mmUVD_GPCOM_VCPU_DATA1), 0)); 1357 amdgpu_ring_write(ring, 0); 1358 amdgpu_ring_write(ring, 1359 PACKET0(SOC15_REG_OFFSET(UVD, 0, mmUVD_GPCOM_VCPU_CMD), 0)); 1360 amdgpu_ring_write(ring, VCN_DEC_CMD_TRAP << 1); 1361 } 1362 1363 /** 1364 * vcn_v1_0_dec_ring_emit_ib - execute indirect buffer 1365 * 1366 * @ring: amdgpu_ring pointer 1367 * @ib: indirect buffer to execute 1368 * 1369 * Write ring commands to execute the indirect buffer 1370 */ 1371 static void vcn_v1_0_dec_ring_emit_ib(struct amdgpu_ring *ring, 1372 struct amdgpu_job *job, 1373 struct amdgpu_ib *ib, 1374 bool ctx_switch) 1375 { 1376 struct amdgpu_device *adev = ring->adev; 1377 unsigned vmid = AMDGPU_JOB_GET_VMID(job); 1378 1379 amdgpu_ring_write(ring, 1380 PACKET0(SOC15_REG_OFFSET(UVD, 0, mmUVD_LMI_RBC_IB_VMID), 0)); 1381 amdgpu_ring_write(ring, vmid); 1382 1383 amdgpu_ring_write(ring, 1384 PACKET0(SOC15_REG_OFFSET(UVD, 0, mmUVD_LMI_RBC_IB_64BIT_BAR_LOW), 0)); 1385 amdgpu_ring_write(ring, lower_32_bits(ib->gpu_addr)); 1386 amdgpu_ring_write(ring, 1387 PACKET0(SOC15_REG_OFFSET(UVD, 0, mmUVD_LMI_RBC_IB_64BIT_BAR_HIGH), 0)); 1388 amdgpu_ring_write(ring, upper_32_bits(ib->gpu_addr)); 1389 amdgpu_ring_write(ring, 1390 PACKET0(SOC15_REG_OFFSET(UVD, 0, mmUVD_RBC_IB_SIZE), 0)); 1391 amdgpu_ring_write(ring, ib->length_dw); 1392 } 1393 1394 static void vcn_v1_0_dec_ring_emit_reg_wait(struct amdgpu_ring *ring, 1395 uint32_t reg, uint32_t val, 1396 uint32_t mask) 1397 { 1398 struct amdgpu_device *adev = ring->adev; 1399 1400 amdgpu_ring_write(ring, 1401 PACKET0(SOC15_REG_OFFSET(UVD, 0, mmUVD_GPCOM_VCPU_DATA0), 0)); 1402 amdgpu_ring_write(ring, reg << 2); 1403 amdgpu_ring_write(ring, 1404 PACKET0(SOC15_REG_OFFSET(UVD, 0, mmUVD_GPCOM_VCPU_DATA1), 0)); 1405 amdgpu_ring_write(ring, val); 1406 amdgpu_ring_write(ring, 1407 PACKET0(SOC15_REG_OFFSET(UVD, 0, mmUVD_GP_SCRATCH8), 0)); 1408 amdgpu_ring_write(ring, mask); 1409 amdgpu_ring_write(ring, 1410 PACKET0(SOC15_REG_OFFSET(UVD, 0, mmUVD_GPCOM_VCPU_CMD), 0)); 1411 amdgpu_ring_write(ring, VCN_DEC_CMD_REG_READ_COND_WAIT << 1); 1412 } 1413 1414 static void vcn_v1_0_dec_ring_emit_vm_flush(struct amdgpu_ring *ring, 1415 unsigned vmid, uint64_t pd_addr) 1416 { 1417 struct amdgpu_vmhub *hub = &ring->adev->vmhub[ring->funcs->vmhub]; 1418 uint32_t data0, data1, mask; 1419 1420 pd_addr = amdgpu_gmc_emit_flush_gpu_tlb(ring, vmid, pd_addr); 1421 1422 /* wait for register write */ 1423 data0 = hub->ctx0_ptb_addr_lo32 + vmid * 2; 1424 data1 = lower_32_bits(pd_addr); 1425 mask = 0xffffffff; 1426 vcn_v1_0_dec_ring_emit_reg_wait(ring, data0, data1, mask); 1427 } 1428 1429 static void vcn_v1_0_dec_ring_emit_wreg(struct amdgpu_ring *ring, 1430 uint32_t reg, uint32_t val) 1431 { 1432 struct amdgpu_device *adev = ring->adev; 1433 1434 amdgpu_ring_write(ring, 1435 PACKET0(SOC15_REG_OFFSET(UVD, 0, mmUVD_GPCOM_VCPU_DATA0), 0)); 1436 amdgpu_ring_write(ring, reg << 2); 1437 amdgpu_ring_write(ring, 1438 PACKET0(SOC15_REG_OFFSET(UVD, 0, mmUVD_GPCOM_VCPU_DATA1), 0)); 1439 amdgpu_ring_write(ring, val); 1440 amdgpu_ring_write(ring, 1441 PACKET0(SOC15_REG_OFFSET(UVD, 0, mmUVD_GPCOM_VCPU_CMD), 0)); 1442 amdgpu_ring_write(ring, VCN_DEC_CMD_WRITE_REG << 1); 1443 } 1444 1445 /** 1446 * vcn_v1_0_enc_ring_get_rptr - get enc read pointer 1447 * 1448 * @ring: amdgpu_ring pointer 1449 * 1450 * Returns the current hardware enc read pointer 1451 */ 1452 static uint64_t vcn_v1_0_enc_ring_get_rptr(struct amdgpu_ring *ring) 1453 { 1454 struct amdgpu_device *adev = ring->adev; 1455 1456 if (ring == &adev->vcn.ring_enc[0]) 1457 return RREG32_SOC15(UVD, 0, mmUVD_RB_RPTR); 1458 else 1459 return RREG32_SOC15(UVD, 0, mmUVD_RB_RPTR2); 1460 } 1461 1462 /** 1463 * vcn_v1_0_enc_ring_get_wptr - get enc write pointer 1464 * 1465 * @ring: amdgpu_ring pointer 1466 * 1467 * Returns the current hardware enc write pointer 1468 */ 1469 static uint64_t vcn_v1_0_enc_ring_get_wptr(struct amdgpu_ring *ring) 1470 { 1471 struct amdgpu_device *adev = ring->adev; 1472 1473 if (ring == &adev->vcn.ring_enc[0]) 1474 return RREG32_SOC15(UVD, 0, mmUVD_RB_WPTR); 1475 else 1476 return RREG32_SOC15(UVD, 0, mmUVD_RB_WPTR2); 1477 } 1478 1479 /** 1480 * vcn_v1_0_enc_ring_set_wptr - set enc write pointer 1481 * 1482 * @ring: amdgpu_ring pointer 1483 * 1484 * Commits the enc write pointer to the hardware 1485 */ 1486 static void vcn_v1_0_enc_ring_set_wptr(struct amdgpu_ring *ring) 1487 { 1488 struct amdgpu_device *adev = ring->adev; 1489 1490 if (ring == &adev->vcn.ring_enc[0]) 1491 WREG32_SOC15(UVD, 0, mmUVD_RB_WPTR, 1492 lower_32_bits(ring->wptr)); 1493 else 1494 WREG32_SOC15(UVD, 0, mmUVD_RB_WPTR2, 1495 lower_32_bits(ring->wptr)); 1496 } 1497 1498 /** 1499 * vcn_v1_0_enc_ring_emit_fence - emit an enc fence & trap command 1500 * 1501 * @ring: amdgpu_ring pointer 1502 * @fence: fence to emit 1503 * 1504 * Write enc a fence and a trap command to the ring. 1505 */ 1506 static void vcn_v1_0_enc_ring_emit_fence(struct amdgpu_ring *ring, u64 addr, 1507 u64 seq, unsigned flags) 1508 { 1509 WARN_ON(flags & AMDGPU_FENCE_FLAG_64BIT); 1510 1511 amdgpu_ring_write(ring, VCN_ENC_CMD_FENCE); 1512 amdgpu_ring_write(ring, addr); 1513 amdgpu_ring_write(ring, upper_32_bits(addr)); 1514 amdgpu_ring_write(ring, seq); 1515 amdgpu_ring_write(ring, VCN_ENC_CMD_TRAP); 1516 } 1517 1518 static void vcn_v1_0_enc_ring_insert_end(struct amdgpu_ring *ring) 1519 { 1520 amdgpu_ring_write(ring, VCN_ENC_CMD_END); 1521 } 1522 1523 /** 1524 * vcn_v1_0_enc_ring_emit_ib - enc execute indirect buffer 1525 * 1526 * @ring: amdgpu_ring pointer 1527 * @ib: indirect buffer to execute 1528 * 1529 * Write enc ring commands to execute the indirect buffer 1530 */ 1531 static void vcn_v1_0_enc_ring_emit_ib(struct amdgpu_ring *ring, 1532 struct amdgpu_job *job, 1533 struct amdgpu_ib *ib, 1534 bool ctx_switch) 1535 { 1536 unsigned vmid = AMDGPU_JOB_GET_VMID(job); 1537 1538 amdgpu_ring_write(ring, VCN_ENC_CMD_IB); 1539 amdgpu_ring_write(ring, vmid); 1540 amdgpu_ring_write(ring, lower_32_bits(ib->gpu_addr)); 1541 amdgpu_ring_write(ring, upper_32_bits(ib->gpu_addr)); 1542 amdgpu_ring_write(ring, ib->length_dw); 1543 } 1544 1545 static void vcn_v1_0_enc_ring_emit_reg_wait(struct amdgpu_ring *ring, 1546 uint32_t reg, uint32_t val, 1547 uint32_t mask) 1548 { 1549 amdgpu_ring_write(ring, VCN_ENC_CMD_REG_WAIT); 1550 amdgpu_ring_write(ring, reg << 2); 1551 amdgpu_ring_write(ring, mask); 1552 amdgpu_ring_write(ring, val); 1553 } 1554 1555 static void vcn_v1_0_enc_ring_emit_vm_flush(struct amdgpu_ring *ring, 1556 unsigned int vmid, uint64_t pd_addr) 1557 { 1558 struct amdgpu_vmhub *hub = &ring->adev->vmhub[ring->funcs->vmhub]; 1559 1560 pd_addr = amdgpu_gmc_emit_flush_gpu_tlb(ring, vmid, pd_addr); 1561 1562 /* wait for reg writes */ 1563 vcn_v1_0_enc_ring_emit_reg_wait(ring, hub->ctx0_ptb_addr_lo32 + vmid * 2, 1564 lower_32_bits(pd_addr), 0xffffffff); 1565 } 1566 1567 static void vcn_v1_0_enc_ring_emit_wreg(struct amdgpu_ring *ring, 1568 uint32_t reg, uint32_t val) 1569 { 1570 amdgpu_ring_write(ring, VCN_ENC_CMD_REG_WRITE); 1571 amdgpu_ring_write(ring, reg << 2); 1572 amdgpu_ring_write(ring, val); 1573 } 1574 1575 1576 /** 1577 * vcn_v1_0_jpeg_ring_get_rptr - get read pointer 1578 * 1579 * @ring: amdgpu_ring pointer 1580 * 1581 * Returns the current hardware read pointer 1582 */ 1583 static uint64_t vcn_v1_0_jpeg_ring_get_rptr(struct amdgpu_ring *ring) 1584 { 1585 struct amdgpu_device *adev = ring->adev; 1586 1587 return RREG32_SOC15(UVD, 0, mmUVD_JRBC_RB_RPTR); 1588 } 1589 1590 /** 1591 * vcn_v1_0_jpeg_ring_get_wptr - get write pointer 1592 * 1593 * @ring: amdgpu_ring pointer 1594 * 1595 * Returns the current hardware write pointer 1596 */ 1597 static uint64_t vcn_v1_0_jpeg_ring_get_wptr(struct amdgpu_ring *ring) 1598 { 1599 struct amdgpu_device *adev = ring->adev; 1600 1601 return RREG32_SOC15(UVD, 0, mmUVD_JRBC_RB_WPTR); 1602 } 1603 1604 /** 1605 * vcn_v1_0_jpeg_ring_set_wptr - set write pointer 1606 * 1607 * @ring: amdgpu_ring pointer 1608 * 1609 * Commits the write pointer to the hardware 1610 */ 1611 static void vcn_v1_0_jpeg_ring_set_wptr(struct amdgpu_ring *ring) 1612 { 1613 struct amdgpu_device *adev = ring->adev; 1614 1615 WREG32_SOC15(UVD, 0, mmUVD_JRBC_RB_WPTR, lower_32_bits(ring->wptr)); 1616 } 1617 1618 /** 1619 * vcn_v1_0_jpeg_ring_insert_start - insert a start command 1620 * 1621 * @ring: amdgpu_ring pointer 1622 * 1623 * Write a start command to the ring. 1624 */ 1625 static void vcn_v1_0_jpeg_ring_insert_start(struct amdgpu_ring *ring) 1626 { 1627 struct amdgpu_device *adev = ring->adev; 1628 1629 amdgpu_ring_write(ring, 1630 PACKETJ(SOC15_REG_OFFSET(UVD, 0, mmUVD_JRBC_EXTERNAL_REG_BASE), 0, 0, PACKETJ_TYPE0)); 1631 amdgpu_ring_write(ring, 0x68e04); 1632 1633 amdgpu_ring_write(ring, PACKETJ(0, 0, 0, PACKETJ_TYPE0)); 1634 amdgpu_ring_write(ring, 0x80010000); 1635 } 1636 1637 /** 1638 * vcn_v1_0_jpeg_ring_insert_end - insert a end command 1639 * 1640 * @ring: amdgpu_ring pointer 1641 * 1642 * Write a end command to the ring. 1643 */ 1644 static void vcn_v1_0_jpeg_ring_insert_end(struct amdgpu_ring *ring) 1645 { 1646 struct amdgpu_device *adev = ring->adev; 1647 1648 amdgpu_ring_write(ring, 1649 PACKETJ(SOC15_REG_OFFSET(UVD, 0, mmUVD_JRBC_EXTERNAL_REG_BASE), 0, 0, PACKETJ_TYPE0)); 1650 amdgpu_ring_write(ring, 0x68e04); 1651 1652 amdgpu_ring_write(ring, PACKETJ(0, 0, 0, PACKETJ_TYPE0)); 1653 amdgpu_ring_write(ring, 0x00010000); 1654 } 1655 1656 /** 1657 * vcn_v1_0_jpeg_ring_emit_fence - emit an fence & trap command 1658 * 1659 * @ring: amdgpu_ring pointer 1660 * @fence: fence to emit 1661 * 1662 * Write a fence and a trap command to the ring. 1663 */ 1664 static void vcn_v1_0_jpeg_ring_emit_fence(struct amdgpu_ring *ring, u64 addr, u64 seq, 1665 unsigned flags) 1666 { 1667 struct amdgpu_device *adev = ring->adev; 1668 1669 WARN_ON(flags & AMDGPU_FENCE_FLAG_64BIT); 1670 1671 amdgpu_ring_write(ring, 1672 PACKETJ(SOC15_REG_OFFSET(UVD, 0, mmUVD_JPEG_GPCOM_DATA0), 0, 0, PACKETJ_TYPE0)); 1673 amdgpu_ring_write(ring, seq); 1674 1675 amdgpu_ring_write(ring, 1676 PACKETJ(SOC15_REG_OFFSET(UVD, 0, mmUVD_JPEG_GPCOM_DATA1), 0, 0, PACKETJ_TYPE0)); 1677 amdgpu_ring_write(ring, seq); 1678 1679 amdgpu_ring_write(ring, 1680 PACKETJ(SOC15_REG_OFFSET(UVD, 0, mmUVD_LMI_JRBC_RB_MEM_WR_64BIT_BAR_LOW), 0, 0, PACKETJ_TYPE0)); 1681 amdgpu_ring_write(ring, lower_32_bits(addr)); 1682 1683 amdgpu_ring_write(ring, 1684 PACKETJ(SOC15_REG_OFFSET(UVD, 0, mmUVD_LMI_JRBC_RB_MEM_WR_64BIT_BAR_HIGH), 0, 0, PACKETJ_TYPE0)); 1685 amdgpu_ring_write(ring, upper_32_bits(addr)); 1686 1687 amdgpu_ring_write(ring, 1688 PACKETJ(SOC15_REG_OFFSET(UVD, 0, mmUVD_JPEG_GPCOM_CMD), 0, 0, PACKETJ_TYPE0)); 1689 amdgpu_ring_write(ring, 0x8); 1690 1691 amdgpu_ring_write(ring, 1692 PACKETJ(SOC15_REG_OFFSET(UVD, 0, mmUVD_JPEG_GPCOM_CMD), 0, PACKETJ_CONDITION_CHECK0, PACKETJ_TYPE4)); 1693 amdgpu_ring_write(ring, 0); 1694 1695 amdgpu_ring_write(ring, 1696 PACKETJ(SOC15_REG_OFFSET(UVD, 0, mmUVD_JRBC_RB_COND_RD_TIMER), 0, 0, PACKETJ_TYPE0)); 1697 amdgpu_ring_write(ring, 0x01400200); 1698 1699 amdgpu_ring_write(ring, 1700 PACKETJ(SOC15_REG_OFFSET(UVD, 0, mmUVD_JRBC_RB_REF_DATA), 0, 0, PACKETJ_TYPE0)); 1701 amdgpu_ring_write(ring, seq); 1702 1703 amdgpu_ring_write(ring, 1704 PACKETJ(SOC15_REG_OFFSET(UVD, 0, mmUVD_LMI_JRBC_RB_MEM_RD_64BIT_BAR_LOW), 0, 0, PACKETJ_TYPE0)); 1705 amdgpu_ring_write(ring, lower_32_bits(addr)); 1706 1707 amdgpu_ring_write(ring, 1708 PACKETJ(SOC15_REG_OFFSET(UVD, 0, mmUVD_LMI_JRBC_RB_MEM_RD_64BIT_BAR_HIGH), 0, 0, PACKETJ_TYPE0)); 1709 amdgpu_ring_write(ring, upper_32_bits(addr)); 1710 1711 amdgpu_ring_write(ring, 1712 PACKETJ(0, 0, PACKETJ_CONDITION_CHECK3, PACKETJ_TYPE2)); 1713 amdgpu_ring_write(ring, 0xffffffff); 1714 1715 amdgpu_ring_write(ring, 1716 PACKETJ(SOC15_REG_OFFSET(UVD, 0, mmUVD_JRBC_EXTERNAL_REG_BASE), 0, 0, PACKETJ_TYPE0)); 1717 amdgpu_ring_write(ring, 0x3fbc); 1718 1719 amdgpu_ring_write(ring, 1720 PACKETJ(0, 0, 0, PACKETJ_TYPE0)); 1721 amdgpu_ring_write(ring, 0x1); 1722 1723 /* emit trap */ 1724 amdgpu_ring_write(ring, PACKETJ(0, 0, 0, PACKETJ_TYPE7)); 1725 amdgpu_ring_write(ring, 0); 1726 } 1727 1728 /** 1729 * vcn_v1_0_jpeg_ring_emit_ib - execute indirect buffer 1730 * 1731 * @ring: amdgpu_ring pointer 1732 * @ib: indirect buffer to execute 1733 * 1734 * Write ring commands to execute the indirect buffer. 1735 */ 1736 static void vcn_v1_0_jpeg_ring_emit_ib(struct amdgpu_ring *ring, 1737 struct amdgpu_job *job, 1738 struct amdgpu_ib *ib, 1739 bool ctx_switch) 1740 { 1741 struct amdgpu_device *adev = ring->adev; 1742 unsigned vmid = AMDGPU_JOB_GET_VMID(job); 1743 1744 amdgpu_ring_write(ring, 1745 PACKETJ(SOC15_REG_OFFSET(UVD, 0, mmUVD_LMI_JRBC_IB_VMID), 0, 0, PACKETJ_TYPE0)); 1746 amdgpu_ring_write(ring, (vmid | (vmid << 4))); 1747 1748 amdgpu_ring_write(ring, 1749 PACKETJ(SOC15_REG_OFFSET(UVD, 0, mmUVD_LMI_JPEG_VMID), 0, 0, PACKETJ_TYPE0)); 1750 amdgpu_ring_write(ring, (vmid | (vmid << 4))); 1751 1752 amdgpu_ring_write(ring, 1753 PACKETJ(SOC15_REG_OFFSET(UVD, 0, mmUVD_LMI_JRBC_IB_64BIT_BAR_LOW), 0, 0, PACKETJ_TYPE0)); 1754 amdgpu_ring_write(ring, lower_32_bits(ib->gpu_addr)); 1755 1756 amdgpu_ring_write(ring, 1757 PACKETJ(SOC15_REG_OFFSET(UVD, 0, mmUVD_LMI_JRBC_IB_64BIT_BAR_HIGH), 0, 0, PACKETJ_TYPE0)); 1758 amdgpu_ring_write(ring, upper_32_bits(ib->gpu_addr)); 1759 1760 amdgpu_ring_write(ring, 1761 PACKETJ(SOC15_REG_OFFSET(UVD, 0, mmUVD_JRBC_IB_SIZE), 0, 0, PACKETJ_TYPE0)); 1762 amdgpu_ring_write(ring, ib->length_dw); 1763 1764 amdgpu_ring_write(ring, 1765 PACKETJ(SOC15_REG_OFFSET(UVD, 0, mmUVD_LMI_JRBC_RB_MEM_RD_64BIT_BAR_LOW), 0, 0, PACKETJ_TYPE0)); 1766 amdgpu_ring_write(ring, lower_32_bits(ring->gpu_addr)); 1767 1768 amdgpu_ring_write(ring, 1769 PACKETJ(SOC15_REG_OFFSET(UVD, 0, mmUVD_LMI_JRBC_RB_MEM_RD_64BIT_BAR_HIGH), 0, 0, PACKETJ_TYPE0)); 1770 amdgpu_ring_write(ring, upper_32_bits(ring->gpu_addr)); 1771 1772 amdgpu_ring_write(ring, 1773 PACKETJ(0, 0, PACKETJ_CONDITION_CHECK0, PACKETJ_TYPE2)); 1774 amdgpu_ring_write(ring, 0); 1775 1776 amdgpu_ring_write(ring, 1777 PACKETJ(SOC15_REG_OFFSET(UVD, 0, mmUVD_JRBC_RB_COND_RD_TIMER), 0, 0, PACKETJ_TYPE0)); 1778 amdgpu_ring_write(ring, 0x01400200); 1779 1780 amdgpu_ring_write(ring, 1781 PACKETJ(SOC15_REG_OFFSET(UVD, 0, mmUVD_JRBC_RB_REF_DATA), 0, 0, PACKETJ_TYPE0)); 1782 amdgpu_ring_write(ring, 0x2); 1783 1784 amdgpu_ring_write(ring, 1785 PACKETJ(SOC15_REG_OFFSET(UVD, 0, mmUVD_JRBC_STATUS), 0, PACKETJ_CONDITION_CHECK3, PACKETJ_TYPE3)); 1786 amdgpu_ring_write(ring, 0x2); 1787 } 1788 1789 static void vcn_v1_0_jpeg_ring_emit_reg_wait(struct amdgpu_ring *ring, 1790 uint32_t reg, uint32_t val, 1791 uint32_t mask) 1792 { 1793 struct amdgpu_device *adev = ring->adev; 1794 uint32_t reg_offset = (reg << 2); 1795 1796 amdgpu_ring_write(ring, 1797 PACKETJ(SOC15_REG_OFFSET(UVD, 0, mmUVD_JRBC_RB_COND_RD_TIMER), 0, 0, PACKETJ_TYPE0)); 1798 amdgpu_ring_write(ring, 0x01400200); 1799 1800 amdgpu_ring_write(ring, 1801 PACKETJ(SOC15_REG_OFFSET(UVD, 0, mmUVD_JRBC_RB_REF_DATA), 0, 0, PACKETJ_TYPE0)); 1802 amdgpu_ring_write(ring, val); 1803 1804 amdgpu_ring_write(ring, 1805 PACKETJ(SOC15_REG_OFFSET(UVD, 0, mmUVD_JRBC_EXTERNAL_REG_BASE), 0, 0, PACKETJ_TYPE0)); 1806 if (((reg_offset >= 0x1f800) && (reg_offset <= 0x21fff)) || 1807 ((reg_offset >= 0x1e000) && (reg_offset <= 0x1e1ff))) { 1808 amdgpu_ring_write(ring, 0); 1809 amdgpu_ring_write(ring, 1810 PACKETJ((reg_offset >> 2), 0, 0, PACKETJ_TYPE3)); 1811 } else { 1812 amdgpu_ring_write(ring, reg_offset); 1813 amdgpu_ring_write(ring, 1814 PACKETJ(0, 0, 0, PACKETJ_TYPE3)); 1815 } 1816 amdgpu_ring_write(ring, mask); 1817 } 1818 1819 static void vcn_v1_0_jpeg_ring_emit_vm_flush(struct amdgpu_ring *ring, 1820 unsigned vmid, uint64_t pd_addr) 1821 { 1822 struct amdgpu_vmhub *hub = &ring->adev->vmhub[ring->funcs->vmhub]; 1823 uint32_t data0, data1, mask; 1824 1825 pd_addr = amdgpu_gmc_emit_flush_gpu_tlb(ring, vmid, pd_addr); 1826 1827 /* wait for register write */ 1828 data0 = hub->ctx0_ptb_addr_lo32 + vmid * 2; 1829 data1 = lower_32_bits(pd_addr); 1830 mask = 0xffffffff; 1831 vcn_v1_0_jpeg_ring_emit_reg_wait(ring, data0, data1, mask); 1832 } 1833 1834 static void vcn_v1_0_jpeg_ring_emit_wreg(struct amdgpu_ring *ring, 1835 uint32_t reg, uint32_t val) 1836 { 1837 struct amdgpu_device *adev = ring->adev; 1838 uint32_t reg_offset = (reg << 2); 1839 1840 amdgpu_ring_write(ring, 1841 PACKETJ(SOC15_REG_OFFSET(UVD, 0, mmUVD_JRBC_EXTERNAL_REG_BASE), 0, 0, PACKETJ_TYPE0)); 1842 if (((reg_offset >= 0x1f800) && (reg_offset <= 0x21fff)) || 1843 ((reg_offset >= 0x1e000) && (reg_offset <= 0x1e1ff))) { 1844 amdgpu_ring_write(ring, 0); 1845 amdgpu_ring_write(ring, 1846 PACKETJ((reg_offset >> 2), 0, 0, PACKETJ_TYPE0)); 1847 } else { 1848 amdgpu_ring_write(ring, reg_offset); 1849 amdgpu_ring_write(ring, 1850 PACKETJ(0, 0, 0, PACKETJ_TYPE0)); 1851 } 1852 amdgpu_ring_write(ring, val); 1853 } 1854 1855 static void vcn_v1_0_jpeg_ring_nop(struct amdgpu_ring *ring, uint32_t count) 1856 { 1857 int i; 1858 1859 WARN_ON(ring->wptr % 2 || count % 2); 1860 1861 for (i = 0; i < count / 2; i++) { 1862 amdgpu_ring_write(ring, PACKETJ(0, 0, 0, PACKETJ_TYPE6)); 1863 amdgpu_ring_write(ring, 0); 1864 } 1865 } 1866 1867 static void vcn_v1_0_jpeg_ring_patch_wreg(struct amdgpu_ring *ring, uint32_t *ptr, uint32_t reg_offset, uint32_t val) 1868 { 1869 struct amdgpu_device *adev = ring->adev; 1870 ring->ring[(*ptr)++] = PACKETJ(SOC15_REG_OFFSET(UVD, 0, mmUVD_JRBC_EXTERNAL_REG_BASE), 0, 0, PACKETJ_TYPE0); 1871 if (((reg_offset >= 0x1f800) && (reg_offset <= 0x21fff)) || 1872 ((reg_offset >= 0x1e000) && (reg_offset <= 0x1e1ff))) { 1873 ring->ring[(*ptr)++] = 0; 1874 ring->ring[(*ptr)++] = PACKETJ((reg_offset >> 2), 0, 0, PACKETJ_TYPE0); 1875 } else { 1876 ring->ring[(*ptr)++] = reg_offset; 1877 ring->ring[(*ptr)++] = PACKETJ(0, 0, 0, PACKETJ_TYPE0); 1878 } 1879 ring->ring[(*ptr)++] = val; 1880 } 1881 1882 static void vcn_v1_0_jpeg_ring_set_patch_ring(struct amdgpu_ring *ring, uint32_t ptr) 1883 { 1884 struct amdgpu_device *adev = ring->adev; 1885 1886 uint32_t reg, reg_offset, val, mask, i; 1887 1888 // 1st: program mmUVD_LMI_JRBC_RB_MEM_RD_64BIT_BAR_LOW 1889 reg = SOC15_REG_OFFSET(UVD, 0, mmUVD_LMI_JRBC_RB_MEM_RD_64BIT_BAR_LOW); 1890 reg_offset = (reg << 2); 1891 val = lower_32_bits(ring->gpu_addr); 1892 vcn_v1_0_jpeg_ring_patch_wreg(ring, &ptr, reg_offset, val); 1893 1894 // 2nd: program mmUVD_LMI_JRBC_RB_MEM_RD_64BIT_BAR_HIGH 1895 reg = SOC15_REG_OFFSET(UVD, 0, mmUVD_LMI_JRBC_RB_MEM_RD_64BIT_BAR_HIGH); 1896 reg_offset = (reg << 2); 1897 val = upper_32_bits(ring->gpu_addr); 1898 vcn_v1_0_jpeg_ring_patch_wreg(ring, &ptr, reg_offset, val); 1899 1900 // 3rd to 5th: issue MEM_READ commands 1901 for (i = 0; i <= 2; i++) { 1902 ring->ring[ptr++] = PACKETJ(0, 0, 0, PACKETJ_TYPE2); 1903 ring->ring[ptr++] = 0; 1904 } 1905 1906 // 6th: program mmUVD_JRBC_RB_CNTL register to enable NO_FETCH and RPTR write ability 1907 reg = SOC15_REG_OFFSET(UVD, 0, mmUVD_JRBC_RB_CNTL); 1908 reg_offset = (reg << 2); 1909 val = 0x13; 1910 vcn_v1_0_jpeg_ring_patch_wreg(ring, &ptr, reg_offset, val); 1911 1912 // 7th: program mmUVD_JRBC_RB_REF_DATA 1913 reg = SOC15_REG_OFFSET(UVD, 0, mmUVD_JRBC_RB_REF_DATA); 1914 reg_offset = (reg << 2); 1915 val = 0x1; 1916 vcn_v1_0_jpeg_ring_patch_wreg(ring, &ptr, reg_offset, val); 1917 1918 // 8th: issue conditional register read mmUVD_JRBC_RB_CNTL 1919 reg = SOC15_REG_OFFSET(UVD, 0, mmUVD_JRBC_RB_CNTL); 1920 reg_offset = (reg << 2); 1921 val = 0x1; 1922 mask = 0x1; 1923 1924 ring->ring[ptr++] = PACKETJ(SOC15_REG_OFFSET(UVD, 0, mmUVD_JRBC_RB_COND_RD_TIMER), 0, 0, PACKETJ_TYPE0); 1925 ring->ring[ptr++] = 0x01400200; 1926 ring->ring[ptr++] = PACKETJ(SOC15_REG_OFFSET(UVD, 0, mmUVD_JRBC_RB_REF_DATA), 0, 0, PACKETJ_TYPE0); 1927 ring->ring[ptr++] = val; 1928 ring->ring[ptr++] = PACKETJ(SOC15_REG_OFFSET(UVD, 0, mmUVD_JRBC_EXTERNAL_REG_BASE), 0, 0, PACKETJ_TYPE0); 1929 if (((reg_offset >= 0x1f800) && (reg_offset <= 0x21fff)) || 1930 ((reg_offset >= 0x1e000) && (reg_offset <= 0x1e1ff))) { 1931 ring->ring[ptr++] = 0; 1932 ring->ring[ptr++] = PACKETJ((reg_offset >> 2), 0, 0, PACKETJ_TYPE3); 1933 } else { 1934 ring->ring[ptr++] = reg_offset; 1935 ring->ring[ptr++] = PACKETJ(0, 0, 0, PACKETJ_TYPE3); 1936 } 1937 ring->ring[ptr++] = mask; 1938 1939 //9th to 21st: insert no-op 1940 for (i = 0; i <= 12; i++) { 1941 ring->ring[ptr++] = PACKETJ(0, 0, 0, PACKETJ_TYPE6); 1942 ring->ring[ptr++] = 0; 1943 } 1944 1945 //22nd: reset mmUVD_JRBC_RB_RPTR 1946 reg = SOC15_REG_OFFSET(UVD, 0, mmUVD_JRBC_RB_RPTR); 1947 reg_offset = (reg << 2); 1948 val = 0; 1949 vcn_v1_0_jpeg_ring_patch_wreg(ring, &ptr, reg_offset, val); 1950 1951 //23rd: program mmUVD_JRBC_RB_CNTL to disable no_fetch 1952 reg = SOC15_REG_OFFSET(UVD, 0, mmUVD_JRBC_RB_CNTL); 1953 reg_offset = (reg << 2); 1954 val = 0x12; 1955 vcn_v1_0_jpeg_ring_patch_wreg(ring, &ptr, reg_offset, val); 1956 } 1957 1958 static int vcn_v1_0_set_interrupt_state(struct amdgpu_device *adev, 1959 struct amdgpu_irq_src *source, 1960 unsigned type, 1961 enum amdgpu_interrupt_state state) 1962 { 1963 return 0; 1964 } 1965 1966 static int vcn_v1_0_process_interrupt(struct amdgpu_device *adev, 1967 struct amdgpu_irq_src *source, 1968 struct amdgpu_iv_entry *entry) 1969 { 1970 DRM_DEBUG("IH: VCN TRAP\n"); 1971 1972 switch (entry->src_id) { 1973 case 124: 1974 amdgpu_fence_process(&adev->vcn.ring_dec); 1975 break; 1976 case 119: 1977 amdgpu_fence_process(&adev->vcn.ring_enc[0]); 1978 break; 1979 case 120: 1980 amdgpu_fence_process(&adev->vcn.ring_enc[1]); 1981 break; 1982 case 126: 1983 amdgpu_fence_process(&adev->vcn.ring_jpeg); 1984 break; 1985 default: 1986 DRM_ERROR("Unhandled interrupt: %d %d\n", 1987 entry->src_id, entry->src_data[0]); 1988 break; 1989 } 1990 1991 return 0; 1992 } 1993 1994 static void vcn_v1_0_dec_ring_insert_nop(struct amdgpu_ring *ring, uint32_t count) 1995 { 1996 struct amdgpu_device *adev = ring->adev; 1997 int i; 1998 1999 WARN_ON(ring->wptr % 2 || count % 2); 2000 2001 for (i = 0; i < count / 2; i++) { 2002 amdgpu_ring_write(ring, PACKET0(SOC15_REG_OFFSET(UVD, 0, mmUVD_NO_OP), 0)); 2003 amdgpu_ring_write(ring, 0); 2004 } 2005 } 2006 2007 static int vcn_v1_0_set_powergating_state(void *handle, 2008 enum amd_powergating_state state) 2009 { 2010 /* This doesn't actually powergate the VCN block. 2011 * That's done in the dpm code via the SMC. This 2012 * just re-inits the block as necessary. The actual 2013 * gating still happens in the dpm code. We should 2014 * revisit this when there is a cleaner line between 2015 * the smc and the hw blocks 2016 */ 2017 int ret; 2018 struct amdgpu_device *adev = (struct amdgpu_device *)handle; 2019 2020 if(state == adev->vcn.cur_state) 2021 return 0; 2022 2023 if (state == AMD_PG_STATE_GATE) 2024 ret = vcn_v1_0_stop(adev); 2025 else 2026 ret = vcn_v1_0_start(adev); 2027 2028 if(!ret) 2029 adev->vcn.cur_state = state; 2030 return ret; 2031 } 2032 2033 static const struct amd_ip_funcs vcn_v1_0_ip_funcs = { 2034 .name = "vcn_v1_0", 2035 .early_init = vcn_v1_0_early_init, 2036 .late_init = NULL, 2037 .sw_init = vcn_v1_0_sw_init, 2038 .sw_fini = vcn_v1_0_sw_fini, 2039 .hw_init = vcn_v1_0_hw_init, 2040 .hw_fini = vcn_v1_0_hw_fini, 2041 .suspend = vcn_v1_0_suspend, 2042 .resume = vcn_v1_0_resume, 2043 .is_idle = vcn_v1_0_is_idle, 2044 .wait_for_idle = vcn_v1_0_wait_for_idle, 2045 .check_soft_reset = NULL /* vcn_v1_0_check_soft_reset */, 2046 .pre_soft_reset = NULL /* vcn_v1_0_pre_soft_reset */, 2047 .soft_reset = NULL /* vcn_v1_0_soft_reset */, 2048 .post_soft_reset = NULL /* vcn_v1_0_post_soft_reset */, 2049 .set_clockgating_state = vcn_v1_0_set_clockgating_state, 2050 .set_powergating_state = vcn_v1_0_set_powergating_state, 2051 }; 2052 2053 static const struct amdgpu_ring_funcs vcn_v1_0_dec_ring_vm_funcs = { 2054 .type = AMDGPU_RING_TYPE_VCN_DEC, 2055 .align_mask = 0xf, 2056 .support_64bit_ptrs = false, 2057 .vmhub = AMDGPU_MMHUB, 2058 .get_rptr = vcn_v1_0_dec_ring_get_rptr, 2059 .get_wptr = vcn_v1_0_dec_ring_get_wptr, 2060 .set_wptr = vcn_v1_0_dec_ring_set_wptr, 2061 .emit_frame_size = 2062 6 + 6 + /* hdp invalidate / flush */ 2063 SOC15_FLUSH_GPU_TLB_NUM_WREG * 6 + 2064 SOC15_FLUSH_GPU_TLB_NUM_REG_WAIT * 8 + 2065 8 + /* vcn_v1_0_dec_ring_emit_vm_flush */ 2066 14 + 14 + /* vcn_v1_0_dec_ring_emit_fence x2 vm fence */ 2067 6, 2068 .emit_ib_size = 8, /* vcn_v1_0_dec_ring_emit_ib */ 2069 .emit_ib = vcn_v1_0_dec_ring_emit_ib, 2070 .emit_fence = vcn_v1_0_dec_ring_emit_fence, 2071 .emit_vm_flush = vcn_v1_0_dec_ring_emit_vm_flush, 2072 .test_ring = amdgpu_vcn_dec_ring_test_ring, 2073 .test_ib = amdgpu_vcn_dec_ring_test_ib, 2074 .insert_nop = vcn_v1_0_dec_ring_insert_nop, 2075 .insert_start = vcn_v1_0_dec_ring_insert_start, 2076 .insert_end = vcn_v1_0_dec_ring_insert_end, 2077 .pad_ib = amdgpu_ring_generic_pad_ib, 2078 .begin_use = amdgpu_vcn_ring_begin_use, 2079 .end_use = amdgpu_vcn_ring_end_use, 2080 .emit_wreg = vcn_v1_0_dec_ring_emit_wreg, 2081 .emit_reg_wait = vcn_v1_0_dec_ring_emit_reg_wait, 2082 .emit_reg_write_reg_wait = amdgpu_ring_emit_reg_write_reg_wait_helper, 2083 }; 2084 2085 static const struct amdgpu_ring_funcs vcn_v1_0_enc_ring_vm_funcs = { 2086 .type = AMDGPU_RING_TYPE_VCN_ENC, 2087 .align_mask = 0x3f, 2088 .nop = VCN_ENC_CMD_NO_OP, 2089 .support_64bit_ptrs = false, 2090 .vmhub = AMDGPU_MMHUB, 2091 .get_rptr = vcn_v1_0_enc_ring_get_rptr, 2092 .get_wptr = vcn_v1_0_enc_ring_get_wptr, 2093 .set_wptr = vcn_v1_0_enc_ring_set_wptr, 2094 .emit_frame_size = 2095 SOC15_FLUSH_GPU_TLB_NUM_WREG * 3 + 2096 SOC15_FLUSH_GPU_TLB_NUM_REG_WAIT * 4 + 2097 4 + /* vcn_v1_0_enc_ring_emit_vm_flush */ 2098 5 + 5 + /* vcn_v1_0_enc_ring_emit_fence x2 vm fence */ 2099 1, /* vcn_v1_0_enc_ring_insert_end */ 2100 .emit_ib_size = 5, /* vcn_v1_0_enc_ring_emit_ib */ 2101 .emit_ib = vcn_v1_0_enc_ring_emit_ib, 2102 .emit_fence = vcn_v1_0_enc_ring_emit_fence, 2103 .emit_vm_flush = vcn_v1_0_enc_ring_emit_vm_flush, 2104 .test_ring = amdgpu_vcn_enc_ring_test_ring, 2105 .test_ib = amdgpu_vcn_enc_ring_test_ib, 2106 .insert_nop = amdgpu_ring_insert_nop, 2107 .insert_end = vcn_v1_0_enc_ring_insert_end, 2108 .pad_ib = amdgpu_ring_generic_pad_ib, 2109 .begin_use = amdgpu_vcn_ring_begin_use, 2110 .end_use = amdgpu_vcn_ring_end_use, 2111 .emit_wreg = vcn_v1_0_enc_ring_emit_wreg, 2112 .emit_reg_wait = vcn_v1_0_enc_ring_emit_reg_wait, 2113 .emit_reg_write_reg_wait = amdgpu_ring_emit_reg_write_reg_wait_helper, 2114 }; 2115 2116 static const struct amdgpu_ring_funcs vcn_v1_0_jpeg_ring_vm_funcs = { 2117 .type = AMDGPU_RING_TYPE_VCN_JPEG, 2118 .align_mask = 0xf, 2119 .nop = PACKET0(0x81ff, 0), 2120 .support_64bit_ptrs = false, 2121 .vmhub = AMDGPU_MMHUB, 2122 .extra_dw = 64, 2123 .get_rptr = vcn_v1_0_jpeg_ring_get_rptr, 2124 .get_wptr = vcn_v1_0_jpeg_ring_get_wptr, 2125 .set_wptr = vcn_v1_0_jpeg_ring_set_wptr, 2126 .emit_frame_size = 2127 6 + 6 + /* hdp invalidate / flush */ 2128 SOC15_FLUSH_GPU_TLB_NUM_WREG * 6 + 2129 SOC15_FLUSH_GPU_TLB_NUM_REG_WAIT * 8 + 2130 8 + /* vcn_v1_0_jpeg_ring_emit_vm_flush */ 2131 26 + 26 + /* vcn_v1_0_jpeg_ring_emit_fence x2 vm fence */ 2132 6, 2133 .emit_ib_size = 22, /* vcn_v1_0_jpeg_ring_emit_ib */ 2134 .emit_ib = vcn_v1_0_jpeg_ring_emit_ib, 2135 .emit_fence = vcn_v1_0_jpeg_ring_emit_fence, 2136 .emit_vm_flush = vcn_v1_0_jpeg_ring_emit_vm_flush, 2137 .test_ring = amdgpu_vcn_jpeg_ring_test_ring, 2138 .test_ib = amdgpu_vcn_jpeg_ring_test_ib, 2139 .insert_nop = vcn_v1_0_jpeg_ring_nop, 2140 .insert_start = vcn_v1_0_jpeg_ring_insert_start, 2141 .insert_end = vcn_v1_0_jpeg_ring_insert_end, 2142 .pad_ib = amdgpu_ring_generic_pad_ib, 2143 .begin_use = amdgpu_vcn_ring_begin_use, 2144 .end_use = amdgpu_vcn_ring_end_use, 2145 .emit_wreg = vcn_v1_0_jpeg_ring_emit_wreg, 2146 .emit_reg_wait = vcn_v1_0_jpeg_ring_emit_reg_wait, 2147 .emit_reg_write_reg_wait = amdgpu_ring_emit_reg_write_reg_wait_helper, 2148 }; 2149 2150 static void vcn_v1_0_set_dec_ring_funcs(struct amdgpu_device *adev) 2151 { 2152 adev->vcn.ring_dec.funcs = &vcn_v1_0_dec_ring_vm_funcs; 2153 DRM_INFO("VCN decode is enabled in VM mode\n"); 2154 } 2155 2156 static void vcn_v1_0_set_enc_ring_funcs(struct amdgpu_device *adev) 2157 { 2158 int i; 2159 2160 for (i = 0; i < adev->vcn.num_enc_rings; ++i) 2161 adev->vcn.ring_enc[i].funcs = &vcn_v1_0_enc_ring_vm_funcs; 2162 2163 DRM_INFO("VCN encode is enabled in VM mode\n"); 2164 } 2165 2166 static void vcn_v1_0_set_jpeg_ring_funcs(struct amdgpu_device *adev) 2167 { 2168 adev->vcn.ring_jpeg.funcs = &vcn_v1_0_jpeg_ring_vm_funcs; 2169 DRM_INFO("VCN jpeg decode is enabled in VM mode\n"); 2170 } 2171 2172 static const struct amdgpu_irq_src_funcs vcn_v1_0_irq_funcs = { 2173 .set = vcn_v1_0_set_interrupt_state, 2174 .process = vcn_v1_0_process_interrupt, 2175 }; 2176 2177 static void vcn_v1_0_set_irq_funcs(struct amdgpu_device *adev) 2178 { 2179 adev->vcn.irq.num_types = adev->vcn.num_enc_rings + 2; 2180 adev->vcn.irq.funcs = &vcn_v1_0_irq_funcs; 2181 } 2182 2183 const struct amdgpu_ip_block_version vcn_v1_0_ip_block = 2184 { 2185 .type = AMD_IP_BLOCK_TYPE_VCN, 2186 .major = 1, 2187 .minor = 0, 2188 .rev = 0, 2189 .funcs = &vcn_v1_0_ip_funcs, 2190 }; 2191