1/* 2 * Copyright 2018 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 "amdgpu.h" 25#include "amdgpu_discovery.h" 26#include "soc15_hw_ip.h" 27#include "discovery.h" 28 29#define mmRCC_CONFIG_MEMSIZE 0xde3 30#define mmMM_INDEX 0x0 31#define mmMM_INDEX_HI 0x6 32#define mmMM_DATA 0x1 33#define HW_ID_MAX 300 34 35static const char *hw_id_names[HW_ID_MAX] = { 36 [MP1_HWID] = "MP1", 37 [MP2_HWID] = "MP2", 38 [THM_HWID] = "THM", 39 [SMUIO_HWID] = "SMUIO", 40 [FUSE_HWID] = "FUSE", 41 [CLKA_HWID] = "CLKA", 42 [PWR_HWID] = "PWR", 43 [GC_HWID] = "GC", 44 [UVD_HWID] = "UVD", 45 [AUDIO_AZ_HWID] = "AUDIO_AZ", 46 [ACP_HWID] = "ACP", 47 [DCI_HWID] = "DCI", 48 [DMU_HWID] = "DMU", 49 [DCO_HWID] = "DCO", 50 [DIO_HWID] = "DIO", 51 [XDMA_HWID] = "XDMA", 52 [DCEAZ_HWID] = "DCEAZ", 53 [DAZ_HWID] = "DAZ", 54 [SDPMUX_HWID] = "SDPMUX", 55 [NTB_HWID] = "NTB", 56 [IOHC_HWID] = "IOHC", 57 [L2IMU_HWID] = "L2IMU", 58 [VCE_HWID] = "VCE", 59 [MMHUB_HWID] = "MMHUB", 60 [ATHUB_HWID] = "ATHUB", 61 [DBGU_NBIO_HWID] = "DBGU_NBIO", 62 [DFX_HWID] = "DFX", 63 [DBGU0_HWID] = "DBGU0", 64 [DBGU1_HWID] = "DBGU1", 65 [OSSSYS_HWID] = "OSSSYS", 66 [HDP_HWID] = "HDP", 67 [SDMA0_HWID] = "SDMA0", 68 [SDMA1_HWID] = "SDMA1", 69 [ISP_HWID] = "ISP", 70 [DBGU_IO_HWID] = "DBGU_IO", 71 [DF_HWID] = "DF", 72 [CLKB_HWID] = "CLKB", 73 [FCH_HWID] = "FCH", 74 [DFX_DAP_HWID] = "DFX_DAP", 75 [L1IMU_PCIE_HWID] = "L1IMU_PCIE", 76 [L1IMU_NBIF_HWID] = "L1IMU_NBIF", 77 [L1IMU_IOAGR_HWID] = "L1IMU_IOAGR", 78 [L1IMU3_HWID] = "L1IMU3", 79 [L1IMU4_HWID] = "L1IMU4", 80 [L1IMU5_HWID] = "L1IMU5", 81 [L1IMU6_HWID] = "L1IMU6", 82 [L1IMU7_HWID] = "L1IMU7", 83 [L1IMU8_HWID] = "L1IMU8", 84 [L1IMU9_HWID] = "L1IMU9", 85 [L1IMU10_HWID] = "L1IMU10", 86 [L1IMU11_HWID] = "L1IMU11", 87 [L1IMU12_HWID] = "L1IMU12", 88 [L1IMU13_HWID] = "L1IMU13", 89 [L1IMU14_HWID] = "L1IMU14", 90 [L1IMU15_HWID] = "L1IMU15", 91 [WAFLC_HWID] = "WAFLC", 92 [FCH_USB_PD_HWID] = "FCH_USB_PD", 93 [PCIE_HWID] = "PCIE", 94 [PCS_HWID] = "PCS", 95 [DDCL_HWID] = "DDCL", 96 [SST_HWID] = "SST", 97 [IOAGR_HWID] = "IOAGR", 98 [NBIF_HWID] = "NBIF", 99 [IOAPIC_HWID] = "IOAPIC", 100 [SYSTEMHUB_HWID] = "SYSTEMHUB", 101 [NTBCCP_HWID] = "NTBCCP", 102 [UMC_HWID] = "UMC", 103 [SATA_HWID] = "SATA", 104 [USB_HWID] = "USB", 105 [CCXSEC_HWID] = "CCXSEC", 106 [XGMI_HWID] = "XGMI", 107 [XGBE_HWID] = "XGBE", 108 [MP0_HWID] = "MP0", 109}; 110 111static int hw_id_map[MAX_HWIP] = { 112 [GC_HWIP] = GC_HWID, 113 [HDP_HWIP] = HDP_HWID, 114 [SDMA0_HWIP] = SDMA0_HWID, 115 [SDMA1_HWIP] = SDMA1_HWID, 116 [MMHUB_HWIP] = MMHUB_HWID, 117 [ATHUB_HWIP] = ATHUB_HWID, 118 [NBIO_HWIP] = NBIF_HWID, 119 [MP0_HWIP] = MP0_HWID, 120 [MP1_HWIP] = MP1_HWID, 121 [UVD_HWIP] = UVD_HWID, 122 [VCE_HWIP] = VCE_HWID, 123 [DF_HWIP] = DF_HWID, 124 [DCE_HWIP] = DMU_HWID, 125 [OSSSYS_HWIP] = OSSSYS_HWID, 126 [SMUIO_HWIP] = SMUIO_HWID, 127 [PWR_HWIP] = PWR_HWID, 128 [NBIF_HWIP] = NBIF_HWID, 129 [THM_HWIP] = THM_HWID, 130 [CLK_HWIP] = CLKA_HWID, 131}; 132 133static int amdgpu_discovery_read_binary(struct amdgpu_device *adev, uint8_t *binary) 134{ 135 uint64_t vram_size = (uint64_t)RREG32(mmRCC_CONFIG_MEMSIZE) << 20; 136 uint64_t pos = vram_size - DISCOVERY_TMR_OFFSET; 137 138 amdgpu_device_vram_access(adev, pos, (uint32_t *)binary, 139 adev->mman.discovery_tmr_size, false); 140 return 0; 141} 142 143static uint16_t amdgpu_discovery_calculate_checksum(uint8_t *data, uint32_t size) 144{ 145 uint16_t checksum = 0; 146 int i; 147 148 for (i = 0; i < size; i++) 149 checksum += data[i]; 150 151 return checksum; 152} 153 154static inline bool amdgpu_discovery_verify_checksum(uint8_t *data, uint32_t size, 155 uint16_t expected) 156{ 157 return !!(amdgpu_discovery_calculate_checksum(data, size) == expected); 158} 159 160static int amdgpu_discovery_init(struct amdgpu_device *adev) 161{ 162 struct table_info *info; 163 struct binary_header *bhdr; 164 struct ip_discovery_header *ihdr; 165 struct gpu_info_header *ghdr; 166 uint16_t offset; 167 uint16_t size; 168 uint16_t checksum; 169 int r; 170 171 adev->mman.discovery_tmr_size = DISCOVERY_TMR_SIZE; 172 adev->mman.discovery_bin = kzalloc(adev->mman.discovery_tmr_size, GFP_KERNEL); 173 if (!adev->mman.discovery_bin) 174 return -ENOMEM; 175 176 r = amdgpu_discovery_read_binary(adev, adev->mman.discovery_bin); 177 if (r) { 178 DRM_ERROR("failed to read ip discovery binary\n"); 179 goto out; 180 } 181 182 bhdr = (struct binary_header *)adev->mman.discovery_bin; 183 184 if (le32_to_cpu(bhdr->binary_signature) != BINARY_SIGNATURE) { 185 DRM_ERROR("invalid ip discovery binary signature\n"); 186 r = -EINVAL; 187 goto out; 188 } 189 190 offset = offsetof(struct binary_header, binary_checksum) + 191 sizeof(bhdr->binary_checksum); 192 size = bhdr->binary_size - offset; 193 checksum = bhdr->binary_checksum; 194 195 if (!amdgpu_discovery_verify_checksum(adev->mman.discovery_bin + offset, 196 size, checksum)) { 197 DRM_ERROR("invalid ip discovery binary checksum\n"); 198 r = -EINVAL; 199 goto out; 200 } 201 202 info = &bhdr->table_list[IP_DISCOVERY]; 203 offset = le16_to_cpu(info->offset); 204 checksum = le16_to_cpu(info->checksum); 205 ihdr = (struct ip_discovery_header *)(adev->mman.discovery_bin + offset); 206 207 if (le32_to_cpu(ihdr->signature) != DISCOVERY_TABLE_SIGNATURE) { 208 DRM_ERROR("invalid ip discovery data table signature\n"); 209 r = -EINVAL; 210 goto out; 211 } 212 213 if (!amdgpu_discovery_verify_checksum(adev->mman.discovery_bin + offset, 214 ihdr->size, checksum)) { 215 DRM_ERROR("invalid ip discovery data table checksum\n"); 216 r = -EINVAL; 217 goto out; 218 } 219 220 info = &bhdr->table_list[GC]; 221 offset = le16_to_cpu(info->offset); 222 checksum = le16_to_cpu(info->checksum); 223 ghdr = (struct gpu_info_header *)(adev->mman.discovery_bin + offset); 224 225 if (!amdgpu_discovery_verify_checksum(adev->mman.discovery_bin + offset, 226 ghdr->size, checksum)) { 227 DRM_ERROR("invalid gc data table checksum\n"); 228 r = -EINVAL; 229 goto out; 230 } 231 232 return 0; 233 234out: 235 kfree(adev->mman.discovery_bin); 236 adev->mman.discovery_bin = NULL; 237 238 return r; 239} 240 241void amdgpu_discovery_fini(struct amdgpu_device *adev) 242{ 243 kfree(adev->mman.discovery_bin); 244 adev->mman.discovery_bin = NULL; 245} 246 247int amdgpu_discovery_reg_base_init(struct amdgpu_device *adev) 248{ 249 struct binary_header *bhdr; 250 struct ip_discovery_header *ihdr; 251 struct die_header *dhdr; 252 struct ip *ip; 253 uint16_t die_offset; 254 uint16_t ip_offset; 255 uint16_t num_dies; 256 uint16_t num_ips; 257 uint8_t num_base_address; 258 int hw_ip; 259 int i, j, k; 260 int r; 261 262 r = amdgpu_discovery_init(adev); 263 if (r) { 264 DRM_ERROR("amdgpu_discovery_init failed\n"); 265 return r; 266 } 267 268 bhdr = (struct binary_header *)adev->mman.discovery_bin; 269 ihdr = (struct ip_discovery_header *)(adev->mman.discovery_bin + 270 le16_to_cpu(bhdr->table_list[IP_DISCOVERY].offset)); 271 num_dies = le16_to_cpu(ihdr->num_dies); 272 273 DRM_DEBUG("number of dies: %d\n", num_dies); 274 275 for (i = 0; i < num_dies; i++) { 276 die_offset = le16_to_cpu(ihdr->die_info[i].die_offset); 277 dhdr = (struct die_header *)(adev->mman.discovery_bin + die_offset); 278 num_ips = le16_to_cpu(dhdr->num_ips); 279 ip_offset = die_offset + sizeof(*dhdr); 280 281 if (le16_to_cpu(dhdr->die_id) != i) { 282 DRM_ERROR("invalid die id %d, expected %d\n", 283 le16_to_cpu(dhdr->die_id), i); 284 return -EINVAL; 285 } 286 287 DRM_DEBUG("number of hardware IPs on die%d: %d\n", 288 le16_to_cpu(dhdr->die_id), num_ips); 289 290 for (j = 0; j < num_ips; j++) { 291 ip = (struct ip *)(adev->mman.discovery_bin + ip_offset); 292 num_base_address = ip->num_base_address; 293 294 DRM_DEBUG("%s(%d) #%d v%d.%d.%d:\n", 295 hw_id_names[le16_to_cpu(ip->hw_id)], 296 le16_to_cpu(ip->hw_id), 297 ip->number_instance, 298 ip->major, ip->minor, 299 ip->revision); 300 301 for (k = 0; k < num_base_address; k++) { 302 /* 303 * convert the endianness of base addresses in place, 304 * so that we don't need to convert them when accessing adev->reg_offset. 305 */ 306 ip->base_address[k] = le32_to_cpu(ip->base_address[k]); 307 DRM_DEBUG("\t0x%08x\n", ip->base_address[k]); 308 } 309 310 for (hw_ip = 0; hw_ip < MAX_HWIP; hw_ip++) { 311 if (hw_id_map[hw_ip] == le16_to_cpu(ip->hw_id)) { 312 DRM_DEBUG("set register base offset for %s\n", 313 hw_id_names[le16_to_cpu(ip->hw_id)]); 314 adev->reg_offset[hw_ip][ip->number_instance] = 315 ip->base_address; 316 } 317 318 } 319 320 ip_offset += sizeof(*ip) + 4 * (ip->num_base_address - 1); 321 } 322 } 323 324 return 0; 325} 326 327int amdgpu_discovery_get_ip_version(struct amdgpu_device *adev, int hw_id, 328 int *major, int *minor, int *revision) 329{ 330 struct binary_header *bhdr; 331 struct ip_discovery_header *ihdr; 332 struct die_header *dhdr; 333 struct ip *ip; 334 uint16_t die_offset; 335 uint16_t ip_offset; 336 uint16_t num_dies; 337 uint16_t num_ips; 338 int i, j; 339 340 if (!adev->mman.discovery_bin) { 341 DRM_ERROR("ip discovery uninitialized\n"); 342 return -EINVAL; 343 } 344 345 bhdr = (struct binary_header *)adev->mman.discovery_bin; 346 ihdr = (struct ip_discovery_header *)(adev->mman.discovery_bin + 347 le16_to_cpu(bhdr->table_list[IP_DISCOVERY].offset)); 348 num_dies = le16_to_cpu(ihdr->num_dies); 349 350 for (i = 0; i < num_dies; i++) { 351 die_offset = le16_to_cpu(ihdr->die_info[i].die_offset); 352 dhdr = (struct die_header *)(adev->mman.discovery_bin + die_offset); 353 num_ips = le16_to_cpu(dhdr->num_ips); 354 ip_offset = die_offset + sizeof(*dhdr); 355 356 for (j = 0; j < num_ips; j++) { 357 ip = (struct ip *)(adev->mman.discovery_bin + ip_offset); 358 359 if (le16_to_cpu(ip->hw_id) == hw_id) { 360 if (major) 361 *major = ip->major; 362 if (minor) 363 *minor = ip->minor; 364 if (revision) 365 *revision = ip->revision; 366 return 0; 367 } 368 ip_offset += sizeof(*ip) + 4 * (ip->num_base_address - 1); 369 } 370 } 371 372 return -EINVAL; 373} 374 375union gc_info { 376 struct gc_info_v1_0 v1; 377 struct gc_info_v2_0 v2; 378}; 379 380int amdgpu_discovery_get_gfx_info(struct amdgpu_device *adev) 381{ 382 struct binary_header *bhdr; 383 union gc_info *gc_info; 384 385 if (!adev->mman.discovery_bin) { 386 DRM_ERROR("ip discovery uninitialized\n"); 387 return -EINVAL; 388 } 389 390 bhdr = (struct binary_header *)adev->mman.discovery_bin; 391 gc_info = (union gc_info *)(adev->mman.discovery_bin + 392 le16_to_cpu(bhdr->table_list[GC].offset)); 393 switch (gc_info->v1.header.version_major) { 394 case 1: 395 adev->gfx.config.max_shader_engines = le32_to_cpu(gc_info->v1.gc_num_se); 396 adev->gfx.config.max_cu_per_sh = 2 * (le32_to_cpu(gc_info->v1.gc_num_wgp0_per_sa) + 397 le32_to_cpu(gc_info->v1.gc_num_wgp1_per_sa)); 398 adev->gfx.config.max_sh_per_se = le32_to_cpu(gc_info->v1.gc_num_sa_per_se); 399 adev->gfx.config.max_backends_per_se = le32_to_cpu(gc_info->v1.gc_num_rb_per_se); 400 adev->gfx.config.max_texture_channel_caches = le32_to_cpu(gc_info->v1.gc_num_gl2c); 401 adev->gfx.config.max_gprs = le32_to_cpu(gc_info->v1.gc_num_gprs); 402 adev->gfx.config.max_gs_threads = le32_to_cpu(gc_info->v1.gc_num_max_gs_thds); 403 adev->gfx.config.gs_vgt_table_depth = le32_to_cpu(gc_info->v1.gc_gs_table_depth); 404 adev->gfx.config.gs_prim_buffer_depth = le32_to_cpu(gc_info->v1.gc_gsprim_buff_depth); 405 adev->gfx.config.double_offchip_lds_buf = le32_to_cpu(gc_info->v1.gc_double_offchip_lds_buffer); 406 adev->gfx.cu_info.wave_front_size = le32_to_cpu(gc_info->v1.gc_wave_size); 407 adev->gfx.cu_info.max_waves_per_simd = le32_to_cpu(gc_info->v1.gc_max_waves_per_simd); 408 adev->gfx.cu_info.max_scratch_slots_per_cu = le32_to_cpu(gc_info->v1.gc_max_scratch_slots_per_cu); 409 adev->gfx.cu_info.lds_size = le32_to_cpu(gc_info->v1.gc_lds_size); 410 adev->gfx.config.num_sc_per_sh = le32_to_cpu(gc_info->v1.gc_num_sc_per_se) / 411 le32_to_cpu(gc_info->v1.gc_num_sa_per_se); 412 adev->gfx.config.num_packer_per_sc = le32_to_cpu(gc_info->v1.gc_num_packer_per_sc); 413 break; 414 case 2: 415 adev->gfx.config.max_shader_engines = le32_to_cpu(gc_info->v2.gc_num_se); 416 adev->gfx.config.max_cu_per_sh = le32_to_cpu(gc_info->v2.gc_num_cu_per_sh); 417 adev->gfx.config.max_sh_per_se = le32_to_cpu(gc_info->v2.gc_num_sh_per_se); 418 adev->gfx.config.max_backends_per_se = le32_to_cpu(gc_info->v2.gc_num_rb_per_se); 419 adev->gfx.config.max_texture_channel_caches = le32_to_cpu(gc_info->v2.gc_num_tccs); 420 adev->gfx.config.max_gprs = le32_to_cpu(gc_info->v2.gc_num_gprs); 421 adev->gfx.config.max_gs_threads = le32_to_cpu(gc_info->v2.gc_num_max_gs_thds); 422 adev->gfx.config.gs_vgt_table_depth = le32_to_cpu(gc_info->v2.gc_gs_table_depth); 423 adev->gfx.config.gs_prim_buffer_depth = le32_to_cpu(gc_info->v2.gc_gsprim_buff_depth); 424 adev->gfx.config.double_offchip_lds_buf = le32_to_cpu(gc_info->v2.gc_double_offchip_lds_buffer); 425 adev->gfx.cu_info.wave_front_size = le32_to_cpu(gc_info->v2.gc_wave_size); 426 adev->gfx.cu_info.max_waves_per_simd = le32_to_cpu(gc_info->v2.gc_max_waves_per_simd); 427 adev->gfx.cu_info.max_scratch_slots_per_cu = le32_to_cpu(gc_info->v2.gc_max_scratch_slots_per_cu); 428 adev->gfx.cu_info.lds_size = le32_to_cpu(gc_info->v2.gc_lds_size); 429 adev->gfx.config.num_sc_per_sh = le32_to_cpu(gc_info->v2.gc_num_sc_per_se) / 430 le32_to_cpu(gc_info->v2.gc_num_sh_per_se); 431 adev->gfx.config.num_packer_per_sc = le32_to_cpu(gc_info->v2.gc_num_packer_per_sc); 432 break; 433 default: 434 dev_err(adev->dev, 435 "Unhandled GC info table %d.%d\n", 436 gc_info->v1.header.version_major, 437 gc_info->v1.header.version_minor); 438 return -EINVAL; 439 } 440 return 0; 441} 442