FILE
ras_aca_v1_0.c
File preview with syntax highlighting.
linux-6.19.14/drivers/gpu/drm/amd/ras/rascore/ras_aca_v1_0.c
1// SPDX-License-Identifier: MIT 2/* 3 * Copyright 2025 Advanced Micro Devices, Inc. 4 * 5 * Permission is hereby granted, free of charge, to any person obtaining a 6 * copy of this software and associated documentation files (the "Software"), 7 * to deal in the Software without restriction, including without limitation 8 * the rights to use, copy, modify, merge, publish, distribute, sublicense, 9 * and/or sell copies of the Software, and to permit persons to whom the 10 * Software is furnished to do so, subject to the following conditions: 11 * 12 * The above copyright notice and this permission notice shall be included in 13 * all copies or substantial portions of the Software. 14 * 15 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR 16 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, 17 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL 18 * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR 19 * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, 20 * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR 21 * OTHER DEALINGS IN THE SOFTWARE. 22 * 23 */ 24#include "ras.h" 25#include "ras_aca.h" 26#include "ras_core_status.h" 27#include "ras_aca_v1_0.h" 28 29struct ras_aca_hwip { 30 int hwid; 31 int mcatype; 32}; 33 34static struct ras_aca_hwip aca_hwid_mcatypes[ACA_ECC_HWIP_COUNT] = { 35 [ACA_ECC_HWIP__SMU] = {0x01, 0x01}, 36 [ACA_ECC_HWIP__PCS_XGMI] = {0x50, 0x00}, 37 [ACA_ECC_HWIP__UMC] = {0x96, 0x00}, 38}; 39 40static int aca_decode_bank_info(struct aca_block *aca_blk, 41 struct aca_bank_reg *bank, struct aca_ecc_info *info) 42{ 43 u64 ipid; 44 u32 instidhi, instidlo; 45 46 ipid = bank->regs[ACA_REG_IDX__IPID]; 47 info->hwid = ACA_REG_IPID_HARDWAREID(ipid); 48 info->mcatype = ACA_REG_IPID_MCATYPE(ipid); 49 /* 50 * Unified DieID Format: SAASS. A:AID, S:Socket. 51 * Unified DieID[4:4] = InstanceId[0:0] 52 * Unified DieID[0:3] = InstanceIdHi[0:3] 53 */ 54 instidhi = ACA_REG_IPID_INSTANCEIDHI(ipid); 55 instidlo = ACA_REG_IPID_INSTANCEIDLO(ipid); 56 info->die_id = ((instidhi >> 2) & 0x03); 57 info->socket_id = ((instidlo & 0x1) << 2) | (instidhi & 0x03); 58 59 if ((aca_blk->blk_info->hwip == ACA_ECC_HWIP__SMU) && 60 (aca_blk->blk_info->ras_block_id == RAS_BLOCK_ID__GFX)) 61 info->xcd_id = 62 ((instidlo & GENMASK_ULL(31, 1)) == mmSMNAID_XCD0_MCA_SMU) ? 0 : 1; 63 64 return 0; 65} 66 67static bool aca_check_bank_hwip(struct aca_bank_reg *bank, enum aca_ecc_hwip type) 68{ 69 struct ras_aca_hwip *hwip; 70 int hwid, mcatype; 71 u64 ipid; 72 73 if (!bank || (type == ACA_ECC_HWIP__UNKNOWN)) 74 return false; 75 76 hwip = &aca_hwid_mcatypes[type]; 77 if (!hwip->hwid) 78 return false; 79 80 ipid = bank->regs[ACA_REG_IDX__IPID]; 81 hwid = ACA_REG_IPID_HARDWAREID(ipid); 82 mcatype = ACA_REG_IPID_MCATYPE(ipid); 83 84 return hwip->hwid == hwid && hwip->mcatype == mcatype; 85} 86 87static bool aca_match_bank_default(struct aca_block *aca_blk, void *data) 88{ 89 return aca_check_bank_hwip((struct aca_bank_reg *)data, aca_blk->blk_info->hwip); 90} 91 92static bool aca_match_gfx_bank(struct aca_block *aca_blk, void *data) 93{ 94 struct aca_bank_reg *bank = (struct aca_bank_reg *)data; 95 u32 instlo; 96 97 if (!aca_check_bank_hwip(bank, aca_blk->blk_info->hwip)) 98 return false; 99 100 instlo = ACA_REG_IPID_INSTANCEIDLO(bank->regs[ACA_REG_IDX__IPID]); 101 instlo &= GENMASK_ULL(31, 1); 102 switch (instlo) { 103 case mmSMNAID_XCD0_MCA_SMU: 104 case mmSMNAID_XCD1_MCA_SMU: 105 case mmSMNXCD_XCD0_MCA_SMU: 106 return true; 107 default: 108 break; 109 } 110 111 return false; 112} 113 114static bool aca_match_sdma_bank(struct aca_block *aca_blk, void *data) 115{ 116 struct aca_bank_reg *bank = (struct aca_bank_reg *)data; 117 /* CODE_SDMA0 - CODE_SDMA4, reference to smu driver if header file */ 118 static int sdma_err_codes[] = { 33, 34, 35, 36 }; 119 u32 instlo; 120 int errcode, i; 121 122 if (!aca_check_bank_hwip(bank, aca_blk->blk_info->hwip)) 123 return false; 124 125 instlo = ACA_REG_IPID_INSTANCEIDLO(bank->regs[ACA_REG_IDX__IPID]); 126 instlo &= GENMASK_ULL(31, 1); 127 if (instlo != mmSMNAID_AID0_MCA_SMU) 128 return false; 129 130 errcode = ACA_REG_SYND_ERRORINFORMATION(bank->regs[ACA_REG_IDX__SYND]); 131 errcode &= 0xff; 132 133 /* Check SDMA error codes */ 134 for (i = 0; i < ARRAY_SIZE(sdma_err_codes); i++) { 135 if (errcode == sdma_err_codes[i]) 136 return true; 137 } 138 139 return false; 140} 141 142static bool aca_match_mmhub_bank(struct aca_block *aca_blk, void *data) 143{ 144 struct aca_bank_reg *bank = (struct aca_bank_reg *)data; 145 /* reference to smu driver if header file */ 146 const int mmhub_err_codes[] = { 147 0, 1, 2, 3, 4, /* CODE_DAGB0 - 4 */ 148 5, 6, 7, 8, 9, /* CODE_EA0 - 4 */ 149 10, /* CODE_UTCL2_ROUTER */ 150 11, /* CODE_VML2 */ 151 12, /* CODE_VML2_WALKER */ 152 13, /* CODE_MMCANE */ 153 }; 154 u32 instlo; 155 int errcode, i; 156 157 if (!aca_check_bank_hwip(bank, aca_blk->blk_info->hwip)) 158 return false; 159 160 instlo = ACA_REG_IPID_INSTANCEIDLO(bank->regs[ACA_REG_IDX__IPID]); 161 instlo &= GENMASK_ULL(31, 1); 162 if (instlo != mmSMNAID_AID0_MCA_SMU) 163 return false; 164 165 errcode = ACA_REG_SYND_ERRORINFORMATION(bank->regs[ACA_REG_IDX__SYND]); 166 errcode &= 0xff; 167 168 /* Check MMHUB error codes */ 169 for (i = 0; i < ARRAY_SIZE(mmhub_err_codes); i++) { 170 if (errcode == mmhub_err_codes[i]) 171 return true; 172 } 173 174 return false; 175} 176 177static bool aca_check_umc_de(struct ras_core_context *ras_core, uint64_t mc_umc_status) 178{ 179 return (ras_core->poison_supported && 180 ACA_REG_STATUS_VAL(mc_umc_status) && 181 ACA_REG_STATUS_DEFERRED(mc_umc_status)); 182} 183 184static bool aca_check_umc_ue(struct ras_core_context *ras_core, uint64_t mc_umc_status) 185{ 186 if (aca_check_umc_de(ras_core, mc_umc_status)) 187 return false; 188 189 return (ACA_REG_STATUS_VAL(mc_umc_status) && 190 (ACA_REG_STATUS_PCC(mc_umc_status) || 191 ACA_REG_STATUS_UC(mc_umc_status) || 192 ACA_REG_STATUS_TCC(mc_umc_status))); 193} 194 195static bool aca_check_umc_ce(struct ras_core_context *ras_core, uint64_t mc_umc_status) 196{ 197 if (aca_check_umc_de(ras_core, mc_umc_status)) 198 return false; 199 200 return (ACA_REG_STATUS_VAL(mc_umc_status) && 201 (ACA_REG_STATUS_CECC(mc_umc_status) || 202 (ACA_REG_STATUS_UECC(mc_umc_status) && 203 ACA_REG_STATUS_UC(mc_umc_status) == 0) || 204 /* Identify data parity error in replay mode */ 205 ((ACA_REG_STATUS_ERRORCODEEXT(mc_umc_status) == 0x5 || 206 ACA_REG_STATUS_ERRORCODEEXT(mc_umc_status) == 0xb) && 207 !(aca_check_umc_ue(ras_core, mc_umc_status))))); 208} 209 210static int aca_parse_umc_bank(struct ras_core_context *ras_core, 211 struct aca_block *ras_blk, void *data, void *buf) 212{ 213 struct aca_bank_reg *bank = (struct aca_bank_reg *)data; 214 struct aca_bank_ecc *ecc = (struct aca_bank_ecc *)buf; 215 struct aca_ecc_info bank_info; 216 uint32_t ext_error_code; 217 uint64_t status0; 218 219 status0 = bank->regs[ACA_REG_IDX__STATUS]; 220 if (!ACA_REG_STATUS_VAL(status0)) 221 return 0; 222 223 memset(&bank_info, 0, sizeof(bank_info)); 224 aca_decode_bank_info(ras_blk, bank, &bank_info); 225 memcpy(&ecc->bank_info, &bank_info, sizeof(bank_info)); 226 ecc->bank_info.status = bank->regs[ACA_REG_IDX__STATUS]; 227 ecc->bank_info.ipid = bank->regs[ACA_REG_IDX__IPID]; 228 ecc->bank_info.addr = bank->regs[ACA_REG_IDX__ADDR]; 229 230 ext_error_code = ACA_REG_STATUS_ERRORCODEEXT(status0); 231 232 if (aca_check_umc_de(ras_core, status0)) 233 ecc->de_count = 1; 234 else if (aca_check_umc_ue(ras_core, status0)) 235 ecc->ue_count = ext_error_code ? 236 1 : ACA_REG_MISC0_ERRCNT(bank->regs[ACA_REG_IDX__MISC0]); 237 else if (aca_check_umc_ce(ras_core, status0)) 238 ecc->ce_count = ext_error_code ? 239 1 : ACA_REG_MISC0_ERRCNT(bank->regs[ACA_REG_IDX__MISC0]); 240 241 return 0; 242} 243 244static bool aca_check_bank_is_de(struct ras_core_context *ras_core, 245 uint64_t status) 246{ 247 return (ACA_REG_STATUS_POISON(status) || 248 ACA_REG_STATUS_DEFERRED(status)); 249} 250 251static int aca_parse_bank_default(struct ras_core_context *ras_core, 252 struct aca_block *ras_blk, 253 void *data, void *buf) 254{ 255 struct aca_bank_reg *bank = (struct aca_bank_reg *)data; 256 struct aca_bank_ecc *ecc = (struct aca_bank_ecc *)buf; 257 struct aca_ecc_info bank_info; 258 u64 misc0 = bank->regs[ACA_REG_IDX__MISC0]; 259 u64 status = bank->regs[ACA_REG_IDX__STATUS]; 260 261 memset(&bank_info, 0, sizeof(bank_info)); 262 aca_decode_bank_info(ras_blk, bank, &bank_info); 263 memcpy(&ecc->bank_info, &bank_info, sizeof(bank_info)); 264 ecc->bank_info.status = status; 265 ecc->bank_info.ipid = bank->regs[ACA_REG_IDX__IPID]; 266 ecc->bank_info.addr = bank->regs[ACA_REG_IDX__ADDR]; 267 268 if (aca_check_bank_is_de(ras_core, status)) { 269 ecc->de_count = 1; 270 } else { 271 if (bank->ecc_type == RAS_ERR_TYPE__UE) 272 ecc->ue_count = 1; 273 else if (bank->ecc_type == RAS_ERR_TYPE__CE) 274 ecc->ce_count = ACA_REG_MISC0_ERRCNT(misc0); 275 } 276 277 return 0; 278} 279 280static int aca_parse_xgmi_bank(struct ras_core_context *ras_core, 281 struct aca_block *ras_blk, 282 void *data, void *buf) 283{ 284 struct aca_bank_reg *bank = (struct aca_bank_reg *)data; 285 struct aca_bank_ecc *ecc = (struct aca_bank_ecc *)buf; 286 struct aca_ecc_info bank_info; 287 u64 status, count; 288 int ext_error_code; 289 290 memset(&bank_info, 0, sizeof(bank_info)); 291 aca_decode_bank_info(ras_blk, bank, &bank_info); 292 memcpy(&ecc->bank_info, &bank_info, sizeof(bank_info)); 293 ecc->bank_info.status = bank->regs[ACA_REG_IDX__STATUS]; 294 ecc->bank_info.ipid = bank->regs[ACA_REG_IDX__IPID]; 295 ecc->bank_info.addr = bank->regs[ACA_REG_IDX__ADDR]; 296 297 status = bank->regs[ACA_REG_IDX__STATUS]; 298 ext_error_code = ACA_REG_STATUS_ERRORCODEEXT(status); 299 300 count = ACA_REG_MISC0_ERRCNT(bank->regs[ACA_REG_IDX__MISC0]); 301 if (bank->ecc_type == RAS_ERR_TYPE__UE) { 302 if (ext_error_code != 0 && ext_error_code != 9) 303 count = 0ULL; 304 ecc->ue_count = count; 305 } else if (bank->ecc_type == RAS_ERR_TYPE__CE) { 306 count = ext_error_code == 6 ? count : 0ULL; 307 ecc->ce_count = count; 308 } 309 310 return 0; 311} 312 313static const struct aca_block_info aca_v1_0_umc = { 314 .name = "umc", 315 .ras_block_id = RAS_BLOCK_ID__UMC, 316 .hwip = ACA_ECC_HWIP__UMC, 317 .mask = ACA_ERROR__UE_MASK | ACA_ERROR__CE_MASK | ACA_ERROR__DE_MASK, 318 .bank_ops = { 319 .bank_match = aca_match_bank_default, 320 .bank_parse = aca_parse_umc_bank, 321 }, 322}; 323 324static const struct aca_block_info aca_v1_0_gfx = { 325 .name = "gfx", 326 .ras_block_id = RAS_BLOCK_ID__GFX, 327 .hwip = ACA_ECC_HWIP__SMU, 328 .mask = ACA_ERROR__UE_MASK | ACA_ERROR__CE_MASK, 329 .bank_ops = { 330 .bank_match = aca_match_gfx_bank, 331 .bank_parse = aca_parse_bank_default, 332 }, 333}; 334 335static const struct aca_block_info aca_v1_0_sdma = { 336 .name = "sdma", 337 .ras_block_id = RAS_BLOCK_ID__SDMA, 338 .hwip = ACA_ECC_HWIP__SMU, 339 .mask = ACA_ERROR__UE_MASK, 340 .bank_ops = { 341 .bank_match = aca_match_sdma_bank, 342 .bank_parse = aca_parse_bank_default, 343 }, 344}; 345 346static const struct aca_block_info aca_v1_0_mmhub = { 347 .name = "mmhub", 348 .ras_block_id = RAS_BLOCK_ID__MMHUB, 349 .hwip = ACA_ECC_HWIP__SMU, 350 .mask = ACA_ERROR__UE_MASK, 351 .bank_ops = { 352 .bank_match = aca_match_mmhub_bank, 353 .bank_parse = aca_parse_bank_default, 354 }, 355}; 356 357static const struct aca_block_info aca_v1_0_xgmi = { 358 .name = "xgmi", 359 .ras_block_id = RAS_BLOCK_ID__XGMI_WAFL, 360 .hwip = ACA_ECC_HWIP__PCS_XGMI, 361 .mask = ACA_ERROR__UE_MASK | ACA_ERROR__CE_MASK, 362 .bank_ops = { 363 .bank_match = aca_match_bank_default, 364 .bank_parse = aca_parse_xgmi_bank, 365 }, 366}; 367 368static const struct aca_block_info *aca_block_info_v1_0[] = { 369 &aca_v1_0_umc, 370 &aca_v1_0_gfx, 371 &aca_v1_0_sdma, 372 &aca_v1_0_mmhub, 373 &aca_v1_0_xgmi, 374}; 375 376const struct ras_aca_ip_func ras_aca_func_v1_0 = { 377 .block_num = ARRAY_SIZE(aca_block_info_v1_0), 378 .block_info = aca_block_info_v1_0, 379};
~/
Read-only file preview.