1 /* SPDX-License-Identifier: GPL-2.0 */ 2 #ifndef _ASM_X86_MCE_H 3 #define _ASM_X86_MCE_H 4 5 #include <uapi/asm/mce.h> 6 7 /* 8 * Machine Check support for x86 9 */ 10 11 /* MCG_CAP register defines */ 12 #define MCG_BANKCNT_MASK 0xff /* Number of Banks */ 13 #define MCG_CTL_P BIT_ULL(8) /* MCG_CTL register available */ 14 #define MCG_EXT_P BIT_ULL(9) /* Extended registers available */ 15 #define MCG_CMCI_P BIT_ULL(10) /* CMCI supported */ 16 #define MCG_EXT_CNT_MASK 0xff0000 /* Number of Extended registers */ 17 #define MCG_EXT_CNT_SHIFT 16 18 #define MCG_EXT_CNT(c) (((c) & MCG_EXT_CNT_MASK) >> MCG_EXT_CNT_SHIFT) 19 #define MCG_SER_P BIT_ULL(24) /* MCA recovery/new status bits */ 20 #define MCG_ELOG_P BIT_ULL(26) /* Extended error log supported */ 21 #define MCG_LMCE_P BIT_ULL(27) /* Local machine check supported */ 22 23 /* MCG_STATUS register defines */ 24 #define MCG_STATUS_RIPV BIT_ULL(0) /* restart ip valid */ 25 #define MCG_STATUS_EIPV BIT_ULL(1) /* ip points to correct instruction */ 26 #define MCG_STATUS_MCIP BIT_ULL(2) /* machine check in progress */ 27 #define MCG_STATUS_LMCES BIT_ULL(3) /* LMCE signaled */ 28 29 /* MCG_EXT_CTL register defines */ 30 #define MCG_EXT_CTL_LMCE_EN BIT_ULL(0) /* Enable LMCE */ 31 32 /* MCi_STATUS register defines */ 33 #define MCI_STATUS_VAL BIT_ULL(63) /* valid error */ 34 #define MCI_STATUS_OVER BIT_ULL(62) /* previous errors lost */ 35 #define MCI_STATUS_UC BIT_ULL(61) /* uncorrected error */ 36 #define MCI_STATUS_EN BIT_ULL(60) /* error enabled */ 37 #define MCI_STATUS_MISCV BIT_ULL(59) /* misc error reg. valid */ 38 #define MCI_STATUS_ADDRV BIT_ULL(58) /* addr reg. valid */ 39 #define MCI_STATUS_PCC BIT_ULL(57) /* processor context corrupt */ 40 #define MCI_STATUS_S BIT_ULL(56) /* Signaled machine check */ 41 #define MCI_STATUS_AR BIT_ULL(55) /* Action required */ 42 #define MCI_STATUS_CEC_SHIFT 38 /* Corrected Error Count */ 43 #define MCI_STATUS_CEC_MASK GENMASK_ULL(52,38) 44 #define MCI_STATUS_CEC(c) (((c) & MCI_STATUS_CEC_MASK) >> MCI_STATUS_CEC_SHIFT) 45 46 /* AMD-specific bits */ 47 #define MCI_STATUS_TCC BIT_ULL(55) /* Task context corrupt */ 48 #define MCI_STATUS_SYNDV BIT_ULL(53) /* synd reg. valid */ 49 #define MCI_STATUS_DEFERRED BIT_ULL(44) /* uncorrected error, deferred exception */ 50 #define MCI_STATUS_POISON BIT_ULL(43) /* access poisonous data */ 51 #define MCI_STATUS_SCRUB BIT_ULL(40) /* Error detected during scrub operation */ 52 53 /* 54 * McaX field if set indicates a given bank supports MCA extensions: 55 * - Deferred error interrupt type is specifiable by bank. 56 * - MCx_MISC0[BlkPtr] field indicates presence of extended MISC registers, 57 * But should not be used to determine MSR numbers. 58 * - TCC bit is present in MCx_STATUS. 59 */ 60 #define MCI_CONFIG_MCAX 0x1 61 #define MCI_IPID_MCATYPE 0xFFFF0000 62 #define MCI_IPID_HWID 0xFFF 63 64 /* 65 * Note that the full MCACOD field of IA32_MCi_STATUS MSR is 66 * bits 15:0. But bit 12 is the 'F' bit, defined for corrected 67 * errors to indicate that errors are being filtered by hardware. 68 * We should mask out bit 12 when looking for specific signatures 69 * of uncorrected errors - so the F bit is deliberately skipped 70 * in this #define. 71 */ 72 #define MCACOD 0xefff /* MCA Error Code */ 73 74 /* Architecturally defined codes from SDM Vol. 3B Chapter 15 */ 75 #define MCACOD_SCRUB 0x00C0 /* 0xC0-0xCF Memory Scrubbing */ 76 #define MCACOD_SCRUBMSK 0xeff0 /* Skip bit 12 ('F' bit) */ 77 #define MCACOD_L3WB 0x017A /* L3 Explicit Writeback */ 78 #define MCACOD_DATA 0x0134 /* Data Load */ 79 #define MCACOD_INSTR 0x0150 /* Instruction Fetch */ 80 81 /* MCi_MISC register defines */ 82 #define MCI_MISC_ADDR_LSB(m) ((m) & 0x3f) 83 #define MCI_MISC_ADDR_MODE(m) (((m) >> 6) & 7) 84 #define MCI_MISC_ADDR_SEGOFF 0 /* segment offset */ 85 #define MCI_MISC_ADDR_LINEAR 1 /* linear address */ 86 #define MCI_MISC_ADDR_PHYS 2 /* physical address */ 87 #define MCI_MISC_ADDR_MEM 3 /* memory address */ 88 #define MCI_MISC_ADDR_GENERIC 7 /* generic */ 89 90 /* CTL2 register defines */ 91 #define MCI_CTL2_CMCI_EN BIT_ULL(30) 92 #define MCI_CTL2_CMCI_THRESHOLD_MASK 0x7fffULL 93 94 #define MCJ_CTX_MASK 3 95 #define MCJ_CTX(flags) ((flags) & MCJ_CTX_MASK) 96 #define MCJ_CTX_RANDOM 0 /* inject context: random */ 97 #define MCJ_CTX_PROCESS 0x1 /* inject context: process */ 98 #define MCJ_CTX_IRQ 0x2 /* inject context: IRQ */ 99 #define MCJ_NMI_BROADCAST 0x4 /* do NMI broadcasting */ 100 #define MCJ_EXCEPTION 0x8 /* raise as exception */ 101 #define MCJ_IRQ_BROADCAST 0x10 /* do IRQ broadcasting */ 102 103 #define MCE_OVERFLOW 0 /* bit 0 in flags means overflow */ 104 105 #define MCE_LOG_LEN 32 106 #define MCE_LOG_SIGNATURE "MACHINECHECK" 107 108 /* AMD Scalable MCA */ 109 #define MSR_AMD64_SMCA_MC0_CTL 0xc0002000 110 #define MSR_AMD64_SMCA_MC0_STATUS 0xc0002001 111 #define MSR_AMD64_SMCA_MC0_ADDR 0xc0002002 112 #define MSR_AMD64_SMCA_MC0_MISC0 0xc0002003 113 #define MSR_AMD64_SMCA_MC0_CONFIG 0xc0002004 114 #define MSR_AMD64_SMCA_MC0_IPID 0xc0002005 115 #define MSR_AMD64_SMCA_MC0_SYND 0xc0002006 116 #define MSR_AMD64_SMCA_MC0_DESTAT 0xc0002008 117 #define MSR_AMD64_SMCA_MC0_DEADDR 0xc0002009 118 #define MSR_AMD64_SMCA_MC0_MISC1 0xc000200a 119 #define MSR_AMD64_SMCA_MCx_CTL(x) (MSR_AMD64_SMCA_MC0_CTL + 0x10*(x)) 120 #define MSR_AMD64_SMCA_MCx_STATUS(x) (MSR_AMD64_SMCA_MC0_STATUS + 0x10*(x)) 121 #define MSR_AMD64_SMCA_MCx_ADDR(x) (MSR_AMD64_SMCA_MC0_ADDR + 0x10*(x)) 122 #define MSR_AMD64_SMCA_MCx_MISC(x) (MSR_AMD64_SMCA_MC0_MISC0 + 0x10*(x)) 123 #define MSR_AMD64_SMCA_MCx_CONFIG(x) (MSR_AMD64_SMCA_MC0_CONFIG + 0x10*(x)) 124 #define MSR_AMD64_SMCA_MCx_IPID(x) (MSR_AMD64_SMCA_MC0_IPID + 0x10*(x)) 125 #define MSR_AMD64_SMCA_MCx_SYND(x) (MSR_AMD64_SMCA_MC0_SYND + 0x10*(x)) 126 #define MSR_AMD64_SMCA_MCx_DESTAT(x) (MSR_AMD64_SMCA_MC0_DESTAT + 0x10*(x)) 127 #define MSR_AMD64_SMCA_MCx_DEADDR(x) (MSR_AMD64_SMCA_MC0_DEADDR + 0x10*(x)) 128 #define MSR_AMD64_SMCA_MCx_MISCy(x, y) ((MSR_AMD64_SMCA_MC0_MISC1 + y) + (0x10*(x))) 129 130 /* 131 * This structure contains all data related to the MCE log. Also 132 * carries a signature to make it easier to find from external 133 * debugging tools. Each entry is only valid when its finished flag 134 * is set. 135 */ 136 struct mce_log_buffer { 137 char signature[12]; /* "MACHINECHECK" */ 138 unsigned len; /* = MCE_LOG_LEN */ 139 unsigned next; 140 unsigned flags; 141 unsigned recordlen; /* length of struct mce */ 142 struct mce entry[MCE_LOG_LEN]; 143 }; 144 145 enum mce_notifier_prios { 146 MCE_PRIO_FIRST = INT_MAX, 147 MCE_PRIO_UC = INT_MAX - 1, 148 MCE_PRIO_EXTLOG = INT_MAX - 2, 149 MCE_PRIO_NFIT = INT_MAX - 3, 150 MCE_PRIO_EDAC = INT_MAX - 4, 151 MCE_PRIO_MCELOG = 1, 152 MCE_PRIO_LOWEST = 0, 153 }; 154 155 struct notifier_block; 156 extern void mce_register_decode_chain(struct notifier_block *nb); 157 extern void mce_unregister_decode_chain(struct notifier_block *nb); 158 159 #include <linux/percpu.h> 160 #include <linux/atomic.h> 161 162 extern int mce_p5_enabled; 163 164 #ifdef CONFIG_X86_MCE 165 int mcheck_init(void); 166 void mcheck_cpu_init(struct cpuinfo_x86 *c); 167 void mcheck_cpu_clear(struct cpuinfo_x86 *c); 168 void mcheck_vendor_init_severity(void); 169 #else 170 static inline int mcheck_init(void) { return 0; } 171 static inline void mcheck_cpu_init(struct cpuinfo_x86 *c) {} 172 static inline void mcheck_cpu_clear(struct cpuinfo_x86 *c) {} 173 static inline void mcheck_vendor_init_severity(void) {} 174 #endif 175 176 #ifdef CONFIG_X86_ANCIENT_MCE 177 void intel_p5_mcheck_init(struct cpuinfo_x86 *c); 178 void winchip_mcheck_init(struct cpuinfo_x86 *c); 179 static inline void enable_p5_mce(void) { mce_p5_enabled = 1; } 180 #else 181 static inline void intel_p5_mcheck_init(struct cpuinfo_x86 *c) {} 182 static inline void winchip_mcheck_init(struct cpuinfo_x86 *c) {} 183 static inline void enable_p5_mce(void) {} 184 #endif 185 186 void mce_setup(struct mce *m); 187 void mce_log(struct mce *m); 188 DECLARE_PER_CPU(struct device *, mce_device); 189 190 /* 191 * Maximum banks number. 192 * This is the limit of the current register layout on 193 * Intel CPUs. 194 */ 195 #define MAX_NR_BANKS 32 196 197 #ifdef CONFIG_X86_MCE_INTEL 198 void mce_intel_feature_init(struct cpuinfo_x86 *c); 199 void mce_intel_feature_clear(struct cpuinfo_x86 *c); 200 void cmci_clear(void); 201 void cmci_reenable(void); 202 void cmci_rediscover(void); 203 void cmci_recheck(void); 204 #else 205 static inline void mce_intel_feature_init(struct cpuinfo_x86 *c) { } 206 static inline void mce_intel_feature_clear(struct cpuinfo_x86 *c) { } 207 static inline void cmci_clear(void) {} 208 static inline void cmci_reenable(void) {} 209 static inline void cmci_rediscover(void) {} 210 static inline void cmci_recheck(void) {} 211 #endif 212 213 int mce_available(struct cpuinfo_x86 *c); 214 bool mce_is_memory_error(struct mce *m); 215 bool mce_is_correctable(struct mce *m); 216 int mce_usable_address(struct mce *m); 217 218 DECLARE_PER_CPU(unsigned, mce_exception_count); 219 DECLARE_PER_CPU(unsigned, mce_poll_count); 220 221 typedef DECLARE_BITMAP(mce_banks_t, MAX_NR_BANKS); 222 DECLARE_PER_CPU(mce_banks_t, mce_poll_banks); 223 224 enum mcp_flags { 225 MCP_TIMESTAMP = BIT(0), /* log time stamp */ 226 MCP_UC = BIT(1), /* log uncorrected errors */ 227 MCP_DONTLOG = BIT(2), /* only clear, don't log */ 228 }; 229 bool machine_check_poll(enum mcp_flags flags, mce_banks_t *b); 230 231 int mce_notify_irq(void); 232 233 DECLARE_PER_CPU(struct mce, injectm); 234 235 /* Disable CMCI/polling for MCA bank claimed by firmware */ 236 extern void mce_disable_bank(int bank); 237 238 /* 239 * Exception handler 240 */ 241 242 /* Call the installed machine check handler for this CPU setup. */ 243 extern void (*machine_check_vector)(struct pt_regs *, long error_code); 244 void do_machine_check(struct pt_regs *, long); 245 246 /* 247 * Threshold handler 248 */ 249 extern void (*mce_threshold_vector)(void); 250 251 /* Deferred error interrupt handler */ 252 extern void (*deferred_error_int_vector)(void); 253 254 /* 255 * Thermal handler 256 */ 257 258 void intel_init_thermal(struct cpuinfo_x86 *c); 259 260 /* Interrupt Handler for core thermal thresholds */ 261 extern int (*platform_thermal_notify)(__u64 msr_val); 262 263 /* Interrupt Handler for package thermal thresholds */ 264 extern int (*platform_thermal_package_notify)(__u64 msr_val); 265 266 /* Callback support of rate control, return true, if 267 * callback has rate control */ 268 extern bool (*platform_thermal_package_rate_control)(void); 269 270 #ifdef CONFIG_X86_THERMAL_VECTOR 271 extern void mcheck_intel_therm_init(void); 272 #else 273 static inline void mcheck_intel_therm_init(void) { } 274 #endif 275 276 /* 277 * Used by APEI to report memory error via /dev/mcelog 278 */ 279 280 struct cper_sec_mem_err; 281 extern void apei_mce_report_mem_error(int corrected, 282 struct cper_sec_mem_err *mem_err); 283 284 /* 285 * Enumerate new IP types and HWID values in AMD processors which support 286 * Scalable MCA. 287 */ 288 #ifdef CONFIG_X86_MCE_AMD 289 290 /* These may be used by multiple smca_hwid_mcatypes */ 291 enum smca_bank_types { 292 SMCA_LS = 0, /* Load Store */ 293 SMCA_LS_V2, /* Load Store */ 294 SMCA_IF, /* Instruction Fetch */ 295 SMCA_L2_CACHE, /* L2 Cache */ 296 SMCA_DE, /* Decoder Unit */ 297 SMCA_RESERVED, /* Reserved */ 298 SMCA_EX, /* Execution Unit */ 299 SMCA_FP, /* Floating Point */ 300 SMCA_L3_CACHE, /* L3 Cache */ 301 SMCA_CS, /* Coherent Slave */ 302 SMCA_CS_V2, /* Coherent Slave */ 303 SMCA_PIE, /* Power, Interrupts, etc. */ 304 SMCA_UMC, /* Unified Memory Controller */ 305 SMCA_PB, /* Parameter Block */ 306 SMCA_PSP, /* Platform Security Processor */ 307 SMCA_PSP_V2, /* Platform Security Processor */ 308 SMCA_SMU, /* System Management Unit */ 309 SMCA_SMU_V2, /* System Management Unit */ 310 SMCA_MP5, /* Microprocessor 5 Unit */ 311 SMCA_NBIO, /* Northbridge IO Unit */ 312 SMCA_PCIE, /* PCI Express Unit */ 313 N_SMCA_BANK_TYPES 314 }; 315 316 #define HWID_MCATYPE(hwid, mcatype) (((hwid) << 16) | (mcatype)) 317 318 struct smca_hwid { 319 unsigned int bank_type; /* Use with smca_bank_types for easy indexing. */ 320 u32 hwid_mcatype; /* (hwid,mcatype) tuple */ 321 u32 xec_bitmap; /* Bitmap of valid ExtErrorCodes; current max is 21. */ 322 u8 count; /* Number of instances. */ 323 }; 324 325 struct smca_bank { 326 struct smca_hwid *hwid; 327 u32 id; /* Value of MCA_IPID[InstanceId]. */ 328 u8 sysfs_id; /* Value used for sysfs name. */ 329 }; 330 331 extern struct smca_bank smca_banks[MAX_NR_BANKS]; 332 333 extern const char *smca_get_long_name(enum smca_bank_types t); 334 extern bool amd_mce_is_memory_error(struct mce *m); 335 336 extern int mce_threshold_create_device(unsigned int cpu); 337 extern int mce_threshold_remove_device(unsigned int cpu); 338 339 void mce_amd_feature_init(struct cpuinfo_x86 *c); 340 int umc_normaddr_to_sysaddr(u64 norm_addr, u16 nid, u8 umc, u64 *sys_addr); 341 342 #else 343 344 static inline int mce_threshold_create_device(unsigned int cpu) { return 0; }; 345 static inline int mce_threshold_remove_device(unsigned int cpu) { return 0; }; 346 static inline bool amd_mce_is_memory_error(struct mce *m) { return false; }; 347 static inline void mce_amd_feature_init(struct cpuinfo_x86 *c) { } 348 static inline int 349 umc_normaddr_to_sysaddr(u64 norm_addr, u16 nid, u8 umc, u64 *sys_addr) { return -EINVAL; }; 350 #endif 351 352 static inline void mce_hygon_feature_init(struct cpuinfo_x86 *c) { return mce_amd_feature_init(c); } 353 354 #endif /* _ASM_X86_MCE_H */ 355