1 #ifndef _ASM_X86_MCE_H 2 #define _ASM_X86_MCE_H 3 4 #include <uapi/asm/mce.h> 5 6 /* 7 * Machine Check support for x86 8 */ 9 10 /* MCG_CAP register defines */ 11 #define MCG_BANKCNT_MASK 0xff /* Number of Banks */ 12 #define MCG_CTL_P (1ULL<<8) /* MCG_CTL register available */ 13 #define MCG_EXT_P (1ULL<<9) /* Extended registers available */ 14 #define MCG_CMCI_P (1ULL<<10) /* CMCI supported */ 15 #define MCG_EXT_CNT_MASK 0xff0000 /* Number of Extended registers */ 16 #define MCG_EXT_CNT_SHIFT 16 17 #define MCG_EXT_CNT(c) (((c) & MCG_EXT_CNT_MASK) >> MCG_EXT_CNT_SHIFT) 18 #define MCG_SER_P (1ULL<<24) /* MCA recovery/new status bits */ 19 #define MCG_ELOG_P (1ULL<<26) /* Extended error log supported */ 20 21 /* MCG_STATUS register defines */ 22 #define MCG_STATUS_RIPV (1ULL<<0) /* restart ip valid */ 23 #define MCG_STATUS_EIPV (1ULL<<1) /* ip points to correct instruction */ 24 #define MCG_STATUS_MCIP (1ULL<<2) /* machine check in progress */ 25 26 /* MCi_STATUS register defines */ 27 #define MCI_STATUS_VAL (1ULL<<63) /* valid error */ 28 #define MCI_STATUS_OVER (1ULL<<62) /* previous errors lost */ 29 #define MCI_STATUS_UC (1ULL<<61) /* uncorrected error */ 30 #define MCI_STATUS_EN (1ULL<<60) /* error enabled */ 31 #define MCI_STATUS_MISCV (1ULL<<59) /* misc error reg. valid */ 32 #define MCI_STATUS_ADDRV (1ULL<<58) /* addr reg. valid */ 33 #define MCI_STATUS_PCC (1ULL<<57) /* processor context corrupt */ 34 #define MCI_STATUS_S (1ULL<<56) /* Signaled machine check */ 35 #define MCI_STATUS_AR (1ULL<<55) /* Action required */ 36 37 /* AMD-specific bits */ 38 #define MCI_STATUS_DEFERRED (1ULL<<44) /* declare an uncorrected error */ 39 #define MCI_STATUS_POISON (1ULL<<43) /* access poisonous data */ 40 41 /* 42 * Note that the full MCACOD field of IA32_MCi_STATUS MSR is 43 * bits 15:0. But bit 12 is the 'F' bit, defined for corrected 44 * errors to indicate that errors are being filtered by hardware. 45 * We should mask out bit 12 when looking for specific signatures 46 * of uncorrected errors - so the F bit is deliberately skipped 47 * in this #define. 48 */ 49 #define MCACOD 0xefff /* MCA Error Code */ 50 51 /* Architecturally defined codes from SDM Vol. 3B Chapter 15 */ 52 #define MCACOD_SCRUB 0x00C0 /* 0xC0-0xCF Memory Scrubbing */ 53 #define MCACOD_SCRUBMSK 0xeff0 /* Skip bit 12 ('F' bit) */ 54 #define MCACOD_L3WB 0x017A /* L3 Explicit Writeback */ 55 #define MCACOD_DATA 0x0134 /* Data Load */ 56 #define MCACOD_INSTR 0x0150 /* Instruction Fetch */ 57 58 /* MCi_MISC register defines */ 59 #define MCI_MISC_ADDR_LSB(m) ((m) & 0x3f) 60 #define MCI_MISC_ADDR_MODE(m) (((m) >> 6) & 7) 61 #define MCI_MISC_ADDR_SEGOFF 0 /* segment offset */ 62 #define MCI_MISC_ADDR_LINEAR 1 /* linear address */ 63 #define MCI_MISC_ADDR_PHYS 2 /* physical address */ 64 #define MCI_MISC_ADDR_MEM 3 /* memory address */ 65 #define MCI_MISC_ADDR_GENERIC 7 /* generic */ 66 67 /* CTL2 register defines */ 68 #define MCI_CTL2_CMCI_EN (1ULL << 30) 69 #define MCI_CTL2_CMCI_THRESHOLD_MASK 0x7fffULL 70 71 #define MCJ_CTX_MASK 3 72 #define MCJ_CTX(flags) ((flags) & MCJ_CTX_MASK) 73 #define MCJ_CTX_RANDOM 0 /* inject context: random */ 74 #define MCJ_CTX_PROCESS 0x1 /* inject context: process */ 75 #define MCJ_CTX_IRQ 0x2 /* inject context: IRQ */ 76 #define MCJ_NMI_BROADCAST 0x4 /* do NMI broadcasting */ 77 #define MCJ_EXCEPTION 0x8 /* raise as exception */ 78 #define MCJ_IRQ_BROADCAST 0x10 /* do IRQ broadcasting */ 79 80 #define MCE_OVERFLOW 0 /* bit 0 in flags means overflow */ 81 82 /* Software defined banks */ 83 #define MCE_EXTENDED_BANK 128 84 #define MCE_THERMAL_BANK (MCE_EXTENDED_BANK + 0) 85 86 #define MCE_LOG_LEN 32 87 #define MCE_LOG_SIGNATURE "MACHINECHECK" 88 89 /* 90 * This structure contains all data related to the MCE log. Also 91 * carries a signature to make it easier to find from external 92 * debugging tools. Each entry is only valid when its finished flag 93 * is set. 94 */ 95 struct mce_log { 96 char signature[12]; /* "MACHINECHECK" */ 97 unsigned len; /* = MCE_LOG_LEN */ 98 unsigned next; 99 unsigned flags; 100 unsigned recordlen; /* length of struct mce */ 101 struct mce entry[MCE_LOG_LEN]; 102 }; 103 104 struct mca_config { 105 bool dont_log_ce; 106 bool cmci_disabled; 107 bool ignore_ce; 108 bool disabled; 109 bool ser; 110 bool bios_cmci_threshold; 111 u8 banks; 112 s8 bootlog; 113 int tolerant; 114 int monarch_timeout; 115 int panic_timeout; 116 u32 rip_msr; 117 }; 118 119 extern struct mca_config mca_cfg; 120 extern void mce_register_decode_chain(struct notifier_block *nb); 121 extern void mce_unregister_decode_chain(struct notifier_block *nb); 122 123 #include <linux/percpu.h> 124 #include <linux/atomic.h> 125 126 extern int mce_p5_enabled; 127 128 #ifdef CONFIG_X86_MCE 129 int mcheck_init(void); 130 void mcheck_cpu_init(struct cpuinfo_x86 *c); 131 #else 132 static inline int mcheck_init(void) { return 0; } 133 static inline void mcheck_cpu_init(struct cpuinfo_x86 *c) {} 134 #endif 135 136 #ifdef CONFIG_X86_ANCIENT_MCE 137 void intel_p5_mcheck_init(struct cpuinfo_x86 *c); 138 void winchip_mcheck_init(struct cpuinfo_x86 *c); 139 static inline void enable_p5_mce(void) { mce_p5_enabled = 1; } 140 #else 141 static inline void intel_p5_mcheck_init(struct cpuinfo_x86 *c) {} 142 static inline void winchip_mcheck_init(struct cpuinfo_x86 *c) {} 143 static inline void enable_p5_mce(void) {} 144 #endif 145 146 void mce_setup(struct mce *m); 147 void mce_log(struct mce *m); 148 DECLARE_PER_CPU(struct device *, mce_device); 149 150 /* 151 * Maximum banks number. 152 * This is the limit of the current register layout on 153 * Intel CPUs. 154 */ 155 #define MAX_NR_BANKS 32 156 157 #ifdef CONFIG_X86_MCE_INTEL 158 void mce_intel_feature_init(struct cpuinfo_x86 *c); 159 void cmci_clear(void); 160 void cmci_reenable(void); 161 void cmci_rediscover(void); 162 void cmci_recheck(void); 163 #else 164 static inline void mce_intel_feature_init(struct cpuinfo_x86 *c) { } 165 static inline void cmci_clear(void) {} 166 static inline void cmci_reenable(void) {} 167 static inline void cmci_rediscover(void) {} 168 static inline void cmci_recheck(void) {} 169 #endif 170 171 #ifdef CONFIG_X86_MCE_AMD 172 void mce_amd_feature_init(struct cpuinfo_x86 *c); 173 #else 174 static inline void mce_amd_feature_init(struct cpuinfo_x86 *c) { } 175 #endif 176 177 int mce_available(struct cpuinfo_x86 *c); 178 179 DECLARE_PER_CPU(unsigned, mce_exception_count); 180 DECLARE_PER_CPU(unsigned, mce_poll_count); 181 182 typedef DECLARE_BITMAP(mce_banks_t, MAX_NR_BANKS); 183 DECLARE_PER_CPU(mce_banks_t, mce_poll_banks); 184 185 enum mcp_flags { 186 MCP_TIMESTAMP = (1 << 0), /* log time stamp */ 187 MCP_UC = (1 << 1), /* log uncorrected errors */ 188 MCP_DONTLOG = (1 << 2), /* only clear, don't log */ 189 }; 190 void machine_check_poll(enum mcp_flags flags, mce_banks_t *b); 191 192 int mce_notify_irq(void); 193 194 DECLARE_PER_CPU(struct mce, injectm); 195 196 extern void register_mce_write_callback(ssize_t (*)(struct file *filp, 197 const char __user *ubuf, 198 size_t usize, loff_t *off)); 199 200 /* Disable CMCI/polling for MCA bank claimed by firmware */ 201 extern void mce_disable_bank(int bank); 202 203 /* 204 * Exception handler 205 */ 206 207 /* Call the installed machine check handler for this CPU setup. */ 208 extern void (*machine_check_vector)(struct pt_regs *, long error_code); 209 void do_machine_check(struct pt_regs *, long); 210 211 /* 212 * Threshold handler 213 */ 214 215 extern void (*mce_threshold_vector)(void); 216 extern void (*threshold_cpu_callback)(unsigned long action, unsigned int cpu); 217 218 /* 219 * Thermal handler 220 */ 221 222 void intel_init_thermal(struct cpuinfo_x86 *c); 223 224 void mce_log_therm_throt_event(__u64 status); 225 226 /* Interrupt Handler for core thermal thresholds */ 227 extern int (*platform_thermal_notify)(__u64 msr_val); 228 229 /* Interrupt Handler for package thermal thresholds */ 230 extern int (*platform_thermal_package_notify)(__u64 msr_val); 231 232 /* Callback support of rate control, return true, if 233 * callback has rate control */ 234 extern bool (*platform_thermal_package_rate_control)(void); 235 236 #ifdef CONFIG_X86_THERMAL_VECTOR 237 extern void mcheck_intel_therm_init(void); 238 #else 239 static inline void mcheck_intel_therm_init(void) { } 240 #endif 241 242 /* 243 * Used by APEI to report memory error via /dev/mcelog 244 */ 245 246 struct cper_sec_mem_err; 247 extern void apei_mce_report_mem_error(int corrected, 248 struct cper_sec_mem_err *mem_err); 249 250 #endif /* _ASM_X86_MCE_H */ 251