4 #include <uapi/asm/mce.h>
7 * Machine Check support for x86
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 #define MCG_LMCE_P (1ULL<<27) /* Local machine check supported */
22 /* MCG_STATUS register defines */
23 #define MCG_STATUS_RIPV (1ULL<<0) /* restart ip valid */
24 #define MCG_STATUS_EIPV (1ULL<<1) /* ip points to correct instruction */
25 #define MCG_STATUS_MCIP (1ULL<<2) /* machine check in progress */
26 #define MCG_STATUS_LMCES (1ULL<<3) /* LMCE signaled */
28 /* MCG_EXT_CTL register defines */
29 #define MCG_EXT_CTL_LMCE_EN (1ULL<<0) /* Enable LMCE */
31 /* MCi_STATUS register defines */
32 #define MCI_STATUS_VAL (1ULL<<63) /* valid error */
33 #define MCI_STATUS_OVER (1ULL<<62) /* previous errors lost */
34 #define MCI_STATUS_UC (1ULL<<61) /* uncorrected error */
35 #define MCI_STATUS_EN (1ULL<<60) /* error enabled */
36 #define MCI_STATUS_MISCV (1ULL<<59) /* misc error reg. valid */
37 #define MCI_STATUS_ADDRV (1ULL<<58) /* addr reg. valid */
38 #define MCI_STATUS_PCC (1ULL<<57) /* processor context corrupt */
39 #define MCI_STATUS_S (1ULL<<56) /* Signaled machine check */
40 #define MCI_STATUS_AR (1ULL<<55) /* Action required */
42 /* AMD-specific bits */
43 #define MCI_STATUS_DEFERRED (1ULL<<44) /* uncorrected error, deferred exception */
44 #define MCI_STATUS_POISON (1ULL<<43) /* access poisonous data */
45 #define MCI_STATUS_TCC (1ULL<<55) /* Task context corrupt */
48 * McaX field if set indicates a given bank supports MCA extensions:
49 * - Deferred error interrupt type is specifiable by bank.
50 * - MCx_MISC0[BlkPtr] field indicates presence of extended MISC registers,
51 * But should not be used to determine MSR numbers.
52 * - TCC bit is present in MCx_STATUS.
54 #define MCI_CONFIG_MCAX 0x1
55 #define MCI_IPID_MCATYPE 0xFFFF0000
56 #define MCI_IPID_HWID 0xFFF
59 * Note that the full MCACOD field of IA32_MCi_STATUS MSR is
60 * bits 15:0. But bit 12 is the 'F' bit, defined for corrected
61 * errors to indicate that errors are being filtered by hardware.
62 * We should mask out bit 12 when looking for specific signatures
63 * of uncorrected errors - so the F bit is deliberately skipped
66 #define MCACOD 0xefff /* MCA Error Code */
68 /* Architecturally defined codes from SDM Vol. 3B Chapter 15 */
69 #define MCACOD_SCRUB 0x00C0 /* 0xC0-0xCF Memory Scrubbing */
70 #define MCACOD_SCRUBMSK 0xeff0 /* Skip bit 12 ('F' bit) */
71 #define MCACOD_L3WB 0x017A /* L3 Explicit Writeback */
72 #define MCACOD_DATA 0x0134 /* Data Load */
73 #define MCACOD_INSTR 0x0150 /* Instruction Fetch */
75 /* MCi_MISC register defines */
76 #define MCI_MISC_ADDR_LSB(m) ((m) & 0x3f)
77 #define MCI_MISC_ADDR_MODE(m) (((m) >> 6) & 7)
78 #define MCI_MISC_ADDR_SEGOFF 0 /* segment offset */
79 #define MCI_MISC_ADDR_LINEAR 1 /* linear address */
80 #define MCI_MISC_ADDR_PHYS 2 /* physical address */
81 #define MCI_MISC_ADDR_MEM 3 /* memory address */
82 #define MCI_MISC_ADDR_GENERIC 7 /* generic */
84 /* CTL2 register defines */
85 #define MCI_CTL2_CMCI_EN (1ULL << 30)
86 #define MCI_CTL2_CMCI_THRESHOLD_MASK 0x7fffULL
88 #define MCJ_CTX_MASK 3
89 #define MCJ_CTX(flags) ((flags) & MCJ_CTX_MASK)
90 #define MCJ_CTX_RANDOM 0 /* inject context: random */
91 #define MCJ_CTX_PROCESS 0x1 /* inject context: process */
92 #define MCJ_CTX_IRQ 0x2 /* inject context: IRQ */
93 #define MCJ_NMI_BROADCAST 0x4 /* do NMI broadcasting */
94 #define MCJ_EXCEPTION 0x8 /* raise as exception */
95 #define MCJ_IRQ_BROADCAST 0x10 /* do IRQ broadcasting */
97 #define MCE_OVERFLOW 0 /* bit 0 in flags means overflow */
99 /* Software defined banks */
100 #define MCE_EXTENDED_BANK 128
101 #define MCE_THERMAL_BANK (MCE_EXTENDED_BANK + 0)
103 #define MCE_LOG_LEN 32
104 #define MCE_LOG_SIGNATURE "MACHINECHECK"
106 /* AMD Scalable MCA */
107 #define MSR_AMD64_SMCA_MC0_CTL 0xc0002000
108 #define MSR_AMD64_SMCA_MC0_STATUS 0xc0002001
109 #define MSR_AMD64_SMCA_MC0_ADDR 0xc0002002
110 #define MSR_AMD64_SMCA_MC0_MISC0 0xc0002003
111 #define MSR_AMD64_SMCA_MC0_CONFIG 0xc0002004
112 #define MSR_AMD64_SMCA_MC0_IPID 0xc0002005
113 #define MSR_AMD64_SMCA_MC0_MISC1 0xc000200a
114 #define MSR_AMD64_SMCA_MCx_CTL(x) (MSR_AMD64_SMCA_MC0_CTL + 0x10*(x))
115 #define MSR_AMD64_SMCA_MCx_STATUS(x) (MSR_AMD64_SMCA_MC0_STATUS + 0x10*(x))
116 #define MSR_AMD64_SMCA_MCx_ADDR(x) (MSR_AMD64_SMCA_MC0_ADDR + 0x10*(x))
117 #define MSR_AMD64_SMCA_MCx_MISC(x) (MSR_AMD64_SMCA_MC0_MISC0 + 0x10*(x))
118 #define MSR_AMD64_SMCA_MCx_CONFIG(x) (MSR_AMD64_SMCA_MC0_CONFIG + 0x10*(x))
119 #define MSR_AMD64_SMCA_MCx_IPID(x) (MSR_AMD64_SMCA_MC0_IPID + 0x10*(x))
120 #define MSR_AMD64_SMCA_MCx_MISCy(x, y) ((MSR_AMD64_SMCA_MC0_MISC1 + y) + (0x10*(x)))
123 * This structure contains all data related to the MCE log. Also
124 * carries a signature to make it easier to find from external
125 * debugging tools. Each entry is only valid when its finished flag
129 char signature[12]; /* "MACHINECHECK" */
130 unsigned len; /* = MCE_LOG_LEN */
133 unsigned recordlen; /* length of struct mce */
134 struct mce entry[MCE_LOG_LEN];
145 bool bios_cmci_threshold;
154 struct mce_vendor_flags {
156 * Indicates that overflow conditions are not fatal, when set.
158 __u64 overflow_recov : 1,
161 * (AMD) SUCCOR stands for S/W UnCorrectable error COntainment and
162 * Recovery. It indicates support for data poisoning in HW and deferred
168 * (AMD) SMCA: This bit indicates support for Scalable MCA which expands
169 * the register space for each MCA bank and also increases number of
170 * banks. Also, to accommodate the new banks and registers, the MCA
171 * register space is moved to a new MSR range.
178 struct mca_msr_regs {
179 u32 (*ctl) (int bank);
180 u32 (*status) (int bank);
181 u32 (*addr) (int bank);
182 u32 (*misc) (int bank);
185 extern struct mce_vendor_flags mce_flags;
187 extern struct mca_config mca_cfg;
188 extern struct mca_msr_regs msr_ops;
189 extern void mce_register_decode_chain(struct notifier_block *nb);
190 extern void mce_unregister_decode_chain(struct notifier_block *nb);
192 #include <linux/percpu.h>
193 #include <linux/atomic.h>
195 extern int mce_p5_enabled;
197 #ifdef CONFIG_X86_MCE
198 int mcheck_init(void);
199 void mcheck_cpu_init(struct cpuinfo_x86 *c);
200 void mcheck_cpu_clear(struct cpuinfo_x86 *c);
201 void mcheck_vendor_init_severity(void);
203 static inline int mcheck_init(void) { return 0; }
204 static inline void mcheck_cpu_init(struct cpuinfo_x86 *c) {}
205 static inline void mcheck_cpu_clear(struct cpuinfo_x86 *c) {}
206 static inline void mcheck_vendor_init_severity(void) {}
209 #ifdef CONFIG_X86_ANCIENT_MCE
210 void intel_p5_mcheck_init(struct cpuinfo_x86 *c);
211 void winchip_mcheck_init(struct cpuinfo_x86 *c);
212 static inline void enable_p5_mce(void) { mce_p5_enabled = 1; }
214 static inline void intel_p5_mcheck_init(struct cpuinfo_x86 *c) {}
215 static inline void winchip_mcheck_init(struct cpuinfo_x86 *c) {}
216 static inline void enable_p5_mce(void) {}
219 void mce_setup(struct mce *m);
220 void mce_log(struct mce *m);
221 DECLARE_PER_CPU(struct device *, mce_device);
224 * Maximum banks number.
225 * This is the limit of the current register layout on
228 #define MAX_NR_BANKS 32
230 #ifdef CONFIG_X86_MCE_INTEL
231 void mce_intel_feature_init(struct cpuinfo_x86 *c);
232 void mce_intel_feature_clear(struct cpuinfo_x86 *c);
233 void cmci_clear(void);
234 void cmci_reenable(void);
235 void cmci_rediscover(void);
236 void cmci_recheck(void);
238 static inline void mce_intel_feature_init(struct cpuinfo_x86 *c) { }
239 static inline void mce_intel_feature_clear(struct cpuinfo_x86 *c) { }
240 static inline void cmci_clear(void) {}
241 static inline void cmci_reenable(void) {}
242 static inline void cmci_rediscover(void) {}
243 static inline void cmci_recheck(void) {}
246 #ifdef CONFIG_X86_MCE_AMD
247 void mce_amd_feature_init(struct cpuinfo_x86 *c);
249 static inline void mce_amd_feature_init(struct cpuinfo_x86 *c) { }
252 int mce_available(struct cpuinfo_x86 *c);
254 DECLARE_PER_CPU(unsigned, mce_exception_count);
255 DECLARE_PER_CPU(unsigned, mce_poll_count);
257 typedef DECLARE_BITMAP(mce_banks_t, MAX_NR_BANKS);
258 DECLARE_PER_CPU(mce_banks_t, mce_poll_banks);
261 MCP_TIMESTAMP = BIT(0), /* log time stamp */
262 MCP_UC = BIT(1), /* log uncorrected errors */
263 MCP_DONTLOG = BIT(2), /* only clear, don't log */
265 bool machine_check_poll(enum mcp_flags flags, mce_banks_t *b);
267 int mce_notify_irq(void);
269 DECLARE_PER_CPU(struct mce, injectm);
271 extern void register_mce_write_callback(ssize_t (*)(struct file *filp,
272 const char __user *ubuf,
273 size_t usize, loff_t *off));
275 /* Disable CMCI/polling for MCA bank claimed by firmware */
276 extern void mce_disable_bank(int bank);
282 /* Call the installed machine check handler for this CPU setup. */
283 extern void (*machine_check_vector)(struct pt_regs *, long error_code);
284 void do_machine_check(struct pt_regs *, long);
290 extern void (*mce_threshold_vector)(void);
291 extern void (*threshold_cpu_callback)(unsigned long action, unsigned int cpu);
293 /* Deferred error interrupt handler */
294 extern void (*deferred_error_int_vector)(void);
300 void intel_init_thermal(struct cpuinfo_x86 *c);
302 void mce_log_therm_throt_event(__u64 status);
304 /* Interrupt Handler for core thermal thresholds */
305 extern int (*platform_thermal_notify)(__u64 msr_val);
307 /* Interrupt Handler for package thermal thresholds */
308 extern int (*platform_thermal_package_notify)(__u64 msr_val);
310 /* Callback support of rate control, return true, if
311 * callback has rate control */
312 extern bool (*platform_thermal_package_rate_control)(void);
314 #ifdef CONFIG_X86_THERMAL_VECTOR
315 extern void mcheck_intel_therm_init(void);
317 static inline void mcheck_intel_therm_init(void) { }
321 * Used by APEI to report memory error via /dev/mcelog
324 struct cper_sec_mem_err;
325 extern void apei_mce_report_mem_error(int corrected,
326 struct cper_sec_mem_err *mem_err);
329 * Enumerate new IP types and HWID values in AMD processors which support
332 #ifdef CONFIG_X86_MCE_AMD
334 SMCA_F17H_CORE = 0, /* Core errors */
335 SMCA_DF, /* Data Fabric */
336 SMCA_UMC, /* Unified Memory Controller */
337 SMCA_PB, /* Parameter Block */
338 SMCA_PSP, /* Platform Security Processor */
339 SMCA_SMU, /* System Management Unit */
348 extern struct amd_hwid amd_hwids[N_AMD_IP_TYPES];
350 enum amd_core_mca_blocks {
351 SMCA_LS = 0, /* Load Store */
352 SMCA_IF, /* Instruction Fetch */
353 SMCA_L2_CACHE, /* L2 cache */
354 SMCA_DE, /* Decoder unit */
356 SMCA_EX, /* Execution unit */
357 SMCA_FP, /* Floating Point */
358 SMCA_L3_CACHE, /* L3 cache */
362 extern const char * const amd_core_mcablock_names[N_CORE_MCA_BLOCKS];
364 enum amd_df_mca_blocks {
365 SMCA_CS = 0, /* Coherent Slave */
366 SMCA_PIE, /* Power management, Interrupts, etc */
370 extern const char * const amd_df_mcablock_names[N_DF_BLOCKS];
373 #endif /* _ASM_X86_MCE_H */