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_MIN_LEN 32U
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 #define XEC(x, mask) (((x) >> 16) & mask)
131
132 /* mce.kflags flag bits for logging etc. */
133 #define MCE_HANDLED_CEC BIT_ULL(0)
134 #define MCE_HANDLED_UC BIT_ULL(1)
135 #define MCE_HANDLED_EXTLOG BIT_ULL(2)
136 #define MCE_HANDLED_NFIT BIT_ULL(3)
137 #define MCE_HANDLED_EDAC BIT_ULL(4)
138 #define MCE_HANDLED_MCELOG BIT_ULL(5)
139
140 /*
141 * Indicates an MCE which has happened in kernel space but from
142 * which the kernel can recover simply by executing fixup_exception()
143 * so that an error is returned to the caller of the function that
144 * hit the machine check.
145 */
146 #define MCE_IN_KERNEL_RECOV BIT_ULL(6)
147
148 /*
149 * Indicates an MCE that happened in kernel space while copying data
150 * from user. In this case fixup_exception() gets the kernel to the
151 * error exit for the copy function. Machine check handler can then
152 * treat it like a fault taken in user mode.
153 */
154 #define MCE_IN_KERNEL_COPYIN BIT_ULL(7)
155
156 /*
157 * This structure contains all data related to the MCE log. Also
158 * carries a signature to make it easier to find from external
159 * debugging tools. Each entry is only valid when its finished flag
160 * is set.
161 */
162 struct mce_log_buffer {
163 char signature[12]; /* "MACHINECHECK" */
164 unsigned len; /* = elements in .mce_entry[] */
165 unsigned next;
166 unsigned flags;
167 unsigned recordlen; /* length of struct mce */
168 struct mce entry[];
169 };
170
171 /* Highest last */
172 enum mce_notifier_prios {
173 MCE_PRIO_LOWEST,
174 MCE_PRIO_MCELOG,
175 MCE_PRIO_EDAC,
176 MCE_PRIO_NFIT,
177 MCE_PRIO_EXTLOG,
178 MCE_PRIO_UC,
179 MCE_PRIO_EARLY,
180 MCE_PRIO_CEC,
181 MCE_PRIO_HIGHEST = MCE_PRIO_CEC
182 };
183
184 struct notifier_block;
185 extern void mce_register_decode_chain(struct notifier_block *nb);
186 extern void mce_unregister_decode_chain(struct notifier_block *nb);
187
188 #include <linux/percpu.h>
189 #include <linux/atomic.h>
190
191 extern int mce_p5_enabled;
192
193 #ifdef CONFIG_ARCH_HAS_COPY_MC
194 extern void enable_copy_mc_fragile(void);
195 unsigned long __must_check copy_mc_fragile(void *dst, const void *src, unsigned cnt);
196 #else
enable_copy_mc_fragile(void)197 static inline void enable_copy_mc_fragile(void)
198 {
199 }
200 #endif
201
202 #ifdef CONFIG_X86_MCE
203 int mcheck_init(void);
204 void mcheck_cpu_init(struct cpuinfo_x86 *c);
205 void mcheck_cpu_clear(struct cpuinfo_x86 *c);
206 void mcheck_vendor_init_severity(void);
207 #else
mcheck_init(void)208 static inline int mcheck_init(void) { return 0; }
mcheck_cpu_init(struct cpuinfo_x86 * c)209 static inline void mcheck_cpu_init(struct cpuinfo_x86 *c) {}
mcheck_cpu_clear(struct cpuinfo_x86 * c)210 static inline void mcheck_cpu_clear(struct cpuinfo_x86 *c) {}
mcheck_vendor_init_severity(void)211 static inline void mcheck_vendor_init_severity(void) {}
212 #endif
213
214 #ifdef CONFIG_X86_ANCIENT_MCE
215 void intel_p5_mcheck_init(struct cpuinfo_x86 *c);
216 void winchip_mcheck_init(struct cpuinfo_x86 *c);
enable_p5_mce(void)217 static inline void enable_p5_mce(void) { mce_p5_enabled = 1; }
218 #else
intel_p5_mcheck_init(struct cpuinfo_x86 * c)219 static inline void intel_p5_mcheck_init(struct cpuinfo_x86 *c) {}
winchip_mcheck_init(struct cpuinfo_x86 * c)220 static inline void winchip_mcheck_init(struct cpuinfo_x86 *c) {}
enable_p5_mce(void)221 static inline void enable_p5_mce(void) {}
222 #endif
223
224 void mce_setup(struct mce *m);
225 void mce_log(struct mce *m);
226 DECLARE_PER_CPU(struct device *, mce_device);
227
228 /* Maximum number of MCA banks per CPU. */
229 #define MAX_NR_BANKS 64
230
231 #ifdef CONFIG_X86_MCE_INTEL
232 void mce_intel_feature_init(struct cpuinfo_x86 *c);
233 void mce_intel_feature_clear(struct cpuinfo_x86 *c);
234 void cmci_clear(void);
235 void cmci_reenable(void);
236 void cmci_rediscover(void);
237 void cmci_recheck(void);
238 #else
mce_intel_feature_init(struct cpuinfo_x86 * c)239 static inline void mce_intel_feature_init(struct cpuinfo_x86 *c) { }
mce_intel_feature_clear(struct cpuinfo_x86 * c)240 static inline void mce_intel_feature_clear(struct cpuinfo_x86 *c) { }
cmci_clear(void)241 static inline void cmci_clear(void) {}
cmci_reenable(void)242 static inline void cmci_reenable(void) {}
cmci_rediscover(void)243 static inline void cmci_rediscover(void) {}
cmci_recheck(void)244 static inline void cmci_recheck(void) {}
245 #endif
246
247 int mce_available(struct cpuinfo_x86 *c);
248 bool mce_is_memory_error(struct mce *m);
249 bool mce_is_correctable(struct mce *m);
250 int mce_usable_address(struct mce *m);
251
252 DECLARE_PER_CPU(unsigned, mce_exception_count);
253 DECLARE_PER_CPU(unsigned, mce_poll_count);
254
255 typedef DECLARE_BITMAP(mce_banks_t, MAX_NR_BANKS);
256 DECLARE_PER_CPU(mce_banks_t, mce_poll_banks);
257
258 enum mcp_flags {
259 MCP_TIMESTAMP = BIT(0), /* log time stamp */
260 MCP_UC = BIT(1), /* log uncorrected errors */
261 MCP_DONTLOG = BIT(2), /* only clear, don't log */
262 MCP_QUEUE_LOG = BIT(3), /* only queue to genpool */
263 };
264 bool machine_check_poll(enum mcp_flags flags, mce_banks_t *b);
265
266 int mce_notify_irq(void);
267
268 DECLARE_PER_CPU(struct mce, injectm);
269
270 /* Disable CMCI/polling for MCA bank claimed by firmware */
271 extern void mce_disable_bank(int bank);
272
273 /*
274 * Exception handler
275 */
276 void do_machine_check(struct pt_regs *pt_regs);
277
278 /*
279 * Threshold handler
280 */
281 extern void (*mce_threshold_vector)(void);
282
283 /* Deferred error interrupt handler */
284 extern void (*deferred_error_int_vector)(void);
285
286 /*
287 * Thermal handler
288 */
289
290 void intel_init_thermal(struct cpuinfo_x86 *c);
291
292 /* Interrupt Handler for core thermal thresholds */
293 extern int (*platform_thermal_notify)(__u64 msr_val);
294
295 /* Interrupt Handler for package thermal thresholds */
296 extern int (*platform_thermal_package_notify)(__u64 msr_val);
297
298 /* Callback support of rate control, return true, if
299 * callback has rate control */
300 extern bool (*platform_thermal_package_rate_control)(void);
301
302 #ifdef CONFIG_X86_THERMAL_VECTOR
303 extern void mcheck_intel_therm_init(void);
304 #else
mcheck_intel_therm_init(void)305 static inline void mcheck_intel_therm_init(void) { }
306 #endif
307
308 /*
309 * Used by APEI to report memory error via /dev/mcelog
310 */
311
312 struct cper_sec_mem_err;
313 extern void apei_mce_report_mem_error(int corrected,
314 struct cper_sec_mem_err *mem_err);
315
316 /*
317 * Enumerate new IP types and HWID values in AMD processors which support
318 * Scalable MCA.
319 */
320 #ifdef CONFIG_X86_MCE_AMD
321
322 /* These may be used by multiple smca_hwid_mcatypes */
323 enum smca_bank_types {
324 SMCA_LS = 0, /* Load Store */
325 SMCA_LS_V2, /* Load Store */
326 SMCA_IF, /* Instruction Fetch */
327 SMCA_L2_CACHE, /* L2 Cache */
328 SMCA_DE, /* Decoder Unit */
329 SMCA_RESERVED, /* Reserved */
330 SMCA_EX, /* Execution Unit */
331 SMCA_FP, /* Floating Point */
332 SMCA_L3_CACHE, /* L3 Cache */
333 SMCA_CS, /* Coherent Slave */
334 SMCA_CS_V2, /* Coherent Slave */
335 SMCA_PIE, /* Power, Interrupts, etc. */
336 SMCA_UMC, /* Unified Memory Controller */
337 SMCA_PB, /* Parameter Block */
338 SMCA_PSP, /* Platform Security Processor */
339 SMCA_PSP_V2, /* Platform Security Processor */
340 SMCA_SMU, /* System Management Unit */
341 SMCA_SMU_V2, /* System Management Unit */
342 SMCA_MP5, /* Microprocessor 5 Unit */
343 SMCA_NBIO, /* Northbridge IO Unit */
344 SMCA_PCIE, /* PCI Express Unit */
345 N_SMCA_BANK_TYPES
346 };
347
348 #define HWID_MCATYPE(hwid, mcatype) (((hwid) << 16) | (mcatype))
349
350 struct smca_hwid {
351 unsigned int bank_type; /* Use with smca_bank_types for easy indexing. */
352 u32 hwid_mcatype; /* (hwid,mcatype) tuple */
353 u8 count; /* Number of instances. */
354 };
355
356 struct smca_bank {
357 struct smca_hwid *hwid;
358 u32 id; /* Value of MCA_IPID[InstanceId]. */
359 u8 sysfs_id; /* Value used for sysfs name. */
360 };
361
362 extern struct smca_bank smca_banks[MAX_NR_BANKS];
363
364 extern const char *smca_get_long_name(enum smca_bank_types t);
365 extern bool amd_mce_is_memory_error(struct mce *m);
366
367 extern int mce_threshold_create_device(unsigned int cpu);
368 extern int mce_threshold_remove_device(unsigned int cpu);
369
370 void mce_amd_feature_init(struct cpuinfo_x86 *c);
371 int umc_normaddr_to_sysaddr(u64 norm_addr, u16 nid, u8 umc, u64 *sys_addr);
372
373 #else
374
mce_threshold_create_device(unsigned int cpu)375 static inline int mce_threshold_create_device(unsigned int cpu) { return 0; };
mce_threshold_remove_device(unsigned int cpu)376 static inline int mce_threshold_remove_device(unsigned int cpu) { return 0; };
amd_mce_is_memory_error(struct mce * m)377 static inline bool amd_mce_is_memory_error(struct mce *m) { return false; };
mce_amd_feature_init(struct cpuinfo_x86 * c)378 static inline void mce_amd_feature_init(struct cpuinfo_x86 *c) { }
379 static inline int
umc_normaddr_to_sysaddr(u64 norm_addr,u16 nid,u8 umc,u64 * sys_addr)380 umc_normaddr_to_sysaddr(u64 norm_addr, u16 nid, u8 umc, u64 *sys_addr) { return -EINVAL; };
381 #endif
382
mce_hygon_feature_init(struct cpuinfo_x86 * c)383 static inline void mce_hygon_feature_init(struct cpuinfo_x86 *c) { return mce_amd_feature_init(c); }
384 #endif /* _ASM_X86_MCE_H */
385