xref: /openbmc/linux/arch/x86/include/asm/mce.h (revision e9c2a283)
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 #define MCI_STATUS_MSCOD(m)	(((m) >> 16) & 0xffff)
46 
47 /* AMD-specific bits */
48 #define MCI_STATUS_TCC		BIT_ULL(55)  /* Task context corrupt */
49 #define MCI_STATUS_SYNDV	BIT_ULL(53)  /* synd reg. valid */
50 #define MCI_STATUS_DEFERRED	BIT_ULL(44)  /* uncorrected error, deferred exception */
51 #define MCI_STATUS_POISON	BIT_ULL(43)  /* access poisonous data */
52 #define MCI_STATUS_SCRUB	BIT_ULL(40)  /* Error detected during scrub operation */
53 
54 /*
55  * McaX field if set indicates a given bank supports MCA extensions:
56  *  - Deferred error interrupt type is specifiable by bank.
57  *  - MCx_MISC0[BlkPtr] field indicates presence of extended MISC registers,
58  *    But should not be used to determine MSR numbers.
59  *  - TCC bit is present in MCx_STATUS.
60  */
61 #define MCI_CONFIG_MCAX		0x1
62 #define MCI_IPID_MCATYPE	0xFFFF0000
63 #define MCI_IPID_HWID		0xFFF
64 
65 /*
66  * Note that the full MCACOD field of IA32_MCi_STATUS MSR is
67  * bits 15:0.  But bit 12 is the 'F' bit, defined for corrected
68  * errors to indicate that errors are being filtered by hardware.
69  * We should mask out bit 12 when looking for specific signatures
70  * of uncorrected errors - so the F bit is deliberately skipped
71  * in this #define.
72  */
73 #define MCACOD		  0xefff     /* MCA Error Code */
74 
75 /* Architecturally defined codes from SDM Vol. 3B Chapter 15 */
76 #define MCACOD_SCRUB	0x00C0	/* 0xC0-0xCF Memory Scrubbing */
77 #define MCACOD_SCRUBMSK	0xeff0	/* Skip bit 12 ('F' bit) */
78 #define MCACOD_L3WB	0x017A	/* L3 Explicit Writeback */
79 #define MCACOD_DATA	0x0134	/* Data Load */
80 #define MCACOD_INSTR	0x0150	/* Instruction Fetch */
81 
82 /* MCi_MISC register defines */
83 #define MCI_MISC_ADDR_LSB(m)	((m) & 0x3f)
84 #define MCI_MISC_ADDR_MODE(m)	(((m) >> 6) & 7)
85 #define  MCI_MISC_ADDR_SEGOFF	0	/* segment offset */
86 #define  MCI_MISC_ADDR_LINEAR	1	/* linear address */
87 #define  MCI_MISC_ADDR_PHYS	2	/* physical address */
88 #define  MCI_MISC_ADDR_MEM	3	/* memory address */
89 #define  MCI_MISC_ADDR_GENERIC	7	/* generic */
90 
91 /* MCi_ADDR register defines */
92 #define MCI_ADDR_PHYSADDR	GENMASK_ULL(boot_cpu_data.x86_phys_bits - 1, 0)
93 
94 /* CTL2 register defines */
95 #define MCI_CTL2_CMCI_EN		BIT_ULL(30)
96 #define MCI_CTL2_CMCI_THRESHOLD_MASK	0x7fffULL
97 
98 #define MCJ_CTX_MASK		3
99 #define MCJ_CTX(flags)		((flags) & MCJ_CTX_MASK)
100 #define MCJ_CTX_RANDOM		0    /* inject context: random */
101 #define MCJ_CTX_PROCESS		0x1  /* inject context: process */
102 #define MCJ_CTX_IRQ		0x2  /* inject context: IRQ */
103 #define MCJ_NMI_BROADCAST	0x4  /* do NMI broadcasting */
104 #define MCJ_EXCEPTION		0x8  /* raise as exception */
105 #define MCJ_IRQ_BROADCAST	0x10 /* do IRQ broadcasting */
106 
107 #define MCE_OVERFLOW 0		/* bit 0 in flags means overflow */
108 
109 #define MCE_LOG_MIN_LEN 32U
110 #define MCE_LOG_SIGNATURE	"MACHINECHECK"
111 
112 /* AMD Scalable MCA */
113 #define MSR_AMD64_SMCA_MC0_CTL		0xc0002000
114 #define MSR_AMD64_SMCA_MC0_STATUS	0xc0002001
115 #define MSR_AMD64_SMCA_MC0_ADDR		0xc0002002
116 #define MSR_AMD64_SMCA_MC0_MISC0	0xc0002003
117 #define MSR_AMD64_SMCA_MC0_CONFIG	0xc0002004
118 #define MSR_AMD64_SMCA_MC0_IPID		0xc0002005
119 #define MSR_AMD64_SMCA_MC0_SYND		0xc0002006
120 #define MSR_AMD64_SMCA_MC0_DESTAT	0xc0002008
121 #define MSR_AMD64_SMCA_MC0_DEADDR	0xc0002009
122 #define MSR_AMD64_SMCA_MC0_MISC1	0xc000200a
123 #define MSR_AMD64_SMCA_MCx_CTL(x)	(MSR_AMD64_SMCA_MC0_CTL + 0x10*(x))
124 #define MSR_AMD64_SMCA_MCx_STATUS(x)	(MSR_AMD64_SMCA_MC0_STATUS + 0x10*(x))
125 #define MSR_AMD64_SMCA_MCx_ADDR(x)	(MSR_AMD64_SMCA_MC0_ADDR + 0x10*(x))
126 #define MSR_AMD64_SMCA_MCx_MISC(x)	(MSR_AMD64_SMCA_MC0_MISC0 + 0x10*(x))
127 #define MSR_AMD64_SMCA_MCx_CONFIG(x)	(MSR_AMD64_SMCA_MC0_CONFIG + 0x10*(x))
128 #define MSR_AMD64_SMCA_MCx_IPID(x)	(MSR_AMD64_SMCA_MC0_IPID + 0x10*(x))
129 #define MSR_AMD64_SMCA_MCx_SYND(x)	(MSR_AMD64_SMCA_MC0_SYND + 0x10*(x))
130 #define MSR_AMD64_SMCA_MCx_DESTAT(x)	(MSR_AMD64_SMCA_MC0_DESTAT + 0x10*(x))
131 #define MSR_AMD64_SMCA_MCx_DEADDR(x)	(MSR_AMD64_SMCA_MC0_DEADDR + 0x10*(x))
132 #define MSR_AMD64_SMCA_MCx_MISCy(x, y)	((MSR_AMD64_SMCA_MC0_MISC1 + y) + (0x10*(x)))
133 
134 #define XEC(x, mask)			(((x) >> 16) & mask)
135 
136 /* mce.kflags flag bits for logging etc. */
137 #define	MCE_HANDLED_CEC		BIT_ULL(0)
138 #define	MCE_HANDLED_UC		BIT_ULL(1)
139 #define	MCE_HANDLED_EXTLOG	BIT_ULL(2)
140 #define	MCE_HANDLED_NFIT	BIT_ULL(3)
141 #define	MCE_HANDLED_EDAC	BIT_ULL(4)
142 #define	MCE_HANDLED_MCELOG	BIT_ULL(5)
143 
144 /*
145  * Indicates an MCE which has happened in kernel space but from
146  * which the kernel can recover simply by executing fixup_exception()
147  * so that an error is returned to the caller of the function that
148  * hit the machine check.
149  */
150 #define MCE_IN_KERNEL_RECOV	BIT_ULL(6)
151 
152 /*
153  * Indicates an MCE that happened in kernel space while copying data
154  * from user. In this case fixup_exception() gets the kernel to the
155  * error exit for the copy function. Machine check handler can then
156  * treat it like a fault taken in user mode.
157  */
158 #define MCE_IN_KERNEL_COPYIN	BIT_ULL(7)
159 
160 /*
161  * This structure contains all data related to the MCE log.  Also
162  * carries a signature to make it easier to find from external
163  * debugging tools.  Each entry is only valid when its finished flag
164  * is set.
165  */
166 struct mce_log_buffer {
167 	char signature[12]; /* "MACHINECHECK" */
168 	unsigned len;	    /* = elements in .mce_entry[] */
169 	unsigned next;
170 	unsigned flags;
171 	unsigned recordlen;	/* length of struct mce */
172 	struct mce entry[];
173 };
174 
175 /* Highest last */
176 enum mce_notifier_prios {
177 	MCE_PRIO_LOWEST,
178 	MCE_PRIO_MCELOG,
179 	MCE_PRIO_EDAC,
180 	MCE_PRIO_NFIT,
181 	MCE_PRIO_EXTLOG,
182 	MCE_PRIO_UC,
183 	MCE_PRIO_EARLY,
184 	MCE_PRIO_CEC,
185 	MCE_PRIO_HIGHEST = MCE_PRIO_CEC
186 };
187 
188 struct notifier_block;
189 extern void mce_register_decode_chain(struct notifier_block *nb);
190 extern void mce_unregister_decode_chain(struct notifier_block *nb);
191 
192 #include <linux/percpu.h>
193 #include <linux/atomic.h>
194 
195 extern int mce_p5_enabled;
196 
197 #ifdef CONFIG_ARCH_HAS_COPY_MC
198 extern void enable_copy_mc_fragile(void);
199 unsigned long __must_check copy_mc_fragile(void *dst, const void *src, unsigned cnt);
200 #else
enable_copy_mc_fragile(void)201 static inline void enable_copy_mc_fragile(void)
202 {
203 }
204 #endif
205 
206 struct cper_ia_proc_ctx;
207 
208 #ifdef CONFIG_X86_MCE
209 int mcheck_init(void);
210 void mcheck_cpu_init(struct cpuinfo_x86 *c);
211 void mcheck_cpu_clear(struct cpuinfo_x86 *c);
212 int apei_smca_report_x86_error(struct cper_ia_proc_ctx *ctx_info,
213 			       u64 lapic_id);
214 #else
mcheck_init(void)215 static inline int mcheck_init(void) { return 0; }
mcheck_cpu_init(struct cpuinfo_x86 * c)216 static inline void mcheck_cpu_init(struct cpuinfo_x86 *c) {}
mcheck_cpu_clear(struct cpuinfo_x86 * c)217 static inline void mcheck_cpu_clear(struct cpuinfo_x86 *c) {}
apei_smca_report_x86_error(struct cper_ia_proc_ctx * ctx_info,u64 lapic_id)218 static inline int apei_smca_report_x86_error(struct cper_ia_proc_ctx *ctx_info,
219 					     u64 lapic_id) { return -EINVAL; }
220 #endif
221 
222 void mce_setup(struct mce *m);
223 void mce_log(struct mce *m);
224 DECLARE_PER_CPU(struct device *, mce_device);
225 
226 /* Maximum number of MCA banks per CPU. */
227 #define MAX_NR_BANKS 64
228 
229 #ifdef CONFIG_X86_MCE_INTEL
230 void mce_intel_feature_init(struct cpuinfo_x86 *c);
231 void mce_intel_feature_clear(struct cpuinfo_x86 *c);
232 void cmci_clear(void);
233 void cmci_reenable(void);
234 void cmci_rediscover(void);
235 void cmci_recheck(void);
236 #else
mce_intel_feature_init(struct cpuinfo_x86 * c)237 static inline void mce_intel_feature_init(struct cpuinfo_x86 *c) { }
mce_intel_feature_clear(struct cpuinfo_x86 * c)238 static inline void mce_intel_feature_clear(struct cpuinfo_x86 *c) { }
cmci_clear(void)239 static inline void cmci_clear(void) {}
cmci_reenable(void)240 static inline void cmci_reenable(void) {}
cmci_rediscover(void)241 static inline void cmci_rediscover(void) {}
cmci_recheck(void)242 static inline void cmci_recheck(void) {}
243 #endif
244 
245 int mce_available(struct cpuinfo_x86 *c);
246 bool mce_is_memory_error(struct mce *m);
247 bool mce_is_correctable(struct mce *m);
248 int mce_usable_address(struct mce *m);
249 
250 DECLARE_PER_CPU(unsigned, mce_exception_count);
251 DECLARE_PER_CPU(unsigned, mce_poll_count);
252 
253 typedef DECLARE_BITMAP(mce_banks_t, MAX_NR_BANKS);
254 DECLARE_PER_CPU(mce_banks_t, mce_poll_banks);
255 
256 enum mcp_flags {
257 	MCP_TIMESTAMP	= BIT(0),	/* log time stamp */
258 	MCP_UC		= BIT(1),	/* log uncorrected errors */
259 	MCP_DONTLOG	= BIT(2),	/* only clear, don't log */
260 	MCP_QUEUE_LOG	= BIT(3),	/* only queue to genpool */
261 };
262 bool machine_check_poll(enum mcp_flags flags, mce_banks_t *b);
263 
264 int mce_notify_irq(void);
265 
266 DECLARE_PER_CPU(struct mce, injectm);
267 
268 /* Disable CMCI/polling for MCA bank claimed by firmware */
269 extern void mce_disable_bank(int bank);
270 
271 /*
272  * Exception handler
273  */
274 void do_machine_check(struct pt_regs *pt_regs);
275 
276 /*
277  * Threshold handler
278  */
279 extern void (*mce_threshold_vector)(void);
280 
281 /* Deferred error interrupt handler */
282 extern void (*deferred_error_int_vector)(void);
283 
284 /*
285  * Used by APEI to report memory error via /dev/mcelog
286  */
287 
288 struct cper_sec_mem_err;
289 extern void apei_mce_report_mem_error(int corrected,
290 				      struct cper_sec_mem_err *mem_err);
291 
292 /*
293  * Enumerate new IP types and HWID values in AMD processors which support
294  * Scalable MCA.
295  */
296 #ifdef CONFIG_X86_MCE_AMD
297 
298 /* These may be used by multiple smca_hwid_mcatypes */
299 enum smca_bank_types {
300 	SMCA_LS = 0,	/* Load Store */
301 	SMCA_LS_V2,
302 	SMCA_IF,	/* Instruction Fetch */
303 	SMCA_L2_CACHE,	/* L2 Cache */
304 	SMCA_DE,	/* Decoder Unit */
305 	SMCA_RESERVED,	/* Reserved */
306 	SMCA_EX,	/* Execution Unit */
307 	SMCA_FP,	/* Floating Point */
308 	SMCA_L3_CACHE,	/* L3 Cache */
309 	SMCA_CS,	/* Coherent Slave */
310 	SMCA_CS_V2,
311 	SMCA_PIE,	/* Power, Interrupts, etc. */
312 	SMCA_UMC,	/* Unified Memory Controller */
313 	SMCA_UMC_V2,
314 	SMCA_PB,	/* Parameter Block */
315 	SMCA_PSP,	/* Platform Security Processor */
316 	SMCA_PSP_V2,
317 	SMCA_SMU,	/* System Management Unit */
318 	SMCA_SMU_V2,
319 	SMCA_MP5,	/* Microprocessor 5 Unit */
320 	SMCA_MPDMA,	/* MPDMA Unit */
321 	SMCA_NBIO,	/* Northbridge IO Unit */
322 	SMCA_PCIE,	/* PCI Express Unit */
323 	SMCA_PCIE_V2,
324 	SMCA_XGMI_PCS,	/* xGMI PCS Unit */
325 	SMCA_NBIF,	/* NBIF Unit */
326 	SMCA_SHUB,	/* System HUB Unit */
327 	SMCA_SATA,	/* SATA Unit */
328 	SMCA_USB,	/* USB Unit */
329 	SMCA_GMI_PCS,	/* GMI PCS Unit */
330 	SMCA_XGMI_PHY,	/* xGMI PHY Unit */
331 	SMCA_WAFL_PHY,	/* WAFL PHY Unit */
332 	SMCA_GMI_PHY,	/* GMI PHY Unit */
333 	N_SMCA_BANK_TYPES
334 };
335 
336 extern const char *smca_get_long_name(enum smca_bank_types t);
337 extern bool amd_mce_is_memory_error(struct mce *m);
338 
339 extern int mce_threshold_create_device(unsigned int cpu);
340 extern int mce_threshold_remove_device(unsigned int cpu);
341 
342 void mce_amd_feature_init(struct cpuinfo_x86 *c);
343 enum smca_bank_types smca_get_bank_type(unsigned int cpu, unsigned int bank);
344 #else
345 
mce_threshold_create_device(unsigned int cpu)346 static inline int mce_threshold_create_device(unsigned int cpu)		{ return 0; };
mce_threshold_remove_device(unsigned int cpu)347 static inline int mce_threshold_remove_device(unsigned int cpu)		{ return 0; };
amd_mce_is_memory_error(struct mce * m)348 static inline bool amd_mce_is_memory_error(struct mce *m)		{ return false; };
mce_amd_feature_init(struct cpuinfo_x86 * c)349 static inline void mce_amd_feature_init(struct cpuinfo_x86 *c)		{ }
350 #endif
351 
mce_hygon_feature_init(struct cpuinfo_x86 * c)352 static inline void mce_hygon_feature_init(struct cpuinfo_x86 *c)	{ return mce_amd_feature_init(c); }
353 
354 unsigned long copy_mc_fragile_handle_tail(char *to, char *from, unsigned len);
355 
356 #endif /* _ASM_X86_MCE_H */
357