1 /* SPDX-License-Identifier: GPL-2.0 */
2 #ifndef __X86_MCE_INTERNAL_H__
3 #define __X86_MCE_INTERNAL_H__
4 
5 #undef pr_fmt
6 #define pr_fmt(fmt) "mce: " fmt
7 
8 #include <linux/device.h>
9 #include <asm/mce.h>
10 
11 /* Pointer to the installed machine check handler for this CPU setup. */
12 extern void (*machine_check_vector)(struct pt_regs *);
13 
14 enum severity_level {
15 	MCE_NO_SEVERITY,
16 	MCE_DEFERRED_SEVERITY,
17 	MCE_UCNA_SEVERITY = MCE_DEFERRED_SEVERITY,
18 	MCE_KEEP_SEVERITY,
19 	MCE_SOME_SEVERITY,
20 	MCE_AO_SEVERITY,
21 	MCE_UC_SEVERITY,
22 	MCE_AR_SEVERITY,
23 	MCE_PANIC_SEVERITY,
24 };
25 
26 extern struct blocking_notifier_head x86_mce_decoder_chain;
27 
28 #define INITIAL_CHECK_INTERVAL	5 * 60 /* 5 minutes */
29 
30 struct mce_evt_llist {
31 	struct llist_node llnode;
32 	struct mce mce;
33 };
34 
35 void mce_gen_pool_process(struct work_struct *__unused);
36 bool mce_gen_pool_empty(void);
37 int mce_gen_pool_add(struct mce *mce);
38 int mce_gen_pool_init(void);
39 struct llist_node *mce_gen_pool_prepare_records(void);
40 
41 extern int (*mce_severity)(struct mce *a, struct pt_regs *regs,
42 			   int tolerant, char **msg, bool is_excp);
43 struct dentry *mce_get_debugfs_dir(void);
44 
45 extern mce_banks_t mce_banks_ce_disabled;
46 
47 #ifdef CONFIG_X86_MCE_INTEL
48 unsigned long cmci_intel_adjust_timer(unsigned long interval);
49 bool mce_intel_cmci_poll(void);
50 void mce_intel_hcpu_update(unsigned long cpu);
51 void cmci_disable_bank(int bank);
52 void intel_init_cmci(void);
53 void intel_init_lmce(void);
54 void intel_clear_lmce(void);
55 bool intel_filter_mce(struct mce *m);
56 #else
57 # define cmci_intel_adjust_timer mce_adjust_timer_default
mce_intel_cmci_poll(void)58 static inline bool mce_intel_cmci_poll(void) { return false; }
mce_intel_hcpu_update(unsigned long cpu)59 static inline void mce_intel_hcpu_update(unsigned long cpu) { }
cmci_disable_bank(int bank)60 static inline void cmci_disable_bank(int bank) { }
intel_init_cmci(void)61 static inline void intel_init_cmci(void) { }
intel_init_lmce(void)62 static inline void intel_init_lmce(void) { }
intel_clear_lmce(void)63 static inline void intel_clear_lmce(void) { }
intel_filter_mce(struct mce * m)64 static inline bool intel_filter_mce(struct mce *m) { return false; };
65 #endif
66 
67 void mce_timer_kick(unsigned long interval);
68 
69 #ifdef CONFIG_ACPI_APEI
70 int apei_write_mce(struct mce *m);
71 ssize_t apei_read_mce(struct mce *m, u64 *record_id);
72 int apei_check_mce(void);
73 int apei_clear_mce(u64 record_id);
74 #else
apei_write_mce(struct mce * m)75 static inline int apei_write_mce(struct mce *m)
76 {
77 	return -EINVAL;
78 }
apei_read_mce(struct mce * m,u64 * record_id)79 static inline ssize_t apei_read_mce(struct mce *m, u64 *record_id)
80 {
81 	return 0;
82 }
apei_check_mce(void)83 static inline int apei_check_mce(void)
84 {
85 	return 0;
86 }
apei_clear_mce(u64 record_id)87 static inline int apei_clear_mce(u64 record_id)
88 {
89 	return -EINVAL;
90 }
91 #endif
92 
93 /*
94  * We consider records to be equivalent if bank+status+addr+misc all match.
95  * This is only used when the system is going down because of a fatal error
96  * to avoid cluttering the console log with essentially repeated information.
97  * In normal processing all errors seen are logged.
98  */
mce_cmp(struct mce * m1,struct mce * m2)99 static inline bool mce_cmp(struct mce *m1, struct mce *m2)
100 {
101 	return m1->bank != m2->bank ||
102 		m1->status != m2->status ||
103 		m1->addr != m2->addr ||
104 		m1->misc != m2->misc;
105 }
106 
107 extern struct device_attribute dev_attr_trigger;
108 
109 #ifdef CONFIG_X86_MCELOG_LEGACY
110 void mce_work_trigger(void);
111 void mce_register_injector_chain(struct notifier_block *nb);
112 void mce_unregister_injector_chain(struct notifier_block *nb);
113 #else
mce_work_trigger(void)114 static inline void mce_work_trigger(void)	{ }
mce_register_injector_chain(struct notifier_block * nb)115 static inline void mce_register_injector_chain(struct notifier_block *nb)	{ }
mce_unregister_injector_chain(struct notifier_block * nb)116 static inline void mce_unregister_injector_chain(struct notifier_block *nb)	{ }
117 #endif
118 
119 struct mca_config {
120 	bool dont_log_ce;
121 	bool cmci_disabled;
122 	bool ignore_ce;
123 	bool print_all;
124 
125 	__u64 lmce_disabled		: 1,
126 	      disabled			: 1,
127 	      ser			: 1,
128 	      recovery			: 1,
129 	      bios_cmci_threshold	: 1,
130 	      __reserved		: 59;
131 
132 	s8 bootlog;
133 	int tolerant;
134 	int monarch_timeout;
135 	int panic_timeout;
136 	u32 rip_msr;
137 };
138 
139 extern struct mca_config mca_cfg;
140 DECLARE_PER_CPU_READ_MOSTLY(unsigned int, mce_num_banks);
141 
142 struct mce_vendor_flags {
143 	/*
144 	 * Indicates that overflow conditions are not fatal, when set.
145 	 */
146 	__u64 overflow_recov	: 1,
147 
148 	/*
149 	 * (AMD) SUCCOR stands for S/W UnCorrectable error COntainment and
150 	 * Recovery. It indicates support for data poisoning in HW and deferred
151 	 * error interrupts.
152 	 */
153 	succor			: 1,
154 
155 	/*
156 	 * (AMD) SMCA: This bit indicates support for Scalable MCA which expands
157 	 * the register space for each MCA bank and also increases number of
158 	 * banks. Also, to accommodate the new banks and registers, the MCA
159 	 * register space is moved to a new MSR range.
160 	 */
161 	smca			: 1,
162 
163 	/* AMD-style error thresholding banks present. */
164 	amd_threshold		: 1,
165 
166 	__reserved_0		: 60;
167 };
168 
169 extern struct mce_vendor_flags mce_flags;
170 
171 struct mca_msr_regs {
172 	u32 (*ctl)	(int bank);
173 	u32 (*status)	(int bank);
174 	u32 (*addr)	(int bank);
175 	u32 (*misc)	(int bank);
176 };
177 
178 extern struct mca_msr_regs msr_ops;
179 
180 /* Decide whether to add MCE record to MCE event pool or filter it out. */
181 extern bool filter_mce(struct mce *m);
182 
183 #ifdef CONFIG_X86_MCE_AMD
184 extern bool amd_filter_mce(struct mce *m);
185 #else
amd_filter_mce(struct mce * m)186 static inline bool amd_filter_mce(struct mce *m)			{ return false; };
187 #endif
188 
189 __visible bool ex_handler_rdmsr_fault(const struct exception_table_entry *fixup,
190 				      struct pt_regs *regs, int trapnr,
191 				      unsigned long error_code,
192 				      unsigned long fault_addr);
193 
194 __visible bool ex_handler_wrmsr_fault(const struct exception_table_entry *fixup,
195 				      struct pt_regs *regs, int trapnr,
196 				      unsigned long error_code,
197 				      unsigned long fault_addr);
198 
199 #endif /* __X86_MCE_INTERNAL_H__ */
200