xref: /openbmc/linux/arch/x86/kernel/cpu/mce/internal.h (revision 7b73a9c8)
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 enum severity_level {
12 	MCE_NO_SEVERITY,
13 	MCE_DEFERRED_SEVERITY,
14 	MCE_UCNA_SEVERITY = MCE_DEFERRED_SEVERITY,
15 	MCE_KEEP_SEVERITY,
16 	MCE_SOME_SEVERITY,
17 	MCE_AO_SEVERITY,
18 	MCE_UC_SEVERITY,
19 	MCE_AR_SEVERITY,
20 	MCE_PANIC_SEVERITY,
21 };
22 
23 extern struct blocking_notifier_head x86_mce_decoder_chain;
24 
25 #define INITIAL_CHECK_INTERVAL	5 * 60 /* 5 minutes */
26 
27 struct mce_evt_llist {
28 	struct llist_node llnode;
29 	struct mce mce;
30 };
31 
32 void mce_gen_pool_process(struct work_struct *__unused);
33 bool mce_gen_pool_empty(void);
34 int mce_gen_pool_add(struct mce *mce);
35 int mce_gen_pool_init(void);
36 struct llist_node *mce_gen_pool_prepare_records(void);
37 
38 extern int (*mce_severity)(struct mce *a, int tolerant, char **msg, bool is_excp);
39 struct dentry *mce_get_debugfs_dir(void);
40 
41 extern mce_banks_t mce_banks_ce_disabled;
42 
43 #ifdef CONFIG_X86_MCE_INTEL
44 unsigned long cmci_intel_adjust_timer(unsigned long interval);
45 bool mce_intel_cmci_poll(void);
46 void mce_intel_hcpu_update(unsigned long cpu);
47 void cmci_disable_bank(int bank);
48 void intel_init_cmci(void);
49 void intel_init_lmce(void);
50 void intel_clear_lmce(void);
51 #else
52 # define cmci_intel_adjust_timer mce_adjust_timer_default
53 static inline bool mce_intel_cmci_poll(void) { return false; }
54 static inline void mce_intel_hcpu_update(unsigned long cpu) { }
55 static inline void cmci_disable_bank(int bank) { }
56 static inline void intel_init_cmci(void) { }
57 static inline void intel_init_lmce(void) { }
58 static inline void intel_clear_lmce(void) { }
59 #endif
60 
61 void mce_timer_kick(unsigned long interval);
62 
63 #ifdef CONFIG_ACPI_APEI
64 int apei_write_mce(struct mce *m);
65 ssize_t apei_read_mce(struct mce *m, u64 *record_id);
66 int apei_check_mce(void);
67 int apei_clear_mce(u64 record_id);
68 #else
69 static inline int apei_write_mce(struct mce *m)
70 {
71 	return -EINVAL;
72 }
73 static inline ssize_t apei_read_mce(struct mce *m, u64 *record_id)
74 {
75 	return 0;
76 }
77 static inline int apei_check_mce(void)
78 {
79 	return 0;
80 }
81 static inline int apei_clear_mce(u64 record_id)
82 {
83 	return -EINVAL;
84 }
85 #endif
86 
87 void mce_inject_log(struct mce *m);
88 
89 /*
90  * We consider records to be equivalent if bank+status+addr+misc all match.
91  * This is only used when the system is going down because of a fatal error
92  * to avoid cluttering the console log with essentially repeated information.
93  * In normal processing all errors seen are logged.
94  */
95 static inline bool mce_cmp(struct mce *m1, struct mce *m2)
96 {
97 	return m1->bank != m2->bank ||
98 		m1->status != m2->status ||
99 		m1->addr != m2->addr ||
100 		m1->misc != m2->misc;
101 }
102 
103 extern struct device_attribute dev_attr_trigger;
104 
105 #ifdef CONFIG_X86_MCELOG_LEGACY
106 void mce_work_trigger(void);
107 void mce_register_injector_chain(struct notifier_block *nb);
108 void mce_unregister_injector_chain(struct notifier_block *nb);
109 #else
110 static inline void mce_work_trigger(void)	{ }
111 static inline void mce_register_injector_chain(struct notifier_block *nb)	{ }
112 static inline void mce_unregister_injector_chain(struct notifier_block *nb)	{ }
113 #endif
114 
115 struct mca_config {
116 	bool dont_log_ce;
117 	bool cmci_disabled;
118 	bool ignore_ce;
119 
120 	__u64 lmce_disabled		: 1,
121 	      disabled			: 1,
122 	      ser			: 1,
123 	      recovery			: 1,
124 	      bios_cmci_threshold	: 1,
125 	      __reserved		: 59;
126 
127 	s8 bootlog;
128 	int tolerant;
129 	int monarch_timeout;
130 	int panic_timeout;
131 	u32 rip_msr;
132 };
133 
134 extern struct mca_config mca_cfg;
135 DECLARE_PER_CPU_READ_MOSTLY(unsigned int, mce_num_banks);
136 
137 struct mce_vendor_flags {
138 	/*
139 	 * Indicates that overflow conditions are not fatal, when set.
140 	 */
141 	__u64 overflow_recov	: 1,
142 
143 	/*
144 	 * (AMD) SUCCOR stands for S/W UnCorrectable error COntainment and
145 	 * Recovery. It indicates support for data poisoning in HW and deferred
146 	 * error interrupts.
147 	 */
148 	      succor		: 1,
149 
150 	/*
151 	 * (AMD) SMCA: This bit indicates support for Scalable MCA which expands
152 	 * the register space for each MCA bank and also increases number of
153 	 * banks. Also, to accommodate the new banks and registers, the MCA
154 	 * register space is moved to a new MSR range.
155 	 */
156 	      smca		: 1,
157 
158 	      __reserved_0	: 61;
159 };
160 
161 extern struct mce_vendor_flags mce_flags;
162 
163 struct mca_msr_regs {
164 	u32 (*ctl)	(int bank);
165 	u32 (*status)	(int bank);
166 	u32 (*addr)	(int bank);
167 	u32 (*misc)	(int bank);
168 };
169 
170 extern struct mca_msr_regs msr_ops;
171 
172 /* Decide whether to add MCE record to MCE event pool or filter it out. */
173 extern bool filter_mce(struct mce *m);
174 
175 #ifdef CONFIG_X86_MCE_AMD
176 extern bool amd_filter_mce(struct mce *m);
177 #else
178 static inline bool amd_filter_mce(struct mce *m)			{ return false; };
179 #endif
180 
181 #endif /* __X86_MCE_INTERNAL_H__ */
182