xref: /openbmc/linux/arch/x86/kernel/cpu/mce/internal.h (revision e149ca29)
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 *, long error_code);
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, int tolerant, char **msg, bool is_excp);
42 struct dentry *mce_get_debugfs_dir(void);
43 
44 extern mce_banks_t mce_banks_ce_disabled;
45 
46 #ifdef CONFIG_X86_MCE_INTEL
47 unsigned long cmci_intel_adjust_timer(unsigned long interval);
48 bool mce_intel_cmci_poll(void);
49 void mce_intel_hcpu_update(unsigned long cpu);
50 void cmci_disable_bank(int bank);
51 void intel_init_cmci(void);
52 void intel_init_lmce(void);
53 void intel_clear_lmce(void);
54 bool intel_filter_mce(struct mce *m);
55 #else
56 # define cmci_intel_adjust_timer mce_adjust_timer_default
57 static inline bool mce_intel_cmci_poll(void) { return false; }
58 static inline void mce_intel_hcpu_update(unsigned long cpu) { }
59 static inline void cmci_disable_bank(int bank) { }
60 static inline void intel_init_cmci(void) { }
61 static inline void intel_init_lmce(void) { }
62 static inline void intel_clear_lmce(void) { }
63 static inline bool intel_filter_mce(struct mce *m) { return false; };
64 #endif
65 
66 void mce_timer_kick(unsigned long interval);
67 
68 #ifdef CONFIG_ACPI_APEI
69 int apei_write_mce(struct mce *m);
70 ssize_t apei_read_mce(struct mce *m, u64 *record_id);
71 int apei_check_mce(void);
72 int apei_clear_mce(u64 record_id);
73 #else
74 static inline int apei_write_mce(struct mce *m)
75 {
76 	return -EINVAL;
77 }
78 static inline ssize_t apei_read_mce(struct mce *m, u64 *record_id)
79 {
80 	return 0;
81 }
82 static inline int apei_check_mce(void)
83 {
84 	return 0;
85 }
86 static inline int apei_clear_mce(u64 record_id)
87 {
88 	return -EINVAL;
89 }
90 #endif
91 
92 /*
93  * We consider records to be equivalent if bank+status+addr+misc all match.
94  * This is only used when the system is going down because of a fatal error
95  * to avoid cluttering the console log with essentially repeated information.
96  * In normal processing all errors seen are logged.
97  */
98 static inline bool mce_cmp(struct mce *m1, struct mce *m2)
99 {
100 	return m1->bank != m2->bank ||
101 		m1->status != m2->status ||
102 		m1->addr != m2->addr ||
103 		m1->misc != m2->misc;
104 }
105 
106 extern struct device_attribute dev_attr_trigger;
107 
108 #ifdef CONFIG_X86_MCELOG_LEGACY
109 void mce_work_trigger(void);
110 void mce_register_injector_chain(struct notifier_block *nb);
111 void mce_unregister_injector_chain(struct notifier_block *nb);
112 #else
113 static inline void mce_work_trigger(void)	{ }
114 static inline void mce_register_injector_chain(struct notifier_block *nb)	{ }
115 static inline void mce_unregister_injector_chain(struct notifier_block *nb)	{ }
116 #endif
117 
118 struct mca_config {
119 	bool dont_log_ce;
120 	bool cmci_disabled;
121 	bool ignore_ce;
122 
123 	__u64 lmce_disabled		: 1,
124 	      disabled			: 1,
125 	      ser			: 1,
126 	      recovery			: 1,
127 	      bios_cmci_threshold	: 1,
128 	      __reserved		: 59;
129 
130 	s8 bootlog;
131 	int tolerant;
132 	int monarch_timeout;
133 	int panic_timeout;
134 	u32 rip_msr;
135 };
136 
137 extern struct mca_config mca_cfg;
138 DECLARE_PER_CPU_READ_MOSTLY(unsigned int, mce_num_banks);
139 
140 struct mce_vendor_flags {
141 	/*
142 	 * Indicates that overflow conditions are not fatal, when set.
143 	 */
144 	__u64 overflow_recov	: 1,
145 
146 	/*
147 	 * (AMD) SUCCOR stands for S/W UnCorrectable error COntainment and
148 	 * Recovery. It indicates support for data poisoning in HW and deferred
149 	 * error interrupts.
150 	 */
151 	      succor		: 1,
152 
153 	/*
154 	 * (AMD) SMCA: This bit indicates support for Scalable MCA which expands
155 	 * the register space for each MCA bank and also increases number of
156 	 * banks. Also, to accommodate the new banks and registers, the MCA
157 	 * register space is moved to a new MSR range.
158 	 */
159 	      smca		: 1,
160 
161 	      __reserved_0	: 61;
162 };
163 
164 extern struct mce_vendor_flags mce_flags;
165 
166 struct mca_msr_regs {
167 	u32 (*ctl)	(int bank);
168 	u32 (*status)	(int bank);
169 	u32 (*addr)	(int bank);
170 	u32 (*misc)	(int bank);
171 };
172 
173 extern struct mca_msr_regs msr_ops;
174 
175 /* Decide whether to add MCE record to MCE event pool or filter it out. */
176 extern bool filter_mce(struct mce *m);
177 
178 #ifdef CONFIG_X86_MCE_AMD
179 extern bool amd_filter_mce(struct mce *m);
180 #else
181 static inline bool amd_filter_mce(struct mce *m)			{ return false; };
182 #endif
183 
184 #endif /* __X86_MCE_INTERNAL_H__ */
185