xref: /linux/arch/x86/kernel/cpu/mce/internal.h (revision 63307d015b91e626c97bb82e88054af3d0b74643)
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 ATTR_LEN		16
26 #define INITIAL_CHECK_INTERVAL	5 * 60 /* 5 minutes */
27 
28 /* One object for each MCE bank, shared by all CPUs */
29 struct mce_bank {
30 	u64			ctl;			/* subevents to enable */
31 	unsigned char init;				/* initialise bank? */
32 	struct device_attribute attr;			/* device attribute */
33 	char			attrname[ATTR_LEN];	/* attribute name */
34 };
35 
36 struct mce_evt_llist {
37 	struct llist_node llnode;
38 	struct mce mce;
39 };
40 
41 void mce_gen_pool_process(struct work_struct *__unused);
42 bool mce_gen_pool_empty(void);
43 int mce_gen_pool_add(struct mce *mce);
44 int mce_gen_pool_init(void);
45 struct llist_node *mce_gen_pool_prepare_records(void);
46 
47 extern int (*mce_severity)(struct mce *a, int tolerant, char **msg, bool is_excp);
48 struct dentry *mce_get_debugfs_dir(void);
49 
50 extern struct mce_bank *mce_banks;
51 extern mce_banks_t mce_banks_ce_disabled;
52 
53 #ifdef CONFIG_X86_MCE_INTEL
54 unsigned long cmci_intel_adjust_timer(unsigned long interval);
55 bool mce_intel_cmci_poll(void);
56 void mce_intel_hcpu_update(unsigned long cpu);
57 void cmci_disable_bank(int bank);
58 #else
59 # define cmci_intel_adjust_timer mce_adjust_timer_default
60 static inline bool mce_intel_cmci_poll(void) { return false; }
61 static inline void mce_intel_hcpu_update(unsigned long cpu) { }
62 static inline void cmci_disable_bank(int bank) { }
63 #endif
64 
65 void mce_timer_kick(unsigned long interval);
66 
67 #ifdef CONFIG_ACPI_APEI
68 int apei_write_mce(struct mce *m);
69 ssize_t apei_read_mce(struct mce *m, u64 *record_id);
70 int apei_check_mce(void);
71 int apei_clear_mce(u64 record_id);
72 #else
73 static inline int apei_write_mce(struct mce *m)
74 {
75 	return -EINVAL;
76 }
77 static inline ssize_t apei_read_mce(struct mce *m, u64 *record_id)
78 {
79 	return 0;
80 }
81 static inline int apei_check_mce(void)
82 {
83 	return 0;
84 }
85 static inline int apei_clear_mce(u64 record_id)
86 {
87 	return -EINVAL;
88 }
89 #endif
90 
91 void mce_inject_log(struct mce *m);
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  */
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
114 static inline void mce_work_trigger(void)	{ }
115 static inline void mce_register_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 
124 	__u64 lmce_disabled		: 1,
125 	      disabled			: 1,
126 	      ser			: 1,
127 	      recovery			: 1,
128 	      bios_cmci_threshold	: 1,
129 	      __reserved		: 59;
130 
131 	u8 banks;
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 
141 struct mce_vendor_flags {
142 	/*
143 	 * Indicates that overflow conditions are not fatal, when set.
144 	 */
145 	__u64 overflow_recov	: 1,
146 
147 	/*
148 	 * (AMD) SUCCOR stands for S/W UnCorrectable error COntainment and
149 	 * Recovery. It indicates support for data poisoning in HW and deferred
150 	 * error interrupts.
151 	 */
152 	      succor		: 1,
153 
154 	/*
155 	 * (AMD) SMCA: This bit indicates support for Scalable MCA which expands
156 	 * the register space for each MCA bank and also increases number of
157 	 * banks. Also, to accommodate the new banks and registers, the MCA
158 	 * register space is moved to a new MSR range.
159 	 */
160 	      smca		: 1,
161 
162 	      __reserved_0	: 61;
163 };
164 
165 extern struct mce_vendor_flags mce_flags;
166 
167 struct mca_msr_regs {
168 	u32 (*ctl)	(int bank);
169 	u32 (*status)	(int bank);
170 	u32 (*addr)	(int bank);
171 	u32 (*misc)	(int bank);
172 };
173 
174 extern struct mca_msr_regs msr_ops;
175 
176 /* Decide whether to add MCE record to MCE event pool or filter it out. */
177 extern bool filter_mce(struct mce *m);
178 
179 #ifdef CONFIG_X86_MCE_AMD
180 extern bool amd_filter_mce(struct mce *m);
181 #else
182 static inline bool amd_filter_mce(struct mce *m)			{ return false; };
183 #endif
184 
185 #endif /* __X86_MCE_INTERNAL_H__ */
186