xref: /openbmc/linux/drivers/acpi/apei/bert.c (revision 0d4bb5e4)
1 // SPDX-License-Identifier: GPL-2.0-only
2 /*
3  * APEI Boot Error Record Table (BERT) support
4  *
5  * Copyright 2011 Intel Corp.
6  *   Author: Huang Ying <ying.huang@intel.com>
7  *
8  * Under normal circumstances, when a hardware error occurs, the error
9  * handler receives control and processes the error. This gives OSPM a
10  * chance to process the error condition, report it, and optionally attempt
11  * recovery. In some cases, the system is unable to process an error.
12  * For example, system firmware or a management controller may choose to
13  * reset the system or the system might experience an uncontrolled crash
14  * or reset.The boot error source is used to report unhandled errors that
15  * occurred in a previous boot. This mechanism is described in the BERT
16  * table.
17  *
18  * For more information about BERT, please refer to ACPI Specification
19  * version 4.0, section 17.3.1
20  */
21 
22 #include <linux/kernel.h>
23 #include <linux/module.h>
24 #include <linux/init.h>
25 #include <linux/acpi.h>
26 #include <linux/io.h>
27 
28 #include "apei-internal.h"
29 
30 #undef pr_fmt
31 #define pr_fmt(fmt) "BERT: " fmt
32 
33 #define ACPI_BERT_PRINT_MAX_RECORDS 5
34 #define ACPI_BERT_PRINT_MAX_LEN 1024
35 
36 static int bert_disable;
37 
38 /*
39  * Print "all" the error records in the BERT table, but avoid huge spam to
40  * the console if the BIOS included oversize records, or too many records.
41  * Skipping some records here does not lose anything because the full
42  * data is available to user tools in:
43  *	/sys/firmware/acpi/tables/data/BERT
44  */
45 static void __init bert_print_all(struct acpi_bert_region *region,
46 				  unsigned int region_len)
47 {
48 	struct acpi_hest_generic_status *estatus =
49 		(struct acpi_hest_generic_status *)region;
50 	int remain = region_len;
51 	int printed = 0, skipped = 0;
52 	u32 estatus_len;
53 
54 	while (remain >= sizeof(struct acpi_bert_region)) {
55 		estatus_len = cper_estatus_len(estatus);
56 		if (remain < estatus_len) {
57 			pr_err(FW_BUG "Truncated status block (length: %u).\n",
58 			       estatus_len);
59 			break;
60 		}
61 
62 		/* No more error records. */
63 		if (!estatus->block_status)
64 			break;
65 
66 		if (cper_estatus_check(estatus)) {
67 			pr_err(FW_BUG "Invalid error record.\n");
68 			break;
69 		}
70 
71 		if (estatus_len < ACPI_BERT_PRINT_MAX_LEN &&
72 		    printed < ACPI_BERT_PRINT_MAX_RECORDS) {
73 			pr_info_once("Error records from previous boot:\n");
74 			cper_estatus_print(KERN_INFO HW_ERR, estatus);
75 			printed++;
76 		} else {
77 			skipped++;
78 		}
79 
80 		/*
81 		 * Because the boot error source is "one-time polled" type,
82 		 * clear Block Status of current Generic Error Status Block,
83 		 * once it's printed.
84 		 */
85 		estatus->block_status = 0;
86 
87 		estatus = (void *)estatus + estatus_len;
88 		remain -= estatus_len;
89 	}
90 
91 	if (skipped)
92 		pr_info(HW_ERR "Skipped %d error records\n", skipped);
93 }
94 
95 static int __init setup_bert_disable(char *str)
96 {
97 	bert_disable = 1;
98 
99 	return 1;
100 }
101 __setup("bert_disable", setup_bert_disable);
102 
103 static int __init bert_check_table(struct acpi_table_bert *bert_tab)
104 {
105 	if (bert_tab->header.length < sizeof(struct acpi_table_bert) ||
106 	    bert_tab->region_length < sizeof(struct acpi_bert_region))
107 		return -EINVAL;
108 
109 	return 0;
110 }
111 
112 static int __init bert_init(void)
113 {
114 	struct apei_resources bert_resources;
115 	struct acpi_bert_region *boot_error_region;
116 	struct acpi_table_bert *bert_tab;
117 	unsigned int region_len;
118 	acpi_status status;
119 	int rc = 0;
120 
121 	if (acpi_disabled)
122 		return 0;
123 
124 	if (bert_disable) {
125 		pr_info("Boot Error Record Table support is disabled.\n");
126 		return 0;
127 	}
128 
129 	status = acpi_get_table(ACPI_SIG_BERT, 0, (struct acpi_table_header **)&bert_tab);
130 	if (status == AE_NOT_FOUND)
131 		return 0;
132 
133 	if (ACPI_FAILURE(status)) {
134 		pr_err("get table failed, %s.\n", acpi_format_exception(status));
135 		return -EINVAL;
136 	}
137 
138 	rc = bert_check_table(bert_tab);
139 	if (rc) {
140 		pr_err(FW_BUG "table invalid.\n");
141 		goto out_put_bert_tab;
142 	}
143 
144 	region_len = bert_tab->region_length;
145 	apei_resources_init(&bert_resources);
146 	rc = apei_resources_add(&bert_resources, bert_tab->address,
147 				region_len, true);
148 	if (rc)
149 		goto out_put_bert_tab;
150 	rc = apei_resources_request(&bert_resources, "APEI BERT");
151 	if (rc)
152 		goto out_fini;
153 	boot_error_region = ioremap_cache(bert_tab->address, region_len);
154 	if (boot_error_region) {
155 		bert_print_all(boot_error_region, region_len);
156 		iounmap(boot_error_region);
157 	} else {
158 		rc = -ENOMEM;
159 	}
160 
161 	apei_resources_release(&bert_resources);
162 out_fini:
163 	apei_resources_fini(&bert_resources);
164 out_put_bert_tab:
165 	acpi_put_table((struct acpi_table_header *)bert_tab);
166 
167 	return rc;
168 }
169 
170 late_initcall(bert_init);
171