1 // SPDX-License-Identifier: GPL-2.0-only
3 * APEI Boot Error Record Table (BERT) support
5 * Copyright 2011 Intel Corp.
6 * Author: Huang Ying <ying.huang@intel.com>
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
18 * For more information about BERT, please refer to ACPI Specification
19 * version 4.0, section 17.3.1
22 #include <linux/kernel.h>
23 #include <linux/module.h>
24 #include <linux/init.h>
25 #include <linux/acpi.h>
28 #include "apei-internal.h"
31 #define pr_fmt(fmt) "BERT: " fmt
33 #define ACPI_BERT_PRINT_MAX_RECORDS 5
34 #define ACPI_BERT_PRINT_MAX_LEN 1024
36 static int bert_disable;
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
45 static void __init bert_print_all(struct acpi_bert_region *region,
46 unsigned int region_len)
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;
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",
62 /* No more error records. */
63 if (!estatus->block_status)
66 if (cper_estatus_check(estatus)) {
67 pr_err(FW_BUG "Invalid error record.\n");
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);
81 * Because the boot error source is "one-time polled" type,
82 * clear Block Status of current Generic Error Status Block,
85 estatus->block_status = 0;
87 estatus = (void *)estatus + estatus_len;
88 remain -= estatus_len;
92 pr_info(HW_ERR "Skipped %d error records\n", skipped);
95 static int __init setup_bert_disable(char *str)
101 __setup("bert_disable", setup_bert_disable);
103 static int __init bert_check_table(struct acpi_table_bert *bert_tab)
105 if (bert_tab->header.length < sizeof(struct acpi_table_bert) ||
106 bert_tab->region_length < sizeof(struct acpi_bert_region))
112 static int __init bert_init(void)
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;
125 pr_info("Boot Error Record Table support is disabled.\n");
129 status = acpi_get_table(ACPI_SIG_BERT, 0, (struct acpi_table_header **)&bert_tab);
130 if (status == AE_NOT_FOUND)
133 if (ACPI_FAILURE(status)) {
134 pr_err("get table failed, %s.\n", acpi_format_exception(status));
138 rc = bert_check_table(bert_tab);
140 pr_err(FW_BUG "table invalid.\n");
141 goto out_put_bert_tab;
144 region_len = bert_tab->region_length;
145 apei_resources_init(&bert_resources);
146 rc = apei_resources_add(&bert_resources, bert_tab->address,
149 goto out_put_bert_tab;
150 rc = apei_resources_request(&bert_resources, "APEI BERT");
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);
161 apei_resources_release(&bert_resources);
163 apei_resources_fini(&bert_resources);
165 acpi_put_table((struct acpi_table_header *)bert_tab);
170 late_initcall(bert_init);