xref: /openbmc/linux/drivers/acpi/apei/hest.c (revision 85250a24)
1 // SPDX-License-Identifier: GPL-2.0-only
2 /*
3  * APEI Hardware Error Source Table support
4  *
5  * HEST describes error sources in detail; communicates operational
6  * parameters (i.e. severity levels, masking bits, and threshold
7  * values) to Linux as necessary. It also allows the BIOS to report
8  * non-standard error sources to Linux (for example, chipset-specific
9  * error registers).
10  *
11  * For more information about HEST, please refer to ACPI Specification
12  * version 4.0, section 17.3.2.
13  *
14  * Copyright 2009 Intel Corp.
15  *   Author: Huang Ying <ying.huang@intel.com>
16  */
17 
18 #include <linux/kernel.h>
19 #include <linux/module.h>
20 #include <linux/init.h>
21 #include <linux/acpi.h>
22 #include <linux/kdebug.h>
23 #include <linux/highmem.h>
24 #include <linux/io.h>
25 #include <linux/platform_device.h>
26 #include <acpi/apei.h>
27 #include <acpi/ghes.h>
28 
29 #include "apei-internal.h"
30 
31 #define HEST_PFX "HEST: "
32 
33 int hest_disable;
34 EXPORT_SYMBOL_GPL(hest_disable);
35 
36 /* HEST table parsing */
37 
38 static struct acpi_table_hest *__read_mostly hest_tab;
39 
40 static const int hest_esrc_len_tab[ACPI_HEST_TYPE_RESERVED] = {
41 	[ACPI_HEST_TYPE_IA32_CHECK] = -1,	/* need further calculation */
42 	[ACPI_HEST_TYPE_IA32_CORRECTED_CHECK] = -1,
43 	[ACPI_HEST_TYPE_IA32_NMI] = sizeof(struct acpi_hest_ia_nmi),
44 	[ACPI_HEST_TYPE_AER_ROOT_PORT] = sizeof(struct acpi_hest_aer_root),
45 	[ACPI_HEST_TYPE_AER_ENDPOINT] = sizeof(struct acpi_hest_aer),
46 	[ACPI_HEST_TYPE_AER_BRIDGE] = sizeof(struct acpi_hest_aer_bridge),
47 	[ACPI_HEST_TYPE_GENERIC_ERROR] = sizeof(struct acpi_hest_generic),
48 	[ACPI_HEST_TYPE_GENERIC_ERROR_V2] = sizeof(struct acpi_hest_generic_v2),
49 	[ACPI_HEST_TYPE_IA32_DEFERRED_CHECK] = -1,
50 };
51 
52 static inline bool is_generic_error(struct acpi_hest_header *hest_hdr)
53 {
54 	return hest_hdr->type == ACPI_HEST_TYPE_GENERIC_ERROR ||
55 	       hest_hdr->type == ACPI_HEST_TYPE_GENERIC_ERROR_V2;
56 }
57 
58 static int hest_esrc_len(struct acpi_hest_header *hest_hdr)
59 {
60 	u16 hest_type = hest_hdr->type;
61 	int len;
62 
63 	if (hest_type >= ACPI_HEST_TYPE_RESERVED)
64 		return 0;
65 
66 	len = hest_esrc_len_tab[hest_type];
67 
68 	if (hest_type == ACPI_HEST_TYPE_IA32_CORRECTED_CHECK) {
69 		struct acpi_hest_ia_corrected *cmc;
70 		cmc = (struct acpi_hest_ia_corrected *)hest_hdr;
71 		len = sizeof(*cmc) + cmc->num_hardware_banks *
72 			sizeof(struct acpi_hest_ia_error_bank);
73 	} else if (hest_type == ACPI_HEST_TYPE_IA32_CHECK) {
74 		struct acpi_hest_ia_machine_check *mc;
75 		mc = (struct acpi_hest_ia_machine_check *)hest_hdr;
76 		len = sizeof(*mc) + mc->num_hardware_banks *
77 			sizeof(struct acpi_hest_ia_error_bank);
78 	} else if (hest_type == ACPI_HEST_TYPE_IA32_DEFERRED_CHECK) {
79 		struct acpi_hest_ia_deferred_check *mc;
80 		mc = (struct acpi_hest_ia_deferred_check *)hest_hdr;
81 		len = sizeof(*mc) + mc->num_hardware_banks *
82 			sizeof(struct acpi_hest_ia_error_bank);
83 	}
84 	BUG_ON(len == -1);
85 
86 	return len;
87 };
88 
89 typedef int (*apei_hest_func_t)(struct acpi_hest_header *hest_hdr, void *data);
90 
91 static int apei_hest_parse(apei_hest_func_t func, void *data)
92 {
93 	struct acpi_hest_header *hest_hdr;
94 	int i, rc, len;
95 
96 	if (hest_disable || !hest_tab)
97 		return -EINVAL;
98 
99 	hest_hdr = (struct acpi_hest_header *)(hest_tab + 1);
100 	for (i = 0; i < hest_tab->error_source_count; i++) {
101 		len = hest_esrc_len(hest_hdr);
102 		if (!len) {
103 			pr_warn(FW_WARN HEST_PFX
104 				"Unknown or unused hardware error source "
105 				"type: %d for hardware error source: %d.\n",
106 				hest_hdr->type, hest_hdr->source_id);
107 			return -EINVAL;
108 		}
109 		if ((void *)hest_hdr + len >
110 		    (void *)hest_tab + hest_tab->header.length) {
111 			pr_warn(FW_BUG HEST_PFX
112 		"Table contents overflow for hardware error source: %d.\n",
113 				hest_hdr->source_id);
114 			return -EINVAL;
115 		}
116 
117 		rc = func(hest_hdr, data);
118 		if (rc)
119 			return rc;
120 
121 		hest_hdr = (void *)hest_hdr + len;
122 	}
123 
124 	return 0;
125 }
126 
127 /*
128  * Check if firmware advertises firmware first mode. We need FF bit to be set
129  * along with a set of MC banks which work in FF mode.
130  */
131 static int __init hest_parse_cmc(struct acpi_hest_header *hest_hdr, void *data)
132 {
133 	if (hest_hdr->type != ACPI_HEST_TYPE_IA32_CORRECTED_CHECK)
134 		return 0;
135 
136 	if (!acpi_disable_cmcff)
137 		return !arch_apei_enable_cmcff(hest_hdr, data);
138 
139 	return 0;
140 }
141 
142 struct ghes_arr {
143 	struct platform_device **ghes_devs;
144 	unsigned int count;
145 };
146 
147 static int __init hest_parse_ghes_count(struct acpi_hest_header *hest_hdr, void *data)
148 {
149 	int *count = data;
150 
151 	if (is_generic_error(hest_hdr))
152 		(*count)++;
153 	return 0;
154 }
155 
156 static int __init hest_parse_ghes(struct acpi_hest_header *hest_hdr, void *data)
157 {
158 	struct platform_device *ghes_dev;
159 	struct ghes_arr *ghes_arr = data;
160 	int rc, i;
161 
162 	if (!is_generic_error(hest_hdr))
163 		return 0;
164 
165 	if (!((struct acpi_hest_generic *)hest_hdr)->enabled)
166 		return 0;
167 	for (i = 0; i < ghes_arr->count; i++) {
168 		struct acpi_hest_header *hdr;
169 		ghes_dev = ghes_arr->ghes_devs[i];
170 		hdr = *(struct acpi_hest_header **)ghes_dev->dev.platform_data;
171 		if (hdr->source_id == hest_hdr->source_id) {
172 			pr_warn(FW_WARN HEST_PFX "Duplicated hardware error source ID: %d.\n",
173 				hdr->source_id);
174 			return -EIO;
175 		}
176 	}
177 	ghes_dev = platform_device_alloc("GHES", hest_hdr->source_id);
178 	if (!ghes_dev)
179 		return -ENOMEM;
180 
181 	rc = platform_device_add_data(ghes_dev, &hest_hdr, sizeof(void *));
182 	if (rc)
183 		goto err;
184 
185 	rc = platform_device_add(ghes_dev);
186 	if (rc)
187 		goto err;
188 	ghes_arr->ghes_devs[ghes_arr->count++] = ghes_dev;
189 
190 	return 0;
191 err:
192 	platform_device_put(ghes_dev);
193 	return rc;
194 }
195 
196 static int __init hest_ghes_dev_register(unsigned int ghes_count)
197 {
198 	int rc, i;
199 	struct ghes_arr ghes_arr;
200 
201 	ghes_arr.count = 0;
202 	ghes_arr.ghes_devs = kmalloc_array(ghes_count, sizeof(void *),
203 					   GFP_KERNEL);
204 	if (!ghes_arr.ghes_devs)
205 		return -ENOMEM;
206 
207 	rc = apei_hest_parse(hest_parse_ghes, &ghes_arr);
208 	if (rc)
209 		goto err;
210 
211 	rc = ghes_estatus_pool_init(ghes_count);
212 	if (rc)
213 		goto err;
214 
215 out:
216 	kfree(ghes_arr.ghes_devs);
217 	return rc;
218 err:
219 	for (i = 0; i < ghes_arr.count; i++)
220 		platform_device_unregister(ghes_arr.ghes_devs[i]);
221 	goto out;
222 }
223 
224 static int __init setup_hest_disable(char *str)
225 {
226 	hest_disable = HEST_DISABLED;
227 	return 1;
228 }
229 
230 __setup("hest_disable", setup_hest_disable);
231 
232 void __init acpi_hest_init(void)
233 {
234 	acpi_status status;
235 	int rc;
236 	unsigned int ghes_count = 0;
237 
238 	if (hest_disable) {
239 		pr_info(HEST_PFX "Table parsing disabled.\n");
240 		return;
241 	}
242 
243 	status = acpi_get_table(ACPI_SIG_HEST, 0,
244 				(struct acpi_table_header **)&hest_tab);
245 	if (status == AE_NOT_FOUND) {
246 		hest_disable = HEST_NOT_FOUND;
247 		return;
248 	} else if (ACPI_FAILURE(status)) {
249 		const char *msg = acpi_format_exception(status);
250 		pr_err(HEST_PFX "Failed to get table, %s\n", msg);
251 		hest_disable = HEST_DISABLED;
252 		return;
253 	}
254 
255 	rc = apei_hest_parse(hest_parse_cmc, NULL);
256 	if (rc)
257 		goto err;
258 
259 	if (!ghes_disable) {
260 		rc = apei_hest_parse(hest_parse_ghes_count, &ghes_count);
261 		if (rc)
262 			goto err;
263 
264 		if (ghes_count)
265 			rc = hest_ghes_dev_register(ghes_count);
266 		if (rc)
267 			goto err;
268 	}
269 
270 	pr_info(HEST_PFX "Table parsing has been initialized.\n");
271 	return;
272 err:
273 	hest_disable = HEST_DISABLED;
274 	acpi_put_table((struct acpi_table_header *)hest_tab);
275 }
276