2 * Extended Error Log driver
4 * Copyright (C) 2013 Intel Corp.
5 * Author: Chen, Gong <gong.chen@intel.com>
7 * This file is licensed under GPLv2.
10 #include <linux/module.h>
11 #include <linux/acpi.h>
12 #include <acpi/acpi_bus.h>
13 #include <linux/cper.h>
14 #include <linux/ratelimit.h>
15 #include <linux/edac.h>
19 #include "apei/apei-internal.h"
21 #define EXT_ELOG_ENTRY_MASK GENMASK_ULL(51, 0) /* elog entry address mask */
23 #define EXTLOG_DSM_REV 0x0
24 #define EXTLOG_FN_QUERY 0x0
25 #define EXTLOG_FN_ADDR 0x1
27 #define FLAG_OS_OPTIN BIT(0)
28 #define EXTLOG_QUERY_L1_EXIST BIT(1)
29 #define ELOG_ENTRY_VALID (1ULL<<63)
30 #define ELOG_ENTRY_LEN 0x1000
33 "Can not request iomem region <0x%016llx-0x%016llx> - eMCA disabled\n"
35 struct extlog_l1_head {
36 u32 ver; /* Header Version */
37 u32 hdr_len; /* Header Length */
38 u64 total_len; /* entire L1 Directory length including this header */
39 u64 elog_base; /* MCA Error Log Directory base address */
40 u64 elog_len; /* MCA Error Log Directory length */
41 u32 flags; /* bit 0 - OS/VMM Opt-in */
43 u32 entries; /* Valid L1 Directory entries per logical processor */
47 static int old_edac_report_status;
49 static u8 extlog_dsm_uuid[] = "663E35AF-CC10-41A4-88EA-5470AF055295";
51 /* L1 table related physical address */
53 static size_t elog_size;
54 static u64 l1_dirbase;
55 static size_t l1_size;
57 /* L1 table related virtual address */
58 static void __iomem *extlog_l1_addr;
59 static void __iomem *elog_addr;
61 static void *elog_buf;
63 static u64 *l1_entry_base;
64 static u32 l1_percpu_entry;
66 #define ELOG_IDX(cpu, bank) \
67 (cpu_physical_id(cpu) * l1_percpu_entry + (bank))
69 #define ELOG_ENTRY_DATA(idx) \
70 (*(l1_entry_base + (idx)))
72 #define ELOG_ENTRY_ADDR(phyaddr) \
73 (phyaddr - elog_base + (u8 *)elog_addr)
75 static struct acpi_generic_status *extlog_elog_entry_check(int cpu, int bank)
79 struct acpi_generic_status *estatus;
82 idx = ELOG_IDX(cpu, bank);
83 data = ELOG_ENTRY_DATA(idx);
84 if ((data & ELOG_ENTRY_VALID) == 0)
87 data &= EXT_ELOG_ENTRY_MASK;
88 estatus = (struct acpi_generic_status *)ELOG_ENTRY_ADDR(data);
90 /* if no valid data in elog entry, just return */
91 if (estatus->block_status == 0)
97 static void __print_extlog_rcd(const char *pfx,
98 struct acpi_generic_status *estatus, int cpu)
100 static atomic_t seqno;
101 unsigned int curr_seqno;
105 if (estatus->error_severity <= CPER_SEV_CORRECTED)
110 curr_seqno = atomic_inc_return(&seqno);
111 snprintf(pfx_seq, sizeof(pfx_seq), "%s{%u}", pfx, curr_seqno);
112 printk("%s""Hardware error detected on CPU%d\n", pfx_seq, cpu);
113 cper_estatus_print(pfx_seq, estatus);
116 static int print_extlog_rcd(const char *pfx,
117 struct acpi_generic_status *estatus, int cpu)
119 /* Not more than 2 messages every 5 seconds */
120 static DEFINE_RATELIMIT_STATE(ratelimit_corrected, 5*HZ, 2);
121 static DEFINE_RATELIMIT_STATE(ratelimit_uncorrected, 5*HZ, 2);
122 struct ratelimit_state *ratelimit;
124 if (estatus->error_severity == CPER_SEV_CORRECTED ||
125 (estatus->error_severity == CPER_SEV_INFORMATIONAL))
126 ratelimit = &ratelimit_corrected;
128 ratelimit = &ratelimit_uncorrected;
129 if (__ratelimit(ratelimit)) {
130 __print_extlog_rcd(pfx, estatus, cpu);
137 static int extlog_print(struct notifier_block *nb, unsigned long val,
140 struct mce *mce = (struct mce *)data;
141 int bank = mce->bank;
142 int cpu = mce->extcpu;
143 struct acpi_generic_status *estatus;
146 estatus = extlog_elog_entry_check(cpu, bank);
150 memcpy(elog_buf, (void *)estatus, ELOG_ENTRY_LEN);
151 /* clear record status to enable BIOS to update it again */
152 estatus->block_status = 0;
154 rc = print_extlog_rcd(NULL, (struct acpi_generic_status *)elog_buf, cpu);
159 static int extlog_get_dsm(acpi_handle handle, int rev, int func, u64 *ret)
161 struct acpi_buffer buf = {ACPI_ALLOCATE_BUFFER, NULL};
162 struct acpi_object_list input;
163 union acpi_object params[4], *obj;
167 acpi_str_to_uuid(extlog_dsm_uuid, uuid);
169 input.pointer = params;
170 params[0].type = ACPI_TYPE_BUFFER;
171 params[0].buffer.length = 16;
172 params[0].buffer.pointer = uuid;
173 params[1].type = ACPI_TYPE_INTEGER;
174 params[1].integer.value = rev;
175 params[2].type = ACPI_TYPE_INTEGER;
176 params[2].integer.value = func;
177 params[3].type = ACPI_TYPE_PACKAGE;
178 params[3].package.count = 0;
179 params[3].package.elements = NULL;
181 if (ACPI_FAILURE(acpi_evaluate_object(handle, "_DSM", &input, &buf)))
185 obj = (union acpi_object *)buf.pointer;
186 if (obj->type == ACPI_TYPE_INTEGER) {
187 *ret = obj->integer.value;
188 } else if (obj->type == ACPI_TYPE_BUFFER) {
189 if (obj->buffer.length <= 8) {
190 for (i = 0; i < obj->buffer.length; i++)
191 *ret |= (obj->buffer.pointer[i] << (i * 8));
199 static bool extlog_get_l1addr(void)
204 if (ACPI_FAILURE(acpi_get_handle(NULL, "\\_SB", &handle)))
207 if (extlog_get_dsm(handle, EXTLOG_DSM_REV, EXTLOG_FN_QUERY, &ret) ||
208 !(ret & EXTLOG_QUERY_L1_EXIST))
211 if (extlog_get_dsm(handle, EXTLOG_DSM_REV, EXTLOG_FN_ADDR, &ret))
215 /* Spec says L1 directory must be 4K aligned, bail out if it isn't */
216 if (l1_dirbase & ((1 << 12) - 1)) {
217 pr_warn(FW_BUG "L1 Directory is invalid at physical %llx\n",
224 static struct notifier_block extlog_mce_dec = {
225 .notifier_call = extlog_print,
228 static int __init extlog_init(void)
230 struct extlog_l1_head *l1_head;
231 void __iomem *extlog_l1_hdr;
237 if (get_edac_report_status() == EDAC_REPORTING_FORCE) {
238 pr_warn("Not loading eMCA, error reporting force-enabled through EDAC.\n");
243 rdmsrl(MSR_IA32_MCG_CAP, cap);
244 if (!(cap & MCG_ELOG_P))
247 if (!extlog_get_l1addr())
251 /* get L1 header to fetch necessary information */
252 l1_hdr_size = sizeof(struct extlog_l1_head);
253 r = request_mem_region(l1_dirbase, l1_hdr_size, "L1 DIR HDR");
255 pr_warn(FW_BUG EMCA_BUG,
256 (unsigned long long)l1_dirbase,
257 (unsigned long long)l1_dirbase + l1_hdr_size);
261 extlog_l1_hdr = acpi_os_map_memory(l1_dirbase, l1_hdr_size);
262 l1_head = (struct extlog_l1_head *)extlog_l1_hdr;
263 l1_size = l1_head->total_len;
264 l1_percpu_entry = l1_head->entries;
265 elog_base = l1_head->elog_base;
266 elog_size = l1_head->elog_len;
267 acpi_os_unmap_memory(extlog_l1_hdr, l1_hdr_size);
268 release_mem_region(l1_dirbase, l1_hdr_size);
270 /* remap L1 header again based on completed information */
271 r = request_mem_region(l1_dirbase, l1_size, "L1 Table");
273 pr_warn(FW_BUG EMCA_BUG,
274 (unsigned long long)l1_dirbase,
275 (unsigned long long)l1_dirbase + l1_size);
278 extlog_l1_addr = acpi_os_map_memory(l1_dirbase, l1_size);
279 l1_entry_base = (u64 *)((u8 *)extlog_l1_addr + l1_hdr_size);
281 /* remap elog table */
282 r = request_mem_region(elog_base, elog_size, "Elog Table");
284 pr_warn(FW_BUG EMCA_BUG,
285 (unsigned long long)elog_base,
286 (unsigned long long)elog_base + elog_size);
287 goto err_release_l1_dir;
289 elog_addr = acpi_os_map_memory(elog_base, elog_size);
292 /* allocate buffer to save elog record */
293 elog_buf = kmalloc(ELOG_ENTRY_LEN, GFP_KERNEL);
294 if (elog_buf == NULL)
295 goto err_release_elog;
298 * eMCA event report method has higher priority than EDAC method,
299 * unless EDAC event report method is mandatory.
301 old_edac_report_status = get_edac_report_status();
302 set_edac_report_status(EDAC_REPORTING_DISABLED);
303 mce_register_decode_chain(&extlog_mce_dec);
304 /* enable OS to be involved to take over management from BIOS */
305 ((struct extlog_l1_head *)extlog_l1_addr)->flags |= FLAG_OS_OPTIN;
311 acpi_os_unmap_memory(elog_addr, elog_size);
312 release_mem_region(elog_base, elog_size);
315 acpi_os_unmap_memory(extlog_l1_addr, l1_size);
316 release_mem_region(l1_dirbase, l1_size);
318 pr_warn(FW_BUG "Extended error log disabled because of problems parsing f/w tables\n");
322 static void __exit extlog_exit(void)
324 set_edac_report_status(old_edac_report_status);
325 mce_unregister_decode_chain(&extlog_mce_dec);
326 ((struct extlog_l1_head *)extlog_l1_addr)->flags &= ~FLAG_OS_OPTIN;
328 acpi_os_unmap_memory(extlog_l1_addr, l1_size);
330 acpi_os_unmap_memory(elog_addr, elog_size);
331 release_mem_region(elog_base, elog_size);
332 release_mem_region(l1_dirbase, l1_size);
336 module_init(extlog_init);
337 module_exit(extlog_exit);
339 MODULE_AUTHOR("Chen, Gong <gong.chen@intel.com>");
340 MODULE_DESCRIPTION("Extended MCA Error Log Driver");
341 MODULE_LICENSE("GPL");