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 <linux/cper.h>
13 #include <linux/ratelimit.h>
17 #include "apei/apei-internal.h"
19 #define EXT_ELOG_ENTRY_MASK GENMASK_ULL(51, 0) /* elog entry address mask */
21 #define EXTLOG_DSM_REV 0x0
22 #define EXTLOG_FN_ADDR 0x1
24 #define FLAG_OS_OPTIN BIT(0)
25 #define ELOG_ENTRY_VALID (1ULL<<63)
26 #define ELOG_ENTRY_LEN 0x1000
29 "Can not request iomem region <0x%016llx-0x%016llx> - eMCA disabled\n"
31 struct extlog_l1_head {
32 u32 ver; /* Header Version */
33 u32 hdr_len; /* Header Length */
34 u64 total_len; /* entire L1 Directory length including this header */
35 u64 elog_base; /* MCA Error Log Directory base address */
36 u64 elog_len; /* MCA Error Log Directory length */
37 u32 flags; /* bit 0 - OS/VMM Opt-in */
39 u32 entries; /* Valid L1 Directory entries per logical processor */
43 static u8 extlog_dsm_uuid[] __initdata = "663E35AF-CC10-41A4-88EA-5470AF055295";
45 /* L1 table related physical address */
47 static size_t elog_size;
48 static u64 l1_dirbase;
49 static size_t l1_size;
51 /* L1 table related virtual address */
52 static void __iomem *extlog_l1_addr;
53 static void __iomem *elog_addr;
55 static void *elog_buf;
57 static u64 *l1_entry_base;
58 static u32 l1_percpu_entry;
60 #define ELOG_IDX(cpu, bank) \
61 (cpu_physical_id(cpu) * l1_percpu_entry + (bank))
63 #define ELOG_ENTRY_DATA(idx) \
64 (*(l1_entry_base + (idx)))
66 #define ELOG_ENTRY_ADDR(phyaddr) \
67 (phyaddr - elog_base + (u8 *)elog_addr)
69 static struct acpi_generic_status *extlog_elog_entry_check(int cpu, int bank)
73 struct acpi_generic_status *estatus;
76 idx = ELOG_IDX(cpu, bank);
77 data = ELOG_ENTRY_DATA(idx);
78 if ((data & ELOG_ENTRY_VALID) == 0)
81 data &= EXT_ELOG_ENTRY_MASK;
82 estatus = (struct acpi_generic_status *)ELOG_ENTRY_ADDR(data);
84 /* if no valid data in elog entry, just return */
85 if (estatus->block_status == 0)
91 static void __print_extlog_rcd(const char *pfx,
92 struct acpi_generic_status *estatus, int cpu)
94 static atomic_t seqno;
95 unsigned int curr_seqno;
99 if (estatus->error_severity <= CPER_SEV_CORRECTED)
104 curr_seqno = atomic_inc_return(&seqno);
105 snprintf(pfx_seq, sizeof(pfx_seq), "%s{%u}", pfx, curr_seqno);
106 printk("%s""Hardware error detected on CPU%d\n", pfx_seq, cpu);
107 cper_estatus_print(pfx_seq, estatus);
110 static int print_extlog_rcd(const char *pfx,
111 struct acpi_generic_status *estatus, int cpu)
113 /* Not more than 2 messages every 5 seconds */
114 static DEFINE_RATELIMIT_STATE(ratelimit_corrected, 5*HZ, 2);
115 static DEFINE_RATELIMIT_STATE(ratelimit_uncorrected, 5*HZ, 2);
116 struct ratelimit_state *ratelimit;
118 if (estatus->error_severity == CPER_SEV_CORRECTED ||
119 (estatus->error_severity == CPER_SEV_INFORMATIONAL))
120 ratelimit = &ratelimit_corrected;
122 ratelimit = &ratelimit_uncorrected;
123 if (__ratelimit(ratelimit)) {
124 __print_extlog_rcd(pfx, estatus, cpu);
131 static int extlog_print(struct notifier_block *nb, unsigned long val,
134 struct mce *mce = (struct mce *)data;
135 int bank = mce->bank;
136 int cpu = mce->extcpu;
137 struct acpi_generic_status *estatus;
140 estatus = extlog_elog_entry_check(cpu, bank);
144 memcpy(elog_buf, (void *)estatus, ELOG_ENTRY_LEN);
145 /* clear record status to enable BIOS to update it again */
146 estatus->block_status = 0;
148 rc = print_extlog_rcd(NULL, (struct acpi_generic_status *)elog_buf, cpu);
153 static bool __init extlog_get_l1addr(void)
157 union acpi_object *obj;
159 acpi_str_to_uuid(extlog_dsm_uuid, uuid);
161 if (ACPI_FAILURE(acpi_get_handle(NULL, "\\_SB", &handle)))
163 if (!acpi_check_dsm(handle, uuid, EXTLOG_DSM_REV, 1 << EXTLOG_FN_ADDR))
165 obj = acpi_evaluate_dsm_typed(handle, uuid, EXTLOG_DSM_REV,
166 EXTLOG_FN_ADDR, NULL, ACPI_TYPE_INTEGER);
170 l1_dirbase = obj->integer.value;
174 /* Spec says L1 directory must be 4K aligned, bail out if it isn't */
175 if (l1_dirbase & ((1 << 12) - 1)) {
176 pr_warn(FW_BUG "L1 Directory is invalid at physical %llx\n",
183 static struct notifier_block extlog_mce_dec = {
184 .notifier_call = extlog_print,
187 static int __init extlog_init(void)
189 struct extlog_l1_head *l1_head;
190 void __iomem *extlog_l1_hdr;
198 rdmsrl(MSR_IA32_MCG_CAP, cap);
199 if (!(cap & MCG_ELOG_P))
202 if (!extlog_get_l1addr())
206 /* get L1 header to fetch necessary information */
207 l1_hdr_size = sizeof(struct extlog_l1_head);
208 r = request_mem_region(l1_dirbase, l1_hdr_size, "L1 DIR HDR");
210 pr_warn(FW_BUG EMCA_BUG,
211 (unsigned long long)l1_dirbase,
212 (unsigned long long)l1_dirbase + l1_hdr_size);
216 extlog_l1_hdr = acpi_os_map_memory(l1_dirbase, l1_hdr_size);
217 l1_head = (struct extlog_l1_head *)extlog_l1_hdr;
218 l1_size = l1_head->total_len;
219 l1_percpu_entry = l1_head->entries;
220 elog_base = l1_head->elog_base;
221 elog_size = l1_head->elog_len;
222 acpi_os_unmap_memory(extlog_l1_hdr, l1_hdr_size);
223 release_mem_region(l1_dirbase, l1_hdr_size);
225 /* remap L1 header again based on completed information */
226 r = request_mem_region(l1_dirbase, l1_size, "L1 Table");
228 pr_warn(FW_BUG EMCA_BUG,
229 (unsigned long long)l1_dirbase,
230 (unsigned long long)l1_dirbase + l1_size);
233 extlog_l1_addr = acpi_os_map_memory(l1_dirbase, l1_size);
234 l1_entry_base = (u64 *)((u8 *)extlog_l1_addr + l1_hdr_size);
236 /* remap elog table */
237 r = request_mem_region(elog_base, elog_size, "Elog Table");
239 pr_warn(FW_BUG EMCA_BUG,
240 (unsigned long long)elog_base,
241 (unsigned long long)elog_base + elog_size);
242 goto err_release_l1_dir;
244 elog_addr = acpi_os_map_memory(elog_base, elog_size);
247 /* allocate buffer to save elog record */
248 elog_buf = kmalloc(ELOG_ENTRY_LEN, GFP_KERNEL);
249 if (elog_buf == NULL)
250 goto err_release_elog;
252 mce_register_decode_chain(&extlog_mce_dec);
253 /* enable OS to be involved to take over management from BIOS */
254 ((struct extlog_l1_head *)extlog_l1_addr)->flags |= FLAG_OS_OPTIN;
260 acpi_os_unmap_memory(elog_addr, elog_size);
261 release_mem_region(elog_base, elog_size);
264 acpi_os_unmap_memory(extlog_l1_addr, l1_size);
265 release_mem_region(l1_dirbase, l1_size);
267 pr_warn(FW_BUG "Extended error log disabled because of problems parsing f/w tables\n");
271 static void __exit extlog_exit(void)
273 mce_unregister_decode_chain(&extlog_mce_dec);
274 ((struct extlog_l1_head *)extlog_l1_addr)->flags &= ~FLAG_OS_OPTIN;
276 acpi_os_unmap_memory(extlog_l1_addr, l1_size);
278 acpi_os_unmap_memory(elog_addr, elog_size);
279 release_mem_region(elog_base, elog_size);
280 release_mem_region(l1_dirbase, l1_size);
284 module_init(extlog_init);
285 module_exit(extlog_exit);
287 MODULE_AUTHOR("Chen, Gong <gong.chen@intel.com>");
288 MODULE_DESCRIPTION("Extended MCA Error Log Driver");
289 MODULE_LICENSE("GPL");