arch/x86/kernel/cpu/mcheck/mce-apei.c

   1 /*
   2  * Bridge between MCE and APEI
   3  *
   4  * On some machine, corrected memory errors are reported via APEI
   5  * generic hardware error source (GHES) instead of corrected Machine
   6  * Check. These corrected memory errors can be reported to user space
   7  * through /dev/mcelog via faking a corrected Machine Check, so that
   8  * the error memory page can be offlined by /sbin/mcelog if the error
   9  * count for one page is beyond the threshold.
  10  *
  11  * For fatal MCE, save MCE record into persistent storage via ERST, so
  12  * that the MCE record can be logged after reboot via ERST.
  13  *
  14  * Copyright 2010 Intel Corp.
  15  *   Author: Huang Ying <ying.huang@intel.com>
  16  *
  17  * This program is free software; you can redistribute it and/or
  18  * modify it under the terms of the GNU General Public License version
  19  * 2 as published by the Free Software Foundation.
  20  *
  21  * This program is distributed in the hope that it will be useful,
  22  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  23  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
  24  * GNU General Public License for more details.
  25  *
  26  * You should have received a copy of the GNU General Public License
  27  * along with this program; if not, write to the Free Software
  28  * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
  29  */
  30
  31 #include <linux/export.h>
  32 #include <linux/kernel.h>
  33 #include <linux/acpi.h>
  34 #include <linux/cper.h>
  35 #include <acpi/apei.h>
  36 #include <asm/mce.h>
  37
  38 #include "mce-internal.h"
  39
  40 void apei_mce_report_mem_error(int corrected, struct cper_sec_mem_err *mem_err)
  41 {
  42         struct mce m;
  43
  44         /* Only corrected MC is reported */
  45         if (!corrected || !(mem_err->validation_bits &
  46                                 CPER_MEM_VALID_PHYSICAL_ADDRESS))
  47                 return;
  48
  49         mce_setup(&m);
  50         m.bank = 1;
  51         /* Fake a memory read corrected error with unknown channel */
  52         m.status = MCI_STATUS_VAL | MCI_STATUS_EN | MCI_STATUS_ADDRV | 0x9f;
  53         m.addr = mem_err->physical_addr;
  54         mce_log(&m);
  55         mce_notify_irq();
  56 }
  57 EXPORT_SYMBOL_GPL(apei_mce_report_mem_error);
  58
  59 #define CPER_CREATOR_MCE                                                \
  60         UUID_LE(0x75a574e3, 0x5052, 0x4b29, 0x8a, 0x8e, 0xbe, 0x2c,     \
  61                 0x64, 0x90, 0xb8, 0x9d)
  62 #define CPER_SECTION_TYPE_MCE                                           \
  63         UUID_LE(0xfe08ffbe, 0x95e4, 0x4be7, 0xbc, 0x73, 0x40, 0x96,     \
  64                 0x04, 0x4a, 0x38, 0xfc)
  65
  66 /*
  67  * CPER specification (in UEFI specification 2.3 appendix N) requires
  68  * byte-packed.
  69  */
  70 struct cper_mce_record {
  71         struct cper_record_header hdr;
  72         struct cper_section_descriptor sec_hdr;
  73         struct mce mce;
  74 } __packed;
  75
  76 int apei_write_mce(struct mce *m)
  77 {
  78         struct cper_mce_record rcd;
  79
  80         memset(&rcd, 0, sizeof(rcd));
  81         memcpy(rcd.hdr.signature, CPER_SIG_RECORD, CPER_SIG_SIZE);
  82         rcd.hdr.revision = CPER_RECORD_REV;
  83         rcd.hdr.signature_end = CPER_SIG_END;
  84         rcd.hdr.section_count = 1;
  85         rcd.hdr.error_severity = CPER_SEV_FATAL;
  86         /* timestamp, platform_id, partition_id are all invalid */
  87         rcd.hdr.validation_bits = 0;
  88         rcd.hdr.record_length = sizeof(rcd);
  89         rcd.hdr.creator_id = CPER_CREATOR_MCE;
  90         rcd.hdr.notification_type = CPER_NOTIFY_MCE;
  91         rcd.hdr.record_id = cper_next_record_id();
  92         rcd.hdr.flags = CPER_HW_ERROR_FLAGS_PREVERR;
  93
  94         rcd.sec_hdr.section_offset = (void *)&rcd.mce - (void *)&rcd;
  95         rcd.sec_hdr.section_length = sizeof(rcd.mce);
  96         rcd.sec_hdr.revision = CPER_SEC_REV;
  97         /* fru_id and fru_text is invalid */
  98         rcd.sec_hdr.validation_bits = 0;
  99         rcd.sec_hdr.flags = CPER_SEC_PRIMARY;
 100         rcd.sec_hdr.section_type = CPER_SECTION_TYPE_MCE;
 101         rcd.sec_hdr.section_severity = CPER_SEV_FATAL;
 102
 103         memcpy(&rcd.mce, m, sizeof(*m));
 104
 105         return erst_write(&rcd.hdr);
 106 }
 107
 108 ssize_t apei_read_mce(struct mce *m, u64 *record_id)
 109 {
 110         struct cper_mce_record rcd;
 111         int rc, pos;
 112
 113         rc = erst_get_record_id_begin(&pos);
 114         if (rc)
 115                 return rc;
 116 retry:
 117         rc = erst_get_record_id_next(&pos, record_id);
 118         if (rc)
 119                 goto out;
 120         /* no more record */
 121         if (*record_id == APEI_ERST_INVALID_RECORD_ID)
 122                 goto out;
 123         rc = erst_read(*record_id, &rcd.hdr, sizeof(rcd));
 124         /* someone else has cleared the record, try next one */
 125         if (rc == -ENOENT)
 126                 goto retry;
 127         else if (rc < 0)
 128                 goto out;
 129         /* try to skip other type records in storage */
 130         else if (rc != sizeof(rcd) ||
 131                  uuid_le_cmp(rcd.hdr.creator_id, CPER_CREATOR_MCE))
 132                 goto retry;
 133         memcpy(m, &rcd.mce, sizeof(*m));
 134         rc = sizeof(*m);
 135 out:
 136         erst_get_record_id_end();
 137
 138         return rc;
 139 }
 140
 141 /* Check whether there is record in ERST */
 142 int apei_check_mce(void)
 143 {
 144         return erst_get_record_count();
 145 }
 146
 147 int apei_clear_mce(u64 record_id)
 148 {
 149         return erst_clear(record_id);
 150 }