LXR linux/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 <acpi/ghes.h>
  37#include <asm/mce.h>
  38
  39#include "mce-internal.h"
  40
  41void apei_mce_report_mem_error(int severity, struct cper_sec_mem_err *mem_err)
  42{
  43        struct mce m;
  44
  45        if (!(mem_err->validation_bits & CPER_MEM_VALID_PA))
  46                return;
  47
  48        mce_setup(&m);
  49        m.bank = -1;
  50        /* Fake a memory read error with unknown channel */
  51        m.status = MCI_STATUS_VAL | MCI_STATUS_EN | MCI_STATUS_ADDRV | 0x9f;
  52
  53        if (severity >= GHES_SEV_RECOVERABLE)
  54                m.status |= MCI_STATUS_UC;
  55
  56        if (severity >= GHES_SEV_PANIC) {
  57                m.status |= MCI_STATUS_PCC;
  58                m.tsc = rdtsc();
  59        }
  60
  61        m.addr = mem_err->physical_addr;
  62        mce_log(&m);
  63}
  64EXPORT_SYMBOL_GPL(apei_mce_report_mem_error);
  65
  66#define CPER_CREATOR_MCE                                                \
  67        UUID_LE(0x75a574e3, 0x5052, 0x4b29, 0x8a, 0x8e, 0xbe, 0x2c,     \
  68                0x64, 0x90, 0xb8, 0x9d)
  69#define CPER_SECTION_TYPE_MCE                                           \
  70        UUID_LE(0xfe08ffbe, 0x95e4, 0x4be7, 0xbc, 0x73, 0x40, 0x96,     \
  71                0x04, 0x4a, 0x38, 0xfc)
  72
  73/*
  74 * CPER specification (in UEFI specification 2.3 appendix N) requires
  75 * byte-packed.
  76 */
  77struct cper_mce_record {
  78        struct cper_record_header hdr;
  79        struct cper_section_descriptor sec_hdr;
  80        struct mce mce;
  81} __packed;
  82
  83int apei_write_mce(struct mce *m)
  84{
  85        struct cper_mce_record rcd;
  86
  87        memset(&rcd, 0, sizeof(rcd));
  88        memcpy(rcd.hdr.signature, CPER_SIG_RECORD, CPER_SIG_SIZE);
  89        rcd.hdr.revision = CPER_RECORD_REV;
  90        rcd.hdr.signature_end = CPER_SIG_END;
  91        rcd.hdr.section_count = 1;
  92        rcd.hdr.error_severity = CPER_SEV_FATAL;
  93        /* timestamp, platform_id, partition_id are all invalid */
  94        rcd.hdr.validation_bits = 0;
  95        rcd.hdr.record_length = sizeof(rcd);
  96        rcd.hdr.creator_id = CPER_CREATOR_MCE;
  97        rcd.hdr.notification_type = CPER_NOTIFY_MCE;
  98        rcd.hdr.record_id = cper_next_record_id();
  99        rcd.hdr.flags = CPER_HW_ERROR_FLAGS_PREVERR;
 100
 101        rcd.sec_hdr.section_offset = (void *)&rcd.mce - (void *)&rcd;
 102        rcd.sec_hdr.section_length = sizeof(rcd.mce);
 103        rcd.sec_hdr.revision = CPER_SEC_REV;
 104        /* fru_id and fru_text is invalid */
 105        rcd.sec_hdr.validation_bits = 0;
 106        rcd.sec_hdr.flags = CPER_SEC_PRIMARY;
 107        rcd.sec_hdr.section_type = CPER_SECTION_TYPE_MCE;
 108        rcd.sec_hdr.section_severity = CPER_SEV_FATAL;
 109
 110        memcpy(&rcd.mce, m, sizeof(*m));
 111
 112        return erst_write(&rcd.hdr);
 113}
 114
 115ssize_t apei_read_mce(struct mce *m, u64 *record_id)
 116{
 117        struct cper_mce_record rcd;
 118        int rc, pos;
 119
 120        rc = erst_get_record_id_begin(&pos);
 121        if (rc)
 122                return rc;
 123retry:
 124        rc = erst_get_record_id_next(&pos, record_id);
 125        if (rc)
 126                goto out;
 127        /* no more record */
 128        if (*record_id == APEI_ERST_INVALID_RECORD_ID)
 129                goto out;
 130        rc = erst_read(*record_id, &rcd.hdr, sizeof(rcd));
 131        /* someone else has cleared the record, try next one */
 132        if (rc == -ENOENT)
 133                goto retry;
 134        else if (rc < 0)
 135                goto out;
 136        /* try to skip other type records in storage */
 137        else if (rc != sizeof(rcd) ||
 138                 uuid_le_cmp(rcd.hdr.creator_id, CPER_CREATOR_MCE))
 139                goto retry;
 140        memcpy(m, &rcd.mce, sizeof(*m));
 141        rc = sizeof(*m);
 142out:
 143        erst_get_record_id_end();
 144
 145        return rc;
 146}
 147
 148/* Check whether there is record in ERST */
 149int apei_check_mce(void)
 150{
 151        return erst_get_record_count();
 152}
 153
 154int apei_clear_mce(u64 record_id)
 155{
 156        return erst_clear(record_id);
 157}
 158