arch/i386/kernel/cpu/mcheck/non-fatal.c

   1 /*
   2  * Non Fatal Machine Check Exception Reporting
   3  *
   4  * (C) Copyright 2002 Dave Jones. <davej@codemonkey.org.uk>
   5  *
   6  * This file contains routines to check for non-fatal MCEs every 15s
   7  *
   8  */
   9
  10 #include <linux/init.h>
  11 #include <linux/types.h>
  12 #include <linux/kernel.h>
  13 #include <linux/jiffies.h>
  14 #include <linux/config.h>
  15 #include <linux/irq.h>
  16 #include <linux/workqueue.h>
  17 #include <linux/interrupt.h>
  18 #include <linux/smp.h>
  19 #include <linux/module.h>
  20
  21 #include <asm/processor.h>
  22 #include <asm/system.h>
  23 #include <asm/msr.h>
  24
  25 #include "mce.h"
  26
  27 static int firstbank;
  28
  29 #define MCE_RATE        15*HZ   /* timer rate is 15s */
  30
  31 static void mce_checkregs (void *info)
  32 {
  33         u32 low, high;
  34         int i;
  35
  36         for (i=firstbank; i<nr_mce_banks; i++) {
  37                 rdmsr (MSR_IA32_MC0_STATUS+i*4, low, high);
  38
  39                 if (high & (1<<31)) {
  40                         printk(KERN_INFO "MCE: The hardware reports a non "
  41                                 "fatal, correctable incident occurred on "
  42                                 "CPU %d.\n",
  43                                 smp_processor_id());
  44                         printk (KERN_INFO "Bank %d: %08x%08x\n", i, high, low);
  45
  46                         /* Scrub the error so we don't pick it up in MCE_RATE seconds time. */
  47                         wrmsr (MSR_IA32_MC0_STATUS+i*4, 0UL, 0UL);
  48
  49                         /* Serialize */
  50                         wmb();
  51                 }
  52         }
  53 }
  54
  55 static void mce_work_fn(void *data);
  56 static DECLARE_WORK(mce_work, mce_work_fn, NULL);
  57
  58 static void mce_work_fn(void *data)
  59 {
  60         on_each_cpu(mce_checkregs, NULL, 1, 1);
  61         schedule_delayed_work(&mce_work, MCE_RATE);
  62 }
  63
  64 static int __init init_nonfatal_mce_checker(void)
  65 {
  66         struct cpuinfo_x86 *c = &boot_cpu_data;
  67
  68         /* Check for MCE support */
  69         if (!cpu_has(c, X86_FEATURE_MCE))
  70                 return -ENODEV;
  71
  72         /* Check for PPro style MCA */
  73         if (!cpu_has(c, X86_FEATURE_MCA))
  74                 return -ENODEV;
  75
  76         /* Some Athlons misbehave when we frob bank 0 */
  77         if (boot_cpu_data.x86_vendor == X86_VENDOR_AMD &&
  78                 boot_cpu_data.x86 == 6)
  79                         firstbank = 1;
  80         else
  81                         firstbank = 0;
  82
  83         /*
  84          * Check for non-fatal errors every MCE_RATE s
  85          */
  86         schedule_delayed_work(&mce_work, MCE_RATE);
  87         printk(KERN_INFO "Machine check exception polling timer started.\n");
  88         return 0;
  89 }
  90 module_init(init_nonfatal_mce_checker);
  91
  92 MODULE_LICENSE("GPL");