arch/x86/kernel/cpu/mcheck/non-fatal.c

   1 /*
   2  * Non Fatal Machine Check Exception Reporting
   3  *
   4  * (C) Copyright 2002 Dave Jones. <davej@redhat.com>
   5  *
   6  * This file contains routines to check for non-fatal MCEs every 15s
   7  *
   8  */
   9 #include <linux/interrupt.h>
  10 #include <linux/workqueue.h>
  11 #include <linux/jiffies.h>
  12 #include <linux/kernel.h>
  13 #include <linux/module.h>
  14 #include <linux/types.h>
  15 #include <linux/init.h>
  16 #include <linux/smp.h>
  17
  18 #include <asm/processor.h>
  19 #include <asm/system.h>
  20 #include <asm/mce.h>
  21 #include <asm/msr.h>
  22
  23 static int              firstbank;
  24
  25 #define MCE_RATE        (15*HZ) /* timer rate is 15s */
  26
  27 static void mce_checkregs(void *info)
  28 {
  29         u32 low, high;
  30         int i;
  31
  32         for (i = firstbank; i < nr_mce_banks; i++) {
  33                 rdmsr(MSR_IA32_MC0_STATUS+i*4, low, high);
  34
  35                 if (!(high & (1<<31)))
  36                         continue;
  37
  38                 printk(KERN_INFO "MCE: The hardware reports a non fatal, "
  39                         "correctable incident occurred on CPU %d.\n",
  40                                 smp_processor_id());
  41
  42                 printk(KERN_INFO "Bank %d: %08x%08x\n", i, high, low);
  43
  44                 /*
  45                  * Scrub the error so we don't pick it up in MCE_RATE
  46                  * seconds time:
  47                  */
  48                 wrmsr(MSR_IA32_MC0_STATUS+i*4, 0UL, 0UL);
  49
  50                 /* Serialize: */
  51                 wmb();
  52                 add_taint(TAINT_MACHINE_CHECK);
  53         }
  54 }
  55
  56 static void mce_work_fn(struct work_struct *work);
  57 static DECLARE_DELAYED_WORK(mce_work, mce_work_fn);
  58
  59 static void mce_work_fn(struct work_struct *work)
  60 {
  61         on_each_cpu(mce_checkregs, NULL, 1);
  62         schedule_delayed_work(&mce_work, round_jiffies_relative(MCE_RATE));
  63 }
  64
  65 static int __init init_nonfatal_mce_checker(void)
  66 {
  67         struct cpuinfo_x86 *c = &boot_cpu_data;
  68
  69         /* Check for MCE support */
  70         if (!cpu_has(c, X86_FEATURE_MCE))
  71                 return -ENODEV;
  72
  73         /* Check for PPro style MCA */
  74         if (!cpu_has(c, X86_FEATURE_MCA))
  75                 return -ENODEV;
  76
  77         /* Some Athlons misbehave when we frob bank 0 */
  78         if (boot_cpu_data.x86_vendor == X86_VENDOR_AMD &&
  79                                                 boot_cpu_data.x86 == 6)
  80                 firstbank = 1;
  81         else
  82                 firstbank = 0;
  83
  84         /*
  85          * Check for non-fatal errors every MCE_RATE s
  86          */
  87         schedule_delayed_work(&mce_work, round_jiffies_relative(MCE_RATE));
  88         printk(KERN_INFO "Machine check exception polling timer started.\n");
  89
  90         return 0;
  91 }
  92 module_init(init_nonfatal_mce_checker);
  93
  94 MODULE_LICENSE("GPL");