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

   1 /*
   2  * Machine check injection support.
   3  * Copyright 2008 Intel Corporation.
   4  *
   5  * This program is free software; you can redistribute it and/or
   6  * modify it under the terms of the GNU General Public License
   7  * as published by the Free Software Foundation; version 2
   8  * of the License.
   9  *
  10  * Authors:
  11  * Andi Kleen
  12  * Ying Huang
  13  */
  14 #include <linux/uaccess.h>
  15 #include <linux/module.h>
  16 #include <linux/timer.h>
  17 #include <linux/kernel.h>
  18 #include <linux/string.h>
  19 #include <linux/fs.h>
  20 #include <linux/smp.h>
  21 #include <linux/notifier.h>
  22 #include <linux/kdebug.h>
  23 #include <linux/cpu.h>
  24 #include <linux/sched.h>
  25 #include <linux/gfp.h>
  26 #include <asm/mce.h>
  27 #include <asm/apic.h>
  28 #include <asm/nmi.h>
  29
  30 /* Update fake mce registers on current CPU. */
  31 static void inject_mce(struct mce *m)
  32 {
  33         struct mce *i = &per_cpu(injectm, m->extcpu);
  34
  35         /* Make sure no one reads partially written injectm */
  36         i->finished = 0;
  37         mb();
  38         m->finished = 0;
  39         /* First set the fields after finished */
  40         i->extcpu = m->extcpu;
  41         mb();
  42         /* Now write record in order, finished last (except above) */
  43         memcpy(i, m, sizeof(struct mce));
  44         /* Finally activate it */
  45         mb();
  46         i->finished = 1;
  47 }
  48
  49 static void raise_poll(struct mce *m)
  50 {
  51         unsigned long flags;
  52         mce_banks_t b;
  53
  54         memset(&b, 0xff, sizeof(mce_banks_t));
  55         local_irq_save(flags);
  56         machine_check_poll(0, &b);
  57         local_irq_restore(flags);
  58         m->finished = 0;
  59 }
  60
  61 static void raise_exception(struct mce *m, struct pt_regs *pregs)
  62 {
  63         struct pt_regs regs;
  64         unsigned long flags;
  65
  66         if (!pregs) {
  67                 memset(&regs, 0, sizeof(struct pt_regs));
  68                 regs.ip = m->ip;
  69                 regs.cs = m->cs;
  70                 pregs = &regs;
  71         }
  72         /* in mcheck exeception handler, irq will be disabled */
  73         local_irq_save(flags);
  74         do_machine_check(pregs, 0);
  75         local_irq_restore(flags);
  76         m->finished = 0;
  77 }
  78
  79 static cpumask_var_t mce_inject_cpumask;
  80
  81 static int mce_raise_notify(unsigned int cmd, struct pt_regs *regs)
  82 {
  83         int cpu = smp_processor_id();
  84         struct mce *m = &__get_cpu_var(injectm);
  85         if (!cpumask_test_cpu(cpu, mce_inject_cpumask))
  86                 return NMI_DONE;
  87         cpumask_clear_cpu(cpu, mce_inject_cpumask);
  88         if (m->inject_flags & MCJ_EXCEPTION)
  89                 raise_exception(m, regs);
  90         else if (m->status)
  91                 raise_poll(m);
  92         return NMI_HANDLED;
  93 }
  94
  95 /* Inject mce on current CPU */
  96 static int raise_local(void)
  97 {
  98         struct mce *m = &__get_cpu_var(injectm);
  99         int context = MCJ_CTX(m->inject_flags);
 100         int ret = 0;
 101         int cpu = m->extcpu;
 102
 103         if (m->inject_flags & MCJ_EXCEPTION) {
 104                 printk(KERN_INFO "Triggering MCE exception on CPU %d\n", cpu);
 105                 switch (context) {
 106                 case MCJ_CTX_IRQ:
 107                         /*
 108                          * Could do more to fake interrupts like
 109                          * calling irq_enter, but the necessary
 110                          * machinery isn't exported currently.
 111                          */
 112                         /*FALL THROUGH*/
 113                 case MCJ_CTX_PROCESS:
 114                         raise_exception(m, NULL);
 115                         break;
 116                 default:
 117                         printk(KERN_INFO "Invalid MCE context\n");
 118                         ret = -EINVAL;
 119                 }
 120                 printk(KERN_INFO "MCE exception done on CPU %d\n", cpu);
 121         } else if (m->status) {
 122                 printk(KERN_INFO "Starting machine check poll CPU %d\n", cpu);
 123                 raise_poll(m);
 124                 mce_notify_irq();
 125                 printk(KERN_INFO "Machine check poll done on CPU %d\n", cpu);
 126         } else
 127                 m->finished = 0;
 128
 129         return ret;
 130 }
 131
 132 static void raise_mce(struct mce *m)
 133 {
 134         int context = MCJ_CTX(m->inject_flags);
 135
 136         inject_mce(m);
 137
 138         if (context == MCJ_CTX_RANDOM)
 139                 return;
 140
 141 #ifdef CONFIG_X86_LOCAL_APIC
 142         if (m->inject_flags & MCJ_NMI_BROADCAST) {
 143                 unsigned long start;
 144                 int cpu;
 145                 get_online_cpus();
 146                 cpumask_copy(mce_inject_cpumask, cpu_online_mask);
 147                 cpumask_clear_cpu(get_cpu(), mce_inject_cpumask);
 148                 for_each_online_cpu(cpu) {
 149                         struct mce *mcpu = &per_cpu(injectm, cpu);
 150                         if (!mcpu->finished ||
 151                             MCJ_CTX(mcpu->inject_flags) != MCJ_CTX_RANDOM)
 152                                 cpumask_clear_cpu(cpu, mce_inject_cpumask);
 153                 }
 154                 if (!cpumask_empty(mce_inject_cpumask))
 155                         apic->send_IPI_mask(mce_inject_cpumask, NMI_VECTOR);
 156                 start = jiffies;
 157                 while (!cpumask_empty(mce_inject_cpumask)) {
 158                         if (!time_before(jiffies, start + 2*HZ)) {
 159                                 printk(KERN_ERR
 160                                 "Timeout waiting for mce inject NMI %lx\n",
 161                                         *cpumask_bits(mce_inject_cpumask));
 162                                 break;
 163                         }
 164                         cpu_relax();
 165                 }
 166                 raise_local();
 167                 put_cpu();
 168                 put_online_cpus();
 169         } else
 170 #endif
 171                 raise_local();
 172 }
 173
 174 /* Error injection interface */
 175 static ssize_t mce_write(struct file *filp, const char __user *ubuf,
 176                          size_t usize, loff_t *off)
 177 {
 178         struct mce m;
 179
 180         if (!capable(CAP_SYS_ADMIN))
 181                 return -EPERM;
 182         /*
 183          * There are some cases where real MSR reads could slip
 184          * through.
 185          */
 186         if (!boot_cpu_has(X86_FEATURE_MCE) || !boot_cpu_has(X86_FEATURE_MCA))
 187                 return -EIO;
 188
 189         if ((unsigned long)usize > sizeof(struct mce))
 190                 usize = sizeof(struct mce);
 191         if (copy_from_user(&m, ubuf, usize))
 192                 return -EFAULT;
 193
 194         if (m.extcpu >= num_possible_cpus() || !cpu_online(m.extcpu))
 195                 return -EINVAL;
 196
 197         /*
 198          * Need to give user space some time to set everything up,
 199          * so do it a jiffie or two later everywhere.
 200          */
 201         schedule_timeout(2);
 202         raise_mce(&m);
 203         return usize;
 204 }
 205
 206 static int inject_init(void)
 207 {
 208         if (!alloc_cpumask_var(&mce_inject_cpumask, GFP_KERNEL))
 209                 return -ENOMEM;
 210         printk(KERN_INFO "Machine check injector initialized\n");
 211         register_mce_write_callback(mce_write);
 212         register_nmi_handler(NMI_LOCAL, mce_raise_notify, 0,
 213                                 "mce_notify");
 214         return 0;
 215 }
 216
 217 module_init(inject_init);
 218 /*
 219  * Cannot tolerate unloading currently because we cannot
 220  * guarantee all openers of mce_chrdev will get a reference to us.
 221  */
 222 MODULE_LICENSE("GPL");