arch/ia64/kernel/err_inject.c

   1 /*
   2  * err_inject.c -
   3  *      1.) Inject errors to a processor.
   4  *      2.) Query error injection capabilities.
   5  * This driver along with user space code can be acting as an error
   6  * injection tool.
   7  *
   8  * This program is free software; you can redistribute it and/or modify
   9  * it under the terms of the GNU General Public License as published by
  10  * the Free Software Foundation; either version 2 of the License, or
  11  * (at your option) any later version.
  12  *
  13  * This program is distributed in the hope that it will be useful, but
  14  * WITHOUT ANY WARRANTY; without even the implied warranty of
  15  * MERCHANTABILITY OR FITNESS FOR A PARTICULAR PURPOSE, GOOD TITLE or
  16  * NON INFRINGEMENT.  See the GNU General Public License for more
  17  * details.
  18  *
  19  * You should have received a copy of the GNU General Public License
  20  * along with this program; if not, write to the Free Software
  21  * Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
  22  *
  23  * Written by: Fenghua Yu <fenghua.yu@intel.com>, Intel Corporation
  24  * Copyright (C) 2006, Intel Corp.  All rights reserved.
  25  *
  26  */
  27 #include <linux/device.h>
  28 #include <linux/init.h>
  29 #include <linux/mm.h>
  30 #include <linux/cpu.h>
  31 #include <linux/module.h>
  32
  33 #define ERR_INJ_DEBUG
  34
  35 #define ERR_DATA_BUFFER_SIZE 3          // Three 8-byte;
  36
  37 #define define_one_ro(name)                                             \
  38 static DEVICE_ATTR(name, 0444, show_##name, NULL)
  39
  40 #define define_one_rw(name)                                             \
  41 static DEVICE_ATTR(name, 0644, show_##name, store_##name)
  42
  43 static u64 call_start[NR_CPUS];
  44 static u64 phys_addr[NR_CPUS];
  45 static u64 err_type_info[NR_CPUS];
  46 static u64 err_struct_info[NR_CPUS];
  47 static struct {
  48         u64 data1;
  49         u64 data2;
  50         u64 data3;
  51 } __attribute__((__aligned__(16))) err_data_buffer[NR_CPUS];
  52 static s64 status[NR_CPUS];
  53 static u64 capabilities[NR_CPUS];
  54 static u64 resources[NR_CPUS];
  55
  56 #define show(name)                                                      \
  57 static ssize_t                                                          \
  58 show_##name(struct device *dev, struct device_attribute *attr,  \
  59                 char *buf)                                              \
  60 {                                                                       \
  61         u32 cpu=dev->id;                                                \
  62         return sprintf(buf, "%lx\n", name[cpu]);                        \
  63 }
  64
  65 #define store(name)                                                     \
  66 static ssize_t                                                          \
  67 store_##name(struct device *dev, struct device_attribute *attr, \
  68                                         const char *buf, size_t size)   \
  69 {                                                                       \
  70         unsigned int cpu=dev->id;                                       \
  71         name[cpu] = simple_strtoull(buf, NULL, 16);                     \
  72         return size;                                                    \
  73 }
  74
  75 show(call_start)
  76
  77 /* It's user's responsibility to call the PAL procedure on a specific
  78  * processor. The cpu number in driver is only used for storing data.
  79  */
  80 static ssize_t
  81 store_call_start(struct device *dev, struct device_attribute *attr,
  82                 const char *buf, size_t size)
  83 {
  84         unsigned int cpu=dev->id;
  85         unsigned long call_start = simple_strtoull(buf, NULL, 16);
  86
  87 #ifdef ERR_INJ_DEBUG
  88         printk(KERN_DEBUG "pal_mc_err_inject for cpu%d:\n", cpu);
  89         printk(KERN_DEBUG "err_type_info=%lx,\n", err_type_info[cpu]);
  90         printk(KERN_DEBUG "err_struct_info=%lx,\n", err_struct_info[cpu]);
  91         printk(KERN_DEBUG "err_data_buffer=%lx, %lx, %lx.\n",
  92                           err_data_buffer[cpu].data1,
  93                           err_data_buffer[cpu].data2,
  94                           err_data_buffer[cpu].data3);
  95 #endif
  96         switch (call_start) {
  97             case 0: /* Do nothing. */
  98                 break;
  99             case 1: /* Call pal_mc_error_inject in physical mode. */
 100                 status[cpu]=ia64_pal_mc_error_inject_phys(err_type_info[cpu],
 101                                         err_struct_info[cpu],
 102                                         ia64_tpa(&err_data_buffer[cpu]),
 103                                         &capabilities[cpu],
 104                                         &resources[cpu]);
 105                 break;
 106             case 2: /* Call pal_mc_error_inject in virtual mode. */
 107                 status[cpu]=ia64_pal_mc_error_inject_virt(err_type_info[cpu],
 108                                         err_struct_info[cpu],
 109                                         ia64_tpa(&err_data_buffer[cpu]),
 110                                         &capabilities[cpu],
 111                                         &resources[cpu]);
 112                 break;
 113             default:
 114                 status[cpu] = -EINVAL;
 115                 break;
 116         }
 117
 118 #ifdef ERR_INJ_DEBUG
 119         printk(KERN_DEBUG "Returns: status=%d,\n", (int)status[cpu]);
 120         printk(KERN_DEBUG "capapbilities=%lx,\n", capabilities[cpu]);
 121         printk(KERN_DEBUG "resources=%lx\n", resources[cpu]);
 122 #endif
 123         return size;
 124 }
 125
 126 show(err_type_info)
 127 store(err_type_info)
 128
 129 static ssize_t
 130 show_virtual_to_phys(struct device *dev, struct device_attribute *attr,
 131                         char *buf)
 132 {
 133         unsigned int cpu=dev->id;
 134         return sprintf(buf, "%lx\n", phys_addr[cpu]);
 135 }
 136
 137 static ssize_t
 138 store_virtual_to_phys(struct device *dev, struct device_attribute *attr,
 139                         const char *buf, size_t size)
 140 {
 141         unsigned int cpu=dev->id;
 142         u64 virt_addr=simple_strtoull(buf, NULL, 16);
 143         int ret;
 144
 145         ret = get_user_pages(current, current->mm, virt_addr,
 146                         1, VM_READ, 0, NULL, NULL);
 147         if (ret<=0) {
 148 #ifdef ERR_INJ_DEBUG
 149                 printk("Virtual address %lx is not existing.\n",virt_addr);
 150 #endif
 151                 return -EINVAL;
 152         }
 153
 154         phys_addr[cpu] = ia64_tpa(virt_addr);
 155         return size;
 156 }
 157
 158 show(err_struct_info)
 159 store(err_struct_info)
 160
 161 static ssize_t
 162 show_err_data_buffer(struct device *dev,
 163                         struct device_attribute *attr, char *buf)
 164 {
 165         unsigned int cpu=dev->id;
 166
 167         return sprintf(buf, "%lx, %lx, %lx\n",
 168                         err_data_buffer[cpu].data1,
 169                         err_data_buffer[cpu].data2,
 170                         err_data_buffer[cpu].data3);
 171 }
 172
 173 static ssize_t
 174 store_err_data_buffer(struct device *dev,
 175                         struct device_attribute *attr,
 176                         const char *buf, size_t size)
 177 {
 178         unsigned int cpu=dev->id;
 179         int ret;
 180
 181 #ifdef ERR_INJ_DEBUG
 182         printk("write err_data_buffer=[%lx,%lx,%lx] on cpu%d\n",
 183                  err_data_buffer[cpu].data1,
 184                  err_data_buffer[cpu].data2,
 185                  err_data_buffer[cpu].data3,
 186                  cpu);
 187 #endif
 188         ret=sscanf(buf, "%lx, %lx, %lx",
 189                         &err_data_buffer[cpu].data1,
 190                         &err_data_buffer[cpu].data2,
 191                         &err_data_buffer[cpu].data3);
 192         if (ret!=ERR_DATA_BUFFER_SIZE)
 193                 return -EINVAL;
 194
 195         return size;
 196 }
 197
 198 show(status)
 199 show(capabilities)
 200 show(resources)
 201
 202 define_one_rw(call_start);
 203 define_one_rw(err_type_info);
 204 define_one_rw(err_struct_info);
 205 define_one_rw(err_data_buffer);
 206 define_one_rw(virtual_to_phys);
 207 define_one_ro(status);
 208 define_one_ro(capabilities);
 209 define_one_ro(resources);
 210
 211 static struct attribute *default_attrs[] = {
 212         &dev_attr_call_start.attr,
 213         &dev_attr_virtual_to_phys.attr,
 214         &dev_attr_err_type_info.attr,
 215         &dev_attr_err_struct_info.attr,
 216         &dev_attr_err_data_buffer.attr,
 217         &dev_attr_status.attr,
 218         &dev_attr_capabilities.attr,
 219         &dev_attr_resources.attr,
 220         NULL
 221 };
 222
 223 static struct attribute_group err_inject_attr_group = {
 224         .attrs = default_attrs,
 225         .name = "err_inject"
 226 };
 227 /* Add/Remove err_inject interface for CPU device */
 228 static int err_inject_add_dev(struct device *sys_dev)
 229 {
 230         return sysfs_create_group(&sys_dev->kobj, &err_inject_attr_group);
 231 }
 232
 233 static int err_inject_remove_dev(struct device *sys_dev)
 234 {
 235         sysfs_remove_group(&sys_dev->kobj, &err_inject_attr_group);
 236         return 0;
 237 }
 238 static int err_inject_cpu_callback(struct notifier_block *nfb,
 239                 unsigned long action, void *hcpu)
 240 {
 241         unsigned int cpu = (unsigned long)hcpu;
 242         struct device *sys_dev;
 243
 244         sys_dev = get_cpu_device(cpu);
 245         switch (action) {
 246         case CPU_ONLINE:
 247         case CPU_ONLINE_FROZEN:
 248                 err_inject_add_dev(sys_dev);
 249                 break;
 250         case CPU_DEAD:
 251         case CPU_DEAD_FROZEN:
 252                 err_inject_remove_dev(sys_dev);
 253                 break;
 254         }
 255
 256         return NOTIFY_OK;
 257 }
 258
 259 static struct notifier_block err_inject_cpu_notifier =
 260 {
 261         .notifier_call = err_inject_cpu_callback,
 262 };
 263
 264 static int __init
 265 err_inject_init(void)
 266 {
 267         int i;
 268
 269 #ifdef ERR_INJ_DEBUG
 270         printk(KERN_INFO "Enter error injection driver.\n");
 271 #endif
 272
 273         cpu_notifier_register_begin();
 274
 275         for_each_online_cpu(i) {
 276                 err_inject_cpu_callback(&err_inject_cpu_notifier, CPU_ONLINE,
 277                                 (void *)(long)i);
 278         }
 279
 280         __register_hotcpu_notifier(&err_inject_cpu_notifier);
 281
 282         cpu_notifier_register_done();
 283
 284         return 0;
 285 }
 286
 287 static void __exit
 288 err_inject_exit(void)
 289 {
 290         int i;
 291         struct device *sys_dev;
 292
 293 #ifdef ERR_INJ_DEBUG
 294         printk(KERN_INFO "Exit error injection driver.\n");
 295 #endif
 296
 297         cpu_notifier_register_begin();
 298
 299         for_each_online_cpu(i) {
 300                 sys_dev = get_cpu_device(i);
 301                 sysfs_remove_group(&sys_dev->kobj, &err_inject_attr_group);
 302         }
 303
 304         __unregister_hotcpu_notifier(&err_inject_cpu_notifier);
 305
 306         cpu_notifier_register_done();
 307 }
 308
 309 module_init(err_inject_init);
 310 module_exit(err_inject_exit);
 311
 312 MODULE_AUTHOR("Fenghua Yu <fenghua.yu@intel.com>");
 313 MODULE_DESCRIPTION("MC error injection kernel sysfs interface");
 314 MODULE_LICENSE("GPL");