drivers/hwmon/coretemp.c

   1 /*
   2  * coretemp.c - Linux kernel module for hardware monitoring
   3  *
   4  * Copyright (C) 2007 Rudolf Marek <r.marek@assembler.cz>
   5  *
   6  * Inspired from many hwmon drivers
   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; version 2 of the License.
  11  *
  12  * This program is distributed in the hope that it will be useful,
  13  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  14  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
  15  * GNU General Public License for more details.
  16  *
  17  * You should have received a copy of the GNU General Public License
  18  * along with this program; if not, write to the Free Software
  19  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA
  20  * 02110-1301 USA.
  21  */
  22
  23 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
  24
  25 #include <linux/module.h>
  26 #include <linux/init.h>
  27 #include <linux/slab.h>
  28 #include <linux/jiffies.h>
  29 #include <linux/hwmon.h>
  30 #include <linux/sysfs.h>
  31 #include <linux/hwmon-sysfs.h>
  32 #include <linux/err.h>
  33 #include <linux/mutex.h>
  34 #include <linux/list.h>
  35 #include <linux/platform_device.h>
  36 #include <linux/cpu.h>
  37 #include <linux/pci.h>
  38 #include <linux/smp.h>
  39 #include <linux/moduleparam.h>
  40 #include <asm/msr.h>
  41 #include <asm/processor.h>
  42 #include <asm/cpu_device_id.h>
  43
  44 #define DRVNAME "coretemp"
  45
  46 /*
  47  * force_tjmax only matters when TjMax can't be read from the CPU itself.
  48  * When set, it replaces the driver's suboptimal heuristic.
  49  */
  50 static int force_tjmax;
  51 module_param_named(tjmax, force_tjmax, int, 0444);
  52 MODULE_PARM_DESC(tjmax, "TjMax value in degrees Celsius");
  53
  54 #define BASE_SYSFS_ATTR_NO      2       /* Sysfs Base attr no for coretemp */
  55 #define NUM_REAL_CORES          32      /* Number of Real cores per cpu */
  56 #define CORETEMP_NAME_LENGTH    17      /* String Length of attrs */
  57 #define MAX_CORE_ATTRS          4       /* Maximum no of basic attrs */
  58 #define TOTAL_ATTRS             (MAX_CORE_ATTRS + 1)
  59 #define MAX_CORE_DATA           (NUM_REAL_CORES + BASE_SYSFS_ATTR_NO)
  60
  61 #define TO_PHYS_ID(cpu)         (cpu_data(cpu).phys_proc_id)
  62 #define TO_CORE_ID(cpu)         (cpu_data(cpu).cpu_core_id)
  63 #define TO_ATTR_NO(cpu)         (TO_CORE_ID(cpu) + BASE_SYSFS_ATTR_NO)
  64
  65 #ifdef CONFIG_SMP
  66 #define for_each_sibling(i, cpu)        for_each_cpu(i, cpu_sibling_mask(cpu))
  67 #else
  68 #define for_each_sibling(i, cpu)        for (i = 0; false; )
  69 #endif
  70
  71 /*
  72  * Per-Core Temperature Data
  73  * @last_updated: The time when the current temperature value was updated
  74  *              earlier (in jiffies).
  75  * @cpu_core_id: The CPU Core from which temperature values should be read
  76  *              This value is passed as "id" field to rdmsr/wrmsr functions.
  77  * @status_reg: One of IA32_THERM_STATUS or IA32_PACKAGE_THERM_STATUS,
  78  *              from where the temperature values should be read.
  79  * @attr_size:  Total number of pre-core attrs displayed in the sysfs.
  80  * @is_pkg_data: If this is 1, the temp_data holds pkgtemp data.
  81  *              Otherwise, temp_data holds coretemp data.
  82  * @valid: If this is 1, the current temperature is valid.
  83  */
  84 struct temp_data {
  85         int temp;
  86         int ttarget;
  87         int tjmax;
  88         unsigned long last_updated;
  89         unsigned int cpu;
  90         u32 cpu_core_id;
  91         u32 status_reg;
  92         int attr_size;
  93         bool is_pkg_data;
  94         bool valid;
  95         struct sensor_device_attribute sd_attrs[TOTAL_ATTRS];
  96         char attr_name[TOTAL_ATTRS][CORETEMP_NAME_LENGTH];
  97         struct mutex update_lock;
  98 };
  99
 100 /* Platform Data per Physical CPU */
 101 struct platform_data {
 102         struct device *hwmon_dev;
 103         u16 phys_proc_id;
 104         struct temp_data *core_data[MAX_CORE_DATA];
 105         struct device_attribute name_attr;
 106 };
 107
 108 struct pdev_entry {
 109         struct list_head list;
 110         struct platform_device *pdev;
 111         u16 phys_proc_id;
 112 };
 113
 114 static LIST_HEAD(pdev_list);
 115 static DEFINE_MUTEX(pdev_list_mutex);
 116
 117 static ssize_t show_name(struct device *dev,
 118                         struct device_attribute *devattr, char *buf)
 119 {
 120         return sprintf(buf, "%s\n", DRVNAME);
 121 }
 122
 123 static ssize_t show_label(struct device *dev,
 124                                 struct device_attribute *devattr, char *buf)
 125 {
 126         struct sensor_device_attribute *attr = to_sensor_dev_attr(devattr);
 127         struct platform_data *pdata = dev_get_drvdata(dev);
 128         struct temp_data *tdata = pdata->core_data[attr->index];
 129
 130         if (tdata->is_pkg_data)
 131                 return sprintf(buf, "Physical id %u\n", pdata->phys_proc_id);
 132
 133         return sprintf(buf, "Core %u\n", tdata->cpu_core_id);
 134 }
 135
 136 static ssize_t show_crit_alarm(struct device *dev,
 137                                 struct device_attribute *devattr, char *buf)
 138 {
 139         u32 eax, edx;
 140         struct sensor_device_attribute *attr = to_sensor_dev_attr(devattr);
 141         struct platform_data *pdata = dev_get_drvdata(dev);
 142         struct temp_data *tdata = pdata->core_data[attr->index];
 143
 144         rdmsr_on_cpu(tdata->cpu, tdata->status_reg, &eax, &edx);
 145
 146         return sprintf(buf, "%d\n", (eax >> 5) & 1);
 147 }
 148
 149 static ssize_t show_tjmax(struct device *dev,
 150                         struct device_attribute *devattr, char *buf)
 151 {
 152         struct sensor_device_attribute *attr = to_sensor_dev_attr(devattr);
 153         struct platform_data *pdata = dev_get_drvdata(dev);
 154
 155         return sprintf(buf, "%d\n", pdata->core_data[attr->index]->tjmax);
 156 }
 157
 158 static ssize_t show_ttarget(struct device *dev,
 159                                 struct device_attribute *devattr, char *buf)
 160 {
 161         struct sensor_device_attribute *attr = to_sensor_dev_attr(devattr);
 162         struct platform_data *pdata = dev_get_drvdata(dev);
 163
 164         return sprintf(buf, "%d\n", pdata->core_data[attr->index]->ttarget);
 165 }
 166
 167 static ssize_t show_temp(struct device *dev,
 168                         struct device_attribute *devattr, char *buf)
 169 {
 170         u32 eax, edx;
 171         struct sensor_device_attribute *attr = to_sensor_dev_attr(devattr);
 172         struct platform_data *pdata = dev_get_drvdata(dev);
 173         struct temp_data *tdata = pdata->core_data[attr->index];
 174
 175         mutex_lock(&tdata->update_lock);
 176
 177         /* Check whether the time interval has elapsed */
 178         if (!tdata->valid || time_after(jiffies, tdata->last_updated + HZ)) {
 179                 rdmsr_on_cpu(tdata->cpu, tdata->status_reg, &eax, &edx);
 180                 tdata->valid = 0;
 181                 /* Check whether the data is valid */
 182                 if (eax & 0x80000000) {
 183                         tdata->temp = tdata->tjmax -
 184                                         ((eax >> 16) & 0x7f) * 1000;
 185                         tdata->valid = 1;
 186                 }
 187                 tdata->last_updated = jiffies;
 188         }
 189
 190         mutex_unlock(&tdata->update_lock);
 191         return tdata->valid ? sprintf(buf, "%d\n", tdata->temp) : -EAGAIN;
 192 }
 193
 194 static int __cpuinit adjust_tjmax(struct cpuinfo_x86 *c, u32 id,
 195                                   struct device *dev)
 196 {
 197         /* The 100C is default for both mobile and non mobile CPUs */
 198
 199         int tjmax = 100000;
 200         int tjmax_ee = 85000;
 201         int usemsr_ee = 1;
 202         int err;
 203         u32 eax, edx;
 204         struct pci_dev *host_bridge;
 205
 206         /* Early chips have no MSR for TjMax */
 207
 208         if (c->x86_model == 0xf && c->x86_mask < 4)
 209                 usemsr_ee = 0;
 210
 211         /* Atom CPUs */
 212
 213         if (c->x86_model == 0x1c) {
 214                 usemsr_ee = 0;
 215
 216                 host_bridge = pci_get_bus_and_slot(0, PCI_DEVFN(0, 0));
 217
 218                 if (host_bridge && host_bridge->vendor == PCI_VENDOR_ID_INTEL
 219                     && (host_bridge->device == 0xa000   /* NM10 based nettop */
 220                     || host_bridge->device == 0xa010))  /* NM10 based netbook */
 221                         tjmax = 100000;
 222                 else
 223                         tjmax = 90000;
 224
 225                 pci_dev_put(host_bridge);
 226         }
 227
 228         if (c->x86_model > 0xe && usemsr_ee) {
 229                 u8 platform_id;
 230
 231                 /*
 232                  * Now we can detect the mobile CPU using Intel provided table
 233                  * http://softwarecommunity.intel.com/Wiki/Mobility/720.htm
 234                  * For Core2 cores, check MSR 0x17, bit 28 1 = Mobile CPU
 235                  */
 236                 err = rdmsr_safe_on_cpu(id, 0x17, &eax, &edx);
 237                 if (err) {
 238                         dev_warn(dev,
 239                                  "Unable to access MSR 0x17, assuming desktop"
 240                                  " CPU\n");
 241                         usemsr_ee = 0;
 242                 } else if (c->x86_model < 0x17 && !(eax & 0x10000000)) {
 243                         /*
 244                          * Trust bit 28 up to Penryn, I could not find any
 245                          * documentation on that; if you happen to know
 246                          * someone at Intel please ask
 247                          */
 248                         usemsr_ee = 0;
 249                 } else {
 250                         /* Platform ID bits 52:50 (EDX starts at bit 32) */
 251                         platform_id = (edx >> 18) & 0x7;
 252
 253                         /*
 254                          * Mobile Penryn CPU seems to be platform ID 7 or 5
 255                          * (guesswork)
 256                          */
 257                         if (c->x86_model == 0x17 &&
 258                             (platform_id == 5 || platform_id == 7)) {
 259                                 /*
 260                                  * If MSR EE bit is set, set it to 90 degrees C,
 261                                  * otherwise 105 degrees C
 262                                  */
 263                                 tjmax_ee = 90000;
 264                                 tjmax = 105000;
 265                         }
 266                 }
 267         }
 268
 269         if (usemsr_ee) {
 270                 err = rdmsr_safe_on_cpu(id, 0xee, &eax, &edx);
 271                 if (err) {
 272                         dev_warn(dev,
 273                                  "Unable to access MSR 0xEE, for Tjmax, left"
 274                                  " at default\n");
 275                 } else if (eax & 0x40000000) {
 276                         tjmax = tjmax_ee;
 277                 }
 278         } else if (tjmax == 100000) {
 279                 /*
 280                  * If we don't use msr EE it means we are desktop CPU
 281                  * (with exeception of Atom)
 282                  */
 283                 dev_warn(dev, "Using relative temperature scale!\n");
 284         }
 285
 286         return tjmax;
 287 }
 288
 289 static int __cpuinit get_tjmax(struct cpuinfo_x86 *c, u32 id,
 290                                struct device *dev)
 291 {
 292         int err;
 293         u32 eax, edx;
 294         u32 val;
 295
 296         /*
 297          * A new feature of current Intel(R) processors, the
 298          * IA32_TEMPERATURE_TARGET contains the TjMax value
 299          */
 300         err = rdmsr_safe_on_cpu(id, MSR_IA32_TEMPERATURE_TARGET, &eax, &edx);
 301         if (err) {
 302                 if (c->x86_model > 0xe && c->x86_model != 0x1c)
 303                         dev_warn(dev, "Unable to read TjMax from CPU %u\n", id);
 304         } else {
 305                 val = (eax >> 16) & 0xff;
 306                 /*
 307                  * If the TjMax is not plausible, an assumption
 308                  * will be used
 309                  */
 310                 if (val) {
 311                         dev_dbg(dev, "TjMax is %d degrees C\n", val);
 312                         return val * 1000;
 313                 }
 314         }
 315
 316         if (force_tjmax) {
 317                 dev_notice(dev, "TjMax forced to %d degrees C by user\n",
 318                            force_tjmax);
 319                 return force_tjmax * 1000;
 320         }
 321
 322         /*
 323          * An assumption is made for early CPUs and unreadable MSR.
 324          * NOTE: the calculated value may not be correct.
 325          */
 326         return adjust_tjmax(c, id, dev);
 327 }
 328
 329 static int __devinit create_name_attr(struct platform_data *pdata,
 330                                       struct device *dev)
 331 {
 332         sysfs_attr_init(&pdata->name_attr.attr);
 333         pdata->name_attr.attr.name = "name";
 334         pdata->name_attr.attr.mode = S_IRUGO;
 335         pdata->name_attr.show = show_name;
 336         return device_create_file(dev, &pdata->name_attr);
 337 }
 338
 339 static int __cpuinit create_core_attrs(struct temp_data *tdata,
 340                                        struct device *dev, int attr_no)
 341 {
 342         int err, i;
 343         static ssize_t (*const rd_ptr[TOTAL_ATTRS]) (struct device *dev,
 344                         struct device_attribute *devattr, char *buf) = {
 345                         show_label, show_crit_alarm, show_temp, show_tjmax,
 346                         show_ttarget };
 347         static const char *const names[TOTAL_ATTRS] = {
 348                                         "temp%d_label", "temp%d_crit_alarm",
 349                                         "temp%d_input", "temp%d_crit",
 350                                         "temp%d_max" };
 351
 352         for (i = 0; i < tdata->attr_size; i++) {
 353                 snprintf(tdata->attr_name[i], CORETEMP_NAME_LENGTH, names[i],
 354                         attr_no);
 355                 sysfs_attr_init(&tdata->sd_attrs[i].dev_attr.attr);
 356                 tdata->sd_attrs[i].dev_attr.attr.name = tdata->attr_name[i];
 357                 tdata->sd_attrs[i].dev_attr.attr.mode = S_IRUGO;
 358                 tdata->sd_attrs[i].dev_attr.show = rd_ptr[i];
 359                 tdata->sd_attrs[i].index = attr_no;
 360                 err = device_create_file(dev, &tdata->sd_attrs[i].dev_attr);
 361                 if (err)
 362                         goto exit_free;
 363         }
 364         return 0;
 365
 366 exit_free:
 367         while (--i >= 0)
 368                 device_remove_file(dev, &tdata->sd_attrs[i].dev_attr);
 369         return err;
 370 }
 371
 372
 373 static int __cpuinit chk_ucode_version(unsigned int cpu)
 374 {
 375         struct cpuinfo_x86 *c = &cpu_data(cpu);
 376
 377         /*
 378          * Check if we have problem with errata AE18 of Core processors:
 379          * Readings might stop update when processor visited too deep sleep,
 380          * fixed for stepping D0 (6EC).
 381          */
 382         if (c->x86_model == 0xe && c->x86_mask < 0xc && c->microcode < 0x39) {
 383                 pr_err("Errata AE18 not fixed, update BIOS or "
 384                        "microcode of the CPU!\n");
 385                 return -ENODEV;
 386         }
 387         return 0;
 388 }
 389
 390 static struct platform_device __cpuinit *coretemp_get_pdev(unsigned int cpu)
 391 {
 392         u16 phys_proc_id = TO_PHYS_ID(cpu);
 393         struct pdev_entry *p;
 394
 395         mutex_lock(&pdev_list_mutex);
 396
 397         list_for_each_entry(p, &pdev_list, list)
 398                 if (p->phys_proc_id == phys_proc_id) {
 399                         mutex_unlock(&pdev_list_mutex);
 400                         return p->pdev;
 401                 }
 402
 403         mutex_unlock(&pdev_list_mutex);
 404         return NULL;
 405 }
 406
 407 static struct temp_data __cpuinit *init_temp_data(unsigned int cpu,
 408                                                   int pkg_flag)
 409 {
 410         struct temp_data *tdata;
 411
 412         tdata = kzalloc(sizeof(struct temp_data), GFP_KERNEL);
 413         if (!tdata)
 414                 return NULL;
 415
 416         tdata->status_reg = pkg_flag ? MSR_IA32_PACKAGE_THERM_STATUS :
 417                                                         MSR_IA32_THERM_STATUS;
 418         tdata->is_pkg_data = pkg_flag;
 419         tdata->cpu = cpu;
 420         tdata->cpu_core_id = TO_CORE_ID(cpu);
 421         tdata->attr_size = MAX_CORE_ATTRS;
 422         mutex_init(&tdata->update_lock);
 423         return tdata;
 424 }
 425
 426 static int __cpuinit create_core_data(struct platform_device *pdev,
 427                                 unsigned int cpu, int pkg_flag)
 428 {
 429         struct temp_data *tdata;
 430         struct platform_data *pdata = platform_get_drvdata(pdev);
 431         struct cpuinfo_x86 *c = &cpu_data(cpu);
 432         u32 eax, edx;
 433         int err, attr_no;
 434
 435         /*
 436          * Find attr number for sysfs:
 437          * We map the attr number to core id of the CPU
 438          * The attr number is always core id + 2
 439          * The Pkgtemp will always show up as temp1_*, if available
 440          */
 441         attr_no = pkg_flag ? 1 : TO_ATTR_NO(cpu);
 442
 443         if (attr_no > MAX_CORE_DATA - 1)
 444                 return -ERANGE;
 445
 446         /*
 447          * Provide a single set of attributes for all HT siblings of a core
 448          * to avoid duplicate sensors (the processor ID and core ID of all
 449          * HT siblings of a core are the same).
 450          * Skip if a HT sibling of this core is already registered.
 451          * This is not an error.
 452          */
 453         if (pdata->core_data[attr_no] != NULL)
 454                 return 0;
 455
 456         tdata = init_temp_data(cpu, pkg_flag);
 457         if (!tdata)
 458                 return -ENOMEM;
 459
 460         /* Test if we can access the status register */
 461         err = rdmsr_safe_on_cpu(cpu, tdata->status_reg, &eax, &edx);
 462         if (err)
 463                 goto exit_free;
 464
 465         /* We can access status register. Get Critical Temperature */
 466         tdata->tjmax = get_tjmax(c, cpu, &pdev->dev);
 467
 468         /*
 469          * Read the still undocumented bits 8:15 of IA32_TEMPERATURE_TARGET.
 470          * The target temperature is available on older CPUs but not in this
 471          * register. Atoms don't have the register at all.
 472          */
 473         if (c->x86_model > 0xe && c->x86_model != 0x1c) {
 474                 err = rdmsr_safe_on_cpu(cpu, MSR_IA32_TEMPERATURE_TARGET,
 475                                         &eax, &edx);
 476                 if (!err) {
 477                         tdata->ttarget
 478                           = tdata->tjmax - ((eax >> 8) & 0xff) * 1000;
 479                         tdata->attr_size++;
 480                 }
 481         }
 482
 483         pdata->core_data[attr_no] = tdata;
 484
 485         /* Create sysfs interfaces */
 486         err = create_core_attrs(tdata, &pdev->dev, attr_no);
 487         if (err)
 488                 goto exit_free;
 489
 490         return 0;
 491 exit_free:
 492         pdata->core_data[attr_no] = NULL;
 493         kfree(tdata);
 494         return err;
 495 }
 496
 497 static void __cpuinit coretemp_add_core(unsigned int cpu, int pkg_flag)
 498 {
 499         struct platform_device *pdev = coretemp_get_pdev(cpu);
 500         int err;
 501
 502         if (!pdev)
 503                 return;
 504
 505         err = create_core_data(pdev, cpu, pkg_flag);
 506         if (err)
 507                 dev_err(&pdev->dev, "Adding Core %u failed\n", cpu);
 508 }
 509
 510 static void coretemp_remove_core(struct platform_data *pdata,
 511                                 struct device *dev, int indx)
 512 {
 513         int i;
 514         struct temp_data *tdata = pdata->core_data[indx];
 515
 516         /* Remove the sysfs attributes */
 517         for (i = 0; i < tdata->attr_size; i++)
 518                 device_remove_file(dev, &tdata->sd_attrs[i].dev_attr);
 519
 520         kfree(pdata->core_data[indx]);
 521         pdata->core_data[indx] = NULL;
 522 }
 523
 524 static int __devinit coretemp_probe(struct platform_device *pdev)
 525 {
 526         struct platform_data *pdata;
 527         int err;
 528
 529         /* Initialize the per-package data structures */
 530         pdata = kzalloc(sizeof(struct platform_data), GFP_KERNEL);
 531         if (!pdata)
 532                 return -ENOMEM;
 533
 534         err = create_name_attr(pdata, &pdev->dev);
 535         if (err)
 536                 goto exit_free;
 537
 538         pdata->phys_proc_id = pdev->id;
 539         platform_set_drvdata(pdev, pdata);
 540
 541         pdata->hwmon_dev = hwmon_device_register(&pdev->dev);
 542         if (IS_ERR(pdata->hwmon_dev)) {
 543                 err = PTR_ERR(pdata->hwmon_dev);
 544                 dev_err(&pdev->dev, "Class registration failed (%d)\n", err);
 545                 goto exit_name;
 546         }
 547         return 0;
 548
 549 exit_name:
 550         device_remove_file(&pdev->dev, &pdata->name_attr);
 551         platform_set_drvdata(pdev, NULL);
 552 exit_free:
 553         kfree(pdata);
 554         return err;
 555 }
 556
 557 static int __devexit coretemp_remove(struct platform_device *pdev)
 558 {
 559         struct platform_data *pdata = platform_get_drvdata(pdev);
 560         int i;
 561
 562         for (i = MAX_CORE_DATA - 1; i >= 0; --i)
 563                 if (pdata->core_data[i])
 564                         coretemp_remove_core(pdata, &pdev->dev, i);
 565
 566         device_remove_file(&pdev->dev, &pdata->name_attr);
 567         hwmon_device_unregister(pdata->hwmon_dev);
 568         platform_set_drvdata(pdev, NULL);
 569         kfree(pdata);
 570         return 0;
 571 }
 572
 573 static struct platform_driver coretemp_driver = {
 574         .driver = {
 575                 .owner = THIS_MODULE,
 576                 .name = DRVNAME,
 577         },
 578         .probe = coretemp_probe,
 579         .remove = __devexit_p(coretemp_remove),
 580 };
 581
 582 static int __cpuinit coretemp_device_add(unsigned int cpu)
 583 {
 584         int err;
 585         struct platform_device *pdev;
 586         struct pdev_entry *pdev_entry;
 587
 588         mutex_lock(&pdev_list_mutex);
 589
 590         pdev = platform_device_alloc(DRVNAME, TO_PHYS_ID(cpu));
 591         if (!pdev) {
 592                 err = -ENOMEM;
 593                 pr_err("Device allocation failed\n");
 594                 goto exit;
 595         }
 596
 597         pdev_entry = kzalloc(sizeof(struct pdev_entry), GFP_KERNEL);
 598         if (!pdev_entry) {
 599                 err = -ENOMEM;
 600                 goto exit_device_put;
 601         }
 602
 603         err = platform_device_add(pdev);
 604         if (err) {
 605                 pr_err("Device addition failed (%d)\n", err);
 606                 goto exit_device_free;
 607         }
 608
 609         pdev_entry->pdev = pdev;
 610         pdev_entry->phys_proc_id = pdev->id;
 611
 612         list_add_tail(&pdev_entry->list, &pdev_list);
 613         mutex_unlock(&pdev_list_mutex);
 614
 615         return 0;
 616
 617 exit_device_free:
 618         kfree(pdev_entry);
 619 exit_device_put:
 620         platform_device_put(pdev);
 621 exit:
 622         mutex_unlock(&pdev_list_mutex);
 623         return err;
 624 }
 625
 626 static void __cpuinit coretemp_device_remove(unsigned int cpu)
 627 {
 628         struct pdev_entry *p, *n;
 629         u16 phys_proc_id = TO_PHYS_ID(cpu);
 630
 631         mutex_lock(&pdev_list_mutex);
 632         list_for_each_entry_safe(p, n, &pdev_list, list) {
 633                 if (p->phys_proc_id != phys_proc_id)
 634                         continue;
 635                 platform_device_unregister(p->pdev);
 636                 list_del(&p->list);
 637                 kfree(p);
 638         }
 639         mutex_unlock(&pdev_list_mutex);
 640 }
 641
 642 static bool __cpuinit is_any_core_online(struct platform_data *pdata)
 643 {
 644         int i;
 645
 646         /* Find online cores, except pkgtemp data */
 647         for (i = MAX_CORE_DATA - 1; i >= 0; --i) {
 648                 if (pdata->core_data[i] &&
 649                         !pdata->core_data[i]->is_pkg_data) {
 650                         return true;
 651                 }
 652         }
 653         return false;
 654 }
 655
 656 static void __cpuinit get_core_online(unsigned int cpu)
 657 {
 658         struct cpuinfo_x86 *c = &cpu_data(cpu);
 659         struct platform_device *pdev = coretemp_get_pdev(cpu);
 660         int err;
 661
 662         /*
 663          * CPUID.06H.EAX[0] indicates whether the CPU has thermal
 664          * sensors. We check this bit only, all the early CPUs
 665          * without thermal sensors will be filtered out.
 666          */
 667         if (!cpu_has(c, X86_FEATURE_DTS))
 668                 return;
 669
 670         if (!pdev) {
 671                 /* Check the microcode version of the CPU */
 672                 if (chk_ucode_version(cpu))
 673                         return;
 674
 675                 /*
 676                  * Alright, we have DTS support.
 677                  * We are bringing the _first_ core in this pkg
 678                  * online. So, initialize per-pkg data structures and
 679                  * then bring this core online.
 680                  */
 681                 err = coretemp_device_add(cpu);
 682                 if (err)
 683                         return;
 684                 /*
 685                  * Check whether pkgtemp support is available.
 686                  * If so, add interfaces for pkgtemp.
 687                  */
 688                 if (cpu_has(c, X86_FEATURE_PTS))
 689                         coretemp_add_core(cpu, 1);
 690         }
 691         /*
 692          * Physical CPU device already exists.
 693          * So, just add interfaces for this core.
 694          */
 695         coretemp_add_core(cpu, 0);
 696 }
 697
 698 static void __cpuinit put_core_offline(unsigned int cpu)
 699 {
 700         int i, indx;
 701         struct platform_data *pdata;
 702         struct platform_device *pdev = coretemp_get_pdev(cpu);
 703
 704         /* If the physical CPU device does not exist, just return */
 705         if (!pdev)
 706                 return;
 707
 708         pdata = platform_get_drvdata(pdev);
 709
 710         indx = TO_ATTR_NO(cpu);
 711
 712         /* The core id is too big, just return */
 713         if (indx > MAX_CORE_DATA - 1)
 714                 return;
 715
 716         if (pdata->core_data[indx] && pdata->core_data[indx]->cpu == cpu)
 717                 coretemp_remove_core(pdata, &pdev->dev, indx);
 718
 719         /*
 720          * If a HT sibling of a core is taken offline, but another HT sibling
 721          * of the same core is still online, register the alternate sibling.
 722          * This ensures that exactly one set of attributes is provided as long
 723          * as at least one HT sibling of a core is online.
 724          */
 725         for_each_sibling(i, cpu) {
 726                 if (i != cpu) {
 727                         get_core_online(i);
 728                         /*
 729                          * Display temperature sensor data for one HT sibling
 730                          * per core only, so abort the loop after one such
 731                          * sibling has been found.
 732                          */
 733                         break;
 734                 }
 735         }
 736         /*
 737          * If all cores in this pkg are offline, remove the device.
 738          * coretemp_device_remove calls unregister_platform_device,
 739          * which in turn calls coretemp_remove. This removes the
 740          * pkgtemp entry and does other clean ups.
 741          */
 742         if (!is_any_core_online(pdata))
 743                 coretemp_device_remove(cpu);
 744 }
 745
 746 static int __cpuinit coretemp_cpu_callback(struct notifier_block *nfb,
 747                                  unsigned long action, void *hcpu)
 748 {
 749         unsigned int cpu = (unsigned long) hcpu;
 750
 751         switch (action) {
 752         case CPU_ONLINE:
 753         case CPU_DOWN_FAILED:
 754                 get_core_online(cpu);
 755                 break;
 756         case CPU_DOWN_PREPARE:
 757                 put_core_offline(cpu);
 758                 break;
 759         }
 760         return NOTIFY_OK;
 761 }
 762
 763 static struct notifier_block coretemp_cpu_notifier __refdata = {
 764         .notifier_call = coretemp_cpu_callback,
 765 };
 766
 767 static const struct x86_cpu_id coretemp_ids[] = {
 768         { X86_VENDOR_INTEL, X86_FAMILY_ANY, X86_MODEL_ANY, X86_FEATURE_DTS },
 769         {}
 770 };
 771 MODULE_DEVICE_TABLE(x86cpu, coretemp_ids);
 772
 773 static int __init coretemp_init(void)
 774 {
 775         int i, err = -ENODEV;
 776
 777         /*
 778          * CPUID.06H.EAX[0] indicates whether the CPU has thermal
 779          * sensors. We check this bit only, all the early CPUs
 780          * without thermal sensors will be filtered out.
 781          */
 782         if (!x86_match_cpu(coretemp_ids))
 783                 return -ENODEV;
 784
 785         err = platform_driver_register(&coretemp_driver);
 786         if (err)
 787                 goto exit;
 788
 789         for_each_online_cpu(i)
 790                 get_core_online(i);
 791
 792 #ifndef CONFIG_HOTPLUG_CPU
 793         if (list_empty(&pdev_list)) {
 794                 err = -ENODEV;
 795                 goto exit_driver_unreg;
 796         }
 797 #endif
 798
 799         register_hotcpu_notifier(&coretemp_cpu_notifier);
 800         return 0;
 801
 802 #ifndef CONFIG_HOTPLUG_CPU
 803 exit_driver_unreg:
 804         platform_driver_unregister(&coretemp_driver);
 805 #endif
 806 exit:
 807         return err;
 808 }
 809
 810 static void __exit coretemp_exit(void)
 811 {
 812         struct pdev_entry *p, *n;
 813
 814         unregister_hotcpu_notifier(&coretemp_cpu_notifier);
 815         mutex_lock(&pdev_list_mutex);
 816         list_for_each_entry_safe(p, n, &pdev_list, list) {
 817                 platform_device_unregister(p->pdev);
 818                 list_del(&p->list);
 819                 kfree(p);
 820         }
 821         mutex_unlock(&pdev_list_mutex);
 822         platform_driver_unregister(&coretemp_driver);
 823 }
 824
 825 MODULE_AUTHOR("Rudolf Marek <r.marek@assembler.cz>");
 826 MODULE_DESCRIPTION("Intel Core temperature monitor");
 827 MODULE_LICENSE("GPL");
 828
 829 module_init(coretemp_init)
 830 module_exit(coretemp_exit)