k10temp.c 10 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292293294295296297298299300301302303304305306307308309310311312313314315316317318319320321322323324325326327328329330331332333334335336337338339340341342343344345346347348349350351352353354355356357358359360361362363364365366367368369370371372373
  1. /*
  2. * k10temp.c - AMD Family 10h/11h/12h/14h/15h/16h processor hardware monitoring
  3. *
  4. * Copyright (c) 2009 Clemens Ladisch <clemens@ladisch.de>
  5. *
  6. *
  7. * This driver is free software; you can redistribute it and/or
  8. * modify it under the terms of the GNU General Public License; either
  9. * version 2 of the License, or (at your option) any later version.
  10. *
  11. * This driver is distributed in the hope that it will be useful,
  12. * but WITHOUT ANY WARRANTY; without even the implied warranty of
  13. * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
  14. * See the GNU General Public License for more details.
  15. *
  16. * You should have received a copy of the GNU General Public License
  17. * along with this driver; if not, see <http://www.gnu.org/licenses/>.
  18. */
  19. #include <linux/err.h>
  20. #include <linux/hwmon.h>
  21. #include <linux/hwmon-sysfs.h>
  22. #include <linux/init.h>
  23. #include <linux/module.h>
  24. #include <linux/pci.h>
  25. #include <linux/pci_ids.h>
  26. #include <asm/amd_nb.h>
  27. #include <asm/processor.h>
  28. MODULE_DESCRIPTION("AMD Family 10h+ CPU core temperature monitor");
  29. MODULE_AUTHOR("Clemens Ladisch <clemens@ladisch.de>");
  30. MODULE_LICENSE("GPL");
  31. static bool force;
  32. module_param(force, bool, 0444);
  33. MODULE_PARM_DESC(force, "force loading on processors with erratum 319");
  34. /* Provide lock for writing to NB_SMU_IND_ADDR */
  35. static DEFINE_MUTEX(nb_smu_ind_mutex);
  36. #ifndef PCI_DEVICE_ID_AMD_15H_M70H_NB_F3
  37. #define PCI_DEVICE_ID_AMD_15H_M70H_NB_F3 0x15b3
  38. #endif
  39. /* CPUID function 0x80000001, ebx */
  40. #define CPUID_PKGTYPE_MASK 0xf0000000
  41. #define CPUID_PKGTYPE_F 0x00000000
  42. #define CPUID_PKGTYPE_AM2R2_AM3 0x10000000
  43. /* DRAM controller (PCI function 2) */
  44. #define REG_DCT0_CONFIG_HIGH 0x094
  45. #define DDR3_MODE 0x00000100
  46. /* miscellaneous (PCI function 3) */
  47. #define REG_HARDWARE_THERMAL_CONTROL 0x64
  48. #define HTC_ENABLE 0x00000001
  49. #define REG_REPORTED_TEMPERATURE 0xa4
  50. #define REG_NORTHBRIDGE_CAPABILITIES 0xe8
  51. #define NB_CAP_HTC 0x00000400
  52. /*
  53. * For F15h M60h and M70h, REG_HARDWARE_THERMAL_CONTROL
  54. * and REG_REPORTED_TEMPERATURE have been moved to
  55. * D0F0xBC_xD820_0C64 [Hardware Temperature Control]
  56. * D0F0xBC_xD820_0CA4 [Reported Temperature Control]
  57. */
  58. #define F15H_M60H_HARDWARE_TEMP_CTRL_OFFSET 0xd8200c64
  59. #define F15H_M60H_REPORTED_TEMP_CTRL_OFFSET 0xd8200ca4
  60. /* F17h M01h Access througn SMN */
  61. #define F17H_M01H_REPORTED_TEMP_CTRL_OFFSET 0x00059800
  62. struct k10temp_data {
  63. struct pci_dev *pdev;
  64. void (*read_htcreg)(struct pci_dev *pdev, u32 *regval);
  65. void (*read_tempreg)(struct pci_dev *pdev, u32 *regval);
  66. int temp_offset;
  67. u32 temp_adjust_mask;
  68. bool show_tdie;
  69. };
  70. struct tctl_offset {
  71. u8 model;
  72. char const *id;
  73. int offset;
  74. };
  75. static const struct tctl_offset tctl_offset_table[] = {
  76. { 0x17, "AMD Ryzen 5 1600X", 20000 },
  77. { 0x17, "AMD Ryzen 7 1700X", 20000 },
  78. { 0x17, "AMD Ryzen 7 1800X", 20000 },
  79. { 0x17, "AMD Ryzen 7 2700X", 10000 },
  80. { 0x17, "AMD Ryzen Threadripper 19", 27000 }, /* 19{00,20,50}X */
  81. { 0x17, "AMD Ryzen Threadripper 29", 27000 }, /* 29{20,50,70,90}[W]X */
  82. };
  83. static void read_htcreg_pci(struct pci_dev *pdev, u32 *regval)
  84. {
  85. pci_read_config_dword(pdev, REG_HARDWARE_THERMAL_CONTROL, regval);
  86. }
  87. static void read_tempreg_pci(struct pci_dev *pdev, u32 *regval)
  88. {
  89. pci_read_config_dword(pdev, REG_REPORTED_TEMPERATURE, regval);
  90. }
  91. static void amd_nb_index_read(struct pci_dev *pdev, unsigned int devfn,
  92. unsigned int base, int offset, u32 *val)
  93. {
  94. mutex_lock(&nb_smu_ind_mutex);
  95. pci_bus_write_config_dword(pdev->bus, devfn,
  96. base, offset);
  97. pci_bus_read_config_dword(pdev->bus, devfn,
  98. base + 4, val);
  99. mutex_unlock(&nb_smu_ind_mutex);
  100. }
  101. static void read_htcreg_nb_f15(struct pci_dev *pdev, u32 *regval)
  102. {
  103. amd_nb_index_read(pdev, PCI_DEVFN(0, 0), 0xb8,
  104. F15H_M60H_HARDWARE_TEMP_CTRL_OFFSET, regval);
  105. }
  106. static void read_tempreg_nb_f15(struct pci_dev *pdev, u32 *regval)
  107. {
  108. amd_nb_index_read(pdev, PCI_DEVFN(0, 0), 0xb8,
  109. F15H_M60H_REPORTED_TEMP_CTRL_OFFSET, regval);
  110. }
  111. static void read_tempreg_nb_f17(struct pci_dev *pdev, u32 *regval)
  112. {
  113. amd_smn_read(amd_pci_dev_to_node_id(pdev),
  114. F17H_M01H_REPORTED_TEMP_CTRL_OFFSET, regval);
  115. }
  116. static unsigned int get_raw_temp(struct k10temp_data *data)
  117. {
  118. unsigned int temp;
  119. u32 regval;
  120. data->read_tempreg(data->pdev, &regval);
  121. temp = (regval >> 21) * 125;
  122. if (regval & data->temp_adjust_mask)
  123. temp -= 49000;
  124. return temp;
  125. }
  126. static ssize_t temp1_input_show(struct device *dev,
  127. struct device_attribute *attr, char *buf)
  128. {
  129. struct k10temp_data *data = dev_get_drvdata(dev);
  130. unsigned int temp = get_raw_temp(data);
  131. if (temp > data->temp_offset)
  132. temp -= data->temp_offset;
  133. else
  134. temp = 0;
  135. return sprintf(buf, "%u\n", temp);
  136. }
  137. static ssize_t temp2_input_show(struct device *dev,
  138. struct device_attribute *devattr, char *buf)
  139. {
  140. struct k10temp_data *data = dev_get_drvdata(dev);
  141. unsigned int temp = get_raw_temp(data);
  142. return sprintf(buf, "%u\n", temp);
  143. }
  144. static ssize_t temp_label_show(struct device *dev,
  145. struct device_attribute *devattr, char *buf)
  146. {
  147. struct sensor_device_attribute *attr = to_sensor_dev_attr(devattr);
  148. return sprintf(buf, "%s\n", attr->index ? "Tctl" : "Tdie");
  149. }
  150. static ssize_t temp1_max_show(struct device *dev,
  151. struct device_attribute *attr, char *buf)
  152. {
  153. return sprintf(buf, "%d\n", 70 * 1000);
  154. }
  155. static ssize_t show_temp_crit(struct device *dev,
  156. struct device_attribute *devattr, char *buf)
  157. {
  158. struct sensor_device_attribute *attr = to_sensor_dev_attr(devattr);
  159. struct k10temp_data *data = dev_get_drvdata(dev);
  160. int show_hyst = attr->index;
  161. u32 regval;
  162. int value;
  163. data->read_htcreg(data->pdev, &regval);
  164. value = ((regval >> 16) & 0x7f) * 500 + 52000;
  165. if (show_hyst)
  166. value -= ((regval >> 24) & 0xf) * 500;
  167. return sprintf(buf, "%d\n", value);
  168. }
  169. static DEVICE_ATTR_RO(temp1_input);
  170. static DEVICE_ATTR_RO(temp1_max);
  171. static SENSOR_DEVICE_ATTR(temp1_crit, S_IRUGO, show_temp_crit, NULL, 0);
  172. static SENSOR_DEVICE_ATTR(temp1_crit_hyst, S_IRUGO, show_temp_crit, NULL, 1);
  173. static SENSOR_DEVICE_ATTR(temp1_label, 0444, temp_label_show, NULL, 0);
  174. static DEVICE_ATTR_RO(temp2_input);
  175. static SENSOR_DEVICE_ATTR(temp2_label, 0444, temp_label_show, NULL, 1);
  176. static umode_t k10temp_is_visible(struct kobject *kobj,
  177. struct attribute *attr, int index)
  178. {
  179. struct device *dev = container_of(kobj, struct device, kobj);
  180. struct k10temp_data *data = dev_get_drvdata(dev);
  181. struct pci_dev *pdev = data->pdev;
  182. u32 reg;
  183. switch (index) {
  184. case 0 ... 1: /* temp1_input, temp1_max */
  185. default:
  186. break;
  187. case 2 ... 3: /* temp1_crit, temp1_crit_hyst */
  188. if (!data->read_htcreg)
  189. return 0;
  190. pci_read_config_dword(pdev, REG_NORTHBRIDGE_CAPABILITIES,
  191. &reg);
  192. if (!(reg & NB_CAP_HTC))
  193. return 0;
  194. data->read_htcreg(data->pdev, &reg);
  195. if (!(reg & HTC_ENABLE))
  196. return 0;
  197. break;
  198. case 4 ... 6: /* temp1_label, temp2_input, temp2_label */
  199. if (!data->show_tdie)
  200. return 0;
  201. break;
  202. }
  203. return attr->mode;
  204. }
  205. static struct attribute *k10temp_attrs[] = {
  206. &dev_attr_temp1_input.attr,
  207. &dev_attr_temp1_max.attr,
  208. &sensor_dev_attr_temp1_crit.dev_attr.attr,
  209. &sensor_dev_attr_temp1_crit_hyst.dev_attr.attr,
  210. &sensor_dev_attr_temp1_label.dev_attr.attr,
  211. &dev_attr_temp2_input.attr,
  212. &sensor_dev_attr_temp2_label.dev_attr.attr,
  213. NULL
  214. };
  215. static const struct attribute_group k10temp_group = {
  216. .attrs = k10temp_attrs,
  217. .is_visible = k10temp_is_visible,
  218. };
  219. __ATTRIBUTE_GROUPS(k10temp);
  220. static bool has_erratum_319(struct pci_dev *pdev)
  221. {
  222. u32 pkg_type, reg_dram_cfg;
  223. if (boot_cpu_data.x86 != 0x10)
  224. return false;
  225. /*
  226. * Erratum 319: The thermal sensor of Socket F/AM2+ processors
  227. * may be unreliable.
  228. */
  229. pkg_type = cpuid_ebx(0x80000001) & CPUID_PKGTYPE_MASK;
  230. if (pkg_type == CPUID_PKGTYPE_F)
  231. return true;
  232. if (pkg_type != CPUID_PKGTYPE_AM2R2_AM3)
  233. return false;
  234. /* DDR3 memory implies socket AM3, which is good */
  235. pci_bus_read_config_dword(pdev->bus,
  236. PCI_DEVFN(PCI_SLOT(pdev->devfn), 2),
  237. REG_DCT0_CONFIG_HIGH, &reg_dram_cfg);
  238. if (reg_dram_cfg & DDR3_MODE)
  239. return false;
  240. /*
  241. * Unfortunately it is possible to run a socket AM3 CPU with DDR2
  242. * memory. We blacklist all the cores which do exist in socket AM2+
  243. * format. It still isn't perfect, as RB-C2 cores exist in both AM2+
  244. * and AM3 formats, but that's the best we can do.
  245. */
  246. return boot_cpu_data.x86_model < 4 ||
  247. (boot_cpu_data.x86_model == 4 && boot_cpu_data.x86_stepping <= 2);
  248. }
  249. static int k10temp_probe(struct pci_dev *pdev,
  250. const struct pci_device_id *id)
  251. {
  252. int unreliable = has_erratum_319(pdev);
  253. struct device *dev = &pdev->dev;
  254. struct k10temp_data *data;
  255. struct device *hwmon_dev;
  256. int i;
  257. if (unreliable) {
  258. if (!force) {
  259. dev_err(dev,
  260. "unreliable CPU thermal sensor; monitoring disabled\n");
  261. return -ENODEV;
  262. }
  263. dev_warn(dev,
  264. "unreliable CPU thermal sensor; check erratum 319\n");
  265. }
  266. data = devm_kzalloc(dev, sizeof(*data), GFP_KERNEL);
  267. if (!data)
  268. return -ENOMEM;
  269. data->pdev = pdev;
  270. if (boot_cpu_data.x86 == 0x15 &&
  271. ((boot_cpu_data.x86_model & 0xf0) == 0x60 ||
  272. (boot_cpu_data.x86_model & 0xf0) == 0x70)) {
  273. data->read_htcreg = read_htcreg_nb_f15;
  274. data->read_tempreg = read_tempreg_nb_f15;
  275. } else if (boot_cpu_data.x86 == 0x17) {
  276. data->temp_adjust_mask = 0x80000;
  277. data->read_tempreg = read_tempreg_nb_f17;
  278. data->show_tdie = true;
  279. } else {
  280. data->read_htcreg = read_htcreg_pci;
  281. data->read_tempreg = read_tempreg_pci;
  282. }
  283. for (i = 0; i < ARRAY_SIZE(tctl_offset_table); i++) {
  284. const struct tctl_offset *entry = &tctl_offset_table[i];
  285. if (boot_cpu_data.x86 == entry->model &&
  286. strstr(boot_cpu_data.x86_model_id, entry->id)) {
  287. data->temp_offset = entry->offset;
  288. break;
  289. }
  290. }
  291. hwmon_dev = devm_hwmon_device_register_with_groups(dev, "k10temp", data,
  292. k10temp_groups);
  293. return PTR_ERR_OR_ZERO(hwmon_dev);
  294. }
  295. static const struct pci_device_id k10temp_id_table[] = {
  296. { PCI_VDEVICE(AMD, PCI_DEVICE_ID_AMD_10H_NB_MISC) },
  297. { PCI_VDEVICE(AMD, PCI_DEVICE_ID_AMD_11H_NB_MISC) },
  298. { PCI_VDEVICE(AMD, PCI_DEVICE_ID_AMD_CNB17H_F3) },
  299. { PCI_VDEVICE(AMD, PCI_DEVICE_ID_AMD_15H_NB_F3) },
  300. { PCI_VDEVICE(AMD, PCI_DEVICE_ID_AMD_15H_M10H_F3) },
  301. { PCI_VDEVICE(AMD, PCI_DEVICE_ID_AMD_15H_M30H_NB_F3) },
  302. { PCI_VDEVICE(AMD, PCI_DEVICE_ID_AMD_15H_M60H_NB_F3) },
  303. { PCI_VDEVICE(AMD, PCI_DEVICE_ID_AMD_15H_M70H_NB_F3) },
  304. { PCI_VDEVICE(AMD, PCI_DEVICE_ID_AMD_16H_NB_F3) },
  305. { PCI_VDEVICE(AMD, PCI_DEVICE_ID_AMD_16H_M30H_NB_F3) },
  306. { PCI_VDEVICE(AMD, PCI_DEVICE_ID_AMD_17H_DF_F3) },
  307. { PCI_VDEVICE(AMD, PCI_DEVICE_ID_AMD_17H_M10H_DF_F3) },
  308. {}
  309. };
  310. MODULE_DEVICE_TABLE(pci, k10temp_id_table);
  311. static struct pci_driver k10temp_driver = {
  312. .name = "k10temp",
  313. .id_table = k10temp_id_table,
  314. .probe = k10temp_probe,
  315. };
  316. module_pci_driver(k10temp_driver);