cper.h 13 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292293294295296297298299300301302303304305306307308309310311312313314315316317318319320321322323324325326327328329330331332333334335336337338339340341342343344345346347348349350351352353354355356357358359360361362363364365366367368369370371372373374375376377378379380381382383384385386387388389390391392393394395396397398399400401402403404405406407408409410411412413414415416417418419420421422423424425426427428429430431432433434435436437438439440441442443444445446447448449450451452453
  1. /*
  2. * UEFI Common Platform Error Record
  3. *
  4. * Copyright (C) 2010, Intel Corp.
  5. * Author: Huang Ying <ying.huang@intel.com>
  6. *
  7. * This program is free software; you can redistribute it and/or
  8. * modify it under the terms of the GNU General Public License version
  9. * 2 as published by the Free Software Foundation.
  10. *
  11. * This program 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. See the
  14. * 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 program; if not, write to the Free Software
  18. * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
  19. */
  20. #ifndef LINUX_CPER_H
  21. #define LINUX_CPER_H
  22. #include <linux/uuid.h>
  23. #include <linux/trace_seq.h>
  24. /* CPER record signature and the size */
  25. #define CPER_SIG_RECORD "CPER"
  26. #define CPER_SIG_SIZE 4
  27. /* Used in signature_end field in struct cper_record_header */
  28. #define CPER_SIG_END 0xffffffff
  29. /*
  30. * CPER record header revision, used in revision field in struct
  31. * cper_record_header
  32. */
  33. #define CPER_RECORD_REV 0x0100
  34. /*
  35. * CPER record length contains the CPER fields which are relevant for further
  36. * handling of a memory error in userspace (we don't carry all the fields
  37. * defined in the UEFI spec because some of them don't make any sense.)
  38. * Currently, a length of 256 should be more than enough.
  39. */
  40. #define CPER_REC_LEN 256
  41. /*
  42. * Severity difinition for error_severity in struct cper_record_header
  43. * and section_severity in struct cper_section_descriptor
  44. */
  45. enum {
  46. CPER_SEV_RECOVERABLE,
  47. CPER_SEV_FATAL,
  48. CPER_SEV_CORRECTED,
  49. CPER_SEV_INFORMATIONAL,
  50. };
  51. /*
  52. * Validation bits difinition for validation_bits in struct
  53. * cper_record_header. If set, corresponding fields in struct
  54. * cper_record_header contain valid information.
  55. *
  56. * corresponds platform_id
  57. */
  58. #define CPER_VALID_PLATFORM_ID 0x0001
  59. /* corresponds timestamp */
  60. #define CPER_VALID_TIMESTAMP 0x0002
  61. /* corresponds partition_id */
  62. #define CPER_VALID_PARTITION_ID 0x0004
  63. /*
  64. * Notification type used to generate error record, used in
  65. * notification_type in struct cper_record_header
  66. *
  67. * Corrected Machine Check
  68. */
  69. #define CPER_NOTIFY_CMC \
  70. UUID_LE(0x2DCE8BB1, 0xBDD7, 0x450e, 0xB9, 0xAD, 0x9C, 0xF4, \
  71. 0xEB, 0xD4, 0xF8, 0x90)
  72. /* Corrected Platform Error */
  73. #define CPER_NOTIFY_CPE \
  74. UUID_LE(0x4E292F96, 0xD843, 0x4a55, 0xA8, 0xC2, 0xD4, 0x81, \
  75. 0xF2, 0x7E, 0xBE, 0xEE)
  76. /* Machine Check Exception */
  77. #define CPER_NOTIFY_MCE \
  78. UUID_LE(0xE8F56FFE, 0x919C, 0x4cc5, 0xBA, 0x88, 0x65, 0xAB, \
  79. 0xE1, 0x49, 0x13, 0xBB)
  80. /* PCI Express Error */
  81. #define CPER_NOTIFY_PCIE \
  82. UUID_LE(0xCF93C01F, 0x1A16, 0x4dfc, 0xB8, 0xBC, 0x9C, 0x4D, \
  83. 0xAF, 0x67, 0xC1, 0x04)
  84. /* INIT Record (for IPF) */
  85. #define CPER_NOTIFY_INIT \
  86. UUID_LE(0xCC5263E8, 0x9308, 0x454a, 0x89, 0xD0, 0x34, 0x0B, \
  87. 0xD3, 0x9B, 0xC9, 0x8E)
  88. /* Non-Maskable Interrupt */
  89. #define CPER_NOTIFY_NMI \
  90. UUID_LE(0x5BAD89FF, 0xB7E6, 0x42c9, 0x81, 0x4A, 0xCF, 0x24, \
  91. 0x85, 0xD6, 0xE9, 0x8A)
  92. /* BOOT Error Record */
  93. #define CPER_NOTIFY_BOOT \
  94. UUID_LE(0x3D61A466, 0xAB40, 0x409a, 0xA6, 0x98, 0xF3, 0x62, \
  95. 0xD4, 0x64, 0xB3, 0x8F)
  96. /* DMA Remapping Error */
  97. #define CPER_NOTIFY_DMAR \
  98. UUID_LE(0x667DD791, 0xC6B3, 0x4c27, 0x8A, 0x6B, 0x0F, 0x8E, \
  99. 0x72, 0x2D, 0xEB, 0x41)
  100. /*
  101. * Flags bits definitions for flags in struct cper_record_header
  102. * If set, the error has been recovered
  103. */
  104. #define CPER_HW_ERROR_FLAGS_RECOVERED 0x1
  105. /* If set, the error is for previous boot */
  106. #define CPER_HW_ERROR_FLAGS_PREVERR 0x2
  107. /* If set, the error is injected for testing */
  108. #define CPER_HW_ERROR_FLAGS_SIMULATED 0x4
  109. /*
  110. * CPER section header revision, used in revision field in struct
  111. * cper_section_descriptor
  112. */
  113. #define CPER_SEC_REV 0x0100
  114. /*
  115. * Validation bits difinition for validation_bits in struct
  116. * cper_section_descriptor. If set, corresponding fields in struct
  117. * cper_section_descriptor contain valid information.
  118. *
  119. * corresponds fru_id
  120. */
  121. #define CPER_SEC_VALID_FRU_ID 0x1
  122. /* corresponds fru_text */
  123. #define CPER_SEC_VALID_FRU_TEXT 0x2
  124. /*
  125. * Flags bits definitions for flags in struct cper_section_descriptor
  126. *
  127. * If set, the section is associated with the error condition
  128. * directly, and should be focused on
  129. */
  130. #define CPER_SEC_PRIMARY 0x0001
  131. /*
  132. * If set, the error was not contained within the processor or memory
  133. * hierarchy and the error may have propagated to persistent storage
  134. * or network
  135. */
  136. #define CPER_SEC_CONTAINMENT_WARNING 0x0002
  137. /* If set, the component must be re-initialized or re-enabled prior to use */
  138. #define CPER_SEC_RESET 0x0004
  139. /* If set, Linux may choose to discontinue use of the resource */
  140. #define CPER_SEC_ERROR_THRESHOLD_EXCEEDED 0x0008
  141. /*
  142. * If set, resource could not be queried for error information due to
  143. * conflicts with other system software or resources. Some fields of
  144. * the section will be invalid
  145. */
  146. #define CPER_SEC_RESOURCE_NOT_ACCESSIBLE 0x0010
  147. /*
  148. * If set, action has been taken to ensure error containment (such as
  149. * poisoning data), but the error has not been fully corrected and the
  150. * data has not been consumed. Linux may choose to take further
  151. * corrective action before the data is consumed
  152. */
  153. #define CPER_SEC_LATENT_ERROR 0x0020
  154. /*
  155. * Section type definitions, used in section_type field in struct
  156. * cper_section_descriptor
  157. *
  158. * Processor Generic
  159. */
  160. #define CPER_SEC_PROC_GENERIC \
  161. UUID_LE(0x9876CCAD, 0x47B4, 0x4bdb, 0xB6, 0x5E, 0x16, 0xF1, \
  162. 0x93, 0xC4, 0xF3, 0xDB)
  163. /* Processor Specific: X86/X86_64 */
  164. #define CPER_SEC_PROC_IA \
  165. UUID_LE(0xDC3EA0B0, 0xA144, 0x4797, 0xB9, 0x5B, 0x53, 0xFA, \
  166. 0x24, 0x2B, 0x6E, 0x1D)
  167. /* Processor Specific: IA64 */
  168. #define CPER_SEC_PROC_IPF \
  169. UUID_LE(0xE429FAF1, 0x3CB7, 0x11D4, 0x0B, 0xCA, 0x07, 0x00, \
  170. 0x80, 0xC7, 0x3C, 0x88, 0x81)
  171. /* Platform Memory */
  172. #define CPER_SEC_PLATFORM_MEM \
  173. UUID_LE(0xA5BC1114, 0x6F64, 0x4EDE, 0xB8, 0x63, 0x3E, 0x83, \
  174. 0xED, 0x7C, 0x83, 0xB1)
  175. #define CPER_SEC_PCIE \
  176. UUID_LE(0xD995E954, 0xBBC1, 0x430F, 0xAD, 0x91, 0xB4, 0x4D, \
  177. 0xCB, 0x3C, 0x6F, 0x35)
  178. /* Firmware Error Record Reference */
  179. #define CPER_SEC_FW_ERR_REC_REF \
  180. UUID_LE(0x81212A96, 0x09ED, 0x4996, 0x94, 0x71, 0x8D, 0x72, \
  181. 0x9C, 0x8E, 0x69, 0xED)
  182. /* PCI/PCI-X Bus */
  183. #define CPER_SEC_PCI_X_BUS \
  184. UUID_LE(0xC5753963, 0x3B84, 0x4095, 0xBF, 0x78, 0xED, 0xDA, \
  185. 0xD3, 0xF9, 0xC9, 0xDD)
  186. /* PCI Component/Device */
  187. #define CPER_SEC_PCI_DEV \
  188. UUID_LE(0xEB5E4685, 0xCA66, 0x4769, 0xB6, 0xA2, 0x26, 0x06, \
  189. 0x8B, 0x00, 0x13, 0x26)
  190. #define CPER_SEC_DMAR_GENERIC \
  191. UUID_LE(0x5B51FEF7, 0xC79D, 0x4434, 0x8F, 0x1B, 0xAA, 0x62, \
  192. 0xDE, 0x3E, 0x2C, 0x64)
  193. /* Intel VT for Directed I/O specific DMAr */
  194. #define CPER_SEC_DMAR_VT \
  195. UUID_LE(0x71761D37, 0x32B2, 0x45cd, 0xA7, 0xD0, 0xB0, 0xFE, \
  196. 0xDD, 0x93, 0xE8, 0xCF)
  197. /* IOMMU specific DMAr */
  198. #define CPER_SEC_DMAR_IOMMU \
  199. UUID_LE(0x036F84E1, 0x7F37, 0x428c, 0xA7, 0x9E, 0x57, 0x5F, \
  200. 0xDF, 0xAA, 0x84, 0xEC)
  201. #define CPER_PROC_VALID_TYPE 0x0001
  202. #define CPER_PROC_VALID_ISA 0x0002
  203. #define CPER_PROC_VALID_ERROR_TYPE 0x0004
  204. #define CPER_PROC_VALID_OPERATION 0x0008
  205. #define CPER_PROC_VALID_FLAGS 0x0010
  206. #define CPER_PROC_VALID_LEVEL 0x0020
  207. #define CPER_PROC_VALID_VERSION 0x0040
  208. #define CPER_PROC_VALID_BRAND_INFO 0x0080
  209. #define CPER_PROC_VALID_ID 0x0100
  210. #define CPER_PROC_VALID_TARGET_ADDRESS 0x0200
  211. #define CPER_PROC_VALID_REQUESTOR_ID 0x0400
  212. #define CPER_PROC_VALID_RESPONDER_ID 0x0800
  213. #define CPER_PROC_VALID_IP 0x1000
  214. #define CPER_MEM_VALID_ERROR_STATUS 0x0001
  215. #define CPER_MEM_VALID_PA 0x0002
  216. #define CPER_MEM_VALID_PA_MASK 0x0004
  217. #define CPER_MEM_VALID_NODE 0x0008
  218. #define CPER_MEM_VALID_CARD 0x0010
  219. #define CPER_MEM_VALID_MODULE 0x0020
  220. #define CPER_MEM_VALID_BANK 0x0040
  221. #define CPER_MEM_VALID_DEVICE 0x0080
  222. #define CPER_MEM_VALID_ROW 0x0100
  223. #define CPER_MEM_VALID_COLUMN 0x0200
  224. #define CPER_MEM_VALID_BIT_POSITION 0x0400
  225. #define CPER_MEM_VALID_REQUESTOR_ID 0x0800
  226. #define CPER_MEM_VALID_RESPONDER_ID 0x1000
  227. #define CPER_MEM_VALID_TARGET_ID 0x2000
  228. #define CPER_MEM_VALID_ERROR_TYPE 0x4000
  229. #define CPER_MEM_VALID_RANK_NUMBER 0x8000
  230. #define CPER_MEM_VALID_CARD_HANDLE 0x10000
  231. #define CPER_MEM_VALID_MODULE_HANDLE 0x20000
  232. #define CPER_PCIE_VALID_PORT_TYPE 0x0001
  233. #define CPER_PCIE_VALID_VERSION 0x0002
  234. #define CPER_PCIE_VALID_COMMAND_STATUS 0x0004
  235. #define CPER_PCIE_VALID_DEVICE_ID 0x0008
  236. #define CPER_PCIE_VALID_SERIAL_NUMBER 0x0010
  237. #define CPER_PCIE_VALID_BRIDGE_CONTROL_STATUS 0x0020
  238. #define CPER_PCIE_VALID_CAPABILITY 0x0040
  239. #define CPER_PCIE_VALID_AER_INFO 0x0080
  240. #define CPER_PCIE_SLOT_SHIFT 3
  241. /*
  242. * All tables and structs must be byte-packed to match CPER
  243. * specification, since the tables are provided by the system BIOS
  244. */
  245. #pragma pack(1)
  246. struct cper_record_header {
  247. char signature[CPER_SIG_SIZE]; /* must be CPER_SIG_RECORD */
  248. __u16 revision; /* must be CPER_RECORD_REV */
  249. __u32 signature_end; /* must be CPER_SIG_END */
  250. __u16 section_count;
  251. __u32 error_severity;
  252. __u32 validation_bits;
  253. __u32 record_length;
  254. __u64 timestamp;
  255. uuid_le platform_id;
  256. uuid_le partition_id;
  257. uuid_le creator_id;
  258. uuid_le notification_type;
  259. __u64 record_id;
  260. __u32 flags;
  261. __u64 persistence_information;
  262. __u8 reserved[12]; /* must be zero */
  263. };
  264. struct cper_section_descriptor {
  265. __u32 section_offset; /* Offset in bytes of the
  266. * section body from the base
  267. * of the record header */
  268. __u32 section_length;
  269. __u16 revision; /* must be CPER_RECORD_REV */
  270. __u8 validation_bits;
  271. __u8 reserved; /* must be zero */
  272. __u32 flags;
  273. uuid_le section_type;
  274. uuid_le fru_id;
  275. __u32 section_severity;
  276. __u8 fru_text[20];
  277. };
  278. /* Generic Processor Error Section */
  279. struct cper_sec_proc_generic {
  280. __u64 validation_bits;
  281. __u8 proc_type;
  282. __u8 proc_isa;
  283. __u8 proc_error_type;
  284. __u8 operation;
  285. __u8 flags;
  286. __u8 level;
  287. __u16 reserved;
  288. __u64 cpu_version;
  289. char cpu_brand[128];
  290. __u64 proc_id;
  291. __u64 target_addr;
  292. __u64 requestor_id;
  293. __u64 responder_id;
  294. __u64 ip;
  295. };
  296. /* IA32/X64 Processor Error Section */
  297. struct cper_sec_proc_ia {
  298. __u64 validation_bits;
  299. __u8 lapic_id;
  300. __u8 cpuid[48];
  301. };
  302. /* IA32/X64 Processor Error Information Structure */
  303. struct cper_ia_err_info {
  304. uuid_le err_type;
  305. __u64 validation_bits;
  306. __u64 check_info;
  307. __u64 target_id;
  308. __u64 requestor_id;
  309. __u64 responder_id;
  310. __u64 ip;
  311. };
  312. /* IA32/X64 Processor Context Information Structure */
  313. struct cper_ia_proc_ctx {
  314. __u16 reg_ctx_type;
  315. __u16 reg_arr_size;
  316. __u32 msr_addr;
  317. __u64 mm_reg_addr;
  318. };
  319. /* Old Memory Error Section UEFI 2.1, 2.2 */
  320. struct cper_sec_mem_err_old {
  321. __u64 validation_bits;
  322. __u64 error_status;
  323. __u64 physical_addr;
  324. __u64 physical_addr_mask;
  325. __u16 node;
  326. __u16 card;
  327. __u16 module;
  328. __u16 bank;
  329. __u16 device;
  330. __u16 row;
  331. __u16 column;
  332. __u16 bit_pos;
  333. __u64 requestor_id;
  334. __u64 responder_id;
  335. __u64 target_id;
  336. __u8 error_type;
  337. };
  338. /* Memory Error Section UEFI >= 2.3 */
  339. struct cper_sec_mem_err {
  340. __u64 validation_bits;
  341. __u64 error_status;
  342. __u64 physical_addr;
  343. __u64 physical_addr_mask;
  344. __u16 node;
  345. __u16 card;
  346. __u16 module;
  347. __u16 bank;
  348. __u16 device;
  349. __u16 row;
  350. __u16 column;
  351. __u16 bit_pos;
  352. __u64 requestor_id;
  353. __u64 responder_id;
  354. __u64 target_id;
  355. __u8 error_type;
  356. __u8 reserved;
  357. __u16 rank;
  358. __u16 mem_array_handle; /* card handle in UEFI 2.4 */
  359. __u16 mem_dev_handle; /* module handle in UEFI 2.4 */
  360. };
  361. struct cper_mem_err_compact {
  362. __u64 validation_bits;
  363. __u16 node;
  364. __u16 card;
  365. __u16 module;
  366. __u16 bank;
  367. __u16 device;
  368. __u16 row;
  369. __u16 column;
  370. __u16 bit_pos;
  371. __u64 requestor_id;
  372. __u64 responder_id;
  373. __u64 target_id;
  374. __u16 rank;
  375. __u16 mem_array_handle;
  376. __u16 mem_dev_handle;
  377. };
  378. struct cper_sec_pcie {
  379. __u64 validation_bits;
  380. __u32 port_type;
  381. struct {
  382. __u8 minor;
  383. __u8 major;
  384. __u8 reserved[2];
  385. } version;
  386. __u16 command;
  387. __u16 status;
  388. __u32 reserved;
  389. struct {
  390. __u16 vendor_id;
  391. __u16 device_id;
  392. __u8 class_code[3];
  393. __u8 function;
  394. __u8 device;
  395. __u16 segment;
  396. __u8 bus;
  397. __u8 secondary_bus;
  398. __u16 slot;
  399. __u8 reserved;
  400. } device_id;
  401. struct {
  402. __u32 lower;
  403. __u32 upper;
  404. } serial_number;
  405. struct {
  406. __u16 secondary_status;
  407. __u16 control;
  408. } bridge;
  409. __u8 capability[60];
  410. __u8 aer_info[96];
  411. };
  412. /* Reset to default packing */
  413. #pragma pack()
  414. u64 cper_next_record_id(void);
  415. const char *cper_severity_str(unsigned int);
  416. const char *cper_mem_err_type_str(unsigned int);
  417. void cper_print_bits(const char *prefix, unsigned int bits,
  418. const char * const strs[], unsigned int strs_size);
  419. void cper_mem_err_pack(const struct cper_sec_mem_err *,
  420. struct cper_mem_err_compact *);
  421. const char *cper_mem_err_unpack(struct trace_seq *,
  422. struct cper_mem_err_compact *);
  423. #endif