blob: 76abba4b238ece14f8a2bd9bcce5ab53306c61fd [file] [log] [blame]
Huang Ying06d65de2010-05-18 14:35:19 +08001/*
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
21#ifndef LINUX_CPER_H
22#define LINUX_CPER_H
23
24#include <linux/uuid.h>
Chen, Gong2dfb7d52014-06-17 22:33:07 -040025#include <linux/trace_seq.h>
Huang Ying06d65de2010-05-18 14:35:19 +080026
27/* CPER record signature and the size */
28#define CPER_SIG_RECORD "CPER"
29#define CPER_SIG_SIZE 4
30/* Used in signature_end field in struct cper_record_header */
31#define CPER_SIG_END 0xffffffff
32
33/*
34 * CPER record header revision, used in revision field in struct
35 * cper_record_header
36 */
37#define CPER_RECORD_REV 0x0100
38
39/*
Chen, Gong3760cd22014-06-11 13:59:45 -070040 * CPER record length contains the CPER fields which are relevant for further
41 * handling of a memory error in userspace (we don't carry all the fields
42 * defined in the UEFI spec because some of them don't make any sense.)
43 * Currently, a length of 256 should be more than enough.
44 */
45#define CPER_REC_LEN 256
46/*
Huang Ying06d65de2010-05-18 14:35:19 +080047 * Severity difinition for error_severity in struct cper_record_header
48 * and section_severity in struct cper_section_descriptor
49 */
Huang Yingc9aa3082010-12-07 10:22:29 +080050enum {
51 CPER_SEV_RECOVERABLE,
52 CPER_SEV_FATAL,
53 CPER_SEV_CORRECTED,
54 CPER_SEV_INFORMATIONAL,
55};
Huang Ying06d65de2010-05-18 14:35:19 +080056
57/*
58 * Validation bits difinition for validation_bits in struct
59 * cper_record_header. If set, corresponding fields in struct
60 * cper_record_header contain valid information.
61 *
62 * corresponds platform_id
63 */
64#define CPER_VALID_PLATFORM_ID 0x0001
65/* corresponds timestamp */
66#define CPER_VALID_TIMESTAMP 0x0002
67/* corresponds partition_id */
68#define CPER_VALID_PARTITION_ID 0x0004
69
70/*
71 * Notification type used to generate error record, used in
72 * notification_type in struct cper_record_header
73 *
74 * Corrected Machine Check
75 */
76#define CPER_NOTIFY_CMC \
77 UUID_LE(0x2DCE8BB1, 0xBDD7, 0x450e, 0xB9, 0xAD, 0x9C, 0xF4, \
78 0xEB, 0xD4, 0xF8, 0x90)
79/* Corrected Platform Error */
80#define CPER_NOTIFY_CPE \
81 UUID_LE(0x4E292F96, 0xD843, 0x4a55, 0xA8, 0xC2, 0xD4, 0x81, \
82 0xF2, 0x7E, 0xBE, 0xEE)
83/* Machine Check Exception */
84#define CPER_NOTIFY_MCE \
85 UUID_LE(0xE8F56FFE, 0x919C, 0x4cc5, 0xBA, 0x88, 0x65, 0xAB, \
86 0xE1, 0x49, 0x13, 0xBB)
87/* PCI Express Error */
88#define CPER_NOTIFY_PCIE \
89 UUID_LE(0xCF93C01F, 0x1A16, 0x4dfc, 0xB8, 0xBC, 0x9C, 0x4D, \
90 0xAF, 0x67, 0xC1, 0x04)
91/* INIT Record (for IPF) */
92#define CPER_NOTIFY_INIT \
93 UUID_LE(0xCC5263E8, 0x9308, 0x454a, 0x89, 0xD0, 0x34, 0x0B, \
94 0xD3, 0x9B, 0xC9, 0x8E)
95/* Non-Maskable Interrupt */
96#define CPER_NOTIFY_NMI \
97 UUID_LE(0x5BAD89FF, 0xB7E6, 0x42c9, 0x81, 0x4A, 0xCF, 0x24, \
98 0x85, 0xD6, 0xE9, 0x8A)
99/* BOOT Error Record */
100#define CPER_NOTIFY_BOOT \
101 UUID_LE(0x3D61A466, 0xAB40, 0x409a, 0xA6, 0x98, 0xF3, 0x62, \
102 0xD4, 0x64, 0xB3, 0x8F)
103/* DMA Remapping Error */
104#define CPER_NOTIFY_DMAR \
105 UUID_LE(0x667DD791, 0xC6B3, 0x4c27, 0x8A, 0x6B, 0x0F, 0x8E, \
106 0x72, 0x2D, 0xEB, 0x41)
107
108/*
109 * Flags bits definitions for flags in struct cper_record_header
110 * If set, the error has been recovered
111 */
112#define CPER_HW_ERROR_FLAGS_RECOVERED 0x1
113/* If set, the error is for previous boot */
114#define CPER_HW_ERROR_FLAGS_PREVERR 0x2
115/* If set, the error is injected for testing */
116#define CPER_HW_ERROR_FLAGS_SIMULATED 0x4
117
118/*
119 * CPER section header revision, used in revision field in struct
120 * cper_section_descriptor
121 */
122#define CPER_SEC_REV 0x0100
123
124/*
125 * Validation bits difinition for validation_bits in struct
126 * cper_section_descriptor. If set, corresponding fields in struct
127 * cper_section_descriptor contain valid information.
128 *
129 * corresponds fru_id
130 */
131#define CPER_SEC_VALID_FRU_ID 0x1
132/* corresponds fru_text */
133#define CPER_SEC_VALID_FRU_TEXT 0x2
134
135/*
136 * Flags bits definitions for flags in struct cper_section_descriptor
137 *
138 * If set, the section is associated with the error condition
139 * directly, and should be focused on
140 */
141#define CPER_SEC_PRIMARY 0x0001
142/*
143 * If set, the error was not contained within the processor or memory
144 * hierarchy and the error may have propagated to persistent storage
145 * or network
146 */
147#define CPER_SEC_CONTAINMENT_WARNING 0x0002
148/* If set, the component must be re-initialized or re-enabled prior to use */
149#define CPER_SEC_RESET 0x0004
150/* If set, Linux may choose to discontinue use of the resource */
151#define CPER_SEC_ERROR_THRESHOLD_EXCEEDED 0x0008
152/*
153 * If set, resource could not be queried for error information due to
154 * conflicts with other system software or resources. Some fields of
155 * the section will be invalid
156 */
157#define CPER_SEC_RESOURCE_NOT_ACCESSIBLE 0x0010
158/*
159 * If set, action has been taken to ensure error containment (such as
160 * poisoning data), but the error has not been fully corrected and the
161 * data has not been consumed. Linux may choose to take further
162 * corrective action before the data is consumed
163 */
164#define CPER_SEC_LATENT_ERROR 0x0020
165
166/*
167 * Section type definitions, used in section_type field in struct
168 * cper_section_descriptor
169 *
170 * Processor Generic
171 */
172#define CPER_SEC_PROC_GENERIC \
173 UUID_LE(0x9876CCAD, 0x47B4, 0x4bdb, 0xB6, 0x5E, 0x16, 0xF1, \
174 0x93, 0xC4, 0xF3, 0xDB)
175/* Processor Specific: X86/X86_64 */
176#define CPER_SEC_PROC_IA \
177 UUID_LE(0xDC3EA0B0, 0xA144, 0x4797, 0xB9, 0x5B, 0x53, 0xFA, \
178 0x24, 0x2B, 0x6E, 0x1D)
179/* Processor Specific: IA64 */
180#define CPER_SEC_PROC_IPF \
181 UUID_LE(0xE429FAF1, 0x3CB7, 0x11D4, 0x0B, 0xCA, 0x07, 0x00, \
182 0x80, 0xC7, 0x3C, 0x88, 0x81)
183/* Platform Memory */
184#define CPER_SEC_PLATFORM_MEM \
185 UUID_LE(0xA5BC1114, 0x6F64, 0x4EDE, 0xB8, 0x63, 0x3E, 0x83, \
186 0xED, 0x7C, 0x83, 0xB1)
187#define CPER_SEC_PCIE \
188 UUID_LE(0xD995E954, 0xBBC1, 0x430F, 0xAD, 0x91, 0xB4, 0x4D, \
189 0xCB, 0x3C, 0x6F, 0x35)
190/* Firmware Error Record Reference */
191#define CPER_SEC_FW_ERR_REC_REF \
192 UUID_LE(0x81212A96, 0x09ED, 0x4996, 0x94, 0x71, 0x8D, 0x72, \
193 0x9C, 0x8E, 0x69, 0xED)
194/* PCI/PCI-X Bus */
195#define CPER_SEC_PCI_X_BUS \
196 UUID_LE(0xC5753963, 0x3B84, 0x4095, 0xBF, 0x78, 0xED, 0xDA, \
197 0xD3, 0xF9, 0xC9, 0xDD)
198/* PCI Component/Device */
199#define CPER_SEC_PCI_DEV \
200 UUID_LE(0xEB5E4685, 0xCA66, 0x4769, 0xB6, 0xA2, 0x26, 0x06, \
201 0x8B, 0x00, 0x13, 0x26)
202#define CPER_SEC_DMAR_GENERIC \
203 UUID_LE(0x5B51FEF7, 0xC79D, 0x4434, 0x8F, 0x1B, 0xAA, 0x62, \
204 0xDE, 0x3E, 0x2C, 0x64)
205/* Intel VT for Directed I/O specific DMAr */
206#define CPER_SEC_DMAR_VT \
207 UUID_LE(0x71761D37, 0x32B2, 0x45cd, 0xA7, 0xD0, 0xB0, 0xFE, \
208 0xDD, 0x93, 0xE8, 0xCF)
209/* IOMMU specific DMAr */
210#define CPER_SEC_DMAR_IOMMU \
211 UUID_LE(0x036F84E1, 0x7F37, 0x428c, 0xA7, 0x9E, 0x57, 0x5F, \
212 0xDF, 0xAA, 0x84, 0xEC)
213
Huang Yingc9aa3082010-12-07 10:22:29 +0800214#define CPER_PROC_VALID_TYPE 0x0001
215#define CPER_PROC_VALID_ISA 0x0002
216#define CPER_PROC_VALID_ERROR_TYPE 0x0004
217#define CPER_PROC_VALID_OPERATION 0x0008
218#define CPER_PROC_VALID_FLAGS 0x0010
219#define CPER_PROC_VALID_LEVEL 0x0020
220#define CPER_PROC_VALID_VERSION 0x0040
221#define CPER_PROC_VALID_BRAND_INFO 0x0080
222#define CPER_PROC_VALID_ID 0x0100
223#define CPER_PROC_VALID_TARGET_ADDRESS 0x0200
224#define CPER_PROC_VALID_REQUESTOR_ID 0x0400
225#define CPER_PROC_VALID_RESPONDER_ID 0x0800
226#define CPER_PROC_VALID_IP 0x1000
227
228#define CPER_MEM_VALID_ERROR_STATUS 0x0001
Chen, Gong147de142013-10-18 14:30:13 -0700229#define CPER_MEM_VALID_PA 0x0002
230#define CPER_MEM_VALID_PA_MASK 0x0004
Huang Yingc9aa3082010-12-07 10:22:29 +0800231#define CPER_MEM_VALID_NODE 0x0008
232#define CPER_MEM_VALID_CARD 0x0010
233#define CPER_MEM_VALID_MODULE 0x0020
234#define CPER_MEM_VALID_BANK 0x0040
235#define CPER_MEM_VALID_DEVICE 0x0080
236#define CPER_MEM_VALID_ROW 0x0100
237#define CPER_MEM_VALID_COLUMN 0x0200
238#define CPER_MEM_VALID_BIT_POSITION 0x0400
239#define CPER_MEM_VALID_REQUESTOR_ID 0x0800
240#define CPER_MEM_VALID_RESPONDER_ID 0x1000
241#define CPER_MEM_VALID_TARGET_ID 0x2000
242#define CPER_MEM_VALID_ERROR_TYPE 0x4000
Chen, Gong147de142013-10-18 14:30:13 -0700243#define CPER_MEM_VALID_RANK_NUMBER 0x8000
244#define CPER_MEM_VALID_CARD_HANDLE 0x10000
245#define CPER_MEM_VALID_MODULE_HANDLE 0x20000
Huang Yingc9aa3082010-12-07 10:22:29 +0800246
247#define CPER_PCIE_VALID_PORT_TYPE 0x0001
248#define CPER_PCIE_VALID_VERSION 0x0002
249#define CPER_PCIE_VALID_COMMAND_STATUS 0x0004
250#define CPER_PCIE_VALID_DEVICE_ID 0x0008
251#define CPER_PCIE_VALID_SERIAL_NUMBER 0x0010
252#define CPER_PCIE_VALID_BRIDGE_CONTROL_STATUS 0x0020
253#define CPER_PCIE_VALID_CAPABILITY 0x0040
254#define CPER_PCIE_VALID_AER_INFO 0x0080
255
256#define CPER_PCIE_SLOT_SHIFT 3
257
Huang Ying06d65de2010-05-18 14:35:19 +0800258/*
259 * All tables and structs must be byte-packed to match CPER
260 * specification, since the tables are provided by the system BIOS
261 */
262#pragma pack(1)
263
264struct cper_record_header {
265 char signature[CPER_SIG_SIZE]; /* must be CPER_SIG_RECORD */
266 __u16 revision; /* must be CPER_RECORD_REV */
267 __u32 signature_end; /* must be CPER_SIG_END */
268 __u16 section_count;
269 __u32 error_severity;
270 __u32 validation_bits;
271 __u32 record_length;
272 __u64 timestamp;
273 uuid_le platform_id;
274 uuid_le partition_id;
275 uuid_le creator_id;
276 uuid_le notification_type;
277 __u64 record_id;
278 __u32 flags;
279 __u64 persistence_information;
280 __u8 reserved[12]; /* must be zero */
281};
282
283struct cper_section_descriptor {
284 __u32 section_offset; /* Offset in bytes of the
285 * section body from the base
286 * of the record header */
287 __u32 section_length;
288 __u16 revision; /* must be CPER_RECORD_REV */
289 __u8 validation_bits;
290 __u8 reserved; /* must be zero */
291 __u32 flags;
292 uuid_le section_type;
293 uuid_le fru_id;
294 __u32 section_severity;
295 __u8 fru_text[20];
296};
297
298/* Generic Processor Error Section */
299struct cper_sec_proc_generic {
300 __u64 validation_bits;
301 __u8 proc_type;
302 __u8 proc_isa;
303 __u8 proc_error_type;
304 __u8 operation;
305 __u8 flags;
306 __u8 level;
307 __u16 reserved;
308 __u64 cpu_version;
309 char cpu_brand[128];
310 __u64 proc_id;
311 __u64 target_addr;
312 __u64 requestor_id;
313 __u64 responder_id;
314 __u64 ip;
315};
316
317/* IA32/X64 Processor Error Section */
318struct cper_sec_proc_ia {
319 __u64 validation_bits;
320 __u8 lapic_id;
321 __u8 cpuid[48];
322};
323
Lucas De Marchi25985ed2011-03-30 22:57:33 -0300324/* IA32/X64 Processor Error Information Structure */
Huang Ying06d65de2010-05-18 14:35:19 +0800325struct cper_ia_err_info {
326 uuid_le err_type;
327 __u64 validation_bits;
328 __u64 check_info;
329 __u64 target_id;
330 __u64 requestor_id;
331 __u64 responder_id;
332 __u64 ip;
333};
334
335/* IA32/X64 Processor Context Information Structure */
336struct cper_ia_proc_ctx {
337 __u16 reg_ctx_type;
338 __u16 reg_arr_size;
339 __u32 msr_addr;
340 __u64 mm_reg_addr;
341};
342
343/* Memory Error Section */
344struct cper_sec_mem_err {
345 __u64 validation_bits;
346 __u64 error_status;
347 __u64 physical_addr;
348 __u64 physical_addr_mask;
349 __u16 node;
350 __u16 card;
351 __u16 module;
352 __u16 bank;
353 __u16 device;
354 __u16 row;
355 __u16 column;
356 __u16 bit_pos;
357 __u64 requestor_id;
358 __u64 responder_id;
359 __u64 target_id;
360 __u8 error_type;
Chen, Gong147de142013-10-18 14:30:13 -0700361 __u8 reserved;
362 __u16 rank;
363 __u16 mem_array_handle; /* card handle in UEFI 2.4 */
364 __u16 mem_dev_handle; /* module handle in UEFI 2.4 */
Huang Ying06d65de2010-05-18 14:35:19 +0800365};
366
Chen, Gong2dfb7d52014-06-17 22:33:07 -0400367struct cper_mem_err_compact {
368 __u64 validation_bits;
369 __u16 node;
370 __u16 card;
371 __u16 module;
372 __u16 bank;
373 __u16 device;
374 __u16 row;
375 __u16 column;
376 __u16 bit_pos;
377 __u64 requestor_id;
378 __u64 responder_id;
379 __u64 target_id;
380 __u16 rank;
381 __u16 mem_array_handle;
382 __u16 mem_dev_handle;
383};
384
Huang Yingc9aa3082010-12-07 10:22:29 +0800385struct cper_sec_pcie {
386 __u64 validation_bits;
387 __u32 port_type;
388 struct {
389 __u8 minor;
390 __u8 major;
391 __u8 reserved[2];
392 } version;
393 __u16 command;
394 __u16 status;
395 __u32 reserved;
396 struct {
397 __u16 vendor_id;
398 __u16 device_id;
399 __u8 class_code[3];
400 __u8 function;
401 __u8 device;
402 __u16 segment;
403 __u8 bus;
404 __u8 secondary_bus;
405 __u16 slot;
406 __u8 reserved;
407 } device_id;
408 struct {
409 __u32 lower;
410 __u32 upper;
411 } serial_number;
412 struct {
413 __u16 secondary_status;
414 __u16 control;
415 } bridge;
416 __u8 capability[60];
417 __u8 aer_info[96];
418};
419
Huang Ying06d65de2010-05-18 14:35:19 +0800420/* Reset to default packing */
421#pragma pack()
422
423u64 cper_next_record_id(void);
Chen, Gong3760cd22014-06-11 13:59:45 -0700424const char *cper_severity_str(unsigned int);
425const char *cper_mem_err_type_str(unsigned int);
Huang Yingc413d762011-02-21 13:54:43 +0800426void cper_print_bits(const char *prefix, unsigned int bits,
Chen, Gong88f074f2013-10-18 14:28:59 -0700427 const char * const strs[], unsigned int strs_size);
Chen, Gong2dfb7d52014-06-17 22:33:07 -0400428void cper_mem_err_pack(const struct cper_sec_mem_err *,
429 struct cper_mem_err_compact *);
430const char *cper_mem_err_unpack(struct trace_seq *,
431 struct cper_mem_err_compact *);
Huang Ying06d65de2010-05-18 14:35:19 +0800432
433#endif