1 // SPDX-License-Identifier: GPL-2.0-only
2 #include <linux/module.h>
3 #include <linux/slab.h>
4
5 #include <asm/cpu.h>
6
7 #include "mce_amd.h"
8
9 static struct amd_decoder_ops fam_ops;
10
11 static u8 xec_mask = 0xf;
12
13 static void (*decode_dram_ecc)(int node_id, struct mce *m);
14
amd_register_ecc_decoder(void (* f)(int,struct mce *))15 void amd_register_ecc_decoder(void (*f)(int, struct mce *))
16 {
17 decode_dram_ecc = f;
18 }
19 EXPORT_SYMBOL_GPL(amd_register_ecc_decoder);
20
amd_unregister_ecc_decoder(void (* f)(int,struct mce *))21 void amd_unregister_ecc_decoder(void (*f)(int, struct mce *))
22 {
23 if (decode_dram_ecc) {
24 WARN_ON(decode_dram_ecc != f);
25
26 decode_dram_ecc = NULL;
27 }
28 }
29 EXPORT_SYMBOL_GPL(amd_unregister_ecc_decoder);
30
31 /*
32 * string representation for the different MCA reported error types, see F3x48
33 * or MSR0000_0411.
34 */
35
36 /* transaction type */
37 static const char * const tt_msgs[] = { "INSN", "DATA", "GEN", "RESV" };
38
39 /* cache level */
40 static const char * const ll_msgs[] = { "RESV", "L1", "L2", "L3/GEN" };
41
42 /* memory transaction type */
43 static const char * const rrrr_msgs[] = {
44 "GEN", "RD", "WR", "DRD", "DWR", "IRD", "PRF", "EV", "SNP"
45 };
46
47 /* participating processor */
48 const char * const pp_msgs[] = { "SRC", "RES", "OBS", "GEN" };
49 EXPORT_SYMBOL_GPL(pp_msgs);
50
51 /* request timeout */
52 static const char * const to_msgs[] = { "no timeout", "timed out" };
53
54 /* memory or i/o */
55 static const char * const ii_msgs[] = { "MEM", "RESV", "IO", "GEN" };
56
57 /* internal error type */
58 static const char * const uu_msgs[] = { "RESV", "RESV", "HWA", "RESV" };
59
60 static const char * const f15h_mc1_mce_desc[] = {
61 "UC during a demand linefill from L2",
62 "Parity error during data load from IC",
63 "Parity error for IC valid bit",
64 "Main tag parity error",
65 "Parity error in prediction queue",
66 "PFB data/address parity error",
67 "Parity error in the branch status reg",
68 "PFB promotion address error",
69 "Tag error during probe/victimization",
70 "Parity error for IC probe tag valid bit",
71 "PFB non-cacheable bit parity error",
72 "PFB valid bit parity error", /* xec = 0xd */
73 "Microcode Patch Buffer", /* xec = 010 */
74 "uop queue",
75 "insn buffer",
76 "predecode buffer",
77 "fetch address FIFO",
78 "dispatch uop queue"
79 };
80
81 static const char * const f15h_mc2_mce_desc[] = {
82 "Fill ECC error on data fills", /* xec = 0x4 */
83 "Fill parity error on insn fills",
84 "Prefetcher request FIFO parity error",
85 "PRQ address parity error",
86 "PRQ data parity error",
87 "WCC Tag ECC error",
88 "WCC Data ECC error",
89 "WCB Data parity error",
90 "VB Data ECC or parity error",
91 "L2 Tag ECC error", /* xec = 0x10 */
92 "Hard L2 Tag ECC error",
93 "Multiple hits on L2 tag",
94 "XAB parity error",
95 "PRB address parity error"
96 };
97
98 static const char * const mc4_mce_desc[] = {
99 "DRAM ECC error detected on the NB",
100 "CRC error detected on HT link",
101 "Link-defined sync error packets detected on HT link",
102 "HT Master abort",
103 "HT Target abort",
104 "Invalid GART PTE entry during GART table walk",
105 "Unsupported atomic RMW received from an IO link",
106 "Watchdog timeout due to lack of progress",
107 "DRAM ECC error detected on the NB",
108 "SVM DMA Exclusion Vector error",
109 "HT data error detected on link",
110 "Protocol error (link, L3, probe filter)",
111 "NB internal arrays parity error",
112 "DRAM addr/ctl signals parity error",
113 "IO link transmission error",
114 "L3 data cache ECC error", /* xec = 0x1c */
115 "L3 cache tag error",
116 "L3 LRU parity bits error",
117 "ECC Error in the Probe Filter directory"
118 };
119
120 static const char * const mc5_mce_desc[] = {
121 "CPU Watchdog timer expire",
122 "Wakeup array dest tag",
123 "AG payload array",
124 "EX payload array",
125 "IDRF array",
126 "Retire dispatch queue",
127 "Mapper checkpoint array",
128 "Physical register file EX0 port",
129 "Physical register file EX1 port",
130 "Physical register file AG0 port",
131 "Physical register file AG1 port",
132 "Flag register file",
133 "DE error occurred",
134 "Retire status queue"
135 };
136
137 static const char * const mc6_mce_desc[] = {
138 "Hardware Assertion",
139 "Free List",
140 "Physical Register File",
141 "Retire Queue",
142 "Scheduler table",
143 "Status Register File",
144 };
145
146 /* Scalable MCA error strings */
147 static const char * const smca_ls_mce_desc[] = {
148 "Load queue parity error",
149 "Store queue parity error",
150 "Miss address buffer payload parity error",
151 "Level 1 TLB parity error",
152 "DC Tag error type 5",
153 "DC Tag error type 6",
154 "DC Tag error type 1",
155 "Internal error type 1",
156 "Internal error type 2",
157 "System Read Data Error Thread 0",
158 "System Read Data Error Thread 1",
159 "DC Tag error type 2",
160 "DC Data error type 1 and poison consumption",
161 "DC Data error type 2",
162 "DC Data error type 3",
163 "DC Tag error type 4",
164 "Level 2 TLB parity error",
165 "PDC parity error",
166 "DC Tag error type 3",
167 "DC Tag error type 5",
168 "L2 Fill Data error",
169 };
170
171 static const char * const smca_ls2_mce_desc[] = {
172 "An ECC error was detected on a data cache read by a probe or victimization",
173 "An ECC error or L2 poison was detected on a data cache read by a load",
174 "An ECC error was detected on a data cache read-modify-write by a store",
175 "An ECC error or poison bit mismatch was detected on a tag read by a probe or victimization",
176 "An ECC error or poison bit mismatch was detected on a tag read by a load",
177 "An ECC error or poison bit mismatch was detected on a tag read by a store",
178 "An ECC error was detected on an EMEM read by a load",
179 "An ECC error was detected on an EMEM read-modify-write by a store",
180 "A parity error was detected in an L1 TLB entry by any access",
181 "A parity error was detected in an L2 TLB entry by any access",
182 "A parity error was detected in a PWC entry by any access",
183 "A parity error was detected in an STQ entry by any access",
184 "A parity error was detected in an LDQ entry by any access",
185 "A parity error was detected in a MAB entry by any access",
186 "A parity error was detected in an SCB entry state field by any access",
187 "A parity error was detected in an SCB entry address field by any access",
188 "A parity error was detected in an SCB entry data field by any access",
189 "A parity error was detected in a WCB entry by any access",
190 "A poisoned line was detected in an SCB entry by any access",
191 "A SystemReadDataError error was reported on read data returned from L2 for a load",
192 "A SystemReadDataError error was reported on read data returned from L2 for an SCB store",
193 "A SystemReadDataError error was reported on read data returned from L2 for a WCB store",
194 "A hardware assertion error was reported",
195 "A parity error was detected in an STLF, SCB EMEM entry or SRB store data by any access",
196 };
197
198 static const char * const smca_if_mce_desc[] = {
199 "Op Cache Microtag Probe Port Parity Error",
200 "IC Microtag or Full Tag Multi-hit Error",
201 "IC Full Tag Parity Error",
202 "IC Data Array Parity Error",
203 "Decoupling Queue PhysAddr Parity Error",
204 "L0 ITLB Parity Error",
205 "L1 ITLB Parity Error",
206 "L2 ITLB Parity Error",
207 "BPQ Thread 0 Snoop Parity Error",
208 "BPQ Thread 1 Snoop Parity Error",
209 "L1 BTB Multi-Match Error",
210 "L2 BTB Multi-Match Error",
211 "L2 Cache Response Poison Error",
212 "System Read Data Error",
213 "Hardware Assertion Error",
214 "L1-TLB Multi-Hit",
215 "L2-TLB Multi-Hit",
216 "BSR Parity Error",
217 "CT MCE",
218 };
219
220 static const char * const smca_l2_mce_desc[] = {
221 "L2M Tag Multiple-Way-Hit error",
222 "L2M Tag or State Array ECC Error",
223 "L2M Data Array ECC Error",
224 "Hardware Assert Error",
225 };
226
227 static const char * const smca_de_mce_desc[] = {
228 "Micro-op cache tag parity error",
229 "Micro-op cache data parity error",
230 "Instruction buffer parity error",
231 "Micro-op queue parity error",
232 "Instruction dispatch queue parity error",
233 "Fetch address FIFO parity error",
234 "Patch RAM data parity error",
235 "Patch RAM sequencer parity error",
236 "Micro-op buffer parity error",
237 "Hardware Assertion MCA Error",
238 };
239
240 static const char * const smca_ex_mce_desc[] = {
241 "Watchdog Timeout error",
242 "Physical register file parity error",
243 "Flag register file parity error",
244 "Immediate displacement register file parity error",
245 "Address generator payload parity error",
246 "EX payload parity error",
247 "Checkpoint queue parity error",
248 "Retire dispatch queue parity error",
249 "Retire status queue parity error",
250 "Scheduling queue parity error",
251 "Branch buffer queue parity error",
252 "Hardware Assertion error",
253 "Spec Map parity error",
254 "Retire Map parity error",
255 };
256
257 static const char * const smca_fp_mce_desc[] = {
258 "Physical register file (PRF) parity error",
259 "Freelist (FL) parity error",
260 "Schedule queue parity error",
261 "NSQ parity error",
262 "Retire queue (RQ) parity error",
263 "Status register file (SRF) parity error",
264 "Hardware assertion",
265 };
266
267 static const char * const smca_l3_mce_desc[] = {
268 "Shadow Tag Macro ECC Error",
269 "Shadow Tag Macro Multi-way-hit Error",
270 "L3M Tag ECC Error",
271 "L3M Tag Multi-way-hit Error",
272 "L3M Data ECC Error",
273 "SDP Parity Error or SystemReadDataError from XI",
274 "L3 Victim Queue Parity Error",
275 "L3 Hardware Assertion",
276 };
277
278 static const char * const smca_cs_mce_desc[] = {
279 "Illegal Request",
280 "Address Violation",
281 "Security Violation",
282 "Illegal Response",
283 "Unexpected Response",
284 "Request or Probe Parity Error",
285 "Read Response Parity Error",
286 "Atomic Request Parity Error",
287 "Probe Filter ECC Error",
288 };
289
290 static const char * const smca_cs2_mce_desc[] = {
291 "Illegal Request",
292 "Address Violation",
293 "Security Violation",
294 "Illegal Response",
295 "Unexpected Response",
296 "Request or Probe Parity Error",
297 "Read Response Parity Error",
298 "Atomic Request Parity Error",
299 "SDP read response had no match in the CS queue",
300 "Probe Filter Protocol Error",
301 "Probe Filter ECC Error",
302 "SDP read response had an unexpected RETRY error",
303 "Counter overflow error",
304 "Counter underflow error",
305 };
306
307 static const char * const smca_pie_mce_desc[] = {
308 "Hardware Assert",
309 "Register security violation",
310 "Link Error",
311 "Poison data consumption",
312 "A deferred error was detected in the DF"
313 };
314
315 static const char * const smca_umc_mce_desc[] = {
316 "DRAM ECC error",
317 "Data poison error",
318 "SDP parity error",
319 "Advanced peripheral bus error",
320 "Address/Command parity error",
321 "Write data CRC error",
322 "DCQ SRAM ECC error",
323 "AES SRAM ECC error",
324 };
325
326 static const char * const smca_umc2_mce_desc[] = {
327 "DRAM ECC error",
328 "Data poison error",
329 "SDP parity error",
330 "Reserved",
331 "Address/Command parity error",
332 "Write data parity error",
333 "DCQ SRAM ECC error",
334 "Reserved",
335 "Read data parity error",
336 "Rdb SRAM ECC error",
337 "RdRsp SRAM ECC error",
338 "LM32 MP errors",
339 };
340
341 static const char * const smca_pb_mce_desc[] = {
342 "An ECC error in the Parameter Block RAM array",
343 };
344
345 static const char * const smca_psp_mce_desc[] = {
346 "An ECC or parity error in a PSP RAM instance",
347 };
348
349 static const char * const smca_psp2_mce_desc[] = {
350 "High SRAM ECC or parity error",
351 "Low SRAM ECC or parity error",
352 "Instruction Cache Bank 0 ECC or parity error",
353 "Instruction Cache Bank 1 ECC or parity error",
354 "Instruction Tag Ram 0 parity error",
355 "Instruction Tag Ram 1 parity error",
356 "Data Cache Bank 0 ECC or parity error",
357 "Data Cache Bank 1 ECC or parity error",
358 "Data Cache Bank 2 ECC or parity error",
359 "Data Cache Bank 3 ECC or parity error",
360 "Data Tag Bank 0 parity error",
361 "Data Tag Bank 1 parity error",
362 "Data Tag Bank 2 parity error",
363 "Data Tag Bank 3 parity error",
364 "Dirty Data Ram parity error",
365 "TLB Bank 0 parity error",
366 "TLB Bank 1 parity error",
367 "System Hub Read Buffer ECC or parity error",
368 };
369
370 static const char * const smca_smu_mce_desc[] = {
371 "An ECC or parity error in an SMU RAM instance",
372 };
373
374 static const char * const smca_smu2_mce_desc[] = {
375 "High SRAM ECC or parity error",
376 "Low SRAM ECC or parity error",
377 "Data Cache Bank A ECC or parity error",
378 "Data Cache Bank B ECC or parity error",
379 "Data Tag Cache Bank A ECC or parity error",
380 "Data Tag Cache Bank B ECC or parity error",
381 "Instruction Cache Bank A ECC or parity error",
382 "Instruction Cache Bank B ECC or parity error",
383 "Instruction Tag Cache Bank A ECC or parity error",
384 "Instruction Tag Cache Bank B ECC or parity error",
385 "System Hub Read Buffer ECC or parity error",
386 "PHY RAM ECC error",
387 };
388
389 static const char * const smca_mp5_mce_desc[] = {
390 "High SRAM ECC or parity error",
391 "Low SRAM ECC or parity error",
392 "Data Cache Bank A ECC or parity error",
393 "Data Cache Bank B ECC or parity error",
394 "Data Tag Cache Bank A ECC or parity error",
395 "Data Tag Cache Bank B ECC or parity error",
396 "Instruction Cache Bank A ECC or parity error",
397 "Instruction Cache Bank B ECC or parity error",
398 "Instruction Tag Cache Bank A ECC or parity error",
399 "Instruction Tag Cache Bank B ECC or parity error",
400 };
401
402 static const char * const smca_mpdma_mce_desc[] = {
403 "Main SRAM [31:0] bank ECC or parity error",
404 "Main SRAM [63:32] bank ECC or parity error",
405 "Main SRAM [95:64] bank ECC or parity error",
406 "Main SRAM [127:96] bank ECC or parity error",
407 "Data Cache Bank A ECC or parity error",
408 "Data Cache Bank B ECC or parity error",
409 "Data Tag Cache Bank A ECC or parity error",
410 "Data Tag Cache Bank B ECC or parity error",
411 "Instruction Cache Bank A ECC or parity error",
412 "Instruction Cache Bank B ECC or parity error",
413 "Instruction Tag Cache Bank A ECC or parity error",
414 "Instruction Tag Cache Bank B ECC or parity error",
415 "Data Cache Bank A ECC or parity error",
416 "Data Cache Bank B ECC or parity error",
417 "Data Tag Cache Bank A ECC or parity error",
418 "Data Tag Cache Bank B ECC or parity error",
419 "Instruction Cache Bank A ECC or parity error",
420 "Instruction Cache Bank B ECC or parity error",
421 "Instruction Tag Cache Bank A ECC or parity error",
422 "Instruction Tag Cache Bank B ECC or parity error",
423 "Data Cache Bank A ECC or parity error",
424 "Data Cache Bank B ECC or parity error",
425 "Data Tag Cache Bank A ECC or parity error",
426 "Data Tag Cache Bank B ECC or parity error",
427 "Instruction Cache Bank A ECC or parity error",
428 "Instruction Cache Bank B ECC or parity error",
429 "Instruction Tag Cache Bank A ECC or parity error",
430 "Instruction Tag Cache Bank B ECC or parity error",
431 "System Hub Read Buffer ECC or parity error",
432 "MPDMA TVF DVSEC Memory ECC or parity error",
433 "MPDMA TVF MMIO Mailbox0 ECC or parity error",
434 "MPDMA TVF MMIO Mailbox1 ECC or parity error",
435 "MPDMA TVF Doorbell Memory ECC or parity error",
436 "MPDMA TVF SDP Slave Memory 0 ECC or parity error",
437 "MPDMA TVF SDP Slave Memory 1 ECC or parity error",
438 "MPDMA TVF SDP Slave Memory 2 ECC or parity error",
439 "MPDMA TVF SDP Master Memory 0 ECC or parity error",
440 "MPDMA TVF SDP Master Memory 1 ECC or parity error",
441 "MPDMA TVF SDP Master Memory 2 ECC or parity error",
442 "MPDMA TVF SDP Master Memory 3 ECC or parity error",
443 "MPDMA TVF SDP Master Memory 4 ECC or parity error",
444 "MPDMA TVF SDP Master Memory 5 ECC or parity error",
445 "MPDMA TVF SDP Master Memory 6 ECC or parity error",
446 "MPDMA PTE Command FIFO ECC or parity error",
447 "MPDMA PTE Hub Data FIFO ECC or parity error",
448 "MPDMA PTE Internal Data FIFO ECC or parity error",
449 "MPDMA PTE Command Memory DMA ECC or parity error",
450 "MPDMA PTE Command Memory Internal ECC or parity error",
451 "MPDMA PTE DMA Completion FIFO ECC or parity error",
452 "MPDMA PTE Tablewalk Completion FIFO ECC or parity error",
453 "MPDMA PTE Descriptor Completion FIFO ECC or parity error",
454 "MPDMA PTE ReadOnly Completion FIFO ECC or parity error",
455 "MPDMA PTE DirectWrite Completion FIFO ECC or parity error",
456 "SDP Watchdog Timer expired",
457 };
458
459 static const char * const smca_nbio_mce_desc[] = {
460 "ECC or Parity error",
461 "PCIE error",
462 "SDP ErrEvent error",
463 "SDP Egress Poison Error",
464 "IOHC Internal Poison Error",
465 };
466
467 static const char * const smca_pcie_mce_desc[] = {
468 "CCIX PER Message logging",
469 "CCIX Read Response with Status: Non-Data Error",
470 "CCIX Write Response with Status: Non-Data Error",
471 "CCIX Read Response with Status: Data Error",
472 "CCIX Non-okay write response with data error",
473 };
474
475 static const char * const smca_pcie2_mce_desc[] = {
476 "SDP Parity Error logging",
477 };
478
479 static const char * const smca_xgmipcs_mce_desc[] = {
480 "Data Loss Error",
481 "Training Error",
482 "Flow Control Acknowledge Error",
483 "Rx Fifo Underflow Error",
484 "Rx Fifo Overflow Error",
485 "CRC Error",
486 "BER Exceeded Error",
487 "Tx Vcid Data Error",
488 "Replay Buffer Parity Error",
489 "Data Parity Error",
490 "Replay Fifo Overflow Error",
491 "Replay Fifo Underflow Error",
492 "Elastic Fifo Overflow Error",
493 "Deskew Error",
494 "Flow Control CRC Error",
495 "Data Startup Limit Error",
496 "FC Init Timeout Error",
497 "Recovery Timeout Error",
498 "Ready Serial Timeout Error",
499 "Ready Serial Attempt Error",
500 "Recovery Attempt Error",
501 "Recovery Relock Attempt Error",
502 "Replay Attempt Error",
503 "Sync Header Error",
504 "Tx Replay Timeout Error",
505 "Rx Replay Timeout Error",
506 "LinkSub Tx Timeout Error",
507 "LinkSub Rx Timeout Error",
508 "Rx CMD Packet Error",
509 };
510
511 static const char * const smca_xgmiphy_mce_desc[] = {
512 "RAM ECC Error",
513 "ARC instruction buffer parity error",
514 "ARC data buffer parity error",
515 "PHY APB error",
516 };
517
518 static const char * const smca_nbif_mce_desc[] = {
519 "Timeout error from GMI",
520 "SRAM ECC error",
521 "NTB Error Event",
522 "SDP Parity error",
523 };
524
525 static const char * const smca_sata_mce_desc[] = {
526 "Parity error for port 0",
527 "Parity error for port 1",
528 "Parity error for port 2",
529 "Parity error for port 3",
530 "Parity error for port 4",
531 "Parity error for port 5",
532 "Parity error for port 6",
533 "Parity error for port 7",
534 };
535
536 static const char * const smca_usb_mce_desc[] = {
537 "Parity error or ECC error for S0 RAM0",
538 "Parity error or ECC error for S0 RAM1",
539 "Parity error or ECC error for S0 RAM2",
540 "Parity error for PHY RAM0",
541 "Parity error for PHY RAM1",
542 "AXI Slave Response error",
543 };
544
545 static const char * const smca_gmipcs_mce_desc[] = {
546 "Data Loss Error",
547 "Training Error",
548 "Replay Parity Error",
549 "Rx Fifo Underflow Error",
550 "Rx Fifo Overflow Error",
551 "CRC Error",
552 "BER Exceeded Error",
553 "Tx Fifo Underflow Error",
554 "Replay Buffer Parity Error",
555 "Tx Overflow Error",
556 "Replay Fifo Overflow Error",
557 "Replay Fifo Underflow Error",
558 "Elastic Fifo Overflow Error",
559 "Deskew Error",
560 "Offline Error",
561 "Data Startup Limit Error",
562 "FC Init Timeout Error",
563 "Recovery Timeout Error",
564 "Ready Serial Timeout Error",
565 "Ready Serial Attempt Error",
566 "Recovery Attempt Error",
567 "Recovery Relock Attempt Error",
568 "Deskew Abort Error",
569 "Rx Buffer Error",
570 "Rx LFDS Fifo Overflow Error",
571 "Rx LFDS Fifo Underflow Error",
572 "LinkSub Tx Timeout Error",
573 "LinkSub Rx Timeout Error",
574 "Rx CMD Packet Error",
575 "LFDS Training Timeout Error",
576 "LFDS FC Init Timeout Error",
577 "Data Loss Error",
578 };
579
580 struct smca_mce_desc {
581 const char * const *descs;
582 unsigned int num_descs;
583 };
584
585 static struct smca_mce_desc smca_mce_descs[] = {
586 [SMCA_LS] = { smca_ls_mce_desc, ARRAY_SIZE(smca_ls_mce_desc) },
587 [SMCA_LS_V2] = { smca_ls2_mce_desc, ARRAY_SIZE(smca_ls2_mce_desc) },
588 [SMCA_IF] = { smca_if_mce_desc, ARRAY_SIZE(smca_if_mce_desc) },
589 [SMCA_L2_CACHE] = { smca_l2_mce_desc, ARRAY_SIZE(smca_l2_mce_desc) },
590 [SMCA_DE] = { smca_de_mce_desc, ARRAY_SIZE(smca_de_mce_desc) },
591 [SMCA_EX] = { smca_ex_mce_desc, ARRAY_SIZE(smca_ex_mce_desc) },
592 [SMCA_FP] = { smca_fp_mce_desc, ARRAY_SIZE(smca_fp_mce_desc) },
593 [SMCA_L3_CACHE] = { smca_l3_mce_desc, ARRAY_SIZE(smca_l3_mce_desc) },
594 [SMCA_CS] = { smca_cs_mce_desc, ARRAY_SIZE(smca_cs_mce_desc) },
595 [SMCA_CS_V2] = { smca_cs2_mce_desc, ARRAY_SIZE(smca_cs2_mce_desc) },
596 [SMCA_PIE] = { smca_pie_mce_desc, ARRAY_SIZE(smca_pie_mce_desc) },
597 [SMCA_UMC] = { smca_umc_mce_desc, ARRAY_SIZE(smca_umc_mce_desc) },
598 [SMCA_UMC_V2] = { smca_umc2_mce_desc, ARRAY_SIZE(smca_umc2_mce_desc) },
599 [SMCA_PB] = { smca_pb_mce_desc, ARRAY_SIZE(smca_pb_mce_desc) },
600 [SMCA_PSP] = { smca_psp_mce_desc, ARRAY_SIZE(smca_psp_mce_desc) },
601 [SMCA_PSP_V2] = { smca_psp2_mce_desc, ARRAY_SIZE(smca_psp2_mce_desc) },
602 [SMCA_SMU] = { smca_smu_mce_desc, ARRAY_SIZE(smca_smu_mce_desc) },
603 [SMCA_SMU_V2] = { smca_smu2_mce_desc, ARRAY_SIZE(smca_smu2_mce_desc) },
604 [SMCA_MP5] = { smca_mp5_mce_desc, ARRAY_SIZE(smca_mp5_mce_desc) },
605 [SMCA_MPDMA] = { smca_mpdma_mce_desc, ARRAY_SIZE(smca_mpdma_mce_desc) },
606 [SMCA_NBIO] = { smca_nbio_mce_desc, ARRAY_SIZE(smca_nbio_mce_desc) },
607 [SMCA_PCIE] = { smca_pcie_mce_desc, ARRAY_SIZE(smca_pcie_mce_desc) },
608 [SMCA_PCIE_V2] = { smca_pcie2_mce_desc, ARRAY_SIZE(smca_pcie2_mce_desc) },
609 [SMCA_XGMI_PCS] = { smca_xgmipcs_mce_desc, ARRAY_SIZE(smca_xgmipcs_mce_desc) },
610 /* NBIF and SHUB have the same error descriptions, for now. */
611 [SMCA_NBIF] = { smca_nbif_mce_desc, ARRAY_SIZE(smca_nbif_mce_desc) },
612 [SMCA_SHUB] = { smca_nbif_mce_desc, ARRAY_SIZE(smca_nbif_mce_desc) },
613 [SMCA_SATA] = { smca_sata_mce_desc, ARRAY_SIZE(smca_sata_mce_desc) },
614 [SMCA_USB] = { smca_usb_mce_desc, ARRAY_SIZE(smca_usb_mce_desc) },
615 [SMCA_GMI_PCS] = { smca_gmipcs_mce_desc, ARRAY_SIZE(smca_gmipcs_mce_desc) },
616 /* All the PHY bank types have the same error descriptions, for now. */
617 [SMCA_XGMI_PHY] = { smca_xgmiphy_mce_desc, ARRAY_SIZE(smca_xgmiphy_mce_desc) },
618 [SMCA_WAFL_PHY] = { smca_xgmiphy_mce_desc, ARRAY_SIZE(smca_xgmiphy_mce_desc) },
619 [SMCA_GMI_PHY] = { smca_xgmiphy_mce_desc, ARRAY_SIZE(smca_xgmiphy_mce_desc) },
620 };
621
f12h_mc0_mce(u16 ec,u8 xec)622 static bool f12h_mc0_mce(u16 ec, u8 xec)
623 {
624 bool ret = false;
625
626 if (MEM_ERROR(ec)) {
627 u8 ll = LL(ec);
628 ret = true;
629
630 if (ll == LL_L2)
631 pr_cont("during L1 linefill from L2.\n");
632 else if (ll == LL_L1)
633 pr_cont("Data/Tag %s error.\n", R4_MSG(ec));
634 else
635 ret = false;
636 }
637 return ret;
638 }
639
f10h_mc0_mce(u16 ec,u8 xec)640 static bool f10h_mc0_mce(u16 ec, u8 xec)
641 {
642 if (R4(ec) == R4_GEN && LL(ec) == LL_L1) {
643 pr_cont("during data scrub.\n");
644 return true;
645 }
646 return f12h_mc0_mce(ec, xec);
647 }
648
k8_mc0_mce(u16 ec,u8 xec)649 static bool k8_mc0_mce(u16 ec, u8 xec)
650 {
651 if (BUS_ERROR(ec)) {
652 pr_cont("during system linefill.\n");
653 return true;
654 }
655
656 return f10h_mc0_mce(ec, xec);
657 }
658
cat_mc0_mce(u16 ec,u8 xec)659 static bool cat_mc0_mce(u16 ec, u8 xec)
660 {
661 u8 r4 = R4(ec);
662 bool ret = true;
663
664 if (MEM_ERROR(ec)) {
665
666 if (TT(ec) != TT_DATA || LL(ec) != LL_L1)
667 return false;
668
669 switch (r4) {
670 case R4_DRD:
671 case R4_DWR:
672 pr_cont("Data/Tag parity error due to %s.\n",
673 (r4 == R4_DRD ? "load/hw prf" : "store"));
674 break;
675 case R4_EVICT:
676 pr_cont("Copyback parity error on a tag miss.\n");
677 break;
678 case R4_SNOOP:
679 pr_cont("Tag parity error during snoop.\n");
680 break;
681 default:
682 ret = false;
683 }
684 } else if (BUS_ERROR(ec)) {
685
686 if ((II(ec) != II_MEM && II(ec) != II_IO) || LL(ec) != LL_LG)
687 return false;
688
689 pr_cont("System read data error on a ");
690
691 switch (r4) {
692 case R4_RD:
693 pr_cont("TLB reload.\n");
694 break;
695 case R4_DWR:
696 pr_cont("store.\n");
697 break;
698 case R4_DRD:
699 pr_cont("load.\n");
700 break;
701 default:
702 ret = false;
703 }
704 } else {
705 ret = false;
706 }
707
708 return ret;
709 }
710
f15h_mc0_mce(u16 ec,u8 xec)711 static bool f15h_mc0_mce(u16 ec, u8 xec)
712 {
713 bool ret = true;
714
715 if (MEM_ERROR(ec)) {
716
717 switch (xec) {
718 case 0x0:
719 pr_cont("Data Array access error.\n");
720 break;
721
722 case 0x1:
723 pr_cont("UC error during a linefill from L2/NB.\n");
724 break;
725
726 case 0x2:
727 case 0x11:
728 pr_cont("STQ access error.\n");
729 break;
730
731 case 0x3:
732 pr_cont("SCB access error.\n");
733 break;
734
735 case 0x10:
736 pr_cont("Tag error.\n");
737 break;
738
739 case 0x12:
740 pr_cont("LDQ access error.\n");
741 break;
742
743 default:
744 ret = false;
745 }
746 } else if (BUS_ERROR(ec)) {
747
748 if (!xec)
749 pr_cont("System Read Data Error.\n");
750 else
751 pr_cont(" Internal error condition type %d.\n", xec);
752 } else if (INT_ERROR(ec)) {
753 if (xec <= 0x1f)
754 pr_cont("Hardware Assert.\n");
755 else
756 ret = false;
757
758 } else
759 ret = false;
760
761 return ret;
762 }
763
decode_mc0_mce(struct mce * m)764 static void decode_mc0_mce(struct mce *m)
765 {
766 u16 ec = EC(m->status);
767 u8 xec = XEC(m->status, xec_mask);
768
769 pr_emerg(HW_ERR "MC0 Error: ");
770
771 /* TLB error signatures are the same across families */
772 if (TLB_ERROR(ec)) {
773 if (TT(ec) == TT_DATA) {
774 pr_cont("%s TLB %s.\n", LL_MSG(ec),
775 ((xec == 2) ? "locked miss"
776 : (xec ? "multimatch" : "parity")));
777 return;
778 }
779 } else if (fam_ops.mc0_mce(ec, xec))
780 ;
781 else
782 pr_emerg(HW_ERR "Corrupted MC0 MCE info?\n");
783 }
784
k8_mc1_mce(u16 ec,u8 xec)785 static bool k8_mc1_mce(u16 ec, u8 xec)
786 {
787 u8 ll = LL(ec);
788 bool ret = true;
789
790 if (!MEM_ERROR(ec))
791 return false;
792
793 if (ll == 0x2)
794 pr_cont("during a linefill from L2.\n");
795 else if (ll == 0x1) {
796 switch (R4(ec)) {
797 case R4_IRD:
798 pr_cont("Parity error during data load.\n");
799 break;
800
801 case R4_EVICT:
802 pr_cont("Copyback Parity/Victim error.\n");
803 break;
804
805 case R4_SNOOP:
806 pr_cont("Tag Snoop error.\n");
807 break;
808
809 default:
810 ret = false;
811 break;
812 }
813 } else
814 ret = false;
815
816 return ret;
817 }
818
cat_mc1_mce(u16 ec,u8 xec)819 static bool cat_mc1_mce(u16 ec, u8 xec)
820 {
821 u8 r4 = R4(ec);
822 bool ret = true;
823
824 if (!MEM_ERROR(ec))
825 return false;
826
827 if (TT(ec) != TT_INSTR)
828 return false;
829
830 if (r4 == R4_IRD)
831 pr_cont("Data/tag array parity error for a tag hit.\n");
832 else if (r4 == R4_SNOOP)
833 pr_cont("Tag error during snoop/victimization.\n");
834 else if (xec == 0x0)
835 pr_cont("Tag parity error from victim castout.\n");
836 else if (xec == 0x2)
837 pr_cont("Microcode patch RAM parity error.\n");
838 else
839 ret = false;
840
841 return ret;
842 }
843
f15h_mc1_mce(u16 ec,u8 xec)844 static bool f15h_mc1_mce(u16 ec, u8 xec)
845 {
846 bool ret = true;
847
848 if (!MEM_ERROR(ec))
849 return false;
850
851 switch (xec) {
852 case 0x0 ... 0xa:
853 pr_cont("%s.\n", f15h_mc1_mce_desc[xec]);
854 break;
855
856 case 0xd:
857 pr_cont("%s.\n", f15h_mc1_mce_desc[xec-2]);
858 break;
859
860 case 0x10:
861 pr_cont("%s.\n", f15h_mc1_mce_desc[xec-4]);
862 break;
863
864 case 0x11 ... 0x15:
865 pr_cont("Decoder %s parity error.\n", f15h_mc1_mce_desc[xec-4]);
866 break;
867
868 default:
869 ret = false;
870 }
871 return ret;
872 }
873
decode_mc1_mce(struct mce * m)874 static void decode_mc1_mce(struct mce *m)
875 {
876 u16 ec = EC(m->status);
877 u8 xec = XEC(m->status, xec_mask);
878
879 pr_emerg(HW_ERR "MC1 Error: ");
880
881 if (TLB_ERROR(ec))
882 pr_cont("%s TLB %s.\n", LL_MSG(ec),
883 (xec ? "multimatch" : "parity error"));
884 else if (BUS_ERROR(ec)) {
885 bool k8 = (boot_cpu_data.x86 == 0xf && (m->status & BIT_64(58)));
886
887 pr_cont("during %s.\n", (k8 ? "system linefill" : "NB data read"));
888 } else if (INT_ERROR(ec)) {
889 if (xec <= 0x3f)
890 pr_cont("Hardware Assert.\n");
891 else
892 goto wrong_mc1_mce;
893 } else if (fam_ops.mc1_mce(ec, xec))
894 ;
895 else
896 goto wrong_mc1_mce;
897
898 return;
899
900 wrong_mc1_mce:
901 pr_emerg(HW_ERR "Corrupted MC1 MCE info?\n");
902 }
903
k8_mc2_mce(u16 ec,u8 xec)904 static bool k8_mc2_mce(u16 ec, u8 xec)
905 {
906 bool ret = true;
907
908 if (xec == 0x1)
909 pr_cont(" in the write data buffers.\n");
910 else if (xec == 0x3)
911 pr_cont(" in the victim data buffers.\n");
912 else if (xec == 0x2 && MEM_ERROR(ec))
913 pr_cont(": %s error in the L2 cache tags.\n", R4_MSG(ec));
914 else if (xec == 0x0) {
915 if (TLB_ERROR(ec))
916 pr_cont("%s error in a Page Descriptor Cache or Guest TLB.\n",
917 TT_MSG(ec));
918 else if (BUS_ERROR(ec))
919 pr_cont(": %s/ECC error in data read from NB: %s.\n",
920 R4_MSG(ec), PP_MSG(ec));
921 else if (MEM_ERROR(ec)) {
922 u8 r4 = R4(ec);
923
924 if (r4 >= 0x7)
925 pr_cont(": %s error during data copyback.\n",
926 R4_MSG(ec));
927 else if (r4 <= 0x1)
928 pr_cont(": %s parity/ECC error during data "
929 "access from L2.\n", R4_MSG(ec));
930 else
931 ret = false;
932 } else
933 ret = false;
934 } else
935 ret = false;
936
937 return ret;
938 }
939
f15h_mc2_mce(u16 ec,u8 xec)940 static bool f15h_mc2_mce(u16 ec, u8 xec)
941 {
942 bool ret = true;
943
944 if (TLB_ERROR(ec)) {
945 if (xec == 0x0)
946 pr_cont("Data parity TLB read error.\n");
947 else if (xec == 0x1)
948 pr_cont("Poison data provided for TLB fill.\n");
949 else
950 ret = false;
951 } else if (BUS_ERROR(ec)) {
952 if (xec > 2)
953 ret = false;
954
955 pr_cont("Error during attempted NB data read.\n");
956 } else if (MEM_ERROR(ec)) {
957 switch (xec) {
958 case 0x4 ... 0xc:
959 pr_cont("%s.\n", f15h_mc2_mce_desc[xec - 0x4]);
960 break;
961
962 case 0x10 ... 0x14:
963 pr_cont("%s.\n", f15h_mc2_mce_desc[xec - 0x7]);
964 break;
965
966 default:
967 ret = false;
968 }
969 } else if (INT_ERROR(ec)) {
970 if (xec <= 0x3f)
971 pr_cont("Hardware Assert.\n");
972 else
973 ret = false;
974 }
975
976 return ret;
977 }
978
f16h_mc2_mce(u16 ec,u8 xec)979 static bool f16h_mc2_mce(u16 ec, u8 xec)
980 {
981 u8 r4 = R4(ec);
982
983 if (!MEM_ERROR(ec))
984 return false;
985
986 switch (xec) {
987 case 0x04 ... 0x05:
988 pr_cont("%cBUFF parity error.\n", (r4 == R4_RD) ? 'I' : 'O');
989 break;
990
991 case 0x09 ... 0x0b:
992 case 0x0d ... 0x0f:
993 pr_cont("ECC error in L2 tag (%s).\n",
994 ((r4 == R4_GEN) ? "BankReq" :
995 ((r4 == R4_SNOOP) ? "Prb" : "Fill")));
996 break;
997
998 case 0x10 ... 0x19:
999 case 0x1b:
1000 pr_cont("ECC error in L2 data array (%s).\n",
1001 (((r4 == R4_RD) && !(xec & 0x3)) ? "Hit" :
1002 ((r4 == R4_GEN) ? "Attr" :
1003 ((r4 == R4_EVICT) ? "Vict" : "Fill"))));
1004 break;
1005
1006 case 0x1c ... 0x1d:
1007 case 0x1f:
1008 pr_cont("Parity error in L2 attribute bits (%s).\n",
1009 ((r4 == R4_RD) ? "Hit" :
1010 ((r4 == R4_GEN) ? "Attr" : "Fill")));
1011 break;
1012
1013 default:
1014 return false;
1015 }
1016
1017 return true;
1018 }
1019
decode_mc2_mce(struct mce * m)1020 static void decode_mc2_mce(struct mce *m)
1021 {
1022 u16 ec = EC(m->status);
1023 u8 xec = XEC(m->status, xec_mask);
1024
1025 pr_emerg(HW_ERR "MC2 Error: ");
1026
1027 if (!fam_ops.mc2_mce(ec, xec))
1028 pr_cont(HW_ERR "Corrupted MC2 MCE info?\n");
1029 }
1030
decode_mc3_mce(struct mce * m)1031 static void decode_mc3_mce(struct mce *m)
1032 {
1033 u16 ec = EC(m->status);
1034 u8 xec = XEC(m->status, xec_mask);
1035
1036 if (boot_cpu_data.x86 >= 0x14) {
1037 pr_emerg("You shouldn't be seeing MC3 MCE on this cpu family,"
1038 " please report on LKML.\n");
1039 return;
1040 }
1041
1042 pr_emerg(HW_ERR "MC3 Error");
1043
1044 if (xec == 0x0) {
1045 u8 r4 = R4(ec);
1046
1047 if (!BUS_ERROR(ec) || (r4 != R4_DRD && r4 != R4_DWR))
1048 goto wrong_mc3_mce;
1049
1050 pr_cont(" during %s.\n", R4_MSG(ec));
1051 } else
1052 goto wrong_mc3_mce;
1053
1054 return;
1055
1056 wrong_mc3_mce:
1057 pr_emerg(HW_ERR "Corrupted MC3 MCE info?\n");
1058 }
1059
decode_mc4_mce(struct mce * m)1060 static void decode_mc4_mce(struct mce *m)
1061 {
1062 unsigned int fam = x86_family(m->cpuid);
1063 int node_id = topology_die_id(m->extcpu);
1064 u16 ec = EC(m->status);
1065 u8 xec = XEC(m->status, 0x1f);
1066 u8 offset = 0;
1067
1068 pr_emerg(HW_ERR "MC4 Error (node %d): ", node_id);
1069
1070 switch (xec) {
1071 case 0x0 ... 0xe:
1072
1073 /* special handling for DRAM ECCs */
1074 if (xec == 0x0 || xec == 0x8) {
1075 /* no ECCs on F11h */
1076 if (fam == 0x11)
1077 goto wrong_mc4_mce;
1078
1079 pr_cont("%s.\n", mc4_mce_desc[xec]);
1080
1081 if (decode_dram_ecc)
1082 decode_dram_ecc(node_id, m);
1083 return;
1084 }
1085 break;
1086
1087 case 0xf:
1088 if (TLB_ERROR(ec))
1089 pr_cont("GART Table Walk data error.\n");
1090 else if (BUS_ERROR(ec))
1091 pr_cont("DMA Exclusion Vector Table Walk error.\n");
1092 else
1093 goto wrong_mc4_mce;
1094 return;
1095
1096 case 0x19:
1097 if (fam == 0x15 || fam == 0x16)
1098 pr_cont("Compute Unit Data Error.\n");
1099 else
1100 goto wrong_mc4_mce;
1101 return;
1102
1103 case 0x1c ... 0x1f:
1104 offset = 13;
1105 break;
1106
1107 default:
1108 goto wrong_mc4_mce;
1109 }
1110
1111 pr_cont("%s.\n", mc4_mce_desc[xec - offset]);
1112 return;
1113
1114 wrong_mc4_mce:
1115 pr_emerg(HW_ERR "Corrupted MC4 MCE info?\n");
1116 }
1117
decode_mc5_mce(struct mce * m)1118 static void decode_mc5_mce(struct mce *m)
1119 {
1120 unsigned int fam = x86_family(m->cpuid);
1121 u16 ec = EC(m->status);
1122 u8 xec = XEC(m->status, xec_mask);
1123
1124 if (fam == 0xf || fam == 0x11)
1125 goto wrong_mc5_mce;
1126
1127 pr_emerg(HW_ERR "MC5 Error: ");
1128
1129 if (INT_ERROR(ec)) {
1130 if (xec <= 0x1f) {
1131 pr_cont("Hardware Assert.\n");
1132 return;
1133 } else
1134 goto wrong_mc5_mce;
1135 }
1136
1137 if (xec == 0x0 || xec == 0xc)
1138 pr_cont("%s.\n", mc5_mce_desc[xec]);
1139 else if (xec <= 0xd)
1140 pr_cont("%s parity error.\n", mc5_mce_desc[xec]);
1141 else
1142 goto wrong_mc5_mce;
1143
1144 return;
1145
1146 wrong_mc5_mce:
1147 pr_emerg(HW_ERR "Corrupted MC5 MCE info?\n");
1148 }
1149
decode_mc6_mce(struct mce * m)1150 static void decode_mc6_mce(struct mce *m)
1151 {
1152 u8 xec = XEC(m->status, xec_mask);
1153
1154 pr_emerg(HW_ERR "MC6 Error: ");
1155
1156 if (xec > 0x5)
1157 goto wrong_mc6_mce;
1158
1159 pr_cont("%s parity error.\n", mc6_mce_desc[xec]);
1160 return;
1161
1162 wrong_mc6_mce:
1163 pr_emerg(HW_ERR "Corrupted MC6 MCE info?\n");
1164 }
1165
1166 /* Decode errors according to Scalable MCA specification */
decode_smca_error(struct mce * m)1167 static void decode_smca_error(struct mce *m)
1168 {
1169 enum smca_bank_types bank_type = smca_get_bank_type(m->extcpu, m->bank);
1170 const char *ip_name;
1171 u8 xec = XEC(m->status, xec_mask);
1172
1173 if (bank_type >= N_SMCA_BANK_TYPES)
1174 return;
1175
1176 if (bank_type == SMCA_RESERVED) {
1177 pr_emerg(HW_ERR "Bank %d is reserved.\n", m->bank);
1178 return;
1179 }
1180
1181 ip_name = smca_get_long_name(bank_type);
1182
1183 pr_emerg(HW_ERR "%s Ext. Error Code: %d", ip_name, xec);
1184
1185 /* Only print the decode of valid error codes */
1186 if (xec < smca_mce_descs[bank_type].num_descs)
1187 pr_cont(", %s.\n", smca_mce_descs[bank_type].descs[xec]);
1188
1189 if (bank_type == SMCA_UMC && xec == 0 && decode_dram_ecc)
1190 decode_dram_ecc(topology_die_id(m->extcpu), m);
1191 }
1192
amd_decode_err_code(u16 ec)1193 static inline void amd_decode_err_code(u16 ec)
1194 {
1195 if (INT_ERROR(ec)) {
1196 pr_emerg(HW_ERR "internal: %s\n", UU_MSG(ec));
1197 return;
1198 }
1199
1200 pr_emerg(HW_ERR "cache level: %s", LL_MSG(ec));
1201
1202 if (BUS_ERROR(ec))
1203 pr_cont(", mem/io: %s", II_MSG(ec));
1204 else
1205 pr_cont(", tx: %s", TT_MSG(ec));
1206
1207 if (MEM_ERROR(ec) || BUS_ERROR(ec)) {
1208 pr_cont(", mem-tx: %s", R4_MSG(ec));
1209
1210 if (BUS_ERROR(ec))
1211 pr_cont(", part-proc: %s (%s)", PP_MSG(ec), TO_MSG(ec));
1212 }
1213
1214 pr_cont("\n");
1215 }
1216
decode_error_status(struct mce * m)1217 static const char *decode_error_status(struct mce *m)
1218 {
1219 if (m->status & MCI_STATUS_UC) {
1220 if (m->status & MCI_STATUS_PCC)
1221 return "System Fatal error.";
1222 if (m->mcgstatus & MCG_STATUS_RIPV)
1223 return "Uncorrected, software restartable error.";
1224 return "Uncorrected, software containable error.";
1225 }
1226
1227 if (m->status & MCI_STATUS_DEFERRED)
1228 return "Deferred error, no action required.";
1229
1230 return "Corrected error, no action required.";
1231 }
1232
1233 static int
amd_decode_mce(struct notifier_block * nb,unsigned long val,void * data)1234 amd_decode_mce(struct notifier_block *nb, unsigned long val, void *data)
1235 {
1236 struct mce *m = (struct mce *)data;
1237 unsigned int fam = x86_family(m->cpuid);
1238 int ecc;
1239
1240 if (m->kflags & MCE_HANDLED_CEC)
1241 return NOTIFY_DONE;
1242
1243 pr_emerg(HW_ERR "%s\n", decode_error_status(m));
1244
1245 pr_emerg(HW_ERR "CPU:%d (%x:%x:%x) MC%d_STATUS[%s|%s|%s|%s|%s",
1246 m->extcpu,
1247 fam, x86_model(m->cpuid), x86_stepping(m->cpuid),
1248 m->bank,
1249 ((m->status & MCI_STATUS_OVER) ? "Over" : "-"),
1250 ((m->status & MCI_STATUS_UC) ? "UE" :
1251 (m->status & MCI_STATUS_DEFERRED) ? "-" : "CE"),
1252 ((m->status & MCI_STATUS_MISCV) ? "MiscV" : "-"),
1253 ((m->status & MCI_STATUS_ADDRV) ? "AddrV" : "-"),
1254 ((m->status & MCI_STATUS_PCC) ? "PCC" : "-"));
1255
1256 if (boot_cpu_has(X86_FEATURE_SMCA)) {
1257 u32 low, high;
1258 u32 addr = MSR_AMD64_SMCA_MCx_CONFIG(m->bank);
1259
1260 if (!rdmsr_safe(addr, &low, &high) &&
1261 (low & MCI_CONFIG_MCAX))
1262 pr_cont("|%s", ((m->status & MCI_STATUS_TCC) ? "TCC" : "-"));
1263
1264 pr_cont("|%s", ((m->status & MCI_STATUS_SYNDV) ? "SyndV" : "-"));
1265 }
1266
1267 /* do the two bits[14:13] together */
1268 ecc = (m->status >> 45) & 0x3;
1269 if (ecc)
1270 pr_cont("|%sECC", ((ecc == 2) ? "C" : "U"));
1271
1272 if (fam >= 0x15) {
1273 pr_cont("|%s", (m->status & MCI_STATUS_DEFERRED ? "Deferred" : "-"));
1274
1275 /* F15h, bank4, bit 43 is part of McaStatSubCache. */
1276 if (fam != 0x15 || m->bank != 4)
1277 pr_cont("|%s", (m->status & MCI_STATUS_POISON ? "Poison" : "-"));
1278 }
1279
1280 if (fam >= 0x17)
1281 pr_cont("|%s", (m->status & MCI_STATUS_SCRUB ? "Scrub" : "-"));
1282
1283 pr_cont("]: 0x%016llx\n", m->status);
1284
1285 if (m->status & MCI_STATUS_ADDRV)
1286 pr_emerg(HW_ERR "Error Addr: 0x%016llx\n", m->addr);
1287
1288 if (m->ppin)
1289 pr_emerg(HW_ERR "PPIN: 0x%016llx\n", m->ppin);
1290
1291 if (boot_cpu_has(X86_FEATURE_SMCA)) {
1292 pr_emerg(HW_ERR "IPID: 0x%016llx", m->ipid);
1293
1294 if (m->status & MCI_STATUS_SYNDV)
1295 pr_cont(", Syndrome: 0x%016llx", m->synd);
1296
1297 pr_cont("\n");
1298
1299 decode_smca_error(m);
1300 goto err_code;
1301 }
1302
1303 if (m->tsc)
1304 pr_emerg(HW_ERR "TSC: %llu\n", m->tsc);
1305
1306 /* Doesn't matter which member to test. */
1307 if (!fam_ops.mc0_mce)
1308 goto err_code;
1309
1310 switch (m->bank) {
1311 case 0:
1312 decode_mc0_mce(m);
1313 break;
1314
1315 case 1:
1316 decode_mc1_mce(m);
1317 break;
1318
1319 case 2:
1320 decode_mc2_mce(m);
1321 break;
1322
1323 case 3:
1324 decode_mc3_mce(m);
1325 break;
1326
1327 case 4:
1328 decode_mc4_mce(m);
1329 break;
1330
1331 case 5:
1332 decode_mc5_mce(m);
1333 break;
1334
1335 case 6:
1336 decode_mc6_mce(m);
1337 break;
1338
1339 default:
1340 break;
1341 }
1342
1343 err_code:
1344 amd_decode_err_code(m->status & 0xffff);
1345
1346 m->kflags |= MCE_HANDLED_EDAC;
1347 return NOTIFY_OK;
1348 }
1349
1350 static struct notifier_block amd_mce_dec_nb = {
1351 .notifier_call = amd_decode_mce,
1352 .priority = MCE_PRIO_EDAC,
1353 };
1354
mce_amd_init(void)1355 static int __init mce_amd_init(void)
1356 {
1357 struct cpuinfo_x86 *c = &boot_cpu_data;
1358
1359 if (c->x86_vendor != X86_VENDOR_AMD &&
1360 c->x86_vendor != X86_VENDOR_HYGON)
1361 return -ENODEV;
1362
1363 if (cpu_feature_enabled(X86_FEATURE_HYPERVISOR))
1364 return -ENODEV;
1365
1366 if (boot_cpu_has(X86_FEATURE_SMCA)) {
1367 xec_mask = 0x3f;
1368 goto out;
1369 }
1370
1371 switch (c->x86) {
1372 case 0xf:
1373 fam_ops.mc0_mce = k8_mc0_mce;
1374 fam_ops.mc1_mce = k8_mc1_mce;
1375 fam_ops.mc2_mce = k8_mc2_mce;
1376 break;
1377
1378 case 0x10:
1379 fam_ops.mc0_mce = f10h_mc0_mce;
1380 fam_ops.mc1_mce = k8_mc1_mce;
1381 fam_ops.mc2_mce = k8_mc2_mce;
1382 break;
1383
1384 case 0x11:
1385 fam_ops.mc0_mce = k8_mc0_mce;
1386 fam_ops.mc1_mce = k8_mc1_mce;
1387 fam_ops.mc2_mce = k8_mc2_mce;
1388 break;
1389
1390 case 0x12:
1391 fam_ops.mc0_mce = f12h_mc0_mce;
1392 fam_ops.mc1_mce = k8_mc1_mce;
1393 fam_ops.mc2_mce = k8_mc2_mce;
1394 break;
1395
1396 case 0x14:
1397 fam_ops.mc0_mce = cat_mc0_mce;
1398 fam_ops.mc1_mce = cat_mc1_mce;
1399 fam_ops.mc2_mce = k8_mc2_mce;
1400 break;
1401
1402 case 0x15:
1403 xec_mask = c->x86_model == 0x60 ? 0x3f : 0x1f;
1404
1405 fam_ops.mc0_mce = f15h_mc0_mce;
1406 fam_ops.mc1_mce = f15h_mc1_mce;
1407 fam_ops.mc2_mce = f15h_mc2_mce;
1408 break;
1409
1410 case 0x16:
1411 xec_mask = 0x1f;
1412 fam_ops.mc0_mce = cat_mc0_mce;
1413 fam_ops.mc1_mce = cat_mc1_mce;
1414 fam_ops.mc2_mce = f16h_mc2_mce;
1415 break;
1416
1417 case 0x17:
1418 case 0x18:
1419 pr_warn_once("Decoding supported only on Scalable MCA processors.\n");
1420 return -EINVAL;
1421
1422 default:
1423 printk(KERN_WARNING "Huh? What family is it: 0x%x?!\n", c->x86);
1424 return -EINVAL;
1425 }
1426
1427 out:
1428 pr_info("MCE: In-kernel MCE decoding enabled.\n");
1429
1430 mce_register_decode_chain(&amd_mce_dec_nb);
1431
1432 return 0;
1433 }
1434 early_initcall(mce_amd_init);
1435
1436 #ifdef MODULE
mce_amd_exit(void)1437 static void __exit mce_amd_exit(void)
1438 {
1439 mce_unregister_decode_chain(&amd_mce_dec_nb);
1440 }
1441
1442 MODULE_DESCRIPTION("AMD MCE decoder");
1443 MODULE_ALIAS("edac-mce-amd");
1444 MODULE_LICENSE("GPL");
1445 module_exit(mce_amd_exit);
1446 #endif
1447