mce_amd.c 17.8 KB
Newer Older
1
#include <linux/module.h>
2 3
#include <linux/slab.h>

B
Borislav Petkov 已提交
4
#include "mce_amd.h"
D
Doug Thompson 已提交
5

6 7
static struct amd_decoder_ops *fam_ops;

8
static u8 xec_mask	 = 0xf;
9 10
static u8 nb_err_cpumask = 0xf;

11
static bool report_gart_errors;
12
static void (*nb_bus_decoder)(int node_id, struct mce *m);
13 14 15 16 17 18 19

void amd_report_gart_errors(bool v)
{
	report_gart_errors = v;
}
EXPORT_SYMBOL_GPL(amd_report_gart_errors);

20
void amd_register_ecc_decoder(void (*f)(int, struct mce *))
21 22 23 24 25
{
	nb_bus_decoder = f;
}
EXPORT_SYMBOL_GPL(amd_register_ecc_decoder);

26
void amd_unregister_ecc_decoder(void (*f)(int, struct mce *))
27 28 29 30 31 32 33 34 35
{
	if (nb_bus_decoder) {
		WARN_ON(nb_bus_decoder != f);

		nb_bus_decoder = NULL;
	}
}
EXPORT_SYMBOL_GPL(amd_unregister_ecc_decoder);

D
Doug Thompson 已提交
36 37 38 39
/*
 * string representation for the different MCA reported error types, see F3x48
 * or MSR0000_0411.
 */
B
Borislav Petkov 已提交
40 41

/* transaction type */
42
static const char * const tt_msgs[] = { "INSN", "DATA", "GEN", "RESV" };
D
Doug Thompson 已提交
43

B
Borislav Petkov 已提交
44
/* cache level */
45
static const char * const ll_msgs[] = { "RESV", "L1", "L2", "L3/GEN" };
D
Doug Thompson 已提交
46

B
Borislav Petkov 已提交
47
/* memory transaction type */
48
static const char * const rrrr_msgs[] = {
B
Borislav Petkov 已提交
49
       "GEN", "RD", "WR", "DRD", "DWR", "IRD", "PRF", "EV", "SNP"
D
Doug Thompson 已提交
50 51
};

B
Borislav Petkov 已提交
52
/* participating processor */
53
const char * const pp_msgs[] = { "SRC", "RES", "OBS", "GEN" };
54
EXPORT_SYMBOL_GPL(pp_msgs);
D
Doug Thompson 已提交
55

B
Borislav Petkov 已提交
56
/* request timeout */
57
static const char * const to_msgs[] = { "no timeout", "timed out" };
D
Doug Thompson 已提交
58

B
Borislav Petkov 已提交
59
/* memory or i/o */
60
static const char * const ii_msgs[] = { "MEM", "RESV", "IO", "GEN" };
D
Doug Thompson 已提交
61

62
/* internal error type */
63
static const char * const uu_msgs[] = { "RESV", "RESV", "HWA", "RESV" };
64

65
static const char * const f15h_mc1_mce_desc[] = {
66 67 68 69 70 71 72 73 74 75 76 77
	"UC during a demand linefill from L2",
	"Parity error during data load from IC",
	"Parity error for IC valid bit",
	"Main tag parity error",
	"Parity error in prediction queue",
	"PFB data/address parity error",
	"Parity error in the branch status reg",
	"PFB promotion address error",
	"Tag error during probe/victimization",
	"Parity error for IC probe tag valid bit",
	"PFB non-cacheable bit parity error",
	"PFB valid bit parity error",			/* xec = 0xd */
78
	"Microcode Patch Buffer",			/* xec = 010 */
79 80 81 82 83 84
	"uop queue",
	"insn buffer",
	"predecode buffer",
	"fetch address FIFO"
};

85
static const char * const f15h_mc2_mce_desc[] = {
86 87 88 89 90 91 92 93
	"Fill ECC error on data fills",			/* xec = 0x4 */
	"Fill parity error on insn fills",
	"Prefetcher request FIFO parity error",
	"PRQ address parity error",
	"PRQ data parity error",
	"WCC Tag ECC error",
	"WCC Data ECC error",
	"WCB Data parity error",
94
	"VB Data ECC or parity error",
95 96 97 98 99 100 101
	"L2 Tag ECC error",				/* xec = 0x10 */
	"Hard L2 Tag ECC error",
	"Multiple hits on L2 tag",
	"XAB parity error",
	"PRB address parity error"
};

102
static const char * const mc4_mce_desc[] = {
103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123
	"DRAM ECC error detected on the NB",
	"CRC error detected on HT link",
	"Link-defined sync error packets detected on HT link",
	"HT Master abort",
	"HT Target abort",
	"Invalid GART PTE entry during GART table walk",
	"Unsupported atomic RMW received from an IO link",
	"Watchdog timeout due to lack of progress",
	"DRAM ECC error detected on the NB",
	"SVM DMA Exclusion Vector error",
	"HT data error detected on link",
	"Protocol error (link, L3, probe filter)",
	"NB internal arrays parity error",
	"DRAM addr/ctl signals parity error",
	"IO link transmission error",
	"L3 data cache ECC error",			/* xec = 0x1c */
	"L3 cache tag error",
	"L3 LRU parity bits error",
	"ECC Error in the Probe Filter directory"
};

124
static const char * const mc5_mce_desc[] = {
125 126 127 128 129 130 131 132 133 134 135 136
	"CPU Watchdog timer expire",
	"Wakeup array dest tag",
	"AG payload array",
	"EX payload array",
	"IDRF array",
	"Retire dispatch queue",
	"Mapper checkpoint array",
	"Physical register file EX0 port",
	"Physical register file EX1 port",
	"Physical register file AG0 port",
	"Physical register file AG1 port",
	"Flag register file",
137
	"DE error occurred"
138 139
};

140
static bool f12h_mc0_mce(u16 ec, u8 xec)
141
{
142
	bool ret = false;
143

144
	if (MEM_ERROR(ec)) {
145
		u8 ll = LL(ec);
146
		ret = true;
147

148 149 150
		if (ll == LL_L2)
			pr_cont("during L1 linefill from L2.\n");
		else if (ll == LL_L1)
151
			pr_cont("Data/Tag %s error.\n", R4_MSG(ec));
152 153 154 155 156
		else
			ret = false;
	}
	return ret;
}
157

158
static bool f10h_mc0_mce(u16 ec, u8 xec)
159
{
160
	if (R4(ec) == R4_GEN && LL(ec) == LL_L1) {
161 162 163
		pr_cont("during data scrub.\n");
		return true;
	}
164
	return f12h_mc0_mce(ec, xec);
165 166
}

167
static bool k8_mc0_mce(u16 ec, u8 xec)
168 169 170 171 172
{
	if (BUS_ERROR(ec)) {
		pr_cont("during system linefill.\n");
		return true;
	}
173

174
	return f10h_mc0_mce(ec, xec);
175 176
}

177
static bool cat_mc0_mce(u16 ec, u8 xec)
178
{
179
	u8 r4	 = R4(ec);
180 181 182 183
	bool ret = true;

	if (MEM_ERROR(ec)) {

184
		if (TT(ec) != TT_DATA || LL(ec) != LL_L1)
185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203
			return false;

		switch (r4) {
		case R4_DRD:
		case R4_DWR:
			pr_cont("Data/Tag parity error due to %s.\n",
				(r4 == R4_DRD ? "load/hw prf" : "store"));
			break;
		case R4_EVICT:
			pr_cont("Copyback parity error on a tag miss.\n");
			break;
		case R4_SNOOP:
			pr_cont("Tag parity error during snoop.\n");
			break;
		default:
			ret = false;
		}
	} else if (BUS_ERROR(ec)) {

204
		if ((II(ec) != II_MEM && II(ec) != II_IO) || LL(ec) != LL_LG)
205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228
			return false;

		pr_cont("System read data error on a ");

		switch (r4) {
		case R4_RD:
			pr_cont("TLB reload.\n");
			break;
		case R4_DWR:
			pr_cont("store.\n");
			break;
		case R4_DRD:
			pr_cont("load.\n");
			break;
		default:
			ret = false;
		}
	} else {
		ret = false;
	}

	return ret;
}

229
static bool f15h_mc0_mce(u16 ec, u8 xec)
230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266
{
	bool ret = true;

	if (MEM_ERROR(ec)) {

		switch (xec) {
		case 0x0:
			pr_cont("Data Array access error.\n");
			break;

		case 0x1:
			pr_cont("UC error during a linefill from L2/NB.\n");
			break;

		case 0x2:
		case 0x11:
			pr_cont("STQ access error.\n");
			break;

		case 0x3:
			pr_cont("SCB access error.\n");
			break;

		case 0x10:
			pr_cont("Tag error.\n");
			break;

		case 0x12:
			pr_cont("LDQ access error.\n");
			break;

		default:
			ret = false;
		}
	} else if (BUS_ERROR(ec)) {

		if (!xec)
267
			pr_cont("System Read Data Error.\n");
268
		else
269
			pr_cont(" Internal error condition type %d.\n", xec);
270 271 272 273 274 275
	} else
		ret = false;

	return ret;
}

276
static void decode_mc0_mce(struct mce *m)
277
{
278 279
	u16 ec = EC(m->status);
	u8 xec = XEC(m->status, xec_mask);
280

281
	pr_emerg(HW_ERR "MC0 Error: ");
282 283 284

	/* TLB error signatures are the same across families */
	if (TLB_ERROR(ec)) {
285
		if (TT(ec) == TT_DATA) {
286
			pr_cont("%s TLB %s.\n", LL_MSG(ec),
287 288
				((xec == 2) ? "locked miss"
					    : (xec ? "multimatch" : "parity")));
289 290
			return;
		}
291
	} else if (fam_ops->mc0_mce(ec, xec))
292 293
		;
	else
294
		pr_emerg(HW_ERR "Corrupted MC0 MCE info?\n");
295 296
}

297
static bool k8_mc1_mce(u16 ec, u8 xec)
298
{
299
	u8 ll	 = LL(ec);
300
	bool ret = true;
301

302 303
	if (!MEM_ERROR(ec))
		return false;
304

305 306 307
	if (ll == 0x2)
		pr_cont("during a linefill from L2.\n");
	else if (ll == 0x1) {
308
		switch (R4(ec)) {
309 310 311
		case R4_IRD:
			pr_cont("Parity error during data load.\n");
			break;
312

313 314 315 316 317 318 319 320 321 322 323 324
		case R4_EVICT:
			pr_cont("Copyback Parity/Victim error.\n");
			break;

		case R4_SNOOP:
			pr_cont("Tag Snoop error.\n");
			break;

		default:
			ret = false;
			break;
		}
325
	} else
326
		ret = false;
327

328 329 330
	return ret;
}

331
static bool cat_mc1_mce(u16 ec, u8 xec)
332
{
333
	u8 r4    = R4(ec);
334
	bool ret = true;
335

336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351
	if (!MEM_ERROR(ec))
		return false;

	if (TT(ec) != TT_INSTR)
		return false;

	if (r4 == R4_IRD)
		pr_cont("Data/tag array parity error for a tag hit.\n");
	else if (r4 == R4_SNOOP)
		pr_cont("Tag error during snoop/victimization.\n");
	else if (xec == 0x0)
		pr_cont("Tag parity error from victim castout.\n");
	else if (xec == 0x2)
		pr_cont("Microcode patch RAM parity error.\n");
	else
		ret = false;
352 353 354 355

	return ret;
}

356
static bool f15h_mc1_mce(u16 ec, u8 xec)
357 358 359 360 361 362 363 364
{
	bool ret = true;

	if (!MEM_ERROR(ec))
		return false;

	switch (xec) {
	case 0x0 ... 0xa:
365
		pr_cont("%s.\n", f15h_mc1_mce_desc[xec]);
366 367 368
		break;

	case 0xd:
369
		pr_cont("%s.\n", f15h_mc1_mce_desc[xec-2]);
370 371
		break;

372
	case 0x10:
373
		pr_cont("%s.\n", f15h_mc1_mce_desc[xec-4]);
374 375 376
		break;

	case 0x11 ... 0x14:
377
		pr_cont("Decoder %s parity error.\n", f15h_mc1_mce_desc[xec-4]);
378 379 380 381 382 383 384 385
		break;

	default:
		ret = false;
	}
	return ret;
}

386
static void decode_mc1_mce(struct mce *m)
387
{
388 389
	u16 ec = EC(m->status);
	u8 xec = XEC(m->status, xec_mask);
390

391
	pr_emerg(HW_ERR "MC1 Error: ");
392 393 394 395 396

	if (TLB_ERROR(ec))
		pr_cont("%s TLB %s.\n", LL_MSG(ec),
			(xec ? "multimatch" : "parity error"));
	else if (BUS_ERROR(ec)) {
397
		bool k8 = (boot_cpu_data.x86 == 0xf && (m->status & BIT_64(58)));
398 399

		pr_cont("during %s.\n", (k8 ? "system linefill" : "NB data read"));
400
	} else if (fam_ops->mc1_mce(ec, xec))
401 402
		;
	else
403
		pr_emerg(HW_ERR "Corrupted MC1 MCE info?\n");
404 405
}

406
static bool k8_mc2_mce(u16 ec, u8 xec)
407
{
408
	bool ret = true;
409 410 411 412 413 414

	if (xec == 0x1)
		pr_cont(" in the write data buffers.\n");
	else if (xec == 0x3)
		pr_cont(" in the victim data buffers.\n");
	else if (xec == 0x2 && MEM_ERROR(ec))
415
		pr_cont(": %s error in the L2 cache tags.\n", R4_MSG(ec));
416 417 418 419 420 421
	else if (xec == 0x0) {
		if (TLB_ERROR(ec))
			pr_cont(": %s error in a Page Descriptor Cache or "
				"Guest TLB.\n", TT_MSG(ec));
		else if (BUS_ERROR(ec))
			pr_cont(": %s/ECC error in data read from NB: %s.\n",
422
				R4_MSG(ec), PP_MSG(ec));
423
		else if (MEM_ERROR(ec)) {
424
			u8 r4 = R4(ec);
425

426
			if (r4 >= 0x7)
427
				pr_cont(": %s error during data copyback.\n",
428 429
					R4_MSG(ec));
			else if (r4 <= 0x1)
430
				pr_cont(": %s parity/ECC error during data "
431
					"access from L2.\n", R4_MSG(ec));
432
			else
433
				ret = false;
434
		} else
435
			ret = false;
436
	} else
437
		ret = false;
438

439
	return ret;
440 441
}

442
static bool f15h_mc2_mce(u16 ec, u8 xec)
443
{
444
	bool ret = true;
445 446 447 448 449 450 451

	if (TLB_ERROR(ec)) {
		if (xec == 0x0)
			pr_cont("Data parity TLB read error.\n");
		else if (xec == 0x1)
			pr_cont("Poison data provided for TLB fill.\n");
		else
452
			ret = false;
453 454
	} else if (BUS_ERROR(ec)) {
		if (xec > 2)
455
			ret = false;
456 457 458 459 460

		pr_cont("Error during attempted NB data read.\n");
	} else if (MEM_ERROR(ec)) {
		switch (xec) {
		case 0x4 ... 0xc:
461
			pr_cont("%s.\n", f15h_mc2_mce_desc[xec - 0x4]);
462 463 464
			break;

		case 0x10 ... 0x14:
465
			pr_cont("%s.\n", f15h_mc2_mce_desc[xec - 0x7]);
466 467 468
			break;

		default:
469
			ret = false;
470 471 472
		}
	}

473 474 475
	return ret;
}

476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516
static bool f16h_mc2_mce(u16 ec, u8 xec)
{
	u8 r4 = R4(ec);

	if (!MEM_ERROR(ec))
		return false;

	switch (xec) {
	case 0x04 ... 0x05:
		pr_cont("%cBUFF parity error.\n", (r4 == R4_RD) ? 'I' : 'O');
		break;

	case 0x09 ... 0x0b:
	case 0x0d ... 0x0f:
		pr_cont("ECC error in L2 tag (%s).\n",
			((r4 == R4_GEN)   ? "BankReq" :
			((r4 == R4_SNOOP) ? "Prb"     : "Fill")));
		break;

	case 0x10 ... 0x19:
	case 0x1b:
		pr_cont("ECC error in L2 data array (%s).\n",
			(((r4 == R4_RD) && !(xec & 0x3)) ? "Hit"  :
			((r4 == R4_GEN)   ? "Attr" :
			((r4 == R4_EVICT) ? "Vict" : "Fill"))));
		break;

	case 0x1c ... 0x1d:
	case 0x1f:
		pr_cont("Parity error in L2 attribute bits (%s).\n",
			((r4 == R4_RD)  ? "Hit"  :
			((r4 == R4_GEN) ? "Attr" : "Fill")));
		break;

	default:
		return false;
	}

	return true;
}

517 518 519 520
static void decode_mc2_mce(struct mce *m)
{
	u16 ec = EC(m->status);
	u8 xec = XEC(m->status, xec_mask);
521

522 523 524 525
	pr_emerg(HW_ERR "MC2 Error: ");

	if (!fam_ops->mc2_mce(ec, xec))
		pr_cont(HW_ERR "Corrupted MC2 MCE info?\n");
526 527
}

528
static void decode_mc3_mce(struct mce *m)
529
{
530 531
	u16 ec = EC(m->status);
	u8 xec = XEC(m->status, xec_mask);
532

533
	if (boot_cpu_data.x86 >= 0x14) {
534
		pr_emerg("You shouldn't be seeing MC3 MCE on this cpu family,"
535 536 537
			 " please report on LKML.\n");
		return;
	}
538

539
	pr_emerg(HW_ERR "MC3 Error");
540 541

	if (xec == 0x0) {
542
		u8 r4 = R4(ec);
543

544
		if (!BUS_ERROR(ec) || (r4 != R4_DRD && r4 != R4_DWR))
545
			goto wrong_mc3_mce;
546

547
		pr_cont(" during %s.\n", R4_MSG(ec));
548
	} else
549
		goto wrong_mc3_mce;
550

551 552
	return;

553 554
 wrong_mc3_mce:
	pr_emerg(HW_ERR "Corrupted MC3 MCE info?\n");
555 556
}

557
static void decode_mc4_mce(struct mce *m)
558
{
559 560 561 562 563
	struct cpuinfo_x86 *c = &boot_cpu_data;
	int node_id = amd_get_nb_id(m->extcpu);
	u16 ec = EC(m->status);
	u8 xec = XEC(m->status, 0x1f);
	u8 offset = 0;
564

565
	pr_emerg(HW_ERR "MC4 Error (node %d): ", node_id);
566

567 568
	switch (xec) {
	case 0x0 ... 0xe:
569

570 571 572 573
		/* special handling for DRAM ECCs */
		if (xec == 0x0 || xec == 0x8) {
			/* no ECCs on F11h */
			if (c->x86 == 0x11)
574
				goto wrong_mc4_mce;
575

576
			pr_cont("%s.\n", mc4_mce_desc[xec]);
577

578 579 580 581
			if (nb_bus_decoder)
				nb_bus_decoder(node_id, m);
			return;
		}
582 583 584 585 586 587 588 589
		break;

	case 0xf:
		if (TLB_ERROR(ec))
			pr_cont("GART Table Walk data error.\n");
		else if (BUS_ERROR(ec))
			pr_cont("DMA Exclusion Vector Table Walk error.\n");
		else
590
			goto wrong_mc4_mce;
591
		return;
592

593
	case 0x19:
594
		if (boot_cpu_data.x86 == 0x15 || boot_cpu_data.x86 == 0x16)
595 596
			pr_cont("Compute Unit Data Error.\n");
		else
597
			goto wrong_mc4_mce;
598
		return;
599

600
	case 0x1c ... 0x1f:
601
		offset = 13;
602 603 604
		break;

	default:
605
		goto wrong_mc4_mce;
606
	}
607

608
	pr_cont("%s.\n", mc4_mce_desc[xec - offset]);
609 610
	return;

611 612
 wrong_mc4_mce:
	pr_emerg(HW_ERR "Corrupted MC4 MCE info?\n");
613 614
}

615
static void decode_mc5_mce(struct mce *m)
B
Borislav Petkov 已提交
616
{
617
	struct cpuinfo_x86 *c = &boot_cpu_data;
618
	u8 xec = XEC(m->status, xec_mask);
619 620

	if (c->x86 == 0xf || c->x86 == 0x11)
621
		goto wrong_mc5_mce;
B
Borislav Petkov 已提交
622

623
	pr_emerg(HW_ERR "MC5 Error: ");
624 625

	if (xec == 0x0 || xec == 0xc)
626
		pr_cont("%s.\n", mc5_mce_desc[xec]);
627
	else if (xec < 0xd)
628
		pr_cont("%s parity error.\n", mc5_mce_desc[xec]);
629
	else
630
		goto wrong_mc5_mce;
631 632

	return;
B
Borislav Petkov 已提交
633

634 635
 wrong_mc5_mce:
	pr_emerg(HW_ERR "Corrupted MC5 MCE info?\n");
B
Borislav Petkov 已提交
636 637
}

638
static void decode_mc6_mce(struct mce *m)
639
{
640
	u8 xec = XEC(m->status, xec_mask);
641

642
	pr_emerg(HW_ERR "MC6 Error: ");
643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665

	switch (xec) {
	case 0x1:
		pr_cont("Free List");
		break;

	case 0x2:
		pr_cont("Physical Register File");
		break;

	case 0x3:
		pr_cont("Retire Queue");
		break;

	case 0x4:
		pr_cont("Scheduler table");
		break;

	case 0x5:
		pr_cont("Status Register File");
		break;

	default:
666
		goto wrong_mc6_mce;
667 668 669 670 671 672 673
		break;
	}

	pr_cont(" parity error.\n");

	return;

674 675
 wrong_mc6_mce:
	pr_emerg(HW_ERR "Corrupted MC6 MCE info?\n");
676 677
}

B
Borislav Petkov 已提交
678
static inline void amd_decode_err_code(u16 ec)
679
{
680 681 682 683
	if (INT_ERROR(ec)) {
		pr_emerg(HW_ERR "internal: %s\n", UU_MSG(ec));
		return;
	}
684 685 686 687 688 689 690 691 692 693 694 695 696 697 698 699

	pr_emerg(HW_ERR "cache level: %s", LL_MSG(ec));

	if (BUS_ERROR(ec))
		pr_cont(", mem/io: %s", II_MSG(ec));
	else
		pr_cont(", tx: %s", TT_MSG(ec));

	if (MEM_ERROR(ec) || BUS_ERROR(ec)) {
		pr_cont(", mem-tx: %s", R4_MSG(ec));

		if (BUS_ERROR(ec))
			pr_cont(", part-proc: %s (%s)", PP_MSG(ec), TO_MSG(ec));
	}

	pr_cont("\n");
700 701
}

702 703 704 705 706 707 708 709 710 711 712 713 714 715 716 717
/*
 * Filter out unwanted MCE signatures here.
 */
static bool amd_filter_mce(struct mce *m)
{
	u8 xec = (m->status >> 16) & 0x1f;

	/*
	 * NB GART TLB error reporting is disabled by default.
	 */
	if (m->bank == 4 && xec == 0x5 && !report_gart_errors)
		return true;

	return false;
}

B
Borislav Petkov 已提交
718 719 720 721 722 723 724 725 726 727 728 729 730 731 732 733
static const char *decode_error_status(struct mce *m)
{
	if (m->status & MCI_STATUS_UC) {
		if (m->status & MCI_STATUS_PCC)
			return "System Fatal error.";
		if (m->mcgstatus & MCG_STATUS_RIPV)
			return "Uncorrected, software restartable error.";
		return "Uncorrected, software containable error.";
	}

	if (m->status & MCI_STATUS_DEFERRED)
		return "Deferred error.";

	return "Corrected error, no action required.";
}

B
Borislav Petkov 已提交
734
int amd_decode_mce(struct notifier_block *nb, unsigned long val, void *data)
735
{
736
	struct mce *m = (struct mce *)data;
737
	struct cpuinfo_x86 *c = &cpu_data(m->extcpu);
738
	int ecc;
739

740 741 742
	if (amd_filter_mce(m))
		return NOTIFY_STOP;

743 744
	switch (m->bank) {
	case 0:
745
		decode_mc0_mce(m);
746
		break;
747

748
	case 1:
749
		decode_mc1_mce(m);
750 751
		break;

752
	case 2:
753
		decode_mc2_mce(m);
754 755
		break;

756
	case 3:
757
		decode_mc3_mce(m);
758 759
		break;

760
	case 4:
761
		decode_mc4_mce(m);
762 763
		break;

B
Borislav Petkov 已提交
764
	case 5:
765
		decode_mc5_mce(m);
B
Borislav Petkov 已提交
766 767
		break;

768
	case 6:
769
		decode_mc6_mce(m);
770 771
		break;

772 773
	default:
		break;
774
	}
775

B
Borislav Petkov 已提交
776 777
	pr_emerg(HW_ERR "Error Status: %s\n", decode_error_status(m));

778 779 780 781 782 783 784 785 786 787
	pr_emerg(HW_ERR "CPU:%d (%x:%x:%x) MC%d_STATUS[%s|%s|%s|%s|%s",
		m->extcpu,
		c->x86, c->x86_model, c->x86_mask,
		m->bank,
		((m->status & MCI_STATUS_OVER)	? "Over"  : "-"),
		((m->status & MCI_STATUS_UC)	? "UE"	  : "CE"),
		((m->status & MCI_STATUS_MISCV)	? "MiscV" : "-"),
		((m->status & MCI_STATUS_PCC)	? "PCC"	  : "-"),
		((m->status & MCI_STATUS_ADDRV)	? "AddrV" : "-"));

788
	if (c->x86 == 0x15 || c->x86 == 0x16)
789
		pr_cont("|%s|%s",
B
Borislav Petkov 已提交
790 791
			((m->status & MCI_STATUS_DEFERRED) ? "Deferred" : "-"),
			((m->status & MCI_STATUS_POISON)   ? "Poison"   : "-"));
792 793 794 795 796 797 798 799 800 801 802

	/* do the two bits[14:13] together */
	ecc = (m->status >> 45) & 0x3;
	if (ecc)
		pr_cont("|%sECC", ((ecc == 2) ? "C" : "U"));

	pr_cont("]: 0x%016llx\n", m->status);

	if (m->status & MCI_STATUS_ADDRV)
		pr_emerg(HW_ERR "MC%d_ADDR: 0x%016llx\n", m->bank, m->addr);

803
	amd_decode_err_code(m->status & 0xffff);
804 805

	return NOTIFY_STOP;
806
}
B
Borislav Petkov 已提交
807
EXPORT_SYMBOL_GPL(amd_decode_mce);
808

809 810 811 812
static struct notifier_block amd_mce_dec_nb = {
	.notifier_call	= amd_decode_mce,
};

813 814
static int __init mce_amd_init(void)
{
815 816 817
	struct cpuinfo_x86 *c = &boot_cpu_data;

	if (c->x86_vendor != X86_VENDOR_AMD)
818 819
		return 0;

820
	if (c->x86 < 0xf || c->x86 > 0x16)
821 822
		return 0;

823 824 825 826
	fam_ops = kzalloc(sizeof(struct amd_decoder_ops), GFP_KERNEL);
	if (!fam_ops)
		return -ENOMEM;

827
	switch (c->x86) {
828
	case 0xf:
829 830
		fam_ops->mc0_mce = k8_mc0_mce;
		fam_ops->mc1_mce = k8_mc1_mce;
831
		fam_ops->mc2_mce = k8_mc2_mce;
832 833 834
		break;

	case 0x10:
835 836
		fam_ops->mc0_mce = f10h_mc0_mce;
		fam_ops->mc1_mce = k8_mc1_mce;
837
		fam_ops->mc2_mce = k8_mc2_mce;
838 839
		break;

840
	case 0x11:
841 842
		fam_ops->mc0_mce = k8_mc0_mce;
		fam_ops->mc1_mce = k8_mc1_mce;
843
		fam_ops->mc2_mce = k8_mc2_mce;
844 845
		break;

846
	case 0x12:
847 848
		fam_ops->mc0_mce = f12h_mc0_mce;
		fam_ops->mc1_mce = k8_mc1_mce;
849
		fam_ops->mc2_mce = k8_mc2_mce;
850 851
		break;

852
	case 0x14:
853
		nb_err_cpumask  = 0x3;
854 855
		fam_ops->mc0_mce = cat_mc0_mce;
		fam_ops->mc1_mce = cat_mc1_mce;
856
		fam_ops->mc2_mce = k8_mc2_mce;
857 858
		break;

859 860
	case 0x15:
		xec_mask = 0x1f;
861 862
		fam_ops->mc0_mce = f15h_mc0_mce;
		fam_ops->mc1_mce = f15h_mc1_mce;
863
		fam_ops->mc2_mce = f15h_mc2_mce;
864 865
		break;

866 867 868 869 870 871 872
	case 0x16:
		xec_mask = 0x1f;
		fam_ops->mc0_mce = cat_mc0_mce;
		fam_ops->mc1_mce = cat_mc1_mce;
		fam_ops->mc2_mce = f16h_mc2_mce;
		break;

873
	default:
874
		printk(KERN_WARNING "Huh? What family is it: 0x%x?!\n", c->x86);
875 876 877 878
		kfree(fam_ops);
		return -EINVAL;
	}

879 880
	pr_info("MCE: In-kernel MCE decoding enabled.\n");

881
	mce_register_decode_chain(&amd_mce_dec_nb);
882 883 884 885

	return 0;
}
early_initcall(mce_amd_init);
886 887 888 889

#ifdef MODULE
static void __exit mce_amd_exit(void)
{
890
	mce_unregister_decode_chain(&amd_mce_dec_nb);
891
	kfree(fam_ops);
892 893 894 895 896 897 898
}

MODULE_DESCRIPTION("AMD MCE decoder");
MODULE_ALIAS("edac-mce-amd");
MODULE_LICENSE("GPL");
module_exit(mce_amd_exit);
#endif