mce_amd.c 25.4 KB
Newer Older
1
/*
2
 *  (c) 2005-2016 Advanced Micro Devices, Inc.
3 4 5 6 7
 *  Your use of this code is subject to the terms and conditions of the
 *  GNU general public license version 2. See "COPYING" or
 *  http://www.gnu.org/licenses/gpl.html
 *
 *  Written by Jacob Shin - AMD, Inc.
8
 *  Maintained by: Borislav Petkov <bp@alien8.de>
9
 *
B
Borislav Petkov 已提交
10
 *  All MC4_MISCi registers are shared between cores on a node.
11 12 13
 */
#include <linux/interrupt.h>
#include <linux/notifier.h>
I
Ingo Molnar 已提交
14
#include <linux/kobject.h>
15
#include <linux/percpu.h>
I
Ingo Molnar 已提交
16 17
#include <linux/errno.h>
#include <linux/sched.h>
18
#include <linux/sysfs.h>
19
#include <linux/slab.h>
I
Ingo Molnar 已提交
20 21 22 23
#include <linux/init.h>
#include <linux/cpu.h>
#include <linux/smp.h>

24
#include <asm/amd_nb.h>
25
#include <asm/apic.h>
I
Ingo Molnar 已提交
26
#include <asm/idle.h>
27 28
#include <asm/mce.h>
#include <asm/msr.h>
29
#include <asm/trace/irq_vectors.h>
30

31
#define NR_BLOCKS         5
J
Jacob Shin 已提交
32 33 34
#define THRESHOLD_MAX     0xFFF
#define INT_TYPE_APIC     0x00020000
#define MASK_VALID_HI     0x80000000
35 36
#define MASK_CNTP_HI      0x40000000
#define MASK_LOCKED_HI    0x20000000
J
Jacob Shin 已提交
37 38 39 40
#define MASK_LVTOFF_HI    0x00F00000
#define MASK_COUNT_EN_HI  0x00080000
#define MASK_INT_TYPE_HI  0x00060000
#define MASK_OVERFLOW_HI  0x00010000
41
#define MASK_ERR_COUNT_HI 0x00000FFF
42 43
#define MASK_BLKPTR_LO    0xFF000000
#define MCG_XBLK_ADDR     0xC0000400
44

45 46 47 48 49 50 51
/* Deferred error settings */
#define MSR_CU_DEF_ERR		0xC0000410
#define MASK_DEF_LVTOFF		0x000000F0
#define MASK_DEF_INT_TYPE	0x00000006
#define DEF_LVT_OFF		0x2
#define DEF_INT_TYPE_APIC	0x2

52 53 54 55 56
/* Scalable MCA: */

/* Threshold LVT offset is at MSR0xC0000410[15:12] */
#define SMCA_THR_LVT_OFF	0xF000

57 58 59 60 61 62 63 64 65
static const char * const th_names[] = {
	"load_store",
	"insn_fetch",
	"combined_unit",
	"",
	"northbridge",
	"execution_unit",
};

66 67 68 69 70 71 72 73 74 75 76 77 78 79
struct smca_bank_name smca_bank_names[] = {
	[SMCA_LS]	= { "load_store",	"Load Store Unit" },
	[SMCA_IF]	= { "insn_fetch",	"Instruction Fetch Unit" },
	[SMCA_L2_CACHE]	= { "l2_cache",		"L2 Cache" },
	[SMCA_DE]	= { "decode_unit",	"Decode Unit" },
	[SMCA_EX]	= { "execution_unit",	"Execution Unit" },
	[SMCA_FP]	= { "floating_point",	"Floating Point Unit" },
	[SMCA_L3_CACHE]	= { "l3_cache",		"L3 Cache" },
	[SMCA_CS]	= { "coherent_slave",	"Coherent Slave" },
	[SMCA_PIE]	= { "pie",		"Power, Interrupts, etc." },
	[SMCA_UMC]	= { "umc",		"Unified Memory Controller" },
	[SMCA_PB]	= { "param_block",	"Parameter Block" },
	[SMCA_PSP]	= { "psp",		"Platform Security Processor" },
	[SMCA_SMU]	= { "smu",		"System Management Unit" },
80
};
81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104
EXPORT_SYMBOL_GPL(smca_bank_names);

static struct smca_hwid_mcatype smca_hwid_mcatypes[] = {
	/* { bank_type, hwid_mcatype, xec_bitmap } */

	/* ZN Core (HWID=0xB0) MCA types */
	{ SMCA_LS,	 HWID_MCATYPE(0xB0, 0x0), 0x1FFFEF },
	{ SMCA_IF,	 HWID_MCATYPE(0xB0, 0x1), 0x3FFF },
	{ SMCA_L2_CACHE, HWID_MCATYPE(0xB0, 0x2), 0xF },
	{ SMCA_DE,	 HWID_MCATYPE(0xB0, 0x3), 0x1FF },
	/* HWID 0xB0 MCATYPE 0x4 is Reserved */
	{ SMCA_EX,	 HWID_MCATYPE(0xB0, 0x5), 0x7FF },
	{ SMCA_FP,	 HWID_MCATYPE(0xB0, 0x6), 0x7F },
	{ SMCA_L3_CACHE, HWID_MCATYPE(0xB0, 0x7), 0xFF },

	/* Data Fabric MCA types */
	{ SMCA_CS,	 HWID_MCATYPE(0x2E, 0x0), 0x1FF },
	{ SMCA_PIE,	 HWID_MCATYPE(0x2E, 0x1), 0xF },

	/* Unified Memory Controller MCA type */
	{ SMCA_UMC,	 HWID_MCATYPE(0x96, 0x0), 0x3F },

	/* Parameter Block MCA type */
	{ SMCA_PB,	 HWID_MCATYPE(0x05, 0x0), 0x1 },
105

106 107 108 109 110
	/* Platform Security Processor MCA type */
	{ SMCA_PSP,	 HWID_MCATYPE(0xFF, 0x0), 0x1 },

	/* System Management Unit MCA type */
	{ SMCA_SMU,	 HWID_MCATYPE(0x01, 0x0), 0x1 },
111
};
112 113 114

struct smca_bank_info smca_banks[MAX_NR_BANKS];
EXPORT_SYMBOL_GPL(smca_banks);
115

116
static DEFINE_PER_CPU(struct threshold_bank **, threshold_banks);
117
static DEFINE_PER_CPU(unsigned int, bank_map);	/* see which banks are on */
118

119
static void amd_threshold_interrupt(void);
120 121 122 123 124 125 126
static void amd_deferred_error_interrupt(void);

static void default_deferred_error_interrupt(void)
{
	pr_err("Unexpected deferred interrupt at vector %x\n", DEFERRED_ERROR_VECTOR);
}
void (*deferred_error_int_vector)(void) = default_deferred_error_interrupt;
127

128 129 130 131
/*
 * CPU Initialization
 */

132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161
static void get_smca_bank_info(unsigned int bank)
{
	unsigned int i, hwid_mcatype, cpu = smp_processor_id();
	struct smca_hwid_mcatype *type;
	u32 high, instanceId;
	u16 hwid, mcatype;

	/* Collect bank_info using CPU 0 for now. */
	if (cpu)
		return;

	if (rdmsr_safe_on_cpu(cpu, MSR_AMD64_SMCA_MCx_IPID(bank), &instanceId, &high)) {
		pr_warn("Failed to read MCA_IPID for bank %d\n", bank);
		return;
	}

	hwid = high & MCI_IPID_HWID;
	mcatype = (high & MCI_IPID_MCATYPE) >> 16;
	hwid_mcatype = HWID_MCATYPE(hwid, mcatype);

	for (i = 0; i < ARRAY_SIZE(smca_hwid_mcatypes); i++) {
		type = &smca_hwid_mcatypes[i];
		if (hwid_mcatype == type->hwid_mcatype) {
			smca_banks[bank].type = type;
			smca_banks[bank].type_instance = instanceId;
			break;
		}
	}
}

162
struct thresh_restart {
I
Ingo Molnar 已提交
163 164
	struct threshold_block	*b;
	int			reset;
165 166
	int			set_lvt_off;
	int			lvt_off;
I
Ingo Molnar 已提交
167
	u16			old_limit;
168 169
};

170 171
static inline bool is_shared_bank(int bank)
{
172 173 174 175 176 177 178
	/*
	 * Scalable MCA provides for only one core to have access to the MSRs of
	 * a shared bank.
	 */
	if (mce_flags.smca)
		return false;

179 180 181 182
	/* Bank 4 is for northbridge reporting and is thus shared */
	return (bank == 4);
}

183
static const char *bank4_names(const struct threshold_block *b)
184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202
{
	switch (b->address) {
	/* MSR4_MISC0 */
	case 0x00000413:
		return "dram";

	case 0xc0000408:
		return "ht_links";

	case 0xc0000409:
		return "l3_cache";

	default:
		WARN(1, "Funny MSR: 0x%08x\n", b->address);
		return "";
	}
};


203 204 205 206 207 208 209 210 211 212 213 214 215 216 217
static bool lvt_interrupt_supported(unsigned int bank, u32 msr_high_bits)
{
	/*
	 * bank 4 supports APIC LVT interrupts implicitly since forever.
	 */
	if (bank == 4)
		return true;

	/*
	 * IntP: interrupt present; if this bit is set, the thresholding
	 * bank can generate APIC LVT interrupts
	 */
	return msr_high_bits & BIT(28);
}

218 219 220 221 222 223 224 225 226 227 228 229
static int lvt_off_valid(struct threshold_block *b, int apic, u32 lo, u32 hi)
{
	int msr = (hi & MASK_LVTOFF_HI) >> 20;

	if (apic < 0) {
		pr_err(FW_BUG "cpu %d, failed to setup threshold interrupt "
		       "for bank %d, block %d (MSR%08X=0x%x%08x)\n", b->cpu,
		       b->bank, b->block, b->address, hi, lo);
		return 0;
	}

	if (apic != msr) {
230 231 232 233 234 235 236 237
		/*
		 * On SMCA CPUs, LVT offset is programmed at a different MSR, and
		 * the BIOS provides the value. The original field where LVT offset
		 * was set is reserved. Return early here:
		 */
		if (mce_flags.smca)
			return 0;

238 239 240 241 242 243 244 245 246
		pr_err(FW_BUG "cpu %d, invalid threshold interrupt offset %d "
		       "for bank %d, block %d (MSR%08X=0x%x%08x)\n",
		       b->cpu, apic, b->bank, b->block, b->address, hi, lo);
		return 0;
	}

	return 1;
};

247
/* Reprogram MCx_MISC MSR behind this threshold bank. */
248
static void threshold_restart_bank(void *_tr)
249
{
250
	struct thresh_restart *tr = _tr;
251
	u32 hi, lo;
252

253
	rdmsr(tr->b->address, lo, hi);
254

255
	if (tr->b->threshold_limit < (hi & THRESHOLD_MAX))
256
		tr->reset = 1;	/* limit cannot be lower than err count */
257

258
	if (tr->reset) {		/* reset err count and overflow bit */
259 260
		hi =
		    (hi & ~(MASK_ERR_COUNT_HI | MASK_OVERFLOW_HI)) |
261 262
		    (THRESHOLD_MAX - tr->b->threshold_limit);
	} else if (tr->old_limit) {	/* change limit w/o reset */
263
		int new_count = (hi & THRESHOLD_MAX) +
264
		    (tr->old_limit - tr->b->threshold_limit);
I
Ingo Molnar 已提交
265

266
		hi = (hi & ~MASK_ERR_COUNT_HI) |
267 268 269
		    (new_count & THRESHOLD_MAX);
	}

270 271 272 273 274 275
	/* clear IntType */
	hi &= ~MASK_INT_TYPE_HI;

	if (!tr->b->interrupt_capable)
		goto done;

276
	if (tr->set_lvt_off) {
277 278 279 280 281
		if (lvt_off_valid(tr->b, tr->lvt_off, lo, hi)) {
			/* set new lvt offset */
			hi &= ~MASK_LVTOFF_HI;
			hi |= tr->lvt_off << 20;
		}
282 283
	}

284 285 286 287
	if (tr->b->interrupt_enable)
		hi |= INT_TYPE_APIC;

 done:
288

289 290
	hi |= MASK_COUNT_EN_HI;
	wrmsr(tr->b->address, lo, hi);
291 292
}

293 294 295 296 297 298 299 300 301 302 303 304
static void mce_threshold_block_init(struct threshold_block *b, int offset)
{
	struct thresh_restart tr = {
		.b			= b,
		.set_lvt_off		= 1,
		.lvt_off		= offset,
	};

	b->threshold_limit		= THRESHOLD_MAX;
	threshold_restart_bank(&tr);
};

305
static int setup_APIC_mce_threshold(int reserved, int new)
306 307 308 309 310 311 312 313
{
	if (reserved < 0 && !setup_APIC_eilvt(new, THRESHOLD_APIC_VECTOR,
					      APIC_EILVT_MSG_FIX, 0))
		return new;

	return reserved;
}

314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346
static int setup_APIC_deferred_error(int reserved, int new)
{
	if (reserved < 0 && !setup_APIC_eilvt(new, DEFERRED_ERROR_VECTOR,
					      APIC_EILVT_MSG_FIX, 0))
		return new;

	return reserved;
}

static void deferred_error_interrupt_enable(struct cpuinfo_x86 *c)
{
	u32 low = 0, high = 0;
	int def_offset = -1, def_new;

	if (rdmsr_safe(MSR_CU_DEF_ERR, &low, &high))
		return;

	def_new = (low & MASK_DEF_LVTOFF) >> 4;
	if (!(low & MASK_DEF_LVTOFF)) {
		pr_err(FW_BUG "Your BIOS is not setting up LVT offset 0x2 for deferred error IRQs correctly.\n");
		def_new = DEF_LVT_OFF;
		low = (low & ~MASK_DEF_LVTOFF) | (DEF_LVT_OFF << 4);
	}

	def_offset = setup_APIC_deferred_error(def_offset, def_new);
	if ((def_offset == def_new) &&
	    (deferred_error_int_vector != amd_deferred_error_interrupt))
		deferred_error_int_vector = amd_deferred_error_interrupt;

	low = (low & ~MASK_DEF_INT_TYPE) | DEF_INT_TYPE_APIC;
	wrmsr(MSR_CU_DEF_ERR, low, high);
}

347
static u32 get_block_address(unsigned int cpu, u32 current_addr, u32 low, u32 high,
348 349 350 351 352 353 354 355 356 357 358 359 360 361 362
			     unsigned int bank, unsigned int block)
{
	u32 addr = 0, offset = 0;

	if (mce_flags.smca) {
		if (!block) {
			addr = MSR_AMD64_SMCA_MCx_MISC(bank);
		} else {
			/*
			 * For SMCA enabled processors, BLKPTR field of the
			 * first MISC register (MCx_MISC0) indicates presence of
			 * additional MISC register set (MISC1-4).
			 */
			u32 low, high;

363
			if (rdmsr_safe_on_cpu(cpu, MSR_AMD64_SMCA_MCx_CONFIG(bank), &low, &high))
364 365 366 367 368
				return addr;

			if (!(low & MCI_CONFIG_MCAX))
				return addr;

369
			if (!rdmsr_safe_on_cpu(cpu, MSR_AMD64_SMCA_MCx_MISC(bank), &low, &high) &&
370 371 372 373 374 375 376 377 378
			    (low & MASK_BLKPTR_LO))
				addr = MSR_AMD64_SMCA_MCx_MISCy(bank, block - 1);
		}
		return addr;
	}

	/* Fall back to method we used for older processors: */
	switch (block) {
	case 0:
379
		addr = msr_ops.misc(bank);
380 381 382 383 384 385 386 387 388 389 390 391
		break;
	case 1:
		offset = ((low & MASK_BLKPTR_LO) >> 21);
		if (offset)
			addr = MCG_XBLK_ADDR + offset;
		break;
	default:
		addr = ++current_addr;
	}
	return addr;
}

392 393 394 395 396
static int
prepare_threshold_block(unsigned int bank, unsigned int block, u32 addr,
			int offset, u32 misc_high)
{
	unsigned int cpu = smp_processor_id();
397
	u32 smca_low, smca_high, smca_addr;
398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415
	struct threshold_block b;
	int new;

	if (!block)
		per_cpu(bank_map, cpu) |= (1 << bank);

	memset(&b, 0, sizeof(b));
	b.cpu			= cpu;
	b.bank			= bank;
	b.block			= block;
	b.address		= addr;
	b.interrupt_capable	= lvt_interrupt_supported(bank, misc_high);

	if (!b.interrupt_capable)
		goto done;

	b.interrupt_enable = 1;

416 417 418 419
	if (!mce_flags.smca) {
		new = (misc_high & MASK_LVTOFF_HI) >> 20;
		goto set_offset;
	}
420

421
	smca_addr = MSR_AMD64_SMCA_MCx_CONFIG(bank);
422

423 424 425 426 427 428 429 430 431 432 433
	if (!rdmsr_safe(smca_addr, &smca_low, &smca_high)) {
		/*
		 * OS is required to set the MCAX bit to acknowledge that it is
		 * now using the new MSR ranges and new registers under each
		 * bank. It also means that the OS will configure deferred
		 * errors in the new MCx_CONFIG register. If the bit is not set,
		 * uncorrectable errors will cause a system panic.
		 *
		 * MCA_CONFIG[MCAX] is bit 32 (0 in the high portion of the MSR.)
		 */
		smca_high |= BIT(0);
434

435 436 437 438 439 440 441 442 443 444 445 446 447
		/*
		 * SMCA logs Deferred Error information in MCA_DE{STAT,ADDR}
		 * registers with the option of additionally logging to
		 * MCA_{STATUS,ADDR} if MCA_CONFIG[LogDeferredInMcaStat] is set.
		 *
		 * This bit is usually set by BIOS to retain the old behavior
		 * for OSes that don't use the new registers. Linux supports the
		 * new registers so let's disable that additional logging here.
		 *
		 * MCA_CONFIG[LogDeferredInMcaStat] is bit 34 (bit 2 in the high
		 * portion of the MSR).
		 */
		smca_high &= ~BIT(2);
448

449
		wrmsr(smca_addr, smca_low, smca_high);
450 451
	}

452 453 454 455 456 457 458
	/* Gather LVT offset for thresholding: */
	if (rdmsr_safe(MSR_CU_DEF_ERR, &smca_low, &smca_high))
		goto out;

	new = (smca_low & SMCA_THR_LVT_OFF) >> 12;

set_offset:
459 460 461 462 463 464 465 466 467 468 469 470
	offset = setup_APIC_mce_threshold(offset, new);

	if ((offset == new) && (mce_threshold_vector != amd_threshold_interrupt))
		mce_threshold_vector = amd_threshold_interrupt;

done:
	mce_threshold_block_init(&b, offset);

out:
	return offset;
}

471
/* cpu init entry point, called from mce.c with preempt off */
472
void mce_amd_feature_init(struct cpuinfo_x86 *c)
473
{
474
	u32 low = 0, high = 0, address = 0;
475
	unsigned int bank, block, cpu = smp_processor_id();
476
	int offset = -1;
477

478
	for (bank = 0; bank < mca_cfg.banks; ++bank) {
479 480 481
		if (mce_flags.smca)
			get_smca_bank_info(bank);

482
		for (block = 0; block < NR_BLOCKS; ++block) {
483
			address = get_block_address(cpu, address, low, high, bank, block);
484 485
			if (!address)
				break;
486 487

			if (rdmsr_safe(address, &low, &high))
488
				break;
489

490 491
			if (!(high & MASK_VALID_HI))
				continue;
492

493 494
			if (!(high & MASK_CNTP_HI)  ||
			     (high & MASK_LOCKED_HI))
495 496
				continue;

497
			offset = prepare_threshold_block(bank, block, address, offset, high);
498
		}
499
	}
500 501 502

	if (mce_flags.succor)
		deferred_error_interrupt_enable(c);
503 504
}

505 506
static void
__log_error(unsigned int bank, bool deferred_err, bool threshold_err, u64 misc)
507
{
508 509
	u32 msr_status = msr_ops.status(bank);
	u32 msr_addr = msr_ops.addr(bank);
510 511 512
	struct mce m;
	u64 status;

513 514 515 516 517 518 519 520 521
	WARN_ON_ONCE(deferred_err && threshold_err);

	if (deferred_err && mce_flags.smca) {
		msr_status = MSR_AMD64_SMCA_MCx_DESTAT(bank);
		msr_addr = MSR_AMD64_SMCA_MCx_DEADDR(bank);
	}

	rdmsrl(msr_status, status);

522 523 524 525 526 527 528
	if (!(status & MCI_STATUS_VAL))
		return;

	mce_setup(&m);

	m.status = status;
	m.bank = bank;
529

530 531 532
	if (threshold_err)
		m.misc = misc;

533
	if (m.status & MCI_STATUS_ADDRV)
534
		rdmsrl(msr_addr, m.addr);
535

536 537 538
	if (mce_flags.smca && (m.status & MCI_STATUS_SYNDV))
		rdmsrl(MSR_AMD64_SMCA_MCx_SYND(bank), m.synd);

539
	mce_log(&m);
540 541

	wrmsrl(msr_status, 0);
542 543
}

544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569
static inline void __smp_deferred_error_interrupt(void)
{
	inc_irq_stat(irq_deferred_error_count);
	deferred_error_int_vector();
}

asmlinkage __visible void smp_deferred_error_interrupt(void)
{
	entering_irq();
	__smp_deferred_error_interrupt();
	exiting_ack_irq();
}

asmlinkage __visible void smp_trace_deferred_error_interrupt(void)
{
	entering_irq();
	trace_deferred_error_apic_entry(DEFERRED_ERROR_VECTOR);
	__smp_deferred_error_interrupt();
	trace_deferred_error_apic_exit(DEFERRED_ERROR_VECTOR);
	exiting_ack_irq();
}

/* APIC interrupt handler for deferred errors */
static void amd_deferred_error_interrupt(void)
{
	unsigned int bank;
570 571
	u32 msr_status;
	u64 status;
572 573

	for (bank = 0; bank < mca_cfg.banks; ++bank) {
574 575 576 577
		msr_status = (mce_flags.smca) ? MSR_AMD64_SMCA_MCx_DESTAT(bank)
					      : msr_ops.status(bank);

		rdmsrl(msr_status, status);
578 579 580 581 582

		if (!(status & MCI_STATUS_VAL) ||
		    !(status & MCI_STATUS_DEFERRED))
			continue;

583
		__log_error(bank, true, false, 0);
584 585 586 587
		break;
	}
}

588 589 590 591 592 593 594 595 596
/*
 * APIC Interrupt Handler
 */

/*
 * threshold interrupt handler will service THRESHOLD_APIC_VECTOR.
 * the interrupt goes off when error_count reaches threshold_limit.
 * the handler will simply log mcelog w/ software defined bank number.
 */
597

598
static void amd_threshold_interrupt(void)
599
{
I
Ingo Molnar 已提交
600
	u32 low = 0, high = 0, address = 0;
601
	unsigned int bank, block, cpu = smp_processor_id();
602 603

	/* assume first bank caused it */
604
	for (bank = 0; bank < mca_cfg.banks; ++bank) {
605
		if (!(per_cpu(bank_map, cpu) & (1 << bank)))
606
			continue;
607
		for (block = 0; block < NR_BLOCKS; ++block) {
608
			address = get_block_address(cpu, address, low, high, bank, block);
609 610
			if (!address)
				break;
611 612

			if (rdmsr_safe(address, &low, &high))
613
				break;
614 615 616 617 618 619 620 621

			if (!(high & MASK_VALID_HI)) {
				if (block)
					continue;
				else
					break;
			}

622 623
			if (!(high & MASK_CNTP_HI)  ||
			     (high & MASK_LOCKED_HI))
624 625
				continue;

I
Ingo Molnar 已提交
626 627 628 629
			/*
			 * Log the machine check that caused the threshold
			 * event.
			 */
630 631
			if (high & MASK_OVERFLOW_HI)
				goto log;
632 633
		}
	}
634 635 636
	return;

log:
637
	__log_error(bank, false, true, ((u64)high << 32) | low);
638 639 640 641 642 643 644
}

/*
 * Sysfs Interface
 */

struct threshold_attr {
J
Jacob Shin 已提交
645
	struct attribute attr;
I
Ingo Molnar 已提交
646 647
	ssize_t (*show) (struct threshold_block *, char *);
	ssize_t (*store) (struct threshold_block *, const char *, size_t count);
648 649
};

I
Ingo Molnar 已提交
650 651 652
#define SHOW_FIELDS(name)						\
static ssize_t show_ ## name(struct threshold_block *b, char *buf)	\
{									\
653
	return sprintf(buf, "%lu\n", (unsigned long) b->name);		\
J
Jacob Shin 已提交
654
}
655 656 657
SHOW_FIELDS(interrupt_enable)
SHOW_FIELDS(threshold_limit)

I
Ingo Molnar 已提交
658
static ssize_t
H
Hidetoshi Seto 已提交
659
store_interrupt_enable(struct threshold_block *b, const char *buf, size_t size)
660
{
661
	struct thresh_restart tr;
I
Ingo Molnar 已提交
662 663
	unsigned long new;

664 665 666
	if (!b->interrupt_capable)
		return -EINVAL;

667
	if (kstrtoul(buf, 0, &new) < 0)
668
		return -EINVAL;
I
Ingo Molnar 已提交
669

670 671
	b->interrupt_enable = !!new;

672
	memset(&tr, 0, sizeof(tr));
I
Ingo Molnar 已提交
673 674
	tr.b		= b;

675
	smp_call_function_single(b->cpu, threshold_restart_bank, &tr, 1);
676

H
Hidetoshi Seto 已提交
677
	return size;
678 679
}

I
Ingo Molnar 已提交
680
static ssize_t
H
Hidetoshi Seto 已提交
681
store_threshold_limit(struct threshold_block *b, const char *buf, size_t size)
682
{
683
	struct thresh_restart tr;
I
Ingo Molnar 已提交
684 685
	unsigned long new;

686
	if (kstrtoul(buf, 0, &new) < 0)
687
		return -EINVAL;
I
Ingo Molnar 已提交
688

689 690 691 692
	if (new > THRESHOLD_MAX)
		new = THRESHOLD_MAX;
	if (new < 1)
		new = 1;
I
Ingo Molnar 已提交
693

694
	memset(&tr, 0, sizeof(tr));
695
	tr.old_limit = b->threshold_limit;
696
	b->threshold_limit = new;
697
	tr.b = b;
698

699
	smp_call_function_single(b->cpu, threshold_restart_bank, &tr, 1);
700

H
Hidetoshi Seto 已提交
701
	return size;
702 703
}

704 705
static ssize_t show_error_count(struct threshold_block *b, char *buf)
{
706 707 708
	u32 lo, hi;

	rdmsr_on_cpu(b->cpu, b->address, &lo, &hi);
709

710 711
	return sprintf(buf, "%u\n", ((hi & THRESHOLD_MAX) -
				     (THRESHOLD_MAX - b->threshold_limit)));
712 713
}

714 715 716 717
static struct threshold_attr error_count = {
	.attr = {.name = __stringify(error_count), .mode = 0444 },
	.show = show_error_count,
};
718

719 720 721 722 723
#define RW_ATTR(val)							\
static struct threshold_attr val = {					\
	.attr	= {.name = __stringify(val), .mode = 0644 },		\
	.show	= show_## val,						\
	.store	= store_## val,						\
724 725
};

J
Jacob Shin 已提交
726 727
RW_ATTR(interrupt_enable);
RW_ATTR(threshold_limit);
728 729 730 731

static struct attribute *default_attrs[] = {
	&threshold_limit.attr,
	&error_count.attr,
732 733
	NULL,	/* possibly interrupt_enable if supported, see below */
	NULL,
734 735
};

I
Ingo Molnar 已提交
736 737
#define to_block(k)	container_of(k, struct threshold_block, kobj)
#define to_attr(a)	container_of(a, struct threshold_attr, attr)
738 739 740

static ssize_t show(struct kobject *kobj, struct attribute *attr, char *buf)
{
741
	struct threshold_block *b = to_block(kobj);
742 743
	struct threshold_attr *a = to_attr(attr);
	ssize_t ret;
I
Ingo Molnar 已提交
744

745
	ret = a->show ? a->show(b, buf) : -EIO;
I
Ingo Molnar 已提交
746

747 748 749 750 751 752
	return ret;
}

static ssize_t store(struct kobject *kobj, struct attribute *attr,
		     const char *buf, size_t count)
{
753
	struct threshold_block *b = to_block(kobj);
754 755
	struct threshold_attr *a = to_attr(attr);
	ssize_t ret;
I
Ingo Molnar 已提交
756

757
	ret = a->store ? a->store(b, buf, count) : -EIO;
I
Ingo Molnar 已提交
758

759 760 761
	return ret;
}

762
static const struct sysfs_ops threshold_ops = {
I
Ingo Molnar 已提交
763 764
	.show			= show,
	.store			= store,
765 766 767
};

static struct kobj_type threshold_ktype = {
I
Ingo Molnar 已提交
768 769
	.sysfs_ops		= &threshold_ops,
	.default_attrs		= default_attrs,
770 771
};

772 773
static int allocate_threshold_blocks(unsigned int cpu, unsigned int bank,
				     unsigned int block, u32 address)
774 775
{
	struct threshold_block *b = NULL;
I
Ingo Molnar 已提交
776 777
	u32 low, high;
	int err;
778

779
	if ((bank >= mca_cfg.banks) || (block >= NR_BLOCKS))
780 781
		return 0;

782
	if (rdmsr_safe_on_cpu(cpu, address, &low, &high))
783
		return 0;
784 785 786 787 788 789 790 791

	if (!(high & MASK_VALID_HI)) {
		if (block)
			goto recurse;
		else
			return 0;
	}

792 793
	if (!(high & MASK_CNTP_HI)  ||
	     (high & MASK_LOCKED_HI))
794 795 796 797 798 799
		goto recurse;

	b = kzalloc(sizeof(struct threshold_block), GFP_KERNEL);
	if (!b)
		return -ENOMEM;

I
Ingo Molnar 已提交
800 801 802 803 804
	b->block		= block;
	b->bank			= bank;
	b->cpu			= cpu;
	b->address		= address;
	b->interrupt_enable	= 0;
805
	b->interrupt_capable	= lvt_interrupt_supported(bank, high);
I
Ingo Molnar 已提交
806
	b->threshold_limit	= THRESHOLD_MAX;
807

808
	if (b->interrupt_capable) {
809
		threshold_ktype.default_attrs[2] = &interrupt_enable.attr;
810 811
		b->interrupt_enable = 1;
	} else {
812
		threshold_ktype.default_attrs[2] = NULL;
813
	}
814

815 816
	INIT_LIST_HEAD(&b->miscj);

I
Ingo Molnar 已提交
817
	if (per_cpu(threshold_banks, cpu)[bank]->blocks) {
818 819
		list_add(&b->miscj,
			 &per_cpu(threshold_banks, cpu)[bank]->blocks->miscj);
I
Ingo Molnar 已提交
820
	} else {
821
		per_cpu(threshold_banks, cpu)[bank]->blocks = b;
I
Ingo Molnar 已提交
822
	}
823

824 825
	err = kobject_init_and_add(&b->kobj, &threshold_ktype,
				   per_cpu(threshold_banks, cpu)[bank]->kobj,
826
				   (bank == 4 ? bank4_names(b) : th_names[bank]));
827 828 829
	if (err)
		goto out_free;
recurse:
830
	address = get_block_address(cpu, address, low, high, bank, ++block);
831 832
	if (!address)
		return 0;
833

834
	err = allocate_threshold_blocks(cpu, bank, block, address);
835 836 837
	if (err)
		goto out_free;

838 839
	if (b)
		kobject_uevent(&b->kobj, KOBJ_ADD);
840

841 842 843 844
	return err;

out_free:
	if (b) {
845
		kobject_put(&b->kobj);
846
		list_del(&b->miscj);
847 848 849 850 851
		kfree(b);
	}
	return err;
}

852
static int __threshold_add_blocks(struct threshold_bank *b)
853 854 855 856 857 858 859 860 861 862 863 864 865 866 867 868 869 870 871 872 873 874 875
{
	struct list_head *head = &b->blocks->miscj;
	struct threshold_block *pos = NULL;
	struct threshold_block *tmp = NULL;
	int err = 0;

	err = kobject_add(&b->blocks->kobj, b->kobj, b->blocks->kobj.name);
	if (err)
		return err;

	list_for_each_entry_safe(pos, tmp, head, miscj) {

		err = kobject_add(&pos->kobj, b->kobj, pos->kobj.name);
		if (err) {
			list_for_each_entry_safe_reverse(pos, tmp, head, miscj)
				kobject_del(&pos->kobj);

			return err;
		}
	}
	return err;
}

876
static int threshold_create_bank(unsigned int cpu, unsigned int bank)
877
{
878
	struct device *dev = per_cpu(mce_device, cpu);
879
	struct amd_northbridge *nb = NULL;
880
	struct threshold_bank *b = NULL;
881
	const char *name = th_names[bank];
882
	int err = 0;
883

884
	if (is_shared_bank(bank)) {
885 886 887
		nb = node_to_amd_nb(amd_get_nb_id(cpu));

		/* threshold descriptor already initialized on this node? */
888
		if (nb && nb->bank4) {
889 890 891 892 893 894 895 896 897 898 899 900 901 902 903
			/* yes, use it */
			b = nb->bank4;
			err = kobject_add(b->kobj, &dev->kobj, name);
			if (err)
				goto out;

			per_cpu(threshold_banks, cpu)[bank] = b;
			atomic_inc(&b->cpus);

			err = __threshold_add_blocks(b);

			goto out;
		}
	}

904
	b = kzalloc(sizeof(struct threshold_bank), GFP_KERNEL);
905 906 907 908 909
	if (!b) {
		err = -ENOMEM;
		goto out;
	}

910
	b->kobj = kobject_create_and_add(name, &dev->kobj);
911 912
	if (!b->kobj) {
		err = -EINVAL;
913
		goto out_free;
914
	}
915

916
	per_cpu(threshold_banks, cpu)[bank] = b;
917

918
	if (is_shared_bank(bank)) {
919 920 921
		atomic_set(&b->cpus, 1);

		/* nb is already initialized, see above */
922 923 924 925
		if (nb) {
			WARN_ON(nb->bank4);
			nb->bank4 = b;
		}
926 927
	}

928
	err = allocate_threshold_blocks(cpu, bank, 0, msr_ops.misc(bank));
929 930
	if (!err)
		goto out;
931

932
 out_free:
933
	kfree(b);
934 935

 out:
936 937 938 939
	return err;
}

/* create dir/files for all valid threshold banks */
940
static int threshold_create_device(unsigned int cpu)
941
{
J
Jacob Shin 已提交
942
	unsigned int bank;
943
	struct threshold_bank **bp;
944 945
	int err = 0;

946 947 948 949 950 951 952 953
	bp = kzalloc(sizeof(struct threshold_bank *) * mca_cfg.banks,
		     GFP_KERNEL);
	if (!bp)
		return -ENOMEM;

	per_cpu(threshold_banks, cpu) = bp;

	for (bank = 0; bank < mca_cfg.banks; ++bank) {
954
		if (!(per_cpu(bank_map, cpu) & (1 << bank)))
955 956 957
			continue;
		err = threshold_create_bank(cpu, bank);
		if (err)
958
			return err;
959
	}
960

961 962 963
	return err;
}

964
static void deallocate_threshold_block(unsigned int cpu,
965 966 967 968 969 970 971 972 973 974
						 unsigned int bank)
{
	struct threshold_block *pos = NULL;
	struct threshold_block *tmp = NULL;
	struct threshold_bank *head = per_cpu(threshold_banks, cpu)[bank];

	if (!head)
		return;

	list_for_each_entry_safe(pos, tmp, &head->blocks->miscj, miscj) {
975
		kobject_put(&pos->kobj);
976 977 978 979 980 981 982 983
		list_del(&pos->miscj);
		kfree(pos);
	}

	kfree(per_cpu(threshold_banks, cpu)[bank]->blocks);
	per_cpu(threshold_banks, cpu)[bank]->blocks = NULL;
}

984 985 986 987 988 989 990 991 992 993 994
static void __threshold_remove_blocks(struct threshold_bank *b)
{
	struct threshold_block *pos = NULL;
	struct threshold_block *tmp = NULL;

	kobject_del(b->kobj);

	list_for_each_entry_safe(pos, tmp, &b->blocks->miscj, miscj)
		kobject_del(&pos->kobj);
}

995
static void threshold_remove_bank(unsigned int cpu, int bank)
996
{
997
	struct amd_northbridge *nb;
998 999 1000 1001 1002
	struct threshold_bank *b;

	b = per_cpu(threshold_banks, cpu)[bank];
	if (!b)
		return;
1003

1004 1005 1006
	if (!b->blocks)
		goto free_out;

1007
	if (is_shared_bank(bank)) {
1008 1009 1010 1011 1012 1013 1014 1015 1016 1017 1018 1019 1020 1021
		if (!atomic_dec_and_test(&b->cpus)) {
			__threshold_remove_blocks(b);
			per_cpu(threshold_banks, cpu)[bank] = NULL;
			return;
		} else {
			/*
			 * the last CPU on this node using the shared bank is
			 * going away, remove that bank now.
			 */
			nb = node_to_amd_nb(amd_get_nb_id(cpu));
			nb->bank4 = NULL;
		}
	}

1022 1023 1024
	deallocate_threshold_block(cpu, bank);

free_out:
1025
	kobject_del(b->kobj);
1026
	kobject_put(b->kobj);
1027 1028
	kfree(b);
	per_cpu(threshold_banks, cpu)[bank] = NULL;
1029 1030
}

1031
static void threshold_remove_device(unsigned int cpu)
1032
{
J
Jacob Shin 已提交
1033
	unsigned int bank;
1034

1035
	for (bank = 0; bank < mca_cfg.banks; ++bank) {
1036
		if (!(per_cpu(bank_map, cpu) & (1 << bank)))
1037 1038 1039
			continue;
		threshold_remove_bank(cpu, bank);
	}
1040
	kfree(per_cpu(threshold_banks, cpu));
1041 1042 1043
}

/* get notified when a cpu comes on/off */
1044
static void
I
Ingo Molnar 已提交
1045
amd_64_threshold_cpu_callback(unsigned long action, unsigned int cpu)
1046 1047 1048
{
	switch (action) {
	case CPU_ONLINE:
1049
	case CPU_ONLINE_FROZEN:
1050 1051 1052
		threshold_create_device(cpu);
		break;
	case CPU_DEAD:
1053
	case CPU_DEAD_FROZEN:
1054 1055 1056 1057 1058 1059 1060 1061 1062
		threshold_remove_device(cpu);
		break;
	default:
		break;
	}
}

static __init int threshold_init_device(void)
{
J
Jacob Shin 已提交
1063
	unsigned lcpu = 0;
1064 1065 1066

	/* to hit CPUs online before the notifier is up */
	for_each_online_cpu(lcpu) {
1067
		int err = threshold_create_device(lcpu);
I
Ingo Molnar 已提交
1068

1069
		if (err)
1070
			return err;
1071
	}
1072
	threshold_cpu_callback = amd_64_threshold_cpu_callback;
I
Ingo Molnar 已提交
1073

1074
	return 0;
1075
}
1076 1077 1078 1079 1080 1081 1082 1083 1084 1085 1086 1087 1088 1089 1090 1091 1092 1093 1094 1095 1096
/*
 * there are 3 funcs which need to be _initcalled in a logic sequence:
 * 1. xen_late_init_mcelog
 * 2. mcheck_init_device
 * 3. threshold_init_device
 *
 * xen_late_init_mcelog must register xen_mce_chrdev_device before
 * native mce_chrdev_device registration if running under xen platform;
 *
 * mcheck_init_device should be inited before threshold_init_device to
 * initialize mce_device, otherwise a NULL ptr dereference will cause panic.
 *
 * so we use following _initcalls
 * 1. device_initcall(xen_late_init_mcelog);
 * 2. device_initcall_sync(mcheck_init_device);
 * 3. late_initcall(threshold_init_device);
 *
 * when running under xen, the initcall order is 1,2,3;
 * on baremetal, we skip 1 and we do only 2 and 3.
 */
late_initcall(threshold_init_device);