mce_amd.c 22.5 KB
Newer Older
1
/*
B
Borislav Petkov 已提交
2
 *  (c) 2005-2015 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
/*
 * 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.
 */
#define SMCA_MCAX_EN_OFF	0x1

65 66 67 68 69 70 71 72 73
static const char * const th_names[] = {
	"load_store",
	"insn_fetch",
	"combined_unit",
	"",
	"northbridge",
	"execution_unit",
};

74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102
/* Define HWID to IP type mappings for Scalable MCA */
struct amd_hwid amd_hwids[] = {
	[SMCA_F17H_CORE]	= { "f17h_core",	0xB0 },
	[SMCA_DF]		= { "data_fabric",	0x2E },
	[SMCA_UMC]		= { "umc",		0x96 },
	[SMCA_PB]		= { "param_block",	0x5 },
	[SMCA_PSP]		= { "psp",		0xFF },
	[SMCA_SMU]		= { "smu",		0x1 },
};
EXPORT_SYMBOL_GPL(amd_hwids);

const char * const amd_core_mcablock_names[] = {
	[SMCA_LS]		= "load_store",
	[SMCA_IF]		= "insn_fetch",
	[SMCA_L2_CACHE]		= "l2_cache",
	[SMCA_DE]		= "decode_unit",
	[RES]			= "",
	[SMCA_EX]		= "execution_unit",
	[SMCA_FP]		= "floating_point",
	[SMCA_L3_CACHE]		= "l3_cache",
};
EXPORT_SYMBOL_GPL(amd_core_mcablock_names);

const char * const amd_df_mcablock_names[] = {
	[SMCA_CS]		= "coherent_slave",
	[SMCA_PIE]		= "pie",
};
EXPORT_SYMBOL_GPL(amd_df_mcablock_names);

103
static DEFINE_PER_CPU(struct threshold_bank **, threshold_banks);
104 105
static DEFINE_PER_CPU(unsigned char, bank_map);	/* see which banks are on */

106
static void amd_threshold_interrupt(void);
107 108 109 110 111 112 113
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;
114

115 116 117 118
/*
 * CPU Initialization
 */

119
struct thresh_restart {
I
Ingo Molnar 已提交
120 121
	struct threshold_block	*b;
	int			reset;
122 123
	int			set_lvt_off;
	int			lvt_off;
I
Ingo Molnar 已提交
124
	u16			old_limit;
125 126
};

127 128
static inline bool is_shared_bank(int bank)
{
129 130 131 132 133 134 135
	/*
	 * Scalable MCA provides for only one core to have access to the MSRs of
	 * a shared bank.
	 */
	if (mce_flags.smca)
		return false;

136 137 138 139
	/* Bank 4 is for northbridge reporting and is thus shared */
	return (bank == 4);
}

140
static const char *bank4_names(const struct threshold_block *b)
141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159
{
	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 "";
	}
};


160 161 162 163 164 165 166 167 168 169 170 171 172 173 174
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);
}

175 176 177 178 179 180 181 182 183 184 185 186
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) {
187 188 189 190 191 192 193 194
		/*
		 * 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;

195 196 197 198 199 200 201 202 203
		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;
};

204 205 206 207
/*
 * Called via smp_call_function_single(), must be called with correct
 * cpu affinity.
 */
208
static void threshold_restart_bank(void *_tr)
209
{
210
	struct thresh_restart *tr = _tr;
211
	u32 hi, lo;
212

213
	rdmsr(tr->b->address, lo, hi);
214

215
	if (tr->b->threshold_limit < (hi & THRESHOLD_MAX))
216
		tr->reset = 1;	/* limit cannot be lower than err count */
217

218
	if (tr->reset) {		/* reset err count and overflow bit */
219 220
		hi =
		    (hi & ~(MASK_ERR_COUNT_HI | MASK_OVERFLOW_HI)) |
221 222
		    (THRESHOLD_MAX - tr->b->threshold_limit);
	} else if (tr->old_limit) {	/* change limit w/o reset */
223
		int new_count = (hi & THRESHOLD_MAX) +
224
		    (tr->old_limit - tr->b->threshold_limit);
I
Ingo Molnar 已提交
225

226
		hi = (hi & ~MASK_ERR_COUNT_HI) |
227 228 229
		    (new_count & THRESHOLD_MAX);
	}

230 231 232 233 234 235
	/* clear IntType */
	hi &= ~MASK_INT_TYPE_HI;

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

236
	if (tr->set_lvt_off) {
237 238 239 240 241
		if (lvt_off_valid(tr->b, tr->lvt_off, lo, hi)) {
			/* set new lvt offset */
			hi &= ~MASK_LVTOFF_HI;
			hi |= tr->lvt_off << 20;
		}
242 243
	}

244 245 246 247
	if (tr->b->interrupt_enable)
		hi |= INT_TYPE_APIC;

 done:
248

249 250
	hi |= MASK_COUNT_EN_HI;
	wrmsr(tr->b->address, lo, hi);
251 252
}

253 254 255 256 257 258 259 260 261 262 263 264
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);
};

265
static int setup_APIC_mce_threshold(int reserved, int new)
266 267 268 269 270 271 272 273
{
	if (reserved < 0 && !setup_APIC_eilvt(new, THRESHOLD_APIC_VECTOR,
					      APIC_EILVT_MSG_FIX, 0))
		return new;

	return reserved;
}

274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306
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);
}

307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351
static u32 get_block_address(u32 current_addr, u32 low, u32 high,
			     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;

			if (rdmsr_safe(MSR_AMD64_SMCA_MCx_CONFIG(bank), &low, &high))
				return addr;

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

			if (!rdmsr_safe(MSR_AMD64_SMCA_MCx_MISC(bank), &low, &high) &&
			    (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:
		addr = MSR_IA32_MCx_MISC(bank);
		break;
	case 1:
		offset = ((low & MASK_BLKPTR_LO) >> 21);
		if (offset)
			addr = MCG_XBLK_ADDR + offset;
		break;
	default:
		addr = ++current_addr;
	}
	return addr;
}

352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376
static int
prepare_threshold_block(unsigned int bank, unsigned int block, u32 addr,
			int offset, u32 misc_high)
{
	unsigned int cpu = smp_processor_id();
	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;

	if (mce_flags.smca) {
		u32 smca_low, smca_high;
377 378 379 380 381 382
		u32 smca_addr = MSR_AMD64_SMCA_MCx_CONFIG(bank);

		if (!rdmsr_safe(smca_addr, &smca_low, &smca_high)) {
			smca_high |= SMCA_MCAX_EN_OFF;
			wrmsr(smca_addr, smca_low, smca_high);
		}
383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404

		/* 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;
	} else {
		new = (misc_high & MASK_LVTOFF_HI) >> 20;
	}

	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;
}

405
/* cpu init entry point, called from mce.c with preempt off */
406
void mce_amd_feature_init(struct cpuinfo_x86 *c)
407
{
408
	u32 low = 0, high = 0, address = 0;
I
Ingo Molnar 已提交
409
	unsigned int bank, block;
410
	int offset = -1;
411

412
	for (bank = 0; bank < mca_cfg.banks; ++bank) {
413
		for (block = 0; block < NR_BLOCKS; ++block) {
414 415 416
			address = get_block_address(address, low, high, bank, block);
			if (!address)
				break;
417 418

			if (rdmsr_safe(address, &low, &high))
419
				break;
420

421 422
			if (!(high & MASK_VALID_HI))
				continue;
423

424 425
			if (!(high & MASK_CNTP_HI)  ||
			     (high & MASK_LOCKED_HI))
426 427
				continue;

428
			offset = prepare_threshold_block(bank, block, address, offset, high);
429
		}
430
	}
431 432 433

	if (mce_flags.succor)
		deferred_error_interrupt_enable(c);
434 435
}

436 437 438 439 440 441 442 443 444 445 446 447 448
static void __log_error(unsigned int bank, bool threshold_err, u64 misc)
{
	struct mce m;
	u64 status;

	rdmsrl(MSR_IA32_MCx_STATUS(bank), status);
	if (!(status & MCI_STATUS_VAL))
		return;

	mce_setup(&m);

	m.status = status;
	m.bank = bank;
449

450 451 452
	if (threshold_err)
		m.misc = misc;

453 454
	if (m.status & MCI_STATUS_ADDRV)
		rdmsrl(MSR_IA32_MCx_ADDR(bank), m.addr);
455

456
	mce_log(&m);
457 458 459
	wrmsrl(MSR_IA32_MCx_STATUS(bank), 0);
}

460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499
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)
{
	u64 status;
	unsigned int bank;

	for (bank = 0; bank < mca_cfg.banks; ++bank) {
		rdmsrl(MSR_IA32_MCx_STATUS(bank), status);

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

		__log_error(bank, false, 0);
		break;
	}
}

500 501 502 503 504 505 506 507 508
/*
 * 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.
 */
509

510
static void amd_threshold_interrupt(void)
511
{
I
Ingo Molnar 已提交
512
	u32 low = 0, high = 0, address = 0;
513
	int cpu = smp_processor_id();
514
	unsigned int bank, block;
515 516

	/* assume first bank caused it */
517
	for (bank = 0; bank < mca_cfg.banks; ++bank) {
518
		if (!(per_cpu(bank_map, cpu) & (1 << bank)))
519
			continue;
520
		for (block = 0; block < NR_BLOCKS; ++block) {
521 522 523
			address = get_block_address(address, low, high, bank, block);
			if (!address)
				break;
524 525

			if (rdmsr_safe(address, &low, &high))
526
				break;
527 528 529 530 531 532 533 534

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

535 536
			if (!(high & MASK_CNTP_HI)  ||
			     (high & MASK_LOCKED_HI))
537 538
				continue;

I
Ingo Molnar 已提交
539 540 541 542
			/*
			 * Log the machine check that caused the threshold
			 * event.
			 */
543 544
			if (high & MASK_OVERFLOW_HI)
				goto log;
545 546
		}
	}
547 548 549
	return;

log:
550
	__log_error(bank, true, ((u64)high << 32) | low);
551 552 553 554 555 556 557
}

/*
 * Sysfs Interface
 */

struct threshold_attr {
J
Jacob Shin 已提交
558
	struct attribute attr;
I
Ingo Molnar 已提交
559 560
	ssize_t (*show) (struct threshold_block *, char *);
	ssize_t (*store) (struct threshold_block *, const char *, size_t count);
561 562
};

I
Ingo Molnar 已提交
563 564 565
#define SHOW_FIELDS(name)						\
static ssize_t show_ ## name(struct threshold_block *b, char *buf)	\
{									\
566
	return sprintf(buf, "%lu\n", (unsigned long) b->name);		\
J
Jacob Shin 已提交
567
}
568 569 570
SHOW_FIELDS(interrupt_enable)
SHOW_FIELDS(threshold_limit)

I
Ingo Molnar 已提交
571
static ssize_t
H
Hidetoshi Seto 已提交
572
store_interrupt_enable(struct threshold_block *b, const char *buf, size_t size)
573
{
574
	struct thresh_restart tr;
I
Ingo Molnar 已提交
575 576
	unsigned long new;

577 578 579
	if (!b->interrupt_capable)
		return -EINVAL;

580
	if (kstrtoul(buf, 0, &new) < 0)
581
		return -EINVAL;
I
Ingo Molnar 已提交
582

583 584
	b->interrupt_enable = !!new;

585
	memset(&tr, 0, sizeof(tr));
I
Ingo Molnar 已提交
586 587
	tr.b		= b;

588
	smp_call_function_single(b->cpu, threshold_restart_bank, &tr, 1);
589

H
Hidetoshi Seto 已提交
590
	return size;
591 592
}

I
Ingo Molnar 已提交
593
static ssize_t
H
Hidetoshi Seto 已提交
594
store_threshold_limit(struct threshold_block *b, const char *buf, size_t size)
595
{
596
	struct thresh_restart tr;
I
Ingo Molnar 已提交
597 598
	unsigned long new;

599
	if (kstrtoul(buf, 0, &new) < 0)
600
		return -EINVAL;
I
Ingo Molnar 已提交
601

602 603 604 605
	if (new > THRESHOLD_MAX)
		new = THRESHOLD_MAX;
	if (new < 1)
		new = 1;
I
Ingo Molnar 已提交
606

607
	memset(&tr, 0, sizeof(tr));
608
	tr.old_limit = b->threshold_limit;
609
	b->threshold_limit = new;
610
	tr.b = b;
611

612
	smp_call_function_single(b->cpu, threshold_restart_bank, &tr, 1);
613

H
Hidetoshi Seto 已提交
614
	return size;
615 616
}

617 618
static ssize_t show_error_count(struct threshold_block *b, char *buf)
{
619 620 621
	u32 lo, hi;

	rdmsr_on_cpu(b->cpu, b->address, &lo, &hi);
622

623 624
	return sprintf(buf, "%u\n", ((hi & THRESHOLD_MAX) -
				     (THRESHOLD_MAX - b->threshold_limit)));
625 626
}

627 628 629 630
static struct threshold_attr error_count = {
	.attr = {.name = __stringify(error_count), .mode = 0444 },
	.show = show_error_count,
};
631

632 633 634 635 636
#define RW_ATTR(val)							\
static struct threshold_attr val = {					\
	.attr	= {.name = __stringify(val), .mode = 0644 },		\
	.show	= show_## val,						\
	.store	= store_## val,						\
637 638
};

J
Jacob Shin 已提交
639 640
RW_ATTR(interrupt_enable);
RW_ATTR(threshold_limit);
641 642 643 644

static struct attribute *default_attrs[] = {
	&threshold_limit.attr,
	&error_count.attr,
645 646
	NULL,	/* possibly interrupt_enable if supported, see below */
	NULL,
647 648
};

I
Ingo Molnar 已提交
649 650
#define to_block(k)	container_of(k, struct threshold_block, kobj)
#define to_attr(a)	container_of(a, struct threshold_attr, attr)
651 652 653

static ssize_t show(struct kobject *kobj, struct attribute *attr, char *buf)
{
654
	struct threshold_block *b = to_block(kobj);
655 656
	struct threshold_attr *a = to_attr(attr);
	ssize_t ret;
I
Ingo Molnar 已提交
657

658
	ret = a->show ? a->show(b, buf) : -EIO;
I
Ingo Molnar 已提交
659

660 661 662 663 664 665
	return ret;
}

static ssize_t store(struct kobject *kobj, struct attribute *attr,
		     const char *buf, size_t count)
{
666
	struct threshold_block *b = to_block(kobj);
667 668
	struct threshold_attr *a = to_attr(attr);
	ssize_t ret;
I
Ingo Molnar 已提交
669

670
	ret = a->store ? a->store(b, buf, count) : -EIO;
I
Ingo Molnar 已提交
671

672 673 674
	return ret;
}

675
static const struct sysfs_ops threshold_ops = {
I
Ingo Molnar 已提交
676 677
	.show			= show,
	.store			= store,
678 679 680
};

static struct kobj_type threshold_ktype = {
I
Ingo Molnar 已提交
681 682
	.sysfs_ops		= &threshold_ops,
	.default_attrs		= default_attrs,
683 684
};

685 686
static int allocate_threshold_blocks(unsigned int cpu, unsigned int bank,
				     unsigned int block, u32 address)
687 688
{
	struct threshold_block *b = NULL;
I
Ingo Molnar 已提交
689 690
	u32 low, high;
	int err;
691

692
	if ((bank >= mca_cfg.banks) || (block >= NR_BLOCKS))
693 694
		return 0;

695
	if (rdmsr_safe_on_cpu(cpu, address, &low, &high))
696
		return 0;
697 698 699 700 701 702 703 704

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

705 706
	if (!(high & MASK_CNTP_HI)  ||
	     (high & MASK_LOCKED_HI))
707 708 709 710 711 712
		goto recurse;

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

I
Ingo Molnar 已提交
713 714 715 716 717
	b->block		= block;
	b->bank			= bank;
	b->cpu			= cpu;
	b->address		= address;
	b->interrupt_enable	= 0;
718
	b->interrupt_capable	= lvt_interrupt_supported(bank, high);
I
Ingo Molnar 已提交
719
	b->threshold_limit	= THRESHOLD_MAX;
720

721
	if (b->interrupt_capable) {
722
		threshold_ktype.default_attrs[2] = &interrupt_enable.attr;
723 724
		b->interrupt_enable = 1;
	} else {
725
		threshold_ktype.default_attrs[2] = NULL;
726
	}
727

728 729
	INIT_LIST_HEAD(&b->miscj);

I
Ingo Molnar 已提交
730
	if (per_cpu(threshold_banks, cpu)[bank]->blocks) {
731 732
		list_add(&b->miscj,
			 &per_cpu(threshold_banks, cpu)[bank]->blocks->miscj);
I
Ingo Molnar 已提交
733
	} else {
734
		per_cpu(threshold_banks, cpu)[bank]->blocks = b;
I
Ingo Molnar 已提交
735
	}
736

737 738
	err = kobject_init_and_add(&b->kobj, &threshold_ktype,
				   per_cpu(threshold_banks, cpu)[bank]->kobj,
739
				   (bank == 4 ? bank4_names(b) : th_names[bank]));
740 741 742
	if (err)
		goto out_free;
recurse:
743 744 745
	address = get_block_address(address, low, high, bank, ++block);
	if (!address)
		return 0;
746

747
	err = allocate_threshold_blocks(cpu, bank, block, address);
748 749 750
	if (err)
		goto out_free;

751 752
	if (b)
		kobject_uevent(&b->kobj, KOBJ_ADD);
753

754 755 756 757
	return err;

out_free:
	if (b) {
758
		kobject_put(&b->kobj);
759
		list_del(&b->miscj);
760 761 762 763 764
		kfree(b);
	}
	return err;
}

765
static int __threshold_add_blocks(struct threshold_bank *b)
766 767 768 769 770 771 772 773 774 775 776 777 778 779 780 781 782 783 784 785 786 787 788
{
	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;
}

789
static int threshold_create_bank(unsigned int cpu, unsigned int bank)
790
{
791
	struct device *dev = per_cpu(mce_device, cpu);
792
	struct amd_northbridge *nb = NULL;
793
	struct threshold_bank *b = NULL;
794
	const char *name = th_names[bank];
795
	int err = 0;
796

797
	if (is_shared_bank(bank)) {
798 799 800
		nb = node_to_amd_nb(amd_get_nb_id(cpu));

		/* threshold descriptor already initialized on this node? */
801
		if (nb && nb->bank4) {
802 803 804 805 806 807 808 809 810 811 812 813 814 815 816
			/* 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;
		}
	}

817
	b = kzalloc(sizeof(struct threshold_bank), GFP_KERNEL);
818 819 820 821 822
	if (!b) {
		err = -ENOMEM;
		goto out;
	}

823
	b->kobj = kobject_create_and_add(name, &dev->kobj);
824 825
	if (!b->kobj) {
		err = -EINVAL;
826
		goto out_free;
827
	}
828

829
	per_cpu(threshold_banks, cpu)[bank] = b;
830

831
	if (is_shared_bank(bank)) {
832 833 834
		atomic_set(&b->cpus, 1);

		/* nb is already initialized, see above */
835 836 837 838
		if (nb) {
			WARN_ON(nb->bank4);
			nb->bank4 = b;
		}
839 840
	}

841
	err = allocate_threshold_blocks(cpu, bank, 0, MSR_IA32_MCx_MISC(bank));
842 843
	if (!err)
		goto out;
844

845
 out_free:
846
	kfree(b);
847 848

 out:
849 850 851 852
	return err;
}

/* create dir/files for all valid threshold banks */
853
static int threshold_create_device(unsigned int cpu)
854
{
J
Jacob Shin 已提交
855
	unsigned int bank;
856
	struct threshold_bank **bp;
857 858
	int err = 0;

859 860 861 862 863 864 865 866
	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) {
867
		if (!(per_cpu(bank_map, cpu) & (1 << bank)))
868 869 870
			continue;
		err = threshold_create_bank(cpu, bank);
		if (err)
871
			return err;
872
	}
873

874 875 876
	return err;
}

877
static void deallocate_threshold_block(unsigned int cpu,
878 879 880 881 882 883 884 885 886 887
						 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) {
888
		kobject_put(&pos->kobj);
889 890 891 892 893 894 895 896
		list_del(&pos->miscj);
		kfree(pos);
	}

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

897 898 899 900 901 902 903 904 905 906 907
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);
}

908
static void threshold_remove_bank(unsigned int cpu, int bank)
909
{
910
	struct amd_northbridge *nb;
911 912 913 914 915
	struct threshold_bank *b;

	b = per_cpu(threshold_banks, cpu)[bank];
	if (!b)
		return;
916

917 918 919
	if (!b->blocks)
		goto free_out;

920
	if (is_shared_bank(bank)) {
921 922 923 924 925 926 927 928 929 930 931 932 933 934
		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;
		}
	}

935 936 937
	deallocate_threshold_block(cpu, bank);

free_out:
938
	kobject_del(b->kobj);
939
	kobject_put(b->kobj);
940 941
	kfree(b);
	per_cpu(threshold_banks, cpu)[bank] = NULL;
942 943
}

944
static void threshold_remove_device(unsigned int cpu)
945
{
J
Jacob Shin 已提交
946
	unsigned int bank;
947

948
	for (bank = 0; bank < mca_cfg.banks; ++bank) {
949
		if (!(per_cpu(bank_map, cpu) & (1 << bank)))
950 951 952
			continue;
		threshold_remove_bank(cpu, bank);
	}
953
	kfree(per_cpu(threshold_banks, cpu));
954 955 956
}

/* get notified when a cpu comes on/off */
957
static void
I
Ingo Molnar 已提交
958
amd_64_threshold_cpu_callback(unsigned long action, unsigned int cpu)
959 960 961
{
	switch (action) {
	case CPU_ONLINE:
962
	case CPU_ONLINE_FROZEN:
963 964 965
		threshold_create_device(cpu);
		break;
	case CPU_DEAD:
966
	case CPU_DEAD_FROZEN:
967 968 969 970 971 972 973 974 975
		threshold_remove_device(cpu);
		break;
	default:
		break;
	}
}

static __init int threshold_init_device(void)
{
J
Jacob Shin 已提交
976
	unsigned lcpu = 0;
977 978 979

	/* to hit CPUs online before the notifier is up */
	for_each_online_cpu(lcpu) {
980
		int err = threshold_create_device(lcpu);
I
Ingo Molnar 已提交
981

982
		if (err)
983
			return err;
984
	}
985
	threshold_cpu_callback = amd_64_threshold_cpu_callback;
I
Ingo Molnar 已提交
986

987
	return 0;
988
}
989 990 991 992 993 994 995 996 997 998 999 1000 1001 1002 1003 1004 1005 1006 1007 1008 1009
/*
 * 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);