ds.c 40.4 KB
Newer Older
1 2 3
#include <linux/bitops.h>
#include <linux/types.h>
#include <linux/slab.h>
4

5
#include <asm/perf_event.h>
6
#include <asm/insn.h>
7

8
#include "../perf_event.h"
9 10 11 12 13

/* The size of a BTS record in bytes: */
#define BTS_RECORD_SIZE		24

#define BTS_BUFFER_SIZE		(PAGE_SIZE << 4)
14
#define PEBS_BUFFER_SIZE	(PAGE_SIZE << 4)
15
#define PEBS_FIXUP_SIZE		PAGE_SIZE
16 17 18 19 20 21 22 23 24 25 26 27

/*
 * pebs_record_32 for p4 and core not supported

struct pebs_record_32 {
	u32 flags, ip;
	u32 ax, bc, cx, dx;
	u32 si, di, bp, sp;
};

 */

28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51
union intel_x86_pebs_dse {
	u64 val;
	struct {
		unsigned int ld_dse:4;
		unsigned int ld_stlb_miss:1;
		unsigned int ld_locked:1;
		unsigned int ld_reserved:26;
	};
	struct {
		unsigned int st_l1d_hit:1;
		unsigned int st_reserved1:3;
		unsigned int st_stlb_miss:1;
		unsigned int st_locked:1;
		unsigned int st_reserved2:26;
	};
};


/*
 * Map PEBS Load Latency Data Source encodings to generic
 * memory data source information
 */
#define P(a, b) PERF_MEM_S(a, b)
#define OP_LH (P(OP, LOAD) | P(LVL, HIT))
52 53
#define LEVEL(x) P(LVLNUM, x)
#define REM P(REMOTE, REMOTE)
54 55
#define SNOOP_NONE_MISS (P(SNOOP, NONE) | P(SNOOP, MISS))

56 57
/* Version for Sandy Bridge and later */
static u64 pebs_data_source[] = {
58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73
	P(OP, LOAD) | P(LVL, MISS) | LEVEL(L3) | P(SNOOP, NA),/* 0x00:ukn L3 */
	OP_LH | P(LVL, L1)  | LEVEL(L1) | P(SNOOP, NONE),  /* 0x01: L1 local */
	OP_LH | P(LVL, LFB) | LEVEL(LFB) | P(SNOOP, NONE), /* 0x02: LFB hit */
	OP_LH | P(LVL, L2)  | LEVEL(L2) | P(SNOOP, NONE),  /* 0x03: L2 hit */
	OP_LH | P(LVL, L3)  | LEVEL(L3) | P(SNOOP, NONE),  /* 0x04: L3 hit */
	OP_LH | P(LVL, L3)  | LEVEL(L3) | P(SNOOP, MISS),  /* 0x05: L3 hit, snoop miss */
	OP_LH | P(LVL, L3)  | LEVEL(L3) | P(SNOOP, HIT),   /* 0x06: L3 hit, snoop hit */
	OP_LH | P(LVL, L3)  | LEVEL(L3) | P(SNOOP, HITM),  /* 0x07: L3 hit, snoop hitm */
	OP_LH | P(LVL, REM_CCE1) | REM | LEVEL(L3) | P(SNOOP, HIT),  /* 0x08: L3 miss snoop hit */
	OP_LH | P(LVL, REM_CCE1) | REM | LEVEL(L3) | P(SNOOP, HITM), /* 0x09: L3 miss snoop hitm*/
	OP_LH | P(LVL, LOC_RAM)  | LEVEL(RAM) | P(SNOOP, HIT),       /* 0x0a: L3 miss, shared */
	OP_LH | P(LVL, REM_RAM1) | REM | LEVEL(L3) | P(SNOOP, HIT),  /* 0x0b: L3 miss, shared */
	OP_LH | P(LVL, LOC_RAM)  | LEVEL(RAM) | SNOOP_NONE_MISS,     /* 0x0c: L3 miss, excl */
	OP_LH | P(LVL, REM_RAM1) | LEVEL(RAM) | REM | SNOOP_NONE_MISS, /* 0x0d: L3 miss, excl */
	OP_LH | P(LVL, IO)  | LEVEL(NA) | P(SNOOP, NONE), /* 0x0e: I/O */
	OP_LH | P(LVL, UNC) | LEVEL(NA) | P(SNOOP, NONE), /* 0x0f: uncached */
74 75
};

76 77 78
/* Patch up minor differences in the bits */
void __init intel_pmu_pebs_data_source_nhm(void)
{
79 80 81 82 83 84 85 86 87 88 89 90 91 92
	pebs_data_source[0x05] = OP_LH | P(LVL, L3) | LEVEL(L3) | P(SNOOP, HIT);
	pebs_data_source[0x06] = OP_LH | P(LVL, L3) | LEVEL(L3) | P(SNOOP, HITM);
	pebs_data_source[0x07] = OP_LH | P(LVL, L3) | LEVEL(L3) | P(SNOOP, HITM);
}

void __init intel_pmu_pebs_data_source_skl(bool pmem)
{
	u64 pmem_or_l4 = pmem ? LEVEL(PMEM) : LEVEL(L4);

	pebs_data_source[0x08] = OP_LH | pmem_or_l4 | P(SNOOP, HIT);
	pebs_data_source[0x09] = OP_LH | pmem_or_l4 | REM | P(SNOOP, HIT);
	pebs_data_source[0x0b] = OP_LH | LEVEL(RAM) | REM | P(SNOOP, NONE);
	pebs_data_source[0x0c] = OP_LH | LEVEL(ANY_CACHE) | REM | P(SNOOPX, FWD);
	pebs_data_source[0x0d] = OP_LH | LEVEL(ANY_CACHE) | REM | P(SNOOP, HITM);
93 94
}

95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132
static u64 precise_store_data(u64 status)
{
	union intel_x86_pebs_dse dse;
	u64 val = P(OP, STORE) | P(SNOOP, NA) | P(LVL, L1) | P(TLB, L2);

	dse.val = status;

	/*
	 * bit 4: TLB access
	 * 1 = stored missed 2nd level TLB
	 *
	 * so it either hit the walker or the OS
	 * otherwise hit 2nd level TLB
	 */
	if (dse.st_stlb_miss)
		val |= P(TLB, MISS);
	else
		val |= P(TLB, HIT);

	/*
	 * bit 0: hit L1 data cache
	 * if not set, then all we know is that
	 * it missed L1D
	 */
	if (dse.st_l1d_hit)
		val |= P(LVL, HIT);
	else
		val |= P(LVL, MISS);

	/*
	 * bit 5: Locked prefix
	 */
	if (dse.st_locked)
		val |= P(LOCK, LOCKED);

	return val;
}

133
static u64 precise_datala_hsw(struct perf_event *event, u64 status)
134 135 136
{
	union perf_mem_data_src dse;

137 138 139 140 141 142
	dse.val = PERF_MEM_NA;

	if (event->hw.flags & PERF_X86_EVENT_PEBS_ST_HSW)
		dse.mem_op = PERF_MEM_OP_STORE;
	else if (event->hw.flags & PERF_X86_EVENT_PEBS_LD_HSW)
		dse.mem_op = PERF_MEM_OP_LOAD;
143 144 145 146 147 148 149 150 151

	/*
	 * L1 info only valid for following events:
	 *
	 * MEM_UOPS_RETIRED.STLB_MISS_STORES
	 * MEM_UOPS_RETIRED.LOCK_STORES
	 * MEM_UOPS_RETIRED.SPLIT_STORES
	 * MEM_UOPS_RETIRED.ALL_STORES
	 */
152 153 154 155 156 157
	if (event->hw.flags & PERF_X86_EVENT_PEBS_ST_HSW) {
		if (status & 1)
			dse.mem_lvl = PERF_MEM_LVL_L1 | PERF_MEM_LVL_HIT;
		else
			dse.mem_lvl = PERF_MEM_LVL_L1 | PERF_MEM_LVL_MISS;
	}
158 159 160
	return dse.val;
}

161 162 163 164 165 166 167 168 169 170 171 172 173 174 175
static u64 load_latency_data(u64 status)
{
	union intel_x86_pebs_dse dse;
	u64 val;

	dse.val = status;

	/*
	 * use the mapping table for bit 0-3
	 */
	val = pebs_data_source[dse.ld_dse];

	/*
	 * Nehalem models do not support TLB, Lock infos
	 */
176
	if (x86_pmu.pebs_no_tlb) {
177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198
		val |= P(TLB, NA) | P(LOCK, NA);
		return val;
	}
	/*
	 * bit 4: TLB access
	 * 0 = did not miss 2nd level TLB
	 * 1 = missed 2nd level TLB
	 */
	if (dse.ld_stlb_miss)
		val |= P(TLB, MISS) | P(TLB, L2);
	else
		val |= P(TLB, HIT) | P(TLB, L1) | P(TLB, L2);

	/*
	 * bit 5: locked prefix
	 */
	if (dse.ld_locked)
		val |= P(LOCK, LOCKED);

	return val;
}

199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215
struct pebs_record_core {
	u64 flags, ip;
	u64 ax, bx, cx, dx;
	u64 si, di, bp, sp;
	u64 r8,  r9,  r10, r11;
	u64 r12, r13, r14, r15;
};

struct pebs_record_nhm {
	u64 flags, ip;
	u64 ax, bx, cx, dx;
	u64 si, di, bp, sp;
	u64 r8,  r9,  r10, r11;
	u64 r12, r13, r14, r15;
	u64 status, dla, dse, lat;
};

216 217 218 219
/*
 * Same as pebs_record_nhm, with two additional fields.
 */
struct pebs_record_hsw {
220 221 222 223 224 225
	u64 flags, ip;
	u64 ax, bx, cx, dx;
	u64 si, di, bp, sp;
	u64 r8,  r9,  r10, r11;
	u64 r12, r13, r14, r15;
	u64 status, dla, dse, lat;
226
	u64 real_ip, tsx_tuning;
227 228 229 230 231 232 233 234 235 236 237 238 239 240 241
};

union hsw_tsx_tuning {
	struct {
		u32 cycles_last_block     : 32,
		    hle_abort		  : 1,
		    rtm_abort		  : 1,
		    instruction_abort     : 1,
		    non_instruction_abort : 1,
		    retry		  : 1,
		    data_conflict	  : 1,
		    capacity_writes	  : 1,
		    capacity_reads	  : 1;
	};
	u64	    value;
242 243
};

244 245
#define PEBS_HSW_TSX_FLAGS	0xff00000000ULL

246 247 248 249 250 251 252 253 254 255 256 257 258
/* Same as HSW, plus TSC */

struct pebs_record_skl {
	u64 flags, ip;
	u64 ax, bx, cx, dx;
	u64 si, di, bp, sp;
	u64 r8,  r9,  r10, r11;
	u64 r12, r13, r14, r15;
	u64 status, dla, dse, lat;
	u64 real_ip, tsx_tuning;
	u64 tsc;
};

259
void init_debug_store_on_cpu(int cpu)
260 261 262 263 264 265 266 267 268 269 270
{
	struct debug_store *ds = per_cpu(cpu_hw_events, cpu).ds;

	if (!ds)
		return;

	wrmsr_on_cpu(cpu, MSR_IA32_DS_AREA,
		     (u32)((u64)(unsigned long)ds),
		     (u32)((u64)(unsigned long)ds >> 32));
}

271
void fini_debug_store_on_cpu(int cpu)
272 273 274 275 276 277 278
{
	if (!per_cpu(cpu_hw_events, cpu).ds)
		return;

	wrmsr_on_cpu(cpu, MSR_IA32_DS_AREA, 0, 0);
}

279 280
static DEFINE_PER_CPU(void *, insn_buffer);

281 282 283
static int alloc_pebs_buffer(int cpu)
{
	struct debug_store *ds = per_cpu(cpu_hw_events, cpu).ds;
284
	int node = cpu_to_node(cpu);
285
	int max;
286
	void *buffer, *ibuffer;
287 288 289 290

	if (!x86_pmu.pebs)
		return 0;

291
	buffer = kzalloc_node(x86_pmu.pebs_buffer_size, GFP_KERNEL, node);
292 293 294
	if (unlikely(!buffer))
		return -ENOMEM;

295 296 297 298 299 300 301 302 303 304 305 306 307
	/*
	 * HSW+ already provides us the eventing ip; no need to allocate this
	 * buffer then.
	 */
	if (x86_pmu.intel_cap.pebs_format < 2) {
		ibuffer = kzalloc_node(PEBS_FIXUP_SIZE, GFP_KERNEL, node);
		if (!ibuffer) {
			kfree(buffer);
			return -ENOMEM;
		}
		per_cpu(insn_buffer, cpu) = ibuffer;
	}

308
	max = x86_pmu.pebs_buffer_size / x86_pmu.pebs_record_size;
309 310 311 312 313 314 315 316 317

	ds->pebs_buffer_base = (u64)(unsigned long)buffer;
	ds->pebs_index = ds->pebs_buffer_base;
	ds->pebs_absolute_maximum = ds->pebs_buffer_base +
		max * x86_pmu.pebs_record_size;

	return 0;
}

318 319 320 321 322 323 324
static void release_pebs_buffer(int cpu)
{
	struct debug_store *ds = per_cpu(cpu_hw_events, cpu).ds;

	if (!ds || !x86_pmu.pebs)
		return;

325 326 327
	kfree(per_cpu(insn_buffer, cpu));
	per_cpu(insn_buffer, cpu) = NULL;

328 329 330 331
	kfree((void *)(unsigned long)ds->pebs_buffer_base);
	ds->pebs_buffer_base = 0;
}

332 333 334
static int alloc_bts_buffer(int cpu)
{
	struct debug_store *ds = per_cpu(cpu_hw_events, cpu).ds;
335
	int node = cpu_to_node(cpu);
336 337 338 339 340 341
	int max, thresh;
	void *buffer;

	if (!x86_pmu.bts)
		return 0;

342 343 344
	buffer = kzalloc_node(BTS_BUFFER_SIZE, GFP_KERNEL | __GFP_NOWARN, node);
	if (unlikely(!buffer)) {
		WARN_ONCE(1, "%s: BTS buffer allocation failure\n", __func__);
345
		return -ENOMEM;
346
	}
347 348 349 350 351 352 353 354 355 356 357 358 359 360

	max = BTS_BUFFER_SIZE / BTS_RECORD_SIZE;
	thresh = max / 16;

	ds->bts_buffer_base = (u64)(unsigned long)buffer;
	ds->bts_index = ds->bts_buffer_base;
	ds->bts_absolute_maximum = ds->bts_buffer_base +
		max * BTS_RECORD_SIZE;
	ds->bts_interrupt_threshold = ds->bts_absolute_maximum -
		thresh * BTS_RECORD_SIZE;

	return 0;
}

361 362 363 364 365 366 367 368 369 370 371
static void release_bts_buffer(int cpu)
{
	struct debug_store *ds = per_cpu(cpu_hw_events, cpu).ds;

	if (!ds || !x86_pmu.bts)
		return;

	kfree((void *)(unsigned long)ds->bts_buffer_base);
	ds->bts_buffer_base = 0;
}

372 373
static int alloc_ds_buffer(int cpu)
{
374
	int node = cpu_to_node(cpu);
375 376
	struct debug_store *ds;

377
	ds = kzalloc_node(sizeof(*ds), GFP_KERNEL, node);
378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396
	if (unlikely(!ds))
		return -ENOMEM;

	per_cpu(cpu_hw_events, cpu).ds = ds;

	return 0;
}

static void release_ds_buffer(int cpu)
{
	struct debug_store *ds = per_cpu(cpu_hw_events, cpu).ds;

	if (!ds)
		return;

	per_cpu(cpu_hw_events, cpu).ds = NULL;
	kfree(ds);
}

397
void release_ds_buffers(void)
398 399 400 401 402 403 404 405 406 407 408
{
	int cpu;

	if (!x86_pmu.bts && !x86_pmu.pebs)
		return;

	get_online_cpus();
	for_each_online_cpu(cpu)
		fini_debug_store_on_cpu(cpu);

	for_each_possible_cpu(cpu) {
409 410
		release_pebs_buffer(cpu);
		release_bts_buffer(cpu);
411
		release_ds_buffer(cpu);
412 413 414 415
	}
	put_online_cpus();
}

416
void reserve_ds_buffers(void)
417
{
418 419 420 421 422
	int bts_err = 0, pebs_err = 0;
	int cpu;

	x86_pmu.bts_active = 0;
	x86_pmu.pebs_active = 0;
423 424

	if (!x86_pmu.bts && !x86_pmu.pebs)
425
		return;
426

427 428 429 430 431 432
	if (!x86_pmu.bts)
		bts_err = 1;

	if (!x86_pmu.pebs)
		pebs_err = 1;

433 434 435
	get_online_cpus();

	for_each_possible_cpu(cpu) {
436 437 438 439
		if (alloc_ds_buffer(cpu)) {
			bts_err = 1;
			pebs_err = 1;
		}
440

441 442 443 444 445
		if (!bts_err && alloc_bts_buffer(cpu))
			bts_err = 1;

		if (!pebs_err && alloc_pebs_buffer(cpu))
			pebs_err = 1;
446

447
		if (bts_err && pebs_err)
448
			break;
449 450 451 452 453 454
	}

	if (bts_err) {
		for_each_possible_cpu(cpu)
			release_bts_buffer(cpu);
	}
455

456 457 458
	if (pebs_err) {
		for_each_possible_cpu(cpu)
			release_pebs_buffer(cpu);
459 460
	}

461 462 463 464 465 466 467 468 469 470
	if (bts_err && pebs_err) {
		for_each_possible_cpu(cpu)
			release_ds_buffer(cpu);
	} else {
		if (x86_pmu.bts && !bts_err)
			x86_pmu.bts_active = 1;

		if (x86_pmu.pebs && !pebs_err)
			x86_pmu.pebs_active = 1;

471 472 473 474 475 476 477 478 479 480 481
		for_each_online_cpu(cpu)
			init_debug_store_on_cpu(cpu);
	}

	put_online_cpus();
}

/*
 * BTS
 */

482
struct event_constraint bts_constraint =
483
	EVENT_CONSTRAINT(0, 1ULL << INTEL_PMC_IDX_FIXED_BTS, 0);
484

485
void intel_pmu_enable_bts(u64 config)
486 487 488 489 490
{
	unsigned long debugctlmsr;

	debugctlmsr = get_debugctlmsr();

491 492
	debugctlmsr |= DEBUGCTLMSR_TR;
	debugctlmsr |= DEBUGCTLMSR_BTS;
493 494
	if (config & ARCH_PERFMON_EVENTSEL_INT)
		debugctlmsr |= DEBUGCTLMSR_BTINT;
495 496

	if (!(config & ARCH_PERFMON_EVENTSEL_OS))
497
		debugctlmsr |= DEBUGCTLMSR_BTS_OFF_OS;
498 499

	if (!(config & ARCH_PERFMON_EVENTSEL_USR))
500
		debugctlmsr |= DEBUGCTLMSR_BTS_OFF_USR;
501 502 503 504

	update_debugctlmsr(debugctlmsr);
}

505
void intel_pmu_disable_bts(void)
506
{
507
	struct cpu_hw_events *cpuc = this_cpu_ptr(&cpu_hw_events);
508 509 510 511 512 513 514 515
	unsigned long debugctlmsr;

	if (!cpuc->ds)
		return;

	debugctlmsr = get_debugctlmsr();

	debugctlmsr &=
516 517
		~(DEBUGCTLMSR_TR | DEBUGCTLMSR_BTS | DEBUGCTLMSR_BTINT |
		  DEBUGCTLMSR_BTS_OFF_OS | DEBUGCTLMSR_BTS_OFF_USR);
518 519 520 521

	update_debugctlmsr(debugctlmsr);
}

522
int intel_pmu_drain_bts_buffer(void)
523
{
524
	struct cpu_hw_events *cpuc = this_cpu_ptr(&cpu_hw_events);
525 526 527 528 529 530
	struct debug_store *ds = cpuc->ds;
	struct bts_record {
		u64	from;
		u64	to;
		u64	flags;
	};
531
	struct perf_event *event = cpuc->events[INTEL_PMC_IDX_FIXED_BTS];
532
	struct bts_record *at, *base, *top;
533 534 535
	struct perf_output_handle handle;
	struct perf_event_header header;
	struct perf_sample_data data;
536
	unsigned long skip = 0;
537 538 539
	struct pt_regs regs;

	if (!event)
540
		return 0;
541

542
	if (!x86_pmu.bts_active)
543
		return 0;
544

545 546
	base = (struct bts_record *)(unsigned long)ds->bts_buffer_base;
	top  = (struct bts_record *)(unsigned long)ds->bts_index;
547

548
	if (top <= base)
549
		return 0;
550

551 552
	memset(&regs, 0, sizeof(regs));

553 554
	ds->bts_index = ds->bts_buffer_base;

555
	perf_sample_data_init(&data, 0, event->hw.last_period);
556

557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577
	/*
	 * BTS leaks kernel addresses in branches across the cpl boundary,
	 * such as traps or system calls, so unless the user is asking for
	 * kernel tracing (and right now it's not possible), we'd need to
	 * filter them out. But first we need to count how many of those we
	 * have in the current batch. This is an extra O(n) pass, however,
	 * it's much faster than the other one especially considering that
	 * n <= 2560 (BTS_BUFFER_SIZE / BTS_RECORD_SIZE * 15/16; see the
	 * alloc_bts_buffer()).
	 */
	for (at = base; at < top; at++) {
		/*
		 * Note that right now *this* BTS code only works if
		 * attr::exclude_kernel is set, but let's keep this extra
		 * check here in case that changes.
		 */
		if (event->attr.exclude_kernel &&
		    (kernel_ip(at->from) || kernel_ip(at->to)))
			skip++;
	}

578 579 580 581 582
	/*
	 * Prepare a generic sample, i.e. fill in the invariant fields.
	 * We will overwrite the from and to address before we output
	 * the sample.
	 */
P
Peter Zijlstra 已提交
583
	rcu_read_lock();
584 585
	perf_prepare_sample(&header, &data, event, &regs);

586 587
	if (perf_output_begin(&handle, event, header.size *
			      (top - base - skip)))
P
Peter Zijlstra 已提交
588
		goto unlock;
589

590 591 592 593 594 595
	for (at = base; at < top; at++) {
		/* Filter out any records that contain kernel addresses. */
		if (event->attr.exclude_kernel &&
		    (kernel_ip(at->from) || kernel_ip(at->to)))
			continue;

596 597 598 599 600 601 602 603 604 605 606
		data.ip		= at->from;
		data.addr	= at->to;

		perf_output_sample(&handle, &header, &data, event);
	}

	perf_output_end(&handle);

	/* There's new data available. */
	event->hw.interrupts++;
	event->pending_kill = POLL_IN;
P
Peter Zijlstra 已提交
607 608
unlock:
	rcu_read_unlock();
609
	return 1;
610 611
}

612 613 614 615 616 617 618
static inline void intel_pmu_drain_pebs_buffer(void)
{
	struct pt_regs regs;

	x86_pmu.drain_pebs(&regs);
}

619 620 621
/*
 * PEBS
 */
622
struct event_constraint intel_core2_pebs_event_constraints[] = {
623 624 625 626 627
	INTEL_FLAGS_UEVENT_CONSTRAINT(0x00c0, 0x1), /* INST_RETIRED.ANY */
	INTEL_FLAGS_UEVENT_CONSTRAINT(0xfec1, 0x1), /* X87_OPS_RETIRED.ANY */
	INTEL_FLAGS_UEVENT_CONSTRAINT(0x00c5, 0x1), /* BR_INST_RETIRED.MISPRED */
	INTEL_FLAGS_UEVENT_CONSTRAINT(0x1fc7, 0x1), /* SIMD_INST_RETURED.ANY */
	INTEL_FLAGS_EVENT_CONSTRAINT(0xcb, 0x1),    /* MEM_LOAD_RETIRED.* */
628 629
	/* INST_RETIRED.ANY_P, inv=1, cmask=16 (cycles:p). */
	INTEL_FLAGS_EVENT_CONSTRAINT(0x108000c0, 0x01),
630 631 632
	EVENT_CONSTRAINT_END
};

633
struct event_constraint intel_atom_pebs_event_constraints[] = {
634 635 636
	INTEL_FLAGS_UEVENT_CONSTRAINT(0x00c0, 0x1), /* INST_RETIRED.ANY */
	INTEL_FLAGS_UEVENT_CONSTRAINT(0x00c5, 0x1), /* MISPREDICTED_BRANCH_RETIRED */
	INTEL_FLAGS_EVENT_CONSTRAINT(0xcb, 0x1),    /* MEM_LOAD_RETIRED.* */
637 638
	/* INST_RETIRED.ANY_P, inv=1, cmask=16 (cycles:p). */
	INTEL_FLAGS_EVENT_CONSTRAINT(0x108000c0, 0x01),
639 640
	/* Allow all events as PEBS with no flags */
	INTEL_ALL_EVENT_CONSTRAINT(0, 0x1),
641 642 643
	EVENT_CONSTRAINT_END
};

644
struct event_constraint intel_slm_pebs_event_constraints[] = {
645 646
	/* INST_RETIRED.ANY_P, inv=1, cmask=16 (cycles:p). */
	INTEL_FLAGS_EVENT_CONSTRAINT(0x108000c0, 0x1),
647 648
	/* Allow all events as PEBS with no flags */
	INTEL_ALL_EVENT_CONSTRAINT(0, 0x1),
649 650 651
	EVENT_CONSTRAINT_END
};

652 653 654 655 656 657
struct event_constraint intel_glm_pebs_event_constraints[] = {
	/* Allow all events as PEBS with no flags */
	INTEL_ALL_EVENT_CONSTRAINT(0, 0x1),
	EVENT_CONSTRAINT_END
};

658 659 660 661 662 663
struct event_constraint intel_glp_pebs_event_constraints[] = {
	/* Allow all events as PEBS with no flags */
	INTEL_ALL_EVENT_CONSTRAINT(0, 0xf),
	EVENT_CONSTRAINT_END
};

664
struct event_constraint intel_nehalem_pebs_event_constraints[] = {
665
	INTEL_PLD_CONSTRAINT(0x100b, 0xf),      /* MEM_INST_RETIRED.* */
666 667 668
	INTEL_FLAGS_EVENT_CONSTRAINT(0x0f, 0xf),    /* MEM_UNCORE_RETIRED.* */
	INTEL_FLAGS_UEVENT_CONSTRAINT(0x010c, 0xf), /* MEM_STORE_RETIRED.DTLB_MISS */
	INTEL_FLAGS_EVENT_CONSTRAINT(0xc0, 0xf),    /* INST_RETIRED.ANY */
669
	INTEL_EVENT_CONSTRAINT(0xc2, 0xf),    /* UOPS_RETIRED.* */
670 671 672 673 674 675
	INTEL_FLAGS_EVENT_CONSTRAINT(0xc4, 0xf),    /* BR_INST_RETIRED.* */
	INTEL_FLAGS_UEVENT_CONSTRAINT(0x02c5, 0xf), /* BR_MISP_RETIRED.NEAR_CALL */
	INTEL_FLAGS_EVENT_CONSTRAINT(0xc7, 0xf),    /* SSEX_UOPS_RETIRED.* */
	INTEL_FLAGS_UEVENT_CONSTRAINT(0x20c8, 0xf), /* ITLB_MISS_RETIRED */
	INTEL_FLAGS_EVENT_CONSTRAINT(0xcb, 0xf),    /* MEM_LOAD_RETIRED.* */
	INTEL_FLAGS_EVENT_CONSTRAINT(0xf7, 0xf),    /* FP_ASSIST.* */
676 677
	/* INST_RETIRED.ANY_P, inv=1, cmask=16 (cycles:p). */
	INTEL_FLAGS_EVENT_CONSTRAINT(0x108000c0, 0x0f),
678 679 680
	EVENT_CONSTRAINT_END
};

681
struct event_constraint intel_westmere_pebs_event_constraints[] = {
682
	INTEL_PLD_CONSTRAINT(0x100b, 0xf),      /* MEM_INST_RETIRED.* */
683 684 685
	INTEL_FLAGS_EVENT_CONSTRAINT(0x0f, 0xf),    /* MEM_UNCORE_RETIRED.* */
	INTEL_FLAGS_UEVENT_CONSTRAINT(0x010c, 0xf), /* MEM_STORE_RETIRED.DTLB_MISS */
	INTEL_FLAGS_EVENT_CONSTRAINT(0xc0, 0xf),    /* INSTR_RETIRED.* */
686
	INTEL_EVENT_CONSTRAINT(0xc2, 0xf),    /* UOPS_RETIRED.* */
687 688 689 690 691 692
	INTEL_FLAGS_EVENT_CONSTRAINT(0xc4, 0xf),    /* BR_INST_RETIRED.* */
	INTEL_FLAGS_EVENT_CONSTRAINT(0xc5, 0xf),    /* BR_MISP_RETIRED.* */
	INTEL_FLAGS_EVENT_CONSTRAINT(0xc7, 0xf),    /* SSEX_UOPS_RETIRED.* */
	INTEL_FLAGS_UEVENT_CONSTRAINT(0x20c8, 0xf), /* ITLB_MISS_RETIRED */
	INTEL_FLAGS_EVENT_CONSTRAINT(0xcb, 0xf),    /* MEM_LOAD_RETIRED.* */
	INTEL_FLAGS_EVENT_CONSTRAINT(0xf7, 0xf),    /* FP_ASSIST.* */
693 694
	/* INST_RETIRED.ANY_P, inv=1, cmask=16 (cycles:p). */
	INTEL_FLAGS_EVENT_CONSTRAINT(0x108000c0, 0x0f),
695 696 697
	EVENT_CONSTRAINT_END
};

698
struct event_constraint intel_snb_pebs_event_constraints[] = {
699
	INTEL_FLAGS_UEVENT_CONSTRAINT(0x01c0, 0x2), /* INST_RETIRED.PRECDIST */
700
	INTEL_PLD_CONSTRAINT(0x01cd, 0x8),    /* MEM_TRANS_RETIRED.LAT_ABOVE_THR */
701
	INTEL_PST_CONSTRAINT(0x02cd, 0x8),    /* MEM_TRANS_RETIRED.PRECISE_STORES */
702 703
	/* UOPS_RETIRED.ALL, inv=1, cmask=16 (cycles:p). */
	INTEL_FLAGS_EVENT_CONSTRAINT(0x108001c2, 0xf),
704 705 706 707
        INTEL_EXCLEVT_CONSTRAINT(0xd0, 0xf),    /* MEM_UOP_RETIRED.* */
        INTEL_EXCLEVT_CONSTRAINT(0xd1, 0xf),    /* MEM_LOAD_UOPS_RETIRED.* */
        INTEL_EXCLEVT_CONSTRAINT(0xd2, 0xf),    /* MEM_LOAD_UOPS_LLC_HIT_RETIRED.* */
        INTEL_EXCLEVT_CONSTRAINT(0xd3, 0xf),    /* MEM_LOAD_UOPS_LLC_MISS_RETIRED.* */
708 709
	/* Allow all events as PEBS with no flags */
	INTEL_ALL_EVENT_CONSTRAINT(0, 0xf),
710 711 712
	EVENT_CONSTRAINT_END
};

713
struct event_constraint intel_ivb_pebs_event_constraints[] = {
714
        INTEL_FLAGS_UEVENT_CONSTRAINT(0x01c0, 0x2), /* INST_RETIRED.PRECDIST */
715
        INTEL_PLD_CONSTRAINT(0x01cd, 0x8),    /* MEM_TRANS_RETIRED.LAT_ABOVE_THR */
716
	INTEL_PST_CONSTRAINT(0x02cd, 0x8),    /* MEM_TRANS_RETIRED.PRECISE_STORES */
717 718
	/* UOPS_RETIRED.ALL, inv=1, cmask=16 (cycles:p). */
	INTEL_FLAGS_EVENT_CONSTRAINT(0x108001c2, 0xf),
719 720
	/* INST_RETIRED.PREC_DIST, inv=1, cmask=16 (cycles:ppp). */
	INTEL_FLAGS_EVENT_CONSTRAINT(0x108001c0, 0x2),
721 722 723 724
	INTEL_EXCLEVT_CONSTRAINT(0xd0, 0xf),    /* MEM_UOP_RETIRED.* */
	INTEL_EXCLEVT_CONSTRAINT(0xd1, 0xf),    /* MEM_LOAD_UOPS_RETIRED.* */
	INTEL_EXCLEVT_CONSTRAINT(0xd2, 0xf),    /* MEM_LOAD_UOPS_LLC_HIT_RETIRED.* */
	INTEL_EXCLEVT_CONSTRAINT(0xd3, 0xf),    /* MEM_LOAD_UOPS_LLC_MISS_RETIRED.* */
725 726
	/* Allow all events as PEBS with no flags */
	INTEL_ALL_EVENT_CONSTRAINT(0, 0xf),
727 728 729
        EVENT_CONSTRAINT_END
};

730
struct event_constraint intel_hsw_pebs_event_constraints[] = {
731
	INTEL_FLAGS_UEVENT_CONSTRAINT(0x01c0, 0x2), /* INST_RETIRED.PRECDIST */
732 733 734
	INTEL_PLD_CONSTRAINT(0x01cd, 0xf),    /* MEM_TRANS_RETIRED.* */
	/* UOPS_RETIRED.ALL, inv=1, cmask=16 (cycles:p). */
	INTEL_FLAGS_EVENT_CONSTRAINT(0x108001c2, 0xf),
735 736
	/* INST_RETIRED.PREC_DIST, inv=1, cmask=16 (cycles:ppp). */
	INTEL_FLAGS_EVENT_CONSTRAINT(0x108001c0, 0x2),
737
	INTEL_FLAGS_UEVENT_CONSTRAINT_DATALA_NA(0x01c2, 0xf), /* UOPS_RETIRED.ALL */
738 739 740 741 742 743 744 745 746 747
	INTEL_FLAGS_UEVENT_CONSTRAINT_DATALA_XLD(0x11d0, 0xf), /* MEM_UOPS_RETIRED.STLB_MISS_LOADS */
	INTEL_FLAGS_UEVENT_CONSTRAINT_DATALA_XLD(0x21d0, 0xf), /* MEM_UOPS_RETIRED.LOCK_LOADS */
	INTEL_FLAGS_UEVENT_CONSTRAINT_DATALA_XLD(0x41d0, 0xf), /* MEM_UOPS_RETIRED.SPLIT_LOADS */
	INTEL_FLAGS_UEVENT_CONSTRAINT_DATALA_XLD(0x81d0, 0xf), /* MEM_UOPS_RETIRED.ALL_LOADS */
	INTEL_FLAGS_UEVENT_CONSTRAINT_DATALA_XST(0x12d0, 0xf), /* MEM_UOPS_RETIRED.STLB_MISS_STORES */
	INTEL_FLAGS_UEVENT_CONSTRAINT_DATALA_XST(0x42d0, 0xf), /* MEM_UOPS_RETIRED.SPLIT_STORES */
	INTEL_FLAGS_UEVENT_CONSTRAINT_DATALA_XST(0x82d0, 0xf), /* MEM_UOPS_RETIRED.ALL_STORES */
	INTEL_FLAGS_EVENT_CONSTRAINT_DATALA_XLD(0xd1, 0xf),    /* MEM_LOAD_UOPS_RETIRED.* */
	INTEL_FLAGS_EVENT_CONSTRAINT_DATALA_XLD(0xd2, 0xf),    /* MEM_LOAD_UOPS_L3_HIT_RETIRED.* */
	INTEL_FLAGS_EVENT_CONSTRAINT_DATALA_XLD(0xd3, 0xf),    /* MEM_LOAD_UOPS_L3_MISS_RETIRED.* */
748 749 750 751 752
	/* Allow all events as PEBS with no flags */
	INTEL_ALL_EVENT_CONSTRAINT(0, 0xf),
	EVENT_CONSTRAINT_END
};

753 754 755 756 757 758 759 760 761 762 763 764 765 766 767 768 769 770 771 772 773 774 775 776
struct event_constraint intel_bdw_pebs_event_constraints[] = {
	INTEL_FLAGS_UEVENT_CONSTRAINT(0x01c0, 0x2), /* INST_RETIRED.PRECDIST */
	INTEL_PLD_CONSTRAINT(0x01cd, 0xf),    /* MEM_TRANS_RETIRED.* */
	/* UOPS_RETIRED.ALL, inv=1, cmask=16 (cycles:p). */
	INTEL_FLAGS_EVENT_CONSTRAINT(0x108001c2, 0xf),
	/* INST_RETIRED.PREC_DIST, inv=1, cmask=16 (cycles:ppp). */
	INTEL_FLAGS_EVENT_CONSTRAINT(0x108001c0, 0x2),
	INTEL_FLAGS_UEVENT_CONSTRAINT_DATALA_NA(0x01c2, 0xf), /* UOPS_RETIRED.ALL */
	INTEL_FLAGS_UEVENT_CONSTRAINT_DATALA_LD(0x11d0, 0xf), /* MEM_UOPS_RETIRED.STLB_MISS_LOADS */
	INTEL_FLAGS_UEVENT_CONSTRAINT_DATALA_LD(0x21d0, 0xf), /* MEM_UOPS_RETIRED.LOCK_LOADS */
	INTEL_FLAGS_UEVENT_CONSTRAINT_DATALA_LD(0x41d0, 0xf), /* MEM_UOPS_RETIRED.SPLIT_LOADS */
	INTEL_FLAGS_UEVENT_CONSTRAINT_DATALA_LD(0x81d0, 0xf), /* MEM_UOPS_RETIRED.ALL_LOADS */
	INTEL_FLAGS_UEVENT_CONSTRAINT_DATALA_ST(0x12d0, 0xf), /* MEM_UOPS_RETIRED.STLB_MISS_STORES */
	INTEL_FLAGS_UEVENT_CONSTRAINT_DATALA_ST(0x42d0, 0xf), /* MEM_UOPS_RETIRED.SPLIT_STORES */
	INTEL_FLAGS_UEVENT_CONSTRAINT_DATALA_ST(0x82d0, 0xf), /* MEM_UOPS_RETIRED.ALL_STORES */
	INTEL_FLAGS_EVENT_CONSTRAINT_DATALA_LD(0xd1, 0xf),    /* MEM_LOAD_UOPS_RETIRED.* */
	INTEL_FLAGS_EVENT_CONSTRAINT_DATALA_LD(0xd2, 0xf),    /* MEM_LOAD_UOPS_L3_HIT_RETIRED.* */
	INTEL_FLAGS_EVENT_CONSTRAINT_DATALA_LD(0xd3, 0xf),    /* MEM_LOAD_UOPS_L3_MISS_RETIRED.* */
	/* Allow all events as PEBS with no flags */
	INTEL_ALL_EVENT_CONSTRAINT(0, 0xf),
	EVENT_CONSTRAINT_END
};


777 778
struct event_constraint intel_skl_pebs_event_constraints[] = {
	INTEL_FLAGS_UEVENT_CONSTRAINT(0x1c0, 0x2),	/* INST_RETIRED.PREC_DIST */
779 780
	/* INST_RETIRED.PREC_DIST, inv=1, cmask=16 (cycles:ppp). */
	INTEL_FLAGS_EVENT_CONSTRAINT(0x108001c0, 0x2),
781 782
	/* INST_RETIRED.TOTAL_CYCLES_PS (inv=1, cmask=16) (cycles:p). */
	INTEL_FLAGS_EVENT_CONSTRAINT(0x108000c0, 0x0f),
783 784 785 786 787 788 789 790 791 792 793 794
	INTEL_PLD_CONSTRAINT(0x1cd, 0xf),		      /* MEM_TRANS_RETIRED.* */
	INTEL_FLAGS_UEVENT_CONSTRAINT_DATALA_LD(0x11d0, 0xf), /* MEM_INST_RETIRED.STLB_MISS_LOADS */
	INTEL_FLAGS_UEVENT_CONSTRAINT_DATALA_ST(0x12d0, 0xf), /* MEM_INST_RETIRED.STLB_MISS_STORES */
	INTEL_FLAGS_UEVENT_CONSTRAINT_DATALA_LD(0x21d0, 0xf), /* MEM_INST_RETIRED.LOCK_LOADS */
	INTEL_FLAGS_UEVENT_CONSTRAINT_DATALA_ST(0x22d0, 0xf), /* MEM_INST_RETIRED.LOCK_STORES */
	INTEL_FLAGS_UEVENT_CONSTRAINT_DATALA_LD(0x41d0, 0xf), /* MEM_INST_RETIRED.SPLIT_LOADS */
	INTEL_FLAGS_UEVENT_CONSTRAINT_DATALA_ST(0x42d0, 0xf), /* MEM_INST_RETIRED.SPLIT_STORES */
	INTEL_FLAGS_UEVENT_CONSTRAINT_DATALA_LD(0x81d0, 0xf), /* MEM_INST_RETIRED.ALL_LOADS */
	INTEL_FLAGS_UEVENT_CONSTRAINT_DATALA_ST(0x82d0, 0xf), /* MEM_INST_RETIRED.ALL_STORES */
	INTEL_FLAGS_EVENT_CONSTRAINT_DATALA_LD(0xd1, 0xf),    /* MEM_LOAD_RETIRED.* */
	INTEL_FLAGS_EVENT_CONSTRAINT_DATALA_LD(0xd2, 0xf),    /* MEM_LOAD_L3_HIT_RETIRED.* */
	INTEL_FLAGS_EVENT_CONSTRAINT_DATALA_LD(0xd3, 0xf),    /* MEM_LOAD_L3_MISS_RETIRED.* */
795 796
	/* Allow all events as PEBS with no flags */
	INTEL_ALL_EVENT_CONSTRAINT(0, 0xf),
797 798 799
	EVENT_CONSTRAINT_END
};

800
struct event_constraint *intel_pebs_constraints(struct perf_event *event)
801 802 803
{
	struct event_constraint *c;

P
Peter Zijlstra 已提交
804
	if (!event->attr.precise_ip)
805 806 807 808
		return NULL;

	if (x86_pmu.pebs_constraints) {
		for_each_event_constraint(c, x86_pmu.pebs_constraints) {
809 810
			if ((event->hw.config & c->cmask) == c->code) {
				event->hw.flags |= c->flags;
811
				return c;
812
			}
813 814 815 816 817 818
		}
	}

	return &emptyconstraint;
}

819 820 821 822 823 824 825 826 827 828
/*
 * We need the sched_task callback even for per-cpu events when we use
 * the large interrupt threshold, such that we can provide PID and TID
 * to PEBS samples.
 */
static inline bool pebs_needs_sched_cb(struct cpu_hw_events *cpuc)
{
	return cpuc->n_pebs && (cpuc->n_pebs == cpuc->n_large_pebs);
}

829 830 831 832 833 834 835 836
void intel_pmu_pebs_sched_task(struct perf_event_context *ctx, bool sched_in)
{
	struct cpu_hw_events *cpuc = this_cpu_ptr(&cpu_hw_events);

	if (!sched_in && pebs_needs_sched_cb(cpuc))
		intel_pmu_drain_pebs_buffer();
}

837 838 839 840 841 842 843 844 845 846 847 848 849 850 851 852 853 854
static inline void pebs_update_threshold(struct cpu_hw_events *cpuc)
{
	struct debug_store *ds = cpuc->ds;
	u64 threshold;

	if (cpuc->n_pebs == cpuc->n_large_pebs) {
		threshold = ds->pebs_absolute_maximum -
			x86_pmu.max_pebs_events * x86_pmu.pebs_record_size;
	} else {
		threshold = ds->pebs_buffer_base + x86_pmu.pebs_record_size;
	}

	ds->pebs_interrupt_threshold = threshold;
}

static void
pebs_update_state(bool needed_cb, struct cpu_hw_events *cpuc, struct pmu *pmu)
{
855 856 857 858 859 860 861
	/*
	 * Make sure we get updated with the first PEBS
	 * event. It will trigger also during removal, but
	 * that does not hurt:
	 */
	bool update = cpuc->n_pebs == 1;

862 863 864 865 866 867
	if (needed_cb != pebs_needs_sched_cb(cpuc)) {
		if (!needed_cb)
			perf_sched_cb_inc(pmu);
		else
			perf_sched_cb_dec(pmu);

868
		update = true;
869
	}
870 871 872

	if (update)
		pebs_update_threshold(cpuc);
873 874
}

875
void intel_pmu_pebs_add(struct perf_event *event)
876
{
877 878 879 880 881 882 883 884 885
	struct cpu_hw_events *cpuc = this_cpu_ptr(&cpu_hw_events);
	struct hw_perf_event *hwc = &event->hw;
	bool needed_cb = pebs_needs_sched_cb(cpuc);

	cpuc->n_pebs++;
	if (hwc->flags & PERF_X86_EVENT_FREERUNNING)
		cpuc->n_large_pebs++;

	pebs_update_state(needed_cb, cpuc, event->ctx->pmu);
886 887
}

888
void intel_pmu_pebs_enable(struct perf_event *event)
889
{
890
	struct cpu_hw_events *cpuc = this_cpu_ptr(&cpu_hw_events);
891
	struct hw_perf_event *hwc = &event->hw;
892
	struct debug_store *ds = cpuc->ds;
893

894 895
	hwc->config &= ~ARCH_PERFMON_EVENTSEL_INT;

896
	cpuc->pebs_enabled |= 1ULL << hwc->idx;
897 898 899

	if (event->hw.flags & PERF_X86_EVENT_PEBS_LDLAT)
		cpuc->pebs_enabled |= 1ULL << (hwc->idx + 32);
900 901
	else if (event->hw.flags & PERF_X86_EVENT_PEBS_ST)
		cpuc->pebs_enabled |= 1ULL << 63;
902

903
	/*
904 905
	 * Use auto-reload if possible to save a MSR write in the PMI.
	 * This must be done in pmu::start(), because PERF_EVENT_IOC_PERIOD.
906
	 */
907 908 909
	if (hwc->flags & PERF_X86_EVENT_AUTO_RELOAD) {
		ds->pebs_event_reset[hwc->idx] =
			(u64)(-hwc->sample_period) & x86_pmu.cntval_mask;
910 911
	} else {
		ds->pebs_event_reset[hwc->idx] = 0;
912
	}
913 914
}

915
void intel_pmu_pebs_del(struct perf_event *event)
916 917 918 919 920 921 922 923
{
	struct cpu_hw_events *cpuc = this_cpu_ptr(&cpu_hw_events);
	struct hw_perf_event *hwc = &event->hw;
	bool needed_cb = pebs_needs_sched_cb(cpuc);

	cpuc->n_pebs--;
	if (hwc->flags & PERF_X86_EVENT_FREERUNNING)
		cpuc->n_large_pebs--;
924

925
	pebs_update_state(needed_cb, cpuc, event->ctx->pmu);
926 927
}

928
void intel_pmu_pebs_disable(struct perf_event *event)
929
{
930
	struct cpu_hw_events *cpuc = this_cpu_ptr(&cpu_hw_events);
931
	struct hw_perf_event *hwc = &event->hw;
932

933
	if (cpuc->n_pebs == cpuc->n_large_pebs)
934
		intel_pmu_drain_pebs_buffer();
935

936
	cpuc->pebs_enabled &= ~(1ULL << hwc->idx);
937

938
	if (event->hw.flags & PERF_X86_EVENT_PEBS_LDLAT)
939
		cpuc->pebs_enabled &= ~(1ULL << (hwc->idx + 32));
940
	else if (event->hw.flags & PERF_X86_EVENT_PEBS_ST)
941 942
		cpuc->pebs_enabled &= ~(1ULL << 63);

943
	if (cpuc->enabled)
944
		wrmsrl(MSR_IA32_PEBS_ENABLE, cpuc->pebs_enabled);
945 946 947 948

	hwc->config |= ARCH_PERFMON_EVENTSEL_INT;
}

949
void intel_pmu_pebs_enable_all(void)
950
{
951
	struct cpu_hw_events *cpuc = this_cpu_ptr(&cpu_hw_events);
952 953 954 955 956

	if (cpuc->pebs_enabled)
		wrmsrl(MSR_IA32_PEBS_ENABLE, cpuc->pebs_enabled);
}

957
void intel_pmu_pebs_disable_all(void)
958
{
959
	struct cpu_hw_events *cpuc = this_cpu_ptr(&cpu_hw_events);
960 961 962 963 964

	if (cpuc->pebs_enabled)
		wrmsrl(MSR_IA32_PEBS_ENABLE, 0);
}

965 966
static int intel_pmu_pebs_fixup_ip(struct pt_regs *regs)
{
967
	struct cpu_hw_events *cpuc = this_cpu_ptr(&cpu_hw_events);
968 969 970
	unsigned long from = cpuc->lbr_entries[0].from;
	unsigned long old_to, to = cpuc->lbr_entries[0].to;
	unsigned long ip = regs->ip;
971
	int is_64bit = 0;
972
	void *kaddr;
973
	int size;
974

975 976 977 978 979 980
	/*
	 * We don't need to fixup if the PEBS assist is fault like
	 */
	if (!x86_pmu.intel_cap.pebs_trap)
		return 1;

P
Peter Zijlstra 已提交
981 982 983
	/*
	 * No LBR entry, no basic block, no rewinding
	 */
984 985 986
	if (!cpuc->lbr_stack.nr || !from || !to)
		return 0;

P
Peter Zijlstra 已提交
987 988 989 990 991 992 993 994 995 996
	/*
	 * Basic blocks should never cross user/kernel boundaries
	 */
	if (kernel_ip(ip) != kernel_ip(to))
		return 0;

	/*
	 * unsigned math, either ip is before the start (impossible) or
	 * the basic block is larger than 1 page (sanity)
	 */
997
	if ((ip - to) > PEBS_FIXUP_SIZE)
998 999 1000 1001 1002 1003
		return 0;

	/*
	 * We sampled a branch insn, rewind using the LBR stack
	 */
	if (ip == to) {
1004
		set_linear_ip(regs, from);
1005 1006 1007
		return 1;
	}

1008
	size = ip - to;
1009
	if (!kernel_ip(ip)) {
1010
		int bytes;
1011 1012
		u8 *buf = this_cpu_read(insn_buffer);

1013
		/* 'size' must fit our buffer, see above */
1014
		bytes = copy_from_user_nmi(buf, (void __user *)to, size);
1015
		if (bytes != 0)
1016 1017 1018 1019 1020 1021 1022
			return 0;

		kaddr = buf;
	} else {
		kaddr = (void *)to;
	}

1023 1024 1025 1026 1027
	do {
		struct insn insn;

		old_to = to;

1028 1029 1030
#ifdef CONFIG_X86_64
		is_64bit = kernel_ip(to) || !test_thread_flag(TIF_IA32);
#endif
1031
		insn_init(&insn, kaddr, size, is_64bit);
1032
		insn_get_length(&insn);
1033 1034 1035 1036 1037 1038 1039 1040
		/*
		 * Make sure there was not a problem decoding the
		 * instruction and getting the length.  This is
		 * doubly important because we have an infinite
		 * loop if insn.length=0.
		 */
		if (!insn.length)
			break;
1041

1042
		to += insn.length;
1043
		kaddr += insn.length;
1044
		size -= insn.length;
1045 1046 1047
	} while (to < ip);

	if (to == ip) {
1048
		set_linear_ip(regs, old_to);
1049 1050 1051
		return 1;
	}

P
Peter Zijlstra 已提交
1052 1053 1054 1055
	/*
	 * Even though we decoded the basic block, the instruction stream
	 * never matched the given IP, either the TO or the IP got corrupted.
	 */
1056 1057 1058
	return 0;
}

1059
static inline u64 intel_hsw_weight(struct pebs_record_skl *pebs)
1060 1061 1062 1063 1064 1065 1066 1067
{
	if (pebs->tsx_tuning) {
		union hsw_tsx_tuning tsx = { .value = pebs->tsx_tuning };
		return tsx.cycles_last_block;
	}
	return 0;
}

1068
static inline u64 intel_hsw_transaction(struct pebs_record_skl *pebs)
1069 1070 1071 1072 1073 1074 1075 1076 1077
{
	u64 txn = (pebs->tsx_tuning & PEBS_HSW_TSX_FLAGS) >> 32;

	/* For RTM XABORTs also log the abort code from AX */
	if ((txn & PERF_TXN_TRANSACTION) && (pebs->ax & 1))
		txn |= ((pebs->ax >> 24) & 0xff) << PERF_TXN_ABORT_SHIFT;
	return txn;
}

1078 1079 1080 1081
static void setup_pebs_sample_data(struct perf_event *event,
				   struct pt_regs *iregs, void *__pebs,
				   struct perf_sample_data *data,
				   struct pt_regs *regs)
1082
{
1083 1084 1085 1086
#define PERF_X86_EVENT_PEBS_HSW_PREC \
		(PERF_X86_EVENT_PEBS_ST_HSW | \
		 PERF_X86_EVENT_PEBS_LD_HSW | \
		 PERF_X86_EVENT_PEBS_NA_HSW)
1087
	/*
1088 1089
	 * We cast to the biggest pebs_record but are careful not to
	 * unconditionally access the 'extra' entries.
1090
	 */
1091
	struct cpu_hw_events *cpuc = this_cpu_ptr(&cpu_hw_events);
1092
	struct pebs_record_skl *pebs = __pebs;
1093
	u64 sample_type;
1094 1095
	int fll, fst, dsrc;
	int fl = event->hw.flags;
1096

1097 1098 1099
	if (pebs == NULL)
		return;

1100 1101 1102 1103 1104
	sample_type = event->attr.sample_type;
	dsrc = sample_type & PERF_SAMPLE_DATA_SRC;

	fll = fl & PERF_X86_EVENT_PEBS_LDLAT;
	fst = fl & (PERF_X86_EVENT_PEBS_ST | PERF_X86_EVENT_PEBS_HSW_PREC);
1105

1106
	perf_sample_data_init(data, 0, event->hw.last_period);
1107

1108
	data->period = event->hw.last_period;
1109 1110

	/*
1111
	 * Use latency for weight (only avail with PEBS-LL)
1112
	 */
1113
	if (fll && (sample_type & PERF_SAMPLE_WEIGHT))
1114
		data->weight = pebs->lat;
1115 1116 1117 1118 1119 1120 1121 1122 1123 1124 1125 1126

	/*
	 * data.data_src encodes the data source
	 */
	if (dsrc) {
		u64 val = PERF_MEM_NA;
		if (fll)
			val = load_latency_data(pebs->dse);
		else if (fst && (fl & PERF_X86_EVENT_PEBS_HSW_PREC))
			val = precise_datala_hsw(event, pebs->dse);
		else if (fst)
			val = precise_store_data(pebs->dse);
1127
		data->data_src.val = val;
1128 1129
	}

1130
	/*
1131 1132 1133
	 * We use the interrupt regs as a base because the PEBS record does not
	 * contain a full regs set, specifically it seems to lack segment
	 * descriptors, which get used by things like user_mode().
1134
	 *
1135 1136 1137 1138 1139 1140
	 * In the simple case fix up only the IP for PERF_SAMPLE_IP.
	 *
	 * We must however always use BP,SP from iregs for the unwinder to stay
	 * sane; the record BP,SP can point into thin air when the record is
	 * from a previous PMI context or an (I)RET happend between the record
	 * and PMI.
1141
	 */
1142 1143 1144
	*regs = *iregs;
	regs->flags = pebs->flags;
	set_linear_ip(regs, pebs->ip);
1145

1146
	if (sample_type & PERF_SAMPLE_REGS_INTR) {
1147 1148 1149 1150 1151 1152 1153
		regs->ax = pebs->ax;
		regs->bx = pebs->bx;
		regs->cx = pebs->cx;
		regs->dx = pebs->dx;
		regs->si = pebs->si;
		regs->di = pebs->di;

1154 1155 1156 1157 1158 1159 1160 1161 1162 1163 1164 1165 1166 1167
		/*
		 * Per the above; only set BP,SP if we don't need callchains.
		 *
		 * XXX: does this make sense?
		 */
		if (!(sample_type & PERF_SAMPLE_CALLCHAIN)) {
			regs->bp = pebs->bp;
			regs->sp = pebs->sp;
		}

		/*
		 * Preserve PERF_EFLAGS_VM from set_linear_ip().
		 */
		regs->flags = pebs->flags | (regs->flags & PERF_EFLAGS_VM);
1168
#ifndef CONFIG_X86_32
1169 1170 1171 1172 1173 1174 1175 1176
		regs->r8 = pebs->r8;
		regs->r9 = pebs->r9;
		regs->r10 = pebs->r10;
		regs->r11 = pebs->r11;
		regs->r12 = pebs->r12;
		regs->r13 = pebs->r13;
		regs->r14 = pebs->r14;
		regs->r15 = pebs->r15;
1177 1178 1179
#endif
	}

1180
	if (event->attr.precise_ip > 1 && x86_pmu.intel_cap.pebs_format >= 2) {
1181 1182 1183 1184
		regs->ip = pebs->real_ip;
		regs->flags |= PERF_EFLAGS_EXACT;
	} else if (event->attr.precise_ip > 1 && intel_pmu_pebs_fixup_ip(regs))
		regs->flags |= PERF_EFLAGS_EXACT;
1185
	else
1186
		regs->flags &= ~PERF_EFLAGS_EXACT;
1187

1188
	if ((sample_type & PERF_SAMPLE_ADDR) &&
1189
	    x86_pmu.intel_cap.pebs_format >= 1)
1190
		data->addr = pebs->dla;
1191

1192 1193
	if (x86_pmu.intel_cap.pebs_format >= 2) {
		/* Only set the TSX weight when no memory weight. */
1194
		if ((sample_type & PERF_SAMPLE_WEIGHT) && !fll)
1195
			data->weight = intel_hsw_weight(pebs);
1196

1197
		if (sample_type & PERF_SAMPLE_TRANSACTION)
1198
			data->txn = intel_hsw_transaction(pebs);
1199
	}
1200

1201 1202 1203 1204 1205 1206 1207 1208 1209 1210
	/*
	 * v3 supplies an accurate time stamp, so we use that
	 * for the time stamp.
	 *
	 * We can only do this for the default trace clock.
	 */
	if (x86_pmu.intel_cap.pebs_format >= 3 &&
		event->attr.use_clockid == 0)
		data->time = native_sched_clock_from_tsc(pebs->tsc);

1211
	if (has_branch_stack(event))
1212 1213 1214
		data->br_stack = &cpuc->lbr_stack;
}

1215 1216 1217 1218 1219 1220 1221
static inline void *
get_next_pebs_record_by_bit(void *base, void *top, int bit)
{
	struct cpu_hw_events *cpuc = this_cpu_ptr(&cpu_hw_events);
	void *at;
	u64 pebs_status;

1222 1223 1224 1225 1226 1227 1228
	/*
	 * fmt0 does not have a status bitfield (does not use
	 * perf_record_nhm format)
	 */
	if (x86_pmu.intel_cap.pebs_format < 1)
		return base;

1229 1230 1231 1232 1233 1234 1235
	if (base == NULL)
		return NULL;

	for (at = base; at < top; at += x86_pmu.pebs_record_size) {
		struct pebs_record_nhm *p = at;

		if (test_bit(bit, (unsigned long *)&p->status)) {
1236 1237 1238
			/* PEBS v3 has accurate status bits */
			if (x86_pmu.intel_cap.pebs_format >= 3)
				return at;
1239 1240 1241 1242 1243 1244

			if (p->status == (1 << bit))
				return at;

			/* clear non-PEBS bit and re-check */
			pebs_status = p->status & cpuc->pebs_enabled;
1245
			pebs_status &= PEBS_COUNTER_MASK;
1246 1247 1248 1249 1250 1251 1252
			if (pebs_status == (1 << bit))
				return at;
		}
	}
	return NULL;
}

1253
static void __intel_pmu_pebs_event(struct perf_event *event,
1254 1255 1256
				   struct pt_regs *iregs,
				   void *base, void *top,
				   int bit, int count)
1257 1258 1259
{
	struct perf_sample_data data;
	struct pt_regs regs;
1260
	void *at = get_next_pebs_record_by_bit(base, top, bit);
1261

1262 1263
	if (!intel_pmu_save_and_restart(event) &&
	    !(event->hw.flags & PERF_X86_EVENT_AUTO_RELOAD))
1264 1265
		return;

1266 1267 1268 1269 1270 1271
	while (count > 1) {
		setup_pebs_sample_data(event, iregs, at, &data, &regs);
		perf_event_output(event, &data, &regs);
		at += x86_pmu.pebs_record_size;
		at = get_next_pebs_record_by_bit(at, top, bit);
		count--;
1272 1273 1274
	}

	setup_pebs_sample_data(event, iregs, at, &data, &regs);
1275

1276 1277 1278 1279 1280
	/*
	 * All but the last records are processed.
	 * The last one is left to be able to call the overflow handler.
	 */
	if (perf_event_overflow(event, &data, &regs)) {
P
Peter Zijlstra 已提交
1281
		x86_pmu_stop(event, 0);
1282 1283 1284
		return;
	}

1285 1286
}

1287 1288
static void intel_pmu_drain_pebs_core(struct pt_regs *iregs)
{
1289
	struct cpu_hw_events *cpuc = this_cpu_ptr(&cpu_hw_events);
1290 1291 1292 1293 1294
	struct debug_store *ds = cpuc->ds;
	struct perf_event *event = cpuc->events[0]; /* PMC0 only */
	struct pebs_record_core *at, *top;
	int n;

1295
	if (!x86_pmu.pebs_active)
1296 1297 1298 1299 1300
		return;

	at  = (struct pebs_record_core *)(unsigned long)ds->pebs_buffer_base;
	top = (struct pebs_record_core *)(unsigned long)ds->pebs_index;

1301 1302 1303 1304 1305 1306
	/*
	 * Whatever else happens, drain the thing
	 */
	ds->pebs_index = ds->pebs_buffer_base;

	if (!test_bit(0, cpuc->active_mask))
P
Peter Zijlstra 已提交
1307
		return;
1308

1309 1310
	WARN_ON_ONCE(!event);

P
Peter Zijlstra 已提交
1311
	if (!event->attr.precise_ip)
1312 1313
		return;

1314
	n = top - at;
1315 1316
	if (n <= 0)
		return;
1317

1318
	__intel_pmu_pebs_event(event, iregs, at, top, 0, n);
1319 1320
}

1321
static void intel_pmu_drain_pebs_nhm(struct pt_regs *iregs)
1322
{
1323
	struct cpu_hw_events *cpuc = this_cpu_ptr(&cpu_hw_events);
1324
	struct debug_store *ds = cpuc->ds;
1325 1326 1327
	struct perf_event *event;
	void *base, *at, *top;
	short counts[MAX_PEBS_EVENTS] = {};
1328
	short error[MAX_PEBS_EVENTS] = {};
1329
	int bit, i;
1330 1331 1332 1333

	if (!x86_pmu.pebs_active)
		return;

1334
	base = (struct pebs_record_nhm *)(unsigned long)ds->pebs_buffer_base;
1335
	top = (struct pebs_record_nhm *)(unsigned long)ds->pebs_index;
1336 1337 1338

	ds->pebs_index = ds->pebs_buffer_base;

1339
	if (unlikely(base >= top))
1340 1341
		return;

1342
	for (at = base; at < top; at += x86_pmu.pebs_record_size) {
1343
		struct pebs_record_nhm *p = at;
1344
		u64 pebs_status;
1345

1346 1347 1348 1349
		pebs_status = p->status & cpuc->pebs_enabled;
		pebs_status &= (1ULL << x86_pmu.max_pebs_events) - 1;

		/* PEBS v3 has more accurate status bits */
1350
		if (x86_pmu.intel_cap.pebs_format >= 3) {
1351 1352
			for_each_set_bit(bit, (unsigned long *)&pebs_status,
					 x86_pmu.max_pebs_events)
1353 1354 1355 1356 1357
				counts[bit]++;

			continue;
		}

1358 1359 1360 1361 1362 1363 1364 1365 1366 1367 1368 1369
		/*
		 * On some CPUs the PEBS status can be zero when PEBS is
		 * racing with clearing of GLOBAL_STATUS.
		 *
		 * Normally we would drop that record, but in the
		 * case when there is only a single active PEBS event
		 * we can assume it's for that event.
		 */
		if (!pebs_status && cpuc->pebs_enabled &&
			!(cpuc->pebs_enabled & (cpuc->pebs_enabled-1)))
			pebs_status = cpuc->pebs_enabled;

1370
		bit = find_first_bit((unsigned long *)&pebs_status,
1371
					x86_pmu.max_pebs_events);
1372
		if (bit >= x86_pmu.max_pebs_events)
1373
			continue;
1374

1375 1376 1377 1378 1379 1380 1381 1382 1383 1384 1385 1386 1387 1388 1389
		/*
		 * The PEBS hardware does not deal well with the situation
		 * when events happen near to each other and multiple bits
		 * are set. But it should happen rarely.
		 *
		 * If these events include one PEBS and multiple non-PEBS
		 * events, it doesn't impact PEBS record. The record will
		 * be handled normally. (slow path)
		 *
		 * If these events include two or more PEBS events, the
		 * records for the events can be collapsed into a single
		 * one, and it's not possible to reconstruct all events
		 * that caused the PEBS record. It's called collision.
		 * If collision happened, the record will be dropped.
		 */
1390 1391 1392 1393 1394
		if (p->status != (1ULL << bit)) {
			for_each_set_bit(i, (unsigned long *)&pebs_status,
					 x86_pmu.max_pebs_events)
				error[i]++;
			continue;
1395
		}
1396

1397 1398
		counts[bit]++;
	}
1399

1400
	for (bit = 0; bit < x86_pmu.max_pebs_events; bit++) {
1401
		if ((counts[bit] == 0) && (error[bit] == 0))
1402
			continue;
1403

1404
		event = cpuc->events[bit];
1405 1406 1407 1408 1409
		if (WARN_ON_ONCE(!event))
			continue;

		if (WARN_ON_ONCE(!event->attr.precise_ip))
			continue;
1410

1411
		/* log dropped samples number */
1412
		if (error[bit]) {
1413 1414
			perf_log_lost_samples(event, error[bit]);

1415 1416 1417 1418
			if (perf_event_account_interrupt(event))
				x86_pmu_stop(event, 0);
		}

1419 1420 1421 1422
		if (counts[bit]) {
			__intel_pmu_pebs_event(event, iregs, base,
					       top, bit, counts[bit]);
		}
1423 1424 1425 1426 1427 1428 1429
	}
}

/*
 * BTS, PEBS probe and setup
 */

1430
void __init intel_ds_init(void)
1431 1432 1433 1434 1435 1436 1437 1438 1439
{
	/*
	 * No support for 32bit formats
	 */
	if (!boot_cpu_has(X86_FEATURE_DTES64))
		return;

	x86_pmu.bts  = boot_cpu_has(X86_FEATURE_BTS);
	x86_pmu.pebs = boot_cpu_has(X86_FEATURE_PEBS);
1440
	x86_pmu.pebs_buffer_size = PEBS_BUFFER_SIZE;
1441
	if (x86_pmu.pebs) {
1442 1443
		char pebs_type = x86_pmu.intel_cap.pebs_trap ?  '+' : '-';
		int format = x86_pmu.intel_cap.pebs_format;
1444 1445 1446

		switch (format) {
		case 0:
1447
			pr_cont("PEBS fmt0%c, ", pebs_type);
1448
			x86_pmu.pebs_record_size = sizeof(struct pebs_record_core);
1449 1450 1451 1452 1453 1454 1455 1456
			/*
			 * Using >PAGE_SIZE buffers makes the WRMSR to
			 * PERF_GLOBAL_CTRL in intel_pmu_enable_all()
			 * mysteriously hang on Core2.
			 *
			 * As a workaround, we don't do this.
			 */
			x86_pmu.pebs_buffer_size = PAGE_SIZE;
1457 1458 1459 1460
			x86_pmu.drain_pebs = intel_pmu_drain_pebs_core;
			break;

		case 1:
1461
			pr_cont("PEBS fmt1%c, ", pebs_type);
1462 1463 1464 1465
			x86_pmu.pebs_record_size = sizeof(struct pebs_record_nhm);
			x86_pmu.drain_pebs = intel_pmu_drain_pebs_nhm;
			break;

1466 1467 1468
		case 2:
			pr_cont("PEBS fmt2%c, ", pebs_type);
			x86_pmu.pebs_record_size = sizeof(struct pebs_record_hsw);
1469
			x86_pmu.drain_pebs = intel_pmu_drain_pebs_nhm;
1470 1471
			break;

1472 1473 1474 1475 1476
		case 3:
			pr_cont("PEBS fmt3%c, ", pebs_type);
			x86_pmu.pebs_record_size =
						sizeof(struct pebs_record_skl);
			x86_pmu.drain_pebs = intel_pmu_drain_pebs_nhm;
1477
			x86_pmu.free_running_flags |= PERF_SAMPLE_TIME;
1478 1479
			break;

1480
		default:
1481
			pr_cont("no PEBS fmt%d%c, ", format, pebs_type);
1482 1483 1484 1485
			x86_pmu.pebs = 0;
		}
	}
}
1486 1487 1488

void perf_restore_debug_store(void)
{
1489 1490
	struct debug_store *ds = __this_cpu_read(cpu_hw_events.ds);

1491 1492 1493
	if (!x86_pmu.bts && !x86_pmu.pebs)
		return;

1494
	wrmsrl(MSR_IA32_DS_AREA, (unsigned long)ds);
1495
}