ds.c 39.7 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 52 53
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))
#define SNOOP_NONE_MISS (P(SNOOP, NONE) | P(SNOOP, MISS))

54 55
/* Version for Sandy Bridge and later */
static u64 pebs_data_source[] = {
56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73
	P(OP, LOAD) | P(LVL, MISS) | P(LVL, L3) | P(SNOOP, NA),/* 0x00:ukn L3 */
	OP_LH | P(LVL, L1)  | P(SNOOP, NONE),	/* 0x01: L1 local */
	OP_LH | P(LVL, LFB) | P(SNOOP, NONE),	/* 0x02: LFB hit */
	OP_LH | P(LVL, L2)  | P(SNOOP, NONE),	/* 0x03: L2 hit */
	OP_LH | P(LVL, L3)  | P(SNOOP, NONE),	/* 0x04: L3 hit */
	OP_LH | P(LVL, L3)  | P(SNOOP, MISS),	/* 0x05: L3 hit, snoop miss */
	OP_LH | P(LVL, L3)  | P(SNOOP, HIT),	/* 0x06: L3 hit, snoop hit */
	OP_LH | P(LVL, L3)  | P(SNOOP, HITM),	/* 0x07: L3 hit, snoop hitm */
	OP_LH | P(LVL, REM_CCE1) | P(SNOOP, HIT),  /* 0x08: L3 miss snoop hit */
	OP_LH | P(LVL, REM_CCE1) | P(SNOOP, HITM), /* 0x09: L3 miss snoop hitm*/
	OP_LH | P(LVL, LOC_RAM)  | P(SNOOP, HIT),  /* 0x0a: L3 miss, shared */
	OP_LH | P(LVL, REM_RAM1) | P(SNOOP, HIT),  /* 0x0b: L3 miss, shared */
	OP_LH | P(LVL, LOC_RAM)  | SNOOP_NONE_MISS,/* 0x0c: L3 miss, excl */
	OP_LH | P(LVL, REM_RAM1) | SNOOP_NONE_MISS,/* 0x0d: L3 miss, excl */
	OP_LH | P(LVL, IO)  | P(SNOOP, NONE), /* 0x0e: I/O */
	OP_LH | P(LVL, UNC) | P(SNOOP, NONE), /* 0x0f: uncached */
};

74 75 76 77 78 79 80 81
/* Patch up minor differences in the bits */
void __init intel_pmu_pebs_data_source_nhm(void)
{
	pebs_data_source[0x05] = OP_LH | P(LVL, L3)  | P(SNOOP, HIT);
	pebs_data_source[0x06] = OP_LH | P(LVL, L3)  | P(SNOOP, HITM);
	pebs_data_source[0x07] = OP_LH | P(LVL, L3)  | P(SNOOP, HITM);
}

82 83 84 85 86 87 88 89 90 91 92 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
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;
}

120
static u64 precise_datala_hsw(struct perf_event *event, u64 status)
121 122 123
{
	union perf_mem_data_src dse;

124 125 126 127 128 129
	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;
130 131 132 133 134 135 136 137 138

	/*
	 * 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
	 */
139 140 141 142 143 144
	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;
	}
145 146 147
	return dse.val;
}

148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188
static u64 load_latency_data(u64 status)
{
	union intel_x86_pebs_dse dse;
	u64 val;
	int model = boot_cpu_data.x86_model;
	int fam = boot_cpu_data.x86;

	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
	 */
	if (fam == 0x6 && (model == 26 || model == 30
	    || model == 31 || model == 46)) {
		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;
}

189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205
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;
};

206 207 208 209
/*
 * Same as pebs_record_nhm, with two additional fields.
 */
struct pebs_record_hsw {
210 211 212 213 214 215
	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
	u64 real_ip, tsx_tuning;
217 218 219 220 221 222 223 224 225 226 227 228 229 230 231
};

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;
232 233
};

234 235
#define PEBS_HSW_TSX_FLAGS	0xff00000000ULL

236 237 238 239 240 241 242 243 244 245 246 247 248
/* 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;
};

249
void init_debug_store_on_cpu(int cpu)
250 251 252 253 254 255 256 257 258 259 260
{
	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));
}

261
void fini_debug_store_on_cpu(int cpu)
262 263 264 265 266 267 268
{
	if (!per_cpu(cpu_hw_events, cpu).ds)
		return;

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

269 270
static DEFINE_PER_CPU(void *, insn_buffer);

271 272 273
static int alloc_pebs_buffer(int cpu)
{
	struct debug_store *ds = per_cpu(cpu_hw_events, cpu).ds;
274
	int node = cpu_to_node(cpu);
275
	int max;
276
	void *buffer, *ibuffer;
277 278 279 280

	if (!x86_pmu.pebs)
		return 0;

281
	buffer = kzalloc_node(x86_pmu.pebs_buffer_size, GFP_KERNEL, node);
282 283 284
	if (unlikely(!buffer))
		return -ENOMEM;

285 286 287 288 289 290 291 292 293 294 295 296 297
	/*
	 * 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;
	}

298
	max = x86_pmu.pebs_buffer_size / x86_pmu.pebs_record_size;
299 300 301 302 303 304 305 306 307

	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;
}

308 309 310 311 312 313 314
static void release_pebs_buffer(int cpu)
{
	struct debug_store *ds = per_cpu(cpu_hw_events, cpu).ds;

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

315 316 317
	kfree(per_cpu(insn_buffer, cpu));
	per_cpu(insn_buffer, cpu) = NULL;

318 319 320 321
	kfree((void *)(unsigned long)ds->pebs_buffer_base);
	ds->pebs_buffer_base = 0;
}

322 323 324
static int alloc_bts_buffer(int cpu)
{
	struct debug_store *ds = per_cpu(cpu_hw_events, cpu).ds;
325
	int node = cpu_to_node(cpu);
326 327 328 329 330 331
	int max, thresh;
	void *buffer;

	if (!x86_pmu.bts)
		return 0;

332 333 334
	buffer = kzalloc_node(BTS_BUFFER_SIZE, GFP_KERNEL | __GFP_NOWARN, node);
	if (unlikely(!buffer)) {
		WARN_ONCE(1, "%s: BTS buffer allocation failure\n", __func__);
335
		return -ENOMEM;
336
	}
337 338 339 340 341 342 343 344 345 346 347 348 349 350

	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;
}

351 352 353 354 355 356 357 358 359 360 361
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;
}

362 363
static int alloc_ds_buffer(int cpu)
{
364
	int node = cpu_to_node(cpu);
365 366
	struct debug_store *ds;

367
	ds = kzalloc_node(sizeof(*ds), GFP_KERNEL, node);
368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386
	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);
}

387
void release_ds_buffers(void)
388 389 390 391 392 393 394 395 396 397 398
{
	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) {
399 400
		release_pebs_buffer(cpu);
		release_bts_buffer(cpu);
401
		release_ds_buffer(cpu);
402 403 404 405
	}
	put_online_cpus();
}

406
void reserve_ds_buffers(void)
407
{
408 409 410 411 412
	int bts_err = 0, pebs_err = 0;
	int cpu;

	x86_pmu.bts_active = 0;
	x86_pmu.pebs_active = 0;
413 414

	if (!x86_pmu.bts && !x86_pmu.pebs)
415
		return;
416

417 418 419 420 421 422
	if (!x86_pmu.bts)
		bts_err = 1;

	if (!x86_pmu.pebs)
		pebs_err = 1;

423 424 425
	get_online_cpus();

	for_each_possible_cpu(cpu) {
426 427 428 429
		if (alloc_ds_buffer(cpu)) {
			bts_err = 1;
			pebs_err = 1;
		}
430

431 432 433 434 435
		if (!bts_err && alloc_bts_buffer(cpu))
			bts_err = 1;

		if (!pebs_err && alloc_pebs_buffer(cpu))
			pebs_err = 1;
436

437
		if (bts_err && pebs_err)
438
			break;
439 440 441 442 443 444
	}

	if (bts_err) {
		for_each_possible_cpu(cpu)
			release_bts_buffer(cpu);
	}
445

446 447 448
	if (pebs_err) {
		for_each_possible_cpu(cpu)
			release_pebs_buffer(cpu);
449 450
	}

451 452 453 454 455 456 457 458 459 460
	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;

461 462 463 464 465 466 467 468 469 470 471
		for_each_online_cpu(cpu)
			init_debug_store_on_cpu(cpu);
	}

	put_online_cpus();
}

/*
 * BTS
 */

472
struct event_constraint bts_constraint =
473
	EVENT_CONSTRAINT(0, 1ULL << INTEL_PMC_IDX_FIXED_BTS, 0);
474

475
void intel_pmu_enable_bts(u64 config)
476 477 478 479 480
{
	unsigned long debugctlmsr;

	debugctlmsr = get_debugctlmsr();

481 482
	debugctlmsr |= DEBUGCTLMSR_TR;
	debugctlmsr |= DEBUGCTLMSR_BTS;
483 484
	if (config & ARCH_PERFMON_EVENTSEL_INT)
		debugctlmsr |= DEBUGCTLMSR_BTINT;
485 486

	if (!(config & ARCH_PERFMON_EVENTSEL_OS))
487
		debugctlmsr |= DEBUGCTLMSR_BTS_OFF_OS;
488 489

	if (!(config & ARCH_PERFMON_EVENTSEL_USR))
490
		debugctlmsr |= DEBUGCTLMSR_BTS_OFF_USR;
491 492 493 494

	update_debugctlmsr(debugctlmsr);
}

495
void intel_pmu_disable_bts(void)
496
{
497
	struct cpu_hw_events *cpuc = this_cpu_ptr(&cpu_hw_events);
498 499 500 501 502 503 504 505
	unsigned long debugctlmsr;

	if (!cpuc->ds)
		return;

	debugctlmsr = get_debugctlmsr();

	debugctlmsr &=
506 507
		~(DEBUGCTLMSR_TR | DEBUGCTLMSR_BTS | DEBUGCTLMSR_BTINT |
		  DEBUGCTLMSR_BTS_OFF_OS | DEBUGCTLMSR_BTS_OFF_USR);
508 509 510 511

	update_debugctlmsr(debugctlmsr);
}

512
int intel_pmu_drain_bts_buffer(void)
513
{
514
	struct cpu_hw_events *cpuc = this_cpu_ptr(&cpu_hw_events);
515 516 517 518 519 520
	struct debug_store *ds = cpuc->ds;
	struct bts_record {
		u64	from;
		u64	to;
		u64	flags;
	};
521
	struct perf_event *event = cpuc->events[INTEL_PMC_IDX_FIXED_BTS];
522
	struct bts_record *at, *base, *top;
523 524 525
	struct perf_output_handle handle;
	struct perf_event_header header;
	struct perf_sample_data data;
526
	unsigned long skip = 0;
527 528 529
	struct pt_regs regs;

	if (!event)
530
		return 0;
531

532
	if (!x86_pmu.bts_active)
533
		return 0;
534

535 536
	base = (struct bts_record *)(unsigned long)ds->bts_buffer_base;
	top  = (struct bts_record *)(unsigned long)ds->bts_index;
537

538
	if (top <= base)
539
		return 0;
540

541 542
	memset(&regs, 0, sizeof(regs));

543 544
	ds->bts_index = ds->bts_buffer_base;

545
	perf_sample_data_init(&data, 0, event->hw.last_period);
546

547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567
	/*
	 * 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++;
	}

568 569 570 571 572
	/*
	 * 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
573
	rcu_read_lock();
574 575
	perf_prepare_sample(&header, &data, event, &regs);

576 577
	if (perf_output_begin(&handle, event, header.size *
			      (top - base - skip)))
P
Peter Zijlstra 已提交
578
		goto unlock;
579

580 581 582 583 584 585
	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;

586 587 588 589 590 591 592 593 594 595 596
		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
597 598
unlock:
	rcu_read_unlock();
599
	return 1;
600 601
}

602 603 604 605 606 607 608 609 610 611 612 613 614
static inline void intel_pmu_drain_pebs_buffer(void)
{
	struct pt_regs regs;

	x86_pmu.drain_pebs(&regs);
}

void intel_pmu_pebs_sched_task(struct perf_event_context *ctx, bool sched_in)
{
	if (!sched_in)
		intel_pmu_drain_pebs_buffer();
}

615 616 617
/*
 * PEBS
 */
618
struct event_constraint intel_core2_pebs_event_constraints[] = {
619 620 621 622 623
	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.* */
624 625
	/* INST_RETIRED.ANY_P, inv=1, cmask=16 (cycles:p). */
	INTEL_FLAGS_EVENT_CONSTRAINT(0x108000c0, 0x01),
626 627 628
	EVENT_CONSTRAINT_END
};

629
struct event_constraint intel_atom_pebs_event_constraints[] = {
630 631 632
	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.* */
633 634
	/* INST_RETIRED.ANY_P, inv=1, cmask=16 (cycles:p). */
	INTEL_FLAGS_EVENT_CONSTRAINT(0x108000c0, 0x01),
635 636
	/* Allow all events as PEBS with no flags */
	INTEL_ALL_EVENT_CONSTRAINT(0, 0x1),
637 638 639
	EVENT_CONSTRAINT_END
};

640
struct event_constraint intel_slm_pebs_event_constraints[] = {
641 642
	/* INST_RETIRED.ANY_P, inv=1, cmask=16 (cycles:p). */
	INTEL_FLAGS_EVENT_CONSTRAINT(0x108000c0, 0x1),
643 644
	/* Allow all events as PEBS with no flags */
	INTEL_ALL_EVENT_CONSTRAINT(0, 0x1),
645 646 647
	EVENT_CONSTRAINT_END
};

648 649 650 651 652 653
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
};

654 655 656 657 658 659
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
};

660
struct event_constraint intel_nehalem_pebs_event_constraints[] = {
661
	INTEL_PLD_CONSTRAINT(0x100b, 0xf),      /* MEM_INST_RETIRED.* */
662 663 664
	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 */
665
	INTEL_EVENT_CONSTRAINT(0xc2, 0xf),    /* UOPS_RETIRED.* */
666 667 668 669 670 671
	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.* */
672 673
	/* INST_RETIRED.ANY_P, inv=1, cmask=16 (cycles:p). */
	INTEL_FLAGS_EVENT_CONSTRAINT(0x108000c0, 0x0f),
674 675 676
	EVENT_CONSTRAINT_END
};

677
struct event_constraint intel_westmere_pebs_event_constraints[] = {
678
	INTEL_PLD_CONSTRAINT(0x100b, 0xf),      /* MEM_INST_RETIRED.* */
679 680 681
	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.* */
682
	INTEL_EVENT_CONSTRAINT(0xc2, 0xf),    /* UOPS_RETIRED.* */
683 684 685 686 687 688
	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.* */
689 690
	/* INST_RETIRED.ANY_P, inv=1, cmask=16 (cycles:p). */
	INTEL_FLAGS_EVENT_CONSTRAINT(0x108000c0, 0x0f),
691 692 693
	EVENT_CONSTRAINT_END
};

694
struct event_constraint intel_snb_pebs_event_constraints[] = {
695
	INTEL_FLAGS_UEVENT_CONSTRAINT(0x01c0, 0x2), /* INST_RETIRED.PRECDIST */
696
	INTEL_PLD_CONSTRAINT(0x01cd, 0x8),    /* MEM_TRANS_RETIRED.LAT_ABOVE_THR */
697
	INTEL_PST_CONSTRAINT(0x02cd, 0x8),    /* MEM_TRANS_RETIRED.PRECISE_STORES */
698 699
	/* UOPS_RETIRED.ALL, inv=1, cmask=16 (cycles:p). */
	INTEL_FLAGS_EVENT_CONSTRAINT(0x108001c2, 0xf),
700 701 702 703
        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.* */
704 705
	/* Allow all events as PEBS with no flags */
	INTEL_ALL_EVENT_CONSTRAINT(0, 0xf),
706 707 708
	EVENT_CONSTRAINT_END
};

709
struct event_constraint intel_ivb_pebs_event_constraints[] = {
710
        INTEL_FLAGS_UEVENT_CONSTRAINT(0x01c0, 0x2), /* INST_RETIRED.PRECDIST */
711
        INTEL_PLD_CONSTRAINT(0x01cd, 0x8),    /* MEM_TRANS_RETIRED.LAT_ABOVE_THR */
712
	INTEL_PST_CONSTRAINT(0x02cd, 0x8),    /* MEM_TRANS_RETIRED.PRECISE_STORES */
713 714
	/* UOPS_RETIRED.ALL, inv=1, cmask=16 (cycles:p). */
	INTEL_FLAGS_EVENT_CONSTRAINT(0x108001c2, 0xf),
715 716
	/* INST_RETIRED.PREC_DIST, inv=1, cmask=16 (cycles:ppp). */
	INTEL_FLAGS_EVENT_CONSTRAINT(0x108001c0, 0x2),
717 718 719 720
	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.* */
721 722
	/* Allow all events as PEBS with no flags */
	INTEL_ALL_EVENT_CONSTRAINT(0, 0xf),
723 724 725
        EVENT_CONSTRAINT_END
};

726
struct event_constraint intel_hsw_pebs_event_constraints[] = {
727
	INTEL_FLAGS_UEVENT_CONSTRAINT(0x01c0, 0x2), /* INST_RETIRED.PRECDIST */
728 729 730
	INTEL_PLD_CONSTRAINT(0x01cd, 0xf),    /* MEM_TRANS_RETIRED.* */
	/* UOPS_RETIRED.ALL, inv=1, cmask=16 (cycles:p). */
	INTEL_FLAGS_EVENT_CONSTRAINT(0x108001c2, 0xf),
731 732
	/* INST_RETIRED.PREC_DIST, inv=1, cmask=16 (cycles:ppp). */
	INTEL_FLAGS_EVENT_CONSTRAINT(0x108001c0, 0x2),
733
	INTEL_FLAGS_UEVENT_CONSTRAINT_DATALA_NA(0x01c2, 0xf), /* UOPS_RETIRED.ALL */
734 735 736 737 738 739 740 741 742 743
	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.* */
744 745 746 747 748
	/* Allow all events as PEBS with no flags */
	INTEL_ALL_EVENT_CONSTRAINT(0, 0xf),
	EVENT_CONSTRAINT_END
};

749 750 751 752 753 754 755 756 757 758 759 760 761 762 763 764 765 766 767 768 769 770 771 772
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
};


773 774
struct event_constraint intel_skl_pebs_event_constraints[] = {
	INTEL_FLAGS_UEVENT_CONSTRAINT(0x1c0, 0x2),	/* INST_RETIRED.PREC_DIST */
775 776
	/* INST_RETIRED.PREC_DIST, inv=1, cmask=16 (cycles:ppp). */
	INTEL_FLAGS_EVENT_CONSTRAINT(0x108001c0, 0x2),
777 778
	/* INST_RETIRED.TOTAL_CYCLES_PS (inv=1, cmask=16) (cycles:p). */
	INTEL_FLAGS_EVENT_CONSTRAINT(0x108000c0, 0x0f),
779 780 781 782 783 784 785 786 787 788 789 790
	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.* */
791 792
	/* Allow all events as PEBS with no flags */
	INTEL_ALL_EVENT_CONSTRAINT(0, 0xf),
793 794 795
	EVENT_CONSTRAINT_END
};

796
struct event_constraint *intel_pebs_constraints(struct perf_event *event)
797 798 799
{
	struct event_constraint *c;

P
Peter Zijlstra 已提交
800
	if (!event->attr.precise_ip)
801 802 803 804
		return NULL;

	if (x86_pmu.pebs_constraints) {
		for_each_event_constraint(c, x86_pmu.pebs_constraints) {
805 806
			if ((event->hw.config & c->cmask) == c->code) {
				event->hw.flags |= c->flags;
807
				return c;
808
			}
809 810 811 812 813 814
		}
	}

	return &emptyconstraint;
}

815 816 817 818 819 820 821 822 823 824 825 826 827 828 829 830 831 832 833 834 835 836 837 838 839 840 841 842
/*
 * 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);
}

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)
{
843 844 845 846 847 848 849
	/*
	 * 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;

850 851 852 853 854 855
	if (needed_cb != pebs_needs_sched_cb(cpuc)) {
		if (!needed_cb)
			perf_sched_cb_inc(pmu);
		else
			perf_sched_cb_dec(pmu);

856
		update = true;
857
	}
858 859 860

	if (update)
		pebs_update_threshold(cpuc);
861 862
}

863
void intel_pmu_pebs_add(struct perf_event *event)
864
{
865 866 867 868 869 870 871 872 873
	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);
874 875
}

876
void intel_pmu_pebs_enable(struct perf_event *event)
877
{
878
	struct cpu_hw_events *cpuc = this_cpu_ptr(&cpu_hw_events);
879
	struct hw_perf_event *hwc = &event->hw;
880
	struct debug_store *ds = cpuc->ds;
881

882 883
	hwc->config &= ~ARCH_PERFMON_EVENTSEL_INT;

884
	cpuc->pebs_enabled |= 1ULL << hwc->idx;
885 886 887

	if (event->hw.flags & PERF_X86_EVENT_PEBS_LDLAT)
		cpuc->pebs_enabled |= 1ULL << (hwc->idx + 32);
888 889
	else if (event->hw.flags & PERF_X86_EVENT_PEBS_ST)
		cpuc->pebs_enabled |= 1ULL << 63;
890

891
	/*
892 893
	 * 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.
894
	 */
895 896 897
	if (hwc->flags & PERF_X86_EVENT_AUTO_RELOAD) {
		ds->pebs_event_reset[hwc->idx] =
			(u64)(-hwc->sample_period) & x86_pmu.cntval_mask;
898 899
	} else {
		ds->pebs_event_reset[hwc->idx] = 0;
900
	}
901 902
}

903
void intel_pmu_pebs_del(struct perf_event *event)
904 905 906 907 908 909 910 911
{
	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--;
912

913
	pebs_update_state(needed_cb, cpuc, event->ctx->pmu);
914 915
}

916
void intel_pmu_pebs_disable(struct perf_event *event)
917
{
918
	struct cpu_hw_events *cpuc = this_cpu_ptr(&cpu_hw_events);
919
	struct hw_perf_event *hwc = &event->hw;
920

921
	if (cpuc->n_pebs == cpuc->n_large_pebs)
922
		intel_pmu_drain_pebs_buffer();
923

924
	cpuc->pebs_enabled &= ~(1ULL << hwc->idx);
925

926
	if (event->hw.flags & PERF_X86_EVENT_PEBS_LDLAT)
927
		cpuc->pebs_enabled &= ~(1ULL << (hwc->idx + 32));
928
	else if (event->hw.flags & PERF_X86_EVENT_PEBS_ST)
929 930
		cpuc->pebs_enabled &= ~(1ULL << 63);

931
	if (cpuc->enabled)
932
		wrmsrl(MSR_IA32_PEBS_ENABLE, cpuc->pebs_enabled);
933 934 935 936

	hwc->config |= ARCH_PERFMON_EVENTSEL_INT;
}

937
void intel_pmu_pebs_enable_all(void)
938
{
939
	struct cpu_hw_events *cpuc = this_cpu_ptr(&cpu_hw_events);
940 941 942 943 944

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

945
void intel_pmu_pebs_disable_all(void)
946
{
947
	struct cpu_hw_events *cpuc = this_cpu_ptr(&cpu_hw_events);
948 949 950 951 952

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

953 954
static int intel_pmu_pebs_fixup_ip(struct pt_regs *regs)
{
955
	struct cpu_hw_events *cpuc = this_cpu_ptr(&cpu_hw_events);
956 957 958
	unsigned long from = cpuc->lbr_entries[0].from;
	unsigned long old_to, to = cpuc->lbr_entries[0].to;
	unsigned long ip = regs->ip;
959
	int is_64bit = 0;
960
	void *kaddr;
961
	int size;
962

963 964 965 966 967 968
	/*
	 * 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
969 970 971
	/*
	 * No LBR entry, no basic block, no rewinding
	 */
972 973 974
	if (!cpuc->lbr_stack.nr || !from || !to)
		return 0;

P
Peter Zijlstra 已提交
975 976 977 978 979 980 981 982 983 984
	/*
	 * 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)
	 */
985
	if ((ip - to) > PEBS_FIXUP_SIZE)
986 987 988 989 990 991
		return 0;

	/*
	 * We sampled a branch insn, rewind using the LBR stack
	 */
	if (ip == to) {
992
		set_linear_ip(regs, from);
993 994 995
		return 1;
	}

996
	size = ip - to;
997
	if (!kernel_ip(ip)) {
998
		int bytes;
999 1000
		u8 *buf = this_cpu_read(insn_buffer);

1001
		/* 'size' must fit our buffer, see above */
1002
		bytes = copy_from_user_nmi(buf, (void __user *)to, size);
1003
		if (bytes != 0)
1004 1005 1006 1007 1008 1009 1010
			return 0;

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

1011 1012 1013 1014 1015
	do {
		struct insn insn;

		old_to = to;

1016 1017 1018
#ifdef CONFIG_X86_64
		is_64bit = kernel_ip(to) || !test_thread_flag(TIF_IA32);
#endif
1019
		insn_init(&insn, kaddr, size, is_64bit);
1020
		insn_get_length(&insn);
1021 1022 1023 1024 1025 1026 1027 1028
		/*
		 * 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;
1029

1030
		to += insn.length;
1031
		kaddr += insn.length;
1032
		size -= insn.length;
1033 1034 1035
	} while (to < ip);

	if (to == ip) {
1036
		set_linear_ip(regs, old_to);
1037 1038 1039
		return 1;
	}

P
Peter Zijlstra 已提交
1040 1041 1042 1043
	/*
	 * Even though we decoded the basic block, the instruction stream
	 * never matched the given IP, either the TO or the IP got corrupted.
	 */
1044 1045 1046
	return 0;
}

1047
static inline u64 intel_hsw_weight(struct pebs_record_skl *pebs)
1048 1049 1050 1051 1052 1053 1054 1055
{
	if (pebs->tsx_tuning) {
		union hsw_tsx_tuning tsx = { .value = pebs->tsx_tuning };
		return tsx.cycles_last_block;
	}
	return 0;
}

1056
static inline u64 intel_hsw_transaction(struct pebs_record_skl *pebs)
1057 1058 1059 1060 1061 1062 1063 1064 1065
{
	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;
}

1066 1067 1068 1069
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)
1070
{
1071 1072 1073 1074
#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)
1075
	/*
1076 1077
	 * We cast to the biggest pebs_record but are careful not to
	 * unconditionally access the 'extra' entries.
1078
	 */
1079
	struct cpu_hw_events *cpuc = this_cpu_ptr(&cpu_hw_events);
1080
	struct pebs_record_skl *pebs = __pebs;
1081
	u64 sample_type;
1082 1083
	int fll, fst, dsrc;
	int fl = event->hw.flags;
1084

1085 1086 1087
	if (pebs == NULL)
		return;

1088 1089 1090 1091 1092
	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);
1093

1094
	perf_sample_data_init(data, 0, event->hw.last_period);
1095

1096
	data->period = event->hw.last_period;
1097 1098

	/*
1099
	 * Use latency for weight (only avail with PEBS-LL)
1100
	 */
1101
	if (fll && (sample_type & PERF_SAMPLE_WEIGHT))
1102
		data->weight = pebs->lat;
1103 1104 1105 1106 1107 1108 1109 1110 1111 1112 1113 1114

	/*
	 * 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);
1115
		data->data_src.val = val;
1116 1117
	}

1118
	/*
1119 1120 1121
	 * 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().
1122
	 *
1123 1124 1125 1126 1127 1128
	 * 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.
1129
	 */
1130 1131 1132
	*regs = *iregs;
	regs->flags = pebs->flags;
	set_linear_ip(regs, pebs->ip);
1133

1134
	if (sample_type & PERF_SAMPLE_REGS_INTR) {
1135 1136 1137 1138 1139 1140 1141
		regs->ax = pebs->ax;
		regs->bx = pebs->bx;
		regs->cx = pebs->cx;
		regs->dx = pebs->dx;
		regs->si = pebs->si;
		regs->di = pebs->di;

1142 1143 1144 1145 1146 1147 1148 1149 1150 1151 1152 1153 1154 1155
		/*
		 * 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);
1156
#ifndef CONFIG_X86_32
1157 1158 1159 1160 1161 1162 1163 1164
		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;
1165 1166 1167
#endif
	}

1168
	if (event->attr.precise_ip > 1 && x86_pmu.intel_cap.pebs_format >= 2) {
1169 1170 1171 1172
		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;
1173
	else
1174
		regs->flags &= ~PERF_EFLAGS_EXACT;
1175

1176
	if ((sample_type & PERF_SAMPLE_ADDR) &&
1177
	    x86_pmu.intel_cap.pebs_format >= 1)
1178
		data->addr = pebs->dla;
1179

1180 1181
	if (x86_pmu.intel_cap.pebs_format >= 2) {
		/* Only set the TSX weight when no memory weight. */
1182
		if ((sample_type & PERF_SAMPLE_WEIGHT) && !fll)
1183
			data->weight = intel_hsw_weight(pebs);
1184

1185
		if (sample_type & PERF_SAMPLE_TRANSACTION)
1186
			data->txn = intel_hsw_transaction(pebs);
1187
	}
1188

1189 1190 1191 1192 1193 1194 1195 1196 1197 1198
	/*
	 * 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);

1199
	if (has_branch_stack(event))
1200 1201 1202
		data->br_stack = &cpuc->lbr_stack;
}

1203 1204 1205 1206 1207 1208 1209
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;

1210 1211 1212 1213 1214 1215 1216
	/*
	 * fmt0 does not have a status bitfield (does not use
	 * perf_record_nhm format)
	 */
	if (x86_pmu.intel_cap.pebs_format < 1)
		return base;

1217 1218 1219 1220 1221 1222 1223
	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)) {
1224 1225 1226
			/* PEBS v3 has accurate status bits */
			if (x86_pmu.intel_cap.pebs_format >= 3)
				return at;
1227 1228 1229 1230 1231 1232

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

			/* clear non-PEBS bit and re-check */
			pebs_status = p->status & cpuc->pebs_enabled;
1233
			pebs_status &= PEBS_COUNTER_MASK;
1234 1235 1236 1237 1238 1239 1240
			if (pebs_status == (1 << bit))
				return at;
		}
	}
	return NULL;
}

1241
static void __intel_pmu_pebs_event(struct perf_event *event,
1242 1243 1244
				   struct pt_regs *iregs,
				   void *base, void *top,
				   int bit, int count)
1245 1246 1247
{
	struct perf_sample_data data;
	struct pt_regs regs;
1248
	void *at = get_next_pebs_record_by_bit(base, top, bit);
1249

1250 1251
	if (!intel_pmu_save_and_restart(event) &&
	    !(event->hw.flags & PERF_X86_EVENT_AUTO_RELOAD))
1252 1253
		return;

1254 1255 1256 1257 1258 1259
	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--;
1260 1261 1262
	}

	setup_pebs_sample_data(event, iregs, at, &data, &regs);
1263

1264 1265 1266 1267 1268
	/*
	 * 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
1269
		x86_pmu_stop(event, 0);
1270 1271 1272
		return;
	}

1273 1274
}

1275 1276
static void intel_pmu_drain_pebs_core(struct pt_regs *iregs)
{
1277
	struct cpu_hw_events *cpuc = this_cpu_ptr(&cpu_hw_events);
1278 1279 1280 1281 1282
	struct debug_store *ds = cpuc->ds;
	struct perf_event *event = cpuc->events[0]; /* PMC0 only */
	struct pebs_record_core *at, *top;
	int n;

1283
	if (!x86_pmu.pebs_active)
1284 1285 1286 1287 1288
		return;

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

1289 1290 1291 1292 1293 1294
	/*
	 * 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
1295
		return;
1296

1297 1298
	WARN_ON_ONCE(!event);

P
Peter Zijlstra 已提交
1299
	if (!event->attr.precise_ip)
1300 1301
		return;

1302
	n = top - at;
1303 1304
	if (n <= 0)
		return;
1305

1306
	__intel_pmu_pebs_event(event, iregs, at, top, 0, n);
1307 1308
}

1309
static void intel_pmu_drain_pebs_nhm(struct pt_regs *iregs)
1310
{
1311
	struct cpu_hw_events *cpuc = this_cpu_ptr(&cpu_hw_events);
1312
	struct debug_store *ds = cpuc->ds;
1313 1314 1315
	struct perf_event *event;
	void *base, *at, *top;
	short counts[MAX_PEBS_EVENTS] = {};
1316
	short error[MAX_PEBS_EVENTS] = {};
1317
	int bit, i;
1318 1319 1320 1321

	if (!x86_pmu.pebs_active)
		return;

1322
	base = (struct pebs_record_nhm *)(unsigned long)ds->pebs_buffer_base;
1323
	top = (struct pebs_record_nhm *)(unsigned long)ds->pebs_index;
1324 1325 1326

	ds->pebs_index = ds->pebs_buffer_base;

1327
	if (unlikely(base >= top))
1328 1329
		return;

1330
	for (at = base; at < top; at += x86_pmu.pebs_record_size) {
1331
		struct pebs_record_nhm *p = at;
1332
		u64 pebs_status;
1333

1334 1335 1336 1337
		pebs_status = p->status & cpuc->pebs_enabled;
		pebs_status &= (1ULL << x86_pmu.max_pebs_events) - 1;

		/* PEBS v3 has more accurate status bits */
1338
		if (x86_pmu.intel_cap.pebs_format >= 3) {
1339 1340
			for_each_set_bit(bit, (unsigned long *)&pebs_status,
					 x86_pmu.max_pebs_events)
1341 1342 1343 1344 1345
				counts[bit]++;

			continue;
		}

1346 1347 1348 1349 1350 1351 1352 1353 1354 1355 1356 1357
		/*
		 * 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;

1358
		bit = find_first_bit((unsigned long *)&pebs_status,
1359
					x86_pmu.max_pebs_events);
1360
		if (bit >= x86_pmu.max_pebs_events)
1361
			continue;
1362

1363 1364 1365 1366 1367 1368 1369 1370 1371 1372 1373 1374 1375 1376 1377
		/*
		 * 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.
		 */
1378 1379 1380 1381 1382
		if (p->status != (1ULL << bit)) {
			for_each_set_bit(i, (unsigned long *)&pebs_status,
					 x86_pmu.max_pebs_events)
				error[i]++;
			continue;
1383
		}
1384

1385 1386
		counts[bit]++;
	}
1387

1388
	for (bit = 0; bit < x86_pmu.max_pebs_events; bit++) {
1389
		if ((counts[bit] == 0) && (error[bit] == 0))
1390
			continue;
1391

1392
		event = cpuc->events[bit];
1393 1394 1395 1396 1397
		if (WARN_ON_ONCE(!event))
			continue;

		if (WARN_ON_ONCE(!event->attr.precise_ip))
			continue;
1398

1399
		/* log dropped samples number */
1400
		if (error[bit]) {
1401 1402
			perf_log_lost_samples(event, error[bit]);

1403 1404 1405 1406
			if (perf_event_account_interrupt(event))
				x86_pmu_stop(event, 0);
		}

1407 1408 1409 1410
		if (counts[bit]) {
			__intel_pmu_pebs_event(event, iregs, base,
					       top, bit, counts[bit]);
		}
1411 1412 1413 1414 1415 1416 1417
	}
}

/*
 * BTS, PEBS probe and setup
 */

1418
void __init intel_ds_init(void)
1419 1420 1421 1422 1423 1424 1425 1426 1427
{
	/*
	 * 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);
1428
	x86_pmu.pebs_buffer_size = PEBS_BUFFER_SIZE;
1429
	if (x86_pmu.pebs) {
1430 1431
		char pebs_type = x86_pmu.intel_cap.pebs_trap ?  '+' : '-';
		int format = x86_pmu.intel_cap.pebs_format;
1432 1433 1434

		switch (format) {
		case 0:
1435
			pr_cont("PEBS fmt0%c, ", pebs_type);
1436
			x86_pmu.pebs_record_size = sizeof(struct pebs_record_core);
1437 1438 1439 1440 1441 1442 1443 1444
			/*
			 * 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;
1445 1446 1447 1448
			x86_pmu.drain_pebs = intel_pmu_drain_pebs_core;
			break;

		case 1:
1449
			pr_cont("PEBS fmt1%c, ", pebs_type);
1450 1451 1452 1453
			x86_pmu.pebs_record_size = sizeof(struct pebs_record_nhm);
			x86_pmu.drain_pebs = intel_pmu_drain_pebs_nhm;
			break;

1454 1455 1456
		case 2:
			pr_cont("PEBS fmt2%c, ", pebs_type);
			x86_pmu.pebs_record_size = sizeof(struct pebs_record_hsw);
1457
			x86_pmu.drain_pebs = intel_pmu_drain_pebs_nhm;
1458 1459
			break;

1460 1461 1462 1463 1464
		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;
1465
			x86_pmu.free_running_flags |= PERF_SAMPLE_TIME;
1466 1467
			break;

1468
		default:
1469
			pr_cont("no PEBS fmt%d%c, ", format, pebs_type);
1470 1471 1472 1473
			x86_pmu.pebs = 0;
		}
	}
}
1474 1475 1476

void perf_restore_debug_store(void)
{
1477 1478
	struct debug_store *ds = __this_cpu_read(cpu_hw_events.ds);

1479 1480 1481
	if (!x86_pmu.bts && !x86_pmu.pebs)
		return;

1482
	wrmsrl(MSR_IA32_DS_AREA, (unsigned long)ds);
1483
}