perf_counter.h 20.1 KB
Newer Older
T
Thomas Gleixner 已提交
1 2 3
/*
 *  Performance counters:
 *
I
Ingo Molnar 已提交
4 5 6
 *    Copyright (C) 2008-2009, Thomas Gleixner <tglx@linutronix.de>
 *    Copyright (C) 2008-2009, Red Hat, Inc., Ingo Molnar
 *    Copyright (C) 2008-2009, Red Hat, Inc., Peter Zijlstra
T
Thomas Gleixner 已提交
7 8 9
 *
 *  Data type definitions, declarations, prototypes.
 *
I
Ingo Molnar 已提交
10
 *    Started by: Thomas Gleixner and Ingo Molnar
T
Thomas Gleixner 已提交
11 12 13 14 15 16
 *
 *  For licencing details see kernel-base/COPYING
 */
#ifndef _LINUX_PERF_COUNTER_H
#define _LINUX_PERF_COUNTER_H

17 18
#include <linux/types.h>
#include <linux/ioctl.h>
19
#include <asm/byteorder.h>
T
Thomas Gleixner 已提交
20 21

/*
I
Ingo Molnar 已提交
22 23 24 25
 * User-space ABI bits:
 */

/*
26
 * attr.type
T
Thomas Gleixner 已提交
27
 */
P
Peter Zijlstra 已提交
28
enum perf_type_id {
I
Ingo Molnar 已提交
29 30 31 32 33
	PERF_TYPE_HARDWARE			= 0,
	PERF_TYPE_SOFTWARE			= 1,
	PERF_TYPE_TRACEPOINT			= 2,
	PERF_TYPE_HW_CACHE			= 3,
	PERF_TYPE_RAW				= 4,
34

I
Ingo Molnar 已提交
35
	PERF_TYPE_MAX,				/* non-ABI */
36
};
37

38
/*
I
Ingo Molnar 已提交
39 40 41
 * Generalized performance counter event types, used by the
 * attr.event_id parameter of the sys_perf_counter_open()
 * syscall:
42
 */
P
Peter Zijlstra 已提交
43
enum perf_hw_id {
I
Ingo Molnar 已提交
44
	/*
45
	 * Common hardware events, generalized by the kernel:
I
Ingo Molnar 已提交
46
	 */
47 48 49 50 51 52 53 54
	PERF_COUNT_HW_CPU_CYCLES		= 0,
	PERF_COUNT_HW_INSTRUCTIONS		= 1,
	PERF_COUNT_HW_CACHE_REFERENCES		= 2,
	PERF_COUNT_HW_CACHE_MISSES		= 3,
	PERF_COUNT_HW_BRANCH_INSTRUCTIONS	= 4,
	PERF_COUNT_HW_BRANCH_MISSES		= 5,
	PERF_COUNT_HW_BUS_CYCLES		= 6,

I
Ingo Molnar 已提交
55
	PERF_COUNT_HW_MAX,			/* non-ABI */
56
};
57

58 59 60
/*
 * Generalized hardware cache counters:
 *
61
 *       { L1-D, L1-I, LLC, ITLB, DTLB, BPU } x
62 63 64
 *       { read, write, prefetch } x
 *       { accesses, misses }
 */
P
Peter Zijlstra 已提交
65
enum perf_hw_cache_id {
I
Ingo Molnar 已提交
66 67 68 69 70 71 72 73
	PERF_COUNT_HW_CACHE_L1D			= 0,
	PERF_COUNT_HW_CACHE_L1I			= 1,
	PERF_COUNT_HW_CACHE_LL			= 2,
	PERF_COUNT_HW_CACHE_DTLB		= 3,
	PERF_COUNT_HW_CACHE_ITLB		= 4,
	PERF_COUNT_HW_CACHE_BPU			= 5,

	PERF_COUNT_HW_CACHE_MAX,		/* non-ABI */
74 75
};

P
Peter Zijlstra 已提交
76
enum perf_hw_cache_op_id {
I
Ingo Molnar 已提交
77 78 79
	PERF_COUNT_HW_CACHE_OP_READ		= 0,
	PERF_COUNT_HW_CACHE_OP_WRITE		= 1,
	PERF_COUNT_HW_CACHE_OP_PREFETCH		= 2,
80

I
Ingo Molnar 已提交
81
	PERF_COUNT_HW_CACHE_OP_MAX,		/* non-ABI */
82 83
};

P
Peter Zijlstra 已提交
84 85 86
enum perf_hw_cache_op_result_id {
	PERF_COUNT_HW_CACHE_RESULT_ACCESS	= 0,
	PERF_COUNT_HW_CACHE_RESULT_MISS		= 1,
87

I
Ingo Molnar 已提交
88
	PERF_COUNT_HW_CACHE_RESULT_MAX,		/* non-ABI */
89 90
};

91 92 93 94 95 96
/*
 * Special "software" counters provided by the kernel, even if the hardware
 * does not support performance counters. These counters measure various
 * physical and sw events of the kernel (and allow the profiling of them as
 * well):
 */
P
Peter Zijlstra 已提交
97
enum perf_sw_ids {
I
Ingo Molnar 已提交
98 99 100 101 102 103 104 105 106
	PERF_COUNT_SW_CPU_CLOCK			= 0,
	PERF_COUNT_SW_TASK_CLOCK		= 1,
	PERF_COUNT_SW_PAGE_FAULTS		= 2,
	PERF_COUNT_SW_CONTEXT_SWITCHES		= 3,
	PERF_COUNT_SW_CPU_MIGRATIONS		= 4,
	PERF_COUNT_SW_PAGE_FAULTS_MIN		= 5,
	PERF_COUNT_SW_PAGE_FAULTS_MAJ		= 6,

	PERF_COUNT_SW_MAX,			/* non-ABI */
T
Thomas Gleixner 已提交
107 108
};

109
/*
110
 * Bits that can be set in attr.sample_type to request information
111 112
 * in the overflow packets.
 */
113
enum perf_counter_sample_format {
I
Ingo Molnar 已提交
114 115 116 117
	PERF_SAMPLE_IP				= 1U << 0,
	PERF_SAMPLE_TID				= 1U << 1,
	PERF_SAMPLE_TIME			= 1U << 2,
	PERF_SAMPLE_ADDR			= 1U << 3,
118
	PERF_SAMPLE_READ			= 1U << 4,
I
Ingo Molnar 已提交
119 120 121 122
	PERF_SAMPLE_CALLCHAIN			= 1U << 5,
	PERF_SAMPLE_ID				= 1U << 6,
	PERF_SAMPLE_CPU				= 1U << 7,
	PERF_SAMPLE_PERIOD			= 1U << 8,
123
	PERF_SAMPLE_STREAM_ID			= 1U << 9,
124
	PERF_SAMPLE_RAW				= 1U << 10,
125

126
	PERF_SAMPLE_MAX = 1U << 11,		/* non-ABI */
127 128
};

129
/*
130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147
 * The format of the data returned by read() on a perf counter fd,
 * as specified by attr.read_format:
 *
 * struct read_format {
 * 	{ u64		value;
 * 	  { u64		time_enabled; } && PERF_FORMAT_ENABLED
 * 	  { u64		time_running; } && PERF_FORMAT_RUNNING
 * 	  { u64		id;           } && PERF_FORMAT_ID
 * 	} && !PERF_FORMAT_GROUP
 *
 * 	{ u64		nr;
 * 	  { u64		time_enabled; } && PERF_FORMAT_ENABLED
 * 	  { u64		time_running; } && PERF_FORMAT_RUNNING
 * 	  { u64		value;
 * 	    { u64	id;           } && PERF_FORMAT_ID
 * 	  }		cntr[nr];
 * 	} && PERF_FORMAT_GROUP
 * };
148 149
 */
enum perf_counter_read_format {
I
Ingo Molnar 已提交
150 151 152
	PERF_FORMAT_TOTAL_TIME_ENABLED		= 1U << 0,
	PERF_FORMAT_TOTAL_TIME_RUNNING		= 1U << 1,
	PERF_FORMAT_ID				= 1U << 2,
153
	PERF_FORMAT_GROUP			= 1U << 3,
154

155
	PERF_FORMAT_MAX = 1U << 4, 		/* non-ABI */
156 157
};

158 159
#define PERF_ATTR_SIZE_VER0	64	/* sizeof first published struct */

I
Ingo Molnar 已提交
160 161 162
/*
 * Hardware event to monitor via a performance monitoring counter:
 */
163
struct perf_counter_attr {
164

165
	/*
166 167 168
	 * Major type: hardware/software/tracepoint/etc.
	 */
	__u32			type;
169 170 171 172 173

	/*
	 * Size of the attr structure, for fwd/bwd compat.
	 */
	__u32			size;
174 175 176

	/*
	 * Type specific configuration information.
177 178
	 */
	__u64			config;
I
Ingo Molnar 已提交
179

180
	union {
181 182
		__u64		sample_period;
		__u64		sample_freq;
183 184
	};

185 186
	__u64			sample_type;
	__u64			read_format;
I
Ingo Molnar 已提交
187

188
	__u64			disabled       :  1, /* off by default        */
189 190 191 192 193 194
				inherit	       :  1, /* children inherit it   */
				pinned	       :  1, /* must always be on PMU */
				exclusive      :  1, /* only group on PMU     */
				exclude_user   :  1, /* don't count user      */
				exclude_kernel :  1, /* ditto kernel          */
				exclude_hv     :  1, /* ditto hypervisor      */
195
				exclude_idle   :  1, /* don't count when idle */
196
				mmap           :  1, /* include mmap data     */
197
				comm	       :  1, /* include comm data     */
198
				freq           :  1, /* use freq, not period  */
199
				inherit_stat   :  1, /* per task counts       */
200
				enable_on_exec :  1, /* next exec enables     */
P
Peter Zijlstra 已提交
201
				task           :  1, /* trace fork/exit       */
202

P
Peter Zijlstra 已提交
203
				__reserved_1   : 50;
204

P
Peter Zijlstra 已提交
205
	__u32			wakeup_events;	/* wakeup every n events */
206
	__u32			__reserved_2;
I
Ingo Molnar 已提交
207

208
	__u64			__reserved_3;
209 210
};

211 212 213
/*
 * Ioctls that can be done on a perf counter fd:
 */
214 215 216 217 218
#define PERF_COUNTER_IOC_ENABLE		_IO ('$', 0)
#define PERF_COUNTER_IOC_DISABLE	_IO ('$', 1)
#define PERF_COUNTER_IOC_REFRESH	_IO ('$', 2)
#define PERF_COUNTER_IOC_RESET		_IO ('$', 3)
#define PERF_COUNTER_IOC_PERIOD		_IOW('$', 4, u64)
219
#define PERF_COUNTER_IOC_SET_OUTPUT	_IO ('$', 5)
P
Peter Zijlstra 已提交
220 221 222 223

enum perf_counter_ioc_flags {
	PERF_IOC_FLAG_GROUP		= 1U << 0,
};
224

225 226 227 228 229 230
/*
 * Structure of the page that can be mapped via mmap
 */
struct perf_counter_mmap_page {
	__u32	version;		/* version number of this structure */
	__u32	compat_version;		/* lowest version this is compat with */
231 232 233 234

	/*
	 * Bits needed to read the hw counters in user-space.
	 *
235 236
	 *   u32 seq;
	 *   s64 count;
237
	 *
238 239
	 *   do {
	 *     seq = pc->lock;
240
	 *
241 242 243 244 245 246
	 *     barrier()
	 *     if (pc->index) {
	 *       count = pmc_read(pc->index - 1);
	 *       count += pc->offset;
	 *     } else
	 *       goto regular_read;
247
	 *
248 249
	 *     barrier();
	 *   } while (pc->lock != seq);
250
	 *
251 252
	 * NOTE: for obvious reason this only works on self-monitoring
	 *       processes.
253
	 */
254 255 256
	__u32	lock;			/* seqlock for synchronization */
	__u32	index;			/* hardware counter identifier */
	__s64	offset;			/* add to hardware counter value */
257 258
	__u64	time_enabled;		/* time counter active */
	__u64	time_running;		/* time counter on cpu */
259

260 261 262 263
		/*
		 * Hole for extension of the self monitor capabilities
		 */

264
	__u64	__reserved[123];	/* align to 1k */
265

266 267 268
	/*
	 * Control data for the mmap() data buffer.
	 *
269 270 271 272 273 274 275
	 * User-space reading the @data_head value should issue an rmb(), on
	 * SMP capable platforms, after reading this value -- see
	 * perf_counter_wakeup().
	 *
	 * When the mapping is PROT_WRITE the @data_tail value should be
	 * written by userspace to reflect the last read data. In this case
	 * the kernel will not over-write unread data.
276
	 */
277
	__u64   data_head;		/* head in the data section */
278
	__u64	data_tail;		/* user-space written tail */
279 280
};

I
Ingo Molnar 已提交
281 282 283 284 285
#define PERF_EVENT_MISC_CPUMODE_MASK		(3 << 0)
#define PERF_EVENT_MISC_CPUMODE_UNKNOWN		(0 << 0)
#define PERF_EVENT_MISC_KERNEL			(1 << 0)
#define PERF_EVENT_MISC_USER			(2 << 0)
#define PERF_EVENT_MISC_HYPERVISOR		(3 << 0)
286

P
Peter Zijlstra 已提交
287 288
struct perf_event_header {
	__u32	type;
289 290
	__u16	misc;
	__u16	size;
P
Peter Zijlstra 已提交
291 292 293
};

enum perf_event_type {
294

295 296 297 298 299
	/*
	 * The MMAP events record the PROT_EXEC mappings so that we can
	 * correlate userspace IPs to code. They have the following structure:
	 *
	 * struct {
I
Ingo Molnar 已提交
300
	 *	struct perf_event_header	header;
301
	 *
I
Ingo Molnar 已提交
302 303 304 305 306
	 *	u32				pid, tid;
	 *	u64				addr;
	 *	u64				len;
	 *	u64				pgoff;
	 *	char				filename[];
307 308
	 * };
	 */
309
	PERF_EVENT_MMAP			= 1,
310

311 312 313 314 315 316 317 318 319
	/*
	 * struct {
	 * 	struct perf_event_header	header;
	 * 	u64				id;
	 * 	u64				lost;
	 * };
	 */
	PERF_EVENT_LOST			= 2,

320 321
	/*
	 * struct {
I
Ingo Molnar 已提交
322
	 *	struct perf_event_header	header;
323
	 *
I
Ingo Molnar 已提交
324 325
	 *	u32				pid, tid;
	 *	char				comm[];
326 327 328 329
	 * };
	 */
	PERF_EVENT_COMM			= 3,

P
Peter Zijlstra 已提交
330 331 332 333 334 335 336 337 338
	/*
	 * struct {
	 *	struct perf_event_header	header;
	 *	u32				pid, ppid;
	 *	u32				tid, ptid;
	 * };
	 */
	PERF_EVENT_EXIT			= 4,

339 340
	/*
	 * struct {
I
Ingo Molnar 已提交
341 342
	 *	struct perf_event_header	header;
	 *	u64				time;
343
	 *	u64				id;
344
	 *	u64				stream_id;
345 346 347 348 349
	 * };
	 */
	PERF_EVENT_THROTTLE		= 5,
	PERF_EVENT_UNTHROTTLE		= 6,

P
Peter Zijlstra 已提交
350 351
	/*
	 * struct {
352 353
	 *	struct perf_event_header	header;
	 *	u32				pid, ppid;
P
Peter Zijlstra 已提交
354
	 *	u32				tid, ptid;
P
Peter Zijlstra 已提交
355 356 357 358
	 * };
	 */
	PERF_EVENT_FORK			= 7,

359 360 361 362
	/*
	 * struct {
	 * 	struct perf_event_header	header;
	 * 	u32				pid, tid;
363 364
	 *
	 * 	struct read_format		values;
365 366 367 368
	 * };
	 */
	PERF_EVENT_READ			= 8,

369
	/*
370
	 * struct {
I
Ingo Molnar 已提交
371
	 *	struct perf_event_header	header;
372
	 *
373 374 375 376
	 *	{ u64			ip;	  } && PERF_SAMPLE_IP
	 *	{ u32			pid, tid; } && PERF_SAMPLE_TID
	 *	{ u64			time;     } && PERF_SAMPLE_TIME
	 *	{ u64			addr;     } && PERF_SAMPLE_ADDR
377
	 *	{ u64			id;	  } && PERF_SAMPLE_ID
378
	 *	{ u64			stream_id;} && PERF_SAMPLE_STREAM_ID
379
	 *	{ u32			cpu, res; } && PERF_SAMPLE_CPU
380
	 * 	{ u64			period;   } && PERF_SAMPLE_PERIOD
381
	 *
382
	 *	{ struct read_format	values;	  } && PERF_SAMPLE_READ
383
	 *
384
	 *	{ u64			nr,
385
	 *	  u64			ips[nr];  } && PERF_SAMPLE_CALLCHAIN
386 387 388 389 390 391 392 393 394 395 396 397
	 *
	 * 	#
	 * 	# The RAW record below is opaque data wrt the ABI
	 * 	#
	 * 	# That is, the ABI doesn't make any promises wrt to
	 * 	# the stability of its content, it may vary depending
	 * 	# on event, hardware, kernel version and phase of
	 * 	# the moon.
	 * 	#
	 * 	# In other words, PERF_SAMPLE_RAW contents are not an ABI.
	 * 	#
	 *
398 399
	 *	{ u32			size;
	 *	  char                  data[size];}&& PERF_SAMPLE_RAW
400
	 * };
401
	 */
402 403 404
	PERF_EVENT_SAMPLE		= 9,

	PERF_EVENT_MAX,			/* non-ABI */
P
Peter Zijlstra 已提交
405 406
};

407 408 409 410
enum perf_callchain_context {
	PERF_CONTEXT_HV			= (__u64)-32,
	PERF_CONTEXT_KERNEL		= (__u64)-128,
	PERF_CONTEXT_USER		= (__u64)-512,
411

412 413 414 415 416
	PERF_CONTEXT_GUEST		= (__u64)-2048,
	PERF_CONTEXT_GUEST_KERNEL	= (__u64)-2176,
	PERF_CONTEXT_GUEST_USER		= (__u64)-2560,

	PERF_CONTEXT_MAX		= (__u64)-4095,
417 418
};

419 420 421
#define PERF_FLAG_FD_NO_GROUP	(1U << 0)
#define PERF_FLAG_FD_OUTPUT	(1U << 1)

422
#ifdef __KERNEL__
I
Ingo Molnar 已提交
423
/*
424
 * Kernel-internal data types and definitions:
I
Ingo Molnar 已提交
425 426
 */

427 428 429 430 431 432 433 434 435
#ifdef CONFIG_PERF_COUNTERS
# include <asm/perf_counter.h>
#endif

#include <linux/list.h>
#include <linux/mutex.h>
#include <linux/rculist.h>
#include <linux/rcupdate.h>
#include <linux/spinlock.h>
436
#include <linux/hrtimer.h>
P
Peter Zijlstra 已提交
437
#include <linux/fs.h>
438
#include <linux/pid_namespace.h>
439 440
#include <asm/atomic.h>

441 442 443 444 445 446 447
#define PERF_MAX_STACK_DEPTH		255

struct perf_callchain_entry {
	__u64				nr;
	__u64				ip[PERF_MAX_STACK_DEPTH];
};

448 449 450
struct perf_raw_record {
	u32				size;
	void				*data;
451 452
};

453 454
struct task_struct;

T
Thomas Gleixner 已提交
455
/**
I
Ingo Molnar 已提交
456
 * struct hw_perf_counter - performance counter hardware details:
T
Thomas Gleixner 已提交
457 458
 */
struct hw_perf_counter {
459
#ifdef CONFIG_PERF_COUNTERS
460 461
	union {
		struct { /* hardware */
I
Ingo Molnar 已提交
462 463 464 465
			u64		config;
			unsigned long	config_base;
			unsigned long	counter_base;
			int		idx;
466 467
		};
		union { /* software */
I
Ingo Molnar 已提交
468 469
			atomic64_t	count;
			struct hrtimer	hrtimer;
470 471
		};
	};
472
	atomic64_t			prev_count;
473
	u64				sample_period;
474
	u64				last_period;
475
	atomic64_t			period_left;
476
	u64				interrupts;
477 478 479

	u64				freq_count;
	u64				freq_interrupts;
480
	u64				freq_stamp;
481
#endif
T
Thomas Gleixner 已提交
482 483
};

I
Ingo Molnar 已提交
484 485 486
struct perf_counter;

/**
487
 * struct pmu - generic performance monitoring unit
I
Ingo Molnar 已提交
488
 */
489
struct pmu {
490
	int (*enable)			(struct perf_counter *counter);
I
Ingo Molnar 已提交
491 492
	void (*disable)			(struct perf_counter *counter);
	void (*read)			(struct perf_counter *counter);
493
	void (*unthrottle)		(struct perf_counter *counter);
I
Ingo Molnar 已提交
494 495
};

496 497 498 499
/**
 * enum perf_counter_active_state - the states of a counter
 */
enum perf_counter_active_state {
500
	PERF_COUNTER_STATE_ERROR	= -2,
501 502 503 504 505
	PERF_COUNTER_STATE_OFF		= -1,
	PERF_COUNTER_STATE_INACTIVE	=  0,
	PERF_COUNTER_STATE_ACTIVE	=  1,
};

506 507
struct file;

508 509
struct perf_mmap_data {
	struct rcu_head			rcu_head;
P
Peter Zijlstra 已提交
510
	int				nr_pages;	/* nr of data pages  */
511
	int				writable;	/* are we writable   */
512
	int				nr_locked;	/* nr pages mlocked  */
P
Peter Zijlstra 已提交
513

514
	atomic_t			poll;		/* POLL_ for wakeups */
P
Peter Zijlstra 已提交
515 516
	atomic_t			events;		/* event limit       */

517 518 519
	atomic_long_t			head;		/* write position    */
	atomic_long_t			done_head;	/* completed head    */

520
	atomic_t			lock;		/* concurrent writes */
521
	atomic_t			wakeup;		/* needs a wakeup    */
522
	atomic_t			lost;		/* nr records lost   */
523

524
	struct perf_counter_mmap_page   *user_page;
I
Ingo Molnar 已提交
525
	void				*data_pages[0];
526 527
};

528 529 530
struct perf_pending_entry {
	struct perf_pending_entry *next;
	void (*func)(struct perf_pending_entry *);
531 532
};

T
Thomas Gleixner 已提交
533 534 535 536
/**
 * struct perf_counter - performance counter kernel representation:
 */
struct perf_counter {
537
#ifdef CONFIG_PERF_COUNTERS
538
	struct list_head		list_entry;
P
Peter Zijlstra 已提交
539
	struct list_head		event_entry;
540
	struct list_head		sibling_list;
I
Ingo Molnar 已提交
541
	int				nr_siblings;
542
	struct perf_counter		*group_leader;
543
	struct perf_counter		*output;
544
	const struct pmu		*pmu;
545

546
	enum perf_counter_active_state	state;
T
Thomas Gleixner 已提交
547
	atomic64_t			count;
548

549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574
	/*
	 * These are the total time in nanoseconds that the counter
	 * has been enabled (i.e. eligible to run, and the task has
	 * been scheduled in, if this is a per-task counter)
	 * and running (scheduled onto the CPU), respectively.
	 *
	 * They are computed from tstamp_enabled, tstamp_running and
	 * tstamp_stopped when the counter is in INACTIVE or ACTIVE state.
	 */
	u64				total_time_enabled;
	u64				total_time_running;

	/*
	 * These are timestamps used for computing total_time_enabled
	 * and total_time_running when the counter is in INACTIVE or
	 * ACTIVE state, measured in nanoseconds from an arbitrary point
	 * in time.
	 * tstamp_enabled: the notional time when the counter was enabled
	 * tstamp_running: the notional time when the counter was scheduled on
	 * tstamp_stopped: in INACTIVE state, the notional time when the
	 *	counter was scheduled off.
	 */
	u64				tstamp_enabled;
	u64				tstamp_running;
	u64				tstamp_stopped;

575
	struct perf_counter_attr	attr;
T
Thomas Gleixner 已提交
576 577 578
	struct hw_perf_counter		hw;

	struct perf_counter_context	*ctx;
579
	struct file			*filp;
T
Thomas Gleixner 已提交
580

581 582 583 584 585 586 587
	/*
	 * These accumulate total time (in nanoseconds) that children
	 * counters have been enabled and running, respectively.
	 */
	atomic64_t			child_total_time_enabled;
	atomic64_t			child_total_time_running;

T
Thomas Gleixner 已提交
588
	/*
589
	 * Protect attach/detach and child_list:
T
Thomas Gleixner 已提交
590
	 */
591 592 593
	struct mutex			child_mutex;
	struct list_head		child_list;
	struct perf_counter		*parent;
T
Thomas Gleixner 已提交
594 595 596 597

	int				oncpu;
	int				cpu;

598 599 600
	struct list_head		owner_entry;
	struct task_struct		*owner;

601 602 603 604
	/* mmap bits */
	struct mutex			mmap_mutex;
	atomic_t			mmap_count;
	struct perf_mmap_data		*data;
605

606
	/* poll related */
T
Thomas Gleixner 已提交
607
	wait_queue_head_t		waitq;
P
Peter Zijlstra 已提交
608
	struct fasync_struct		*fasync;
609 610 611

	/* delayed work for NMIs and such */
	int				pending_wakeup;
612
	int				pending_kill;
613
	int				pending_disable;
614
	struct perf_pending_entry	pending;
P
Peter Zijlstra 已提交
615

616 617
	atomic_t			event_limit;

618
	void (*destroy)(struct perf_counter *);
P
Peter Zijlstra 已提交
619
	struct rcu_head			rcu_head;
620 621

	struct pid_namespace		*ns;
622
	u64				id;
623
#endif
T
Thomas Gleixner 已提交
624 625 626 627 628 629 630 631 632
};

/**
 * struct perf_counter_context - counter context structure
 *
 * Used as a container for task counters and CPU counters as well:
 */
struct perf_counter_context {
	/*
633 634
	 * Protect the states of the counters in the list,
	 * nr_active, and the list:
T
Thomas Gleixner 已提交
635
	 */
I
Ingo Molnar 已提交
636
	spinlock_t			lock;
637 638 639 640 641
	/*
	 * Protect the list of counters.  Locking either mutex or lock
	 * is sufficient to ensure the list doesn't change; to change
	 * the list you need to lock both the mutex and the spinlock.
	 */
I
Ingo Molnar 已提交
642
	struct mutex			mutex;
643

I
Ingo Molnar 已提交
644 645 646 647 648
	struct list_head		counter_list;
	struct list_head		event_list;
	int				nr_counters;
	int				nr_active;
	int				is_active;
649
	int				nr_stat;
I
Ingo Molnar 已提交
650 651
	atomic_t			refcount;
	struct task_struct		*task;
652 653

	/*
654
	 * Context clock, runs when context enabled.
655
	 */
I
Ingo Molnar 已提交
656 657
	u64				time;
	u64				timestamp;
658 659 660 661 662

	/*
	 * These fields let us detect when two contexts have both
	 * been cloned (inherited) from a common ancestor.
	 */
I
Ingo Molnar 已提交
663 664 665 666 667
	struct perf_counter_context	*parent_ctx;
	u64				parent_gen;
	u64				generation;
	int				pin_count;
	struct rcu_head			rcu_head;
T
Thomas Gleixner 已提交
668 669 670 671 672 673 674 675 676 677
};

/**
 * struct perf_counter_cpu_context - per cpu counter context structure
 */
struct perf_cpu_context {
	struct perf_counter_context	ctx;
	struct perf_counter_context	*task_ctx;
	int				active_oncpu;
	int				max_pertask;
678
	int				exclusive;
P
Peter Zijlstra 已提交
679 680 681 682 683 684

	/*
	 * Recursion avoidance:
	 *
	 * task, softirq, irq, nmi context
	 */
685
	int				recursion[4];
T
Thomas Gleixner 已提交
686 687
};

688 689
#ifdef CONFIG_PERF_COUNTERS

T
Thomas Gleixner 已提交
690 691 692 693 694
/*
 * Set by architecture code:
 */
extern int perf_max_counters;

695
extern const struct pmu *hw_perf_counter_init(struct perf_counter *counter);
I
Ingo Molnar 已提交
696

T
Thomas Gleixner 已提交
697
extern void perf_counter_task_sched_in(struct task_struct *task, int cpu);
698 699
extern void perf_counter_task_sched_out(struct task_struct *task,
					struct task_struct *next, int cpu);
T
Thomas Gleixner 已提交
700
extern void perf_counter_task_tick(struct task_struct *task, int cpu);
701
extern int perf_counter_init_task(struct task_struct *child);
702
extern void perf_counter_exit_task(struct task_struct *child);
703
extern void perf_counter_free_task(struct task_struct *task);
704
extern void set_perf_counter_pending(void);
705
extern void perf_counter_do_pending(void);
T
Thomas Gleixner 已提交
706
extern void perf_counter_print_debug(void);
707 708 709 710
extern void __perf_disable(void);
extern bool __perf_enable(void);
extern void perf_disable(void);
extern void perf_enable(void);
711 712
extern int perf_counter_task_disable(void);
extern int perf_counter_task_enable(void);
713 714 715
extern int hw_perf_group_sched_in(struct perf_counter *group_leader,
	       struct perf_cpu_context *cpuctx,
	       struct perf_counter_context *ctx, int cpu);
716
extern void perf_counter_update_userpage(struct perf_counter *counter);
717

718
struct perf_sample_data {
I
Ingo Molnar 已提交
719 720 721
	struct pt_regs			*regs;
	u64				addr;
	u64				period;
722
	struct perf_raw_record		*raw;
723 724 725 726
};

extern int perf_counter_overflow(struct perf_counter *counter, int nmi,
				 struct perf_sample_data *data);
727 728
extern void perf_counter_output(struct perf_counter *counter, int nmi,
				struct perf_sample_data *data);
729

730 731 732 733 734
/*
 * Return 1 for a software counter, 0 for a hardware counter
 */
static inline int is_software_counter(struct perf_counter *counter)
{
735
	return (counter->attr.type != PERF_TYPE_RAW) &&
736 737
		(counter->attr.type != PERF_TYPE_HARDWARE) &&
		(counter->attr.type != PERF_TYPE_HW_CACHE);
738 739
}

740 741 742 743 744 745 746 747 748 749
extern atomic_t perf_swcounter_enabled[PERF_COUNT_SW_MAX];

extern void __perf_swcounter_event(u32, u64, int, struct pt_regs *, u64);

static inline void
perf_swcounter_event(u32 event, u64 nr, int nmi, struct pt_regs *regs, u64 addr)
{
	if (atomic_read(&perf_swcounter_enabled[event]))
		__perf_swcounter_event(event, nr, nmi, regs, addr);
}
750

751 752 753 754 755 756 757
extern void __perf_counter_mmap(struct vm_area_struct *vma);

static inline void perf_counter_mmap(struct vm_area_struct *vma)
{
	if (vma->vm_flags & VM_EXEC)
		__perf_counter_mmap(vma);
}
758

759
extern void perf_counter_comm(struct task_struct *tsk);
P
Peter Zijlstra 已提交
760
extern void perf_counter_fork(struct task_struct *tsk);
761

762 763
extern struct perf_callchain_entry *perf_callchain(struct pt_regs *regs);

764
extern int sysctl_perf_counter_paranoid;
765
extern int sysctl_perf_counter_mlock;
766
extern int sysctl_perf_counter_sample_rate;
767

768
extern void perf_counter_init(void);
769 770
extern void perf_tpcounter_event(int event_id, u64 addr, u64 count,
				 void *record, int entry_size);
771

772 773 774 775 776 777
#ifndef perf_misc_flags
#define perf_misc_flags(regs)	(user_mode(regs) ? PERF_EVENT_MISC_USER : \
				 PERF_EVENT_MISC_KERNEL)
#define perf_instruction_pointer(regs)	instruction_pointer(regs)
#endif

T
Thomas Gleixner 已提交
778 779 780 781
#else
static inline void
perf_counter_task_sched_in(struct task_struct *task, int cpu)		{ }
static inline void
782 783
perf_counter_task_sched_out(struct task_struct *task,
			    struct task_struct *next, int cpu)		{ }
T
Thomas Gleixner 已提交
784 785
static inline void
perf_counter_task_tick(struct task_struct *task, int cpu)		{ }
786
static inline int perf_counter_init_task(struct task_struct *child)	{ return 0; }
787
static inline void perf_counter_exit_task(struct task_struct *child)	{ }
788
static inline void perf_counter_free_task(struct task_struct *task)	{ }
789
static inline void perf_counter_do_pending(void)			{ }
T
Thomas Gleixner 已提交
790
static inline void perf_counter_print_debug(void)			{ }
791 792
static inline void perf_disable(void)					{ }
static inline void perf_enable(void)					{ }
793 794
static inline int perf_counter_task_disable(void)	{ return -EINVAL; }
static inline int perf_counter_task_enable(void)	{ return -EINVAL; }
795

796
static inline void
797 798
perf_swcounter_event(u32 event, u64 nr, int nmi,
		     struct pt_regs *regs, u64 addr)			{ }
799

800
static inline void perf_counter_mmap(struct vm_area_struct *vma)	{ }
801
static inline void perf_counter_comm(struct task_struct *tsk)		{ }
P
Peter Zijlstra 已提交
802
static inline void perf_counter_fork(struct task_struct *tsk)		{ }
803
static inline void perf_counter_init(void)				{ }
T
Thomas Gleixner 已提交
804 805
#endif

806
#endif /* __KERNEL__ */
T
Thomas Gleixner 已提交
807
#endif /* _LINUX_PERF_COUNTER_H */