[deliverable/linux.git] / include / linux / perf_counter.h

/*
 *  Performance counters:
 *
 *   Copyright(C) 2008, Thomas Gleixner <tglx@linutronix.de>
 *   Copyright(C) 2008, Red Hat, Inc., Ingo Molnar
 *
 *  Data type definitions, declarations, prototypes.
 *
 *  Started by: Thomas Gleixner and Ingo Molnar
 *
 *  For licencing details see kernel-base/COPYING
 */
#ifndef _LINUX_PERF_COUNTER_H
#define _LINUX_PERF_COUNTER_H

#include <linux/types.h>
#include <linux/ioctl.h>
#include <asm/byteorder.h>

/*
 * User-space ABI bits:
 */

/*
 * hw_event.type
 */
enum perf_event_types {
	PERF_TYPE_HARDWARE		= 0,
	PERF_TYPE_SOFTWARE		= 1,
	PERF_TYPE_TRACEPOINT		= 2,

	/*
	 * available TYPE space, raw is the max value.
	 */

	PERF_TYPE_RAW			= 128,
};

/*
 * Generalized performance counter event types, used by the hw_event.event_id
 * parameter of the sys_perf_counter_open() syscall:
 */
enum hw_event_ids {
	/*
	 * Common hardware events, generalized by the kernel:
	 */
	PERF_COUNT_CPU_CYCLES		= 0,
	PERF_COUNT_INSTRUCTIONS		= 1,
	PERF_COUNT_CACHE_REFERENCES	= 2,
	PERF_COUNT_CACHE_MISSES		= 3,
	PERF_COUNT_BRANCH_INSTRUCTIONS	= 4,
	PERF_COUNT_BRANCH_MISSES	= 5,
	PERF_COUNT_BUS_CYCLES		= 6,

	PERF_HW_EVENTS_MAX		= 7,
};

/*
 * 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):
 */
enum sw_event_ids {
	PERF_COUNT_CPU_CLOCK		= 0,
	PERF_COUNT_TASK_CLOCK		= 1,
	PERF_COUNT_PAGE_FAULTS		= 2,
	PERF_COUNT_CONTEXT_SWITCHES	= 3,
	PERF_COUNT_CPU_MIGRATIONS	= 4,
	PERF_COUNT_PAGE_FAULTS_MIN	= 5,
	PERF_COUNT_PAGE_FAULTS_MAJ	= 6,

	PERF_SW_EVENTS_MAX		= 7,
};

#define __PERF_COUNTER_MASK(name) 			\
	(((1ULL << PERF_COUNTER_##name##_BITS) - 1) <<	\
	 PERF_COUNTER_##name##_SHIFT)

#define PERF_COUNTER_RAW_BITS		1
#define PERF_COUNTER_RAW_SHIFT		63
#define PERF_COUNTER_RAW_MASK		__PERF_COUNTER_MASK(RAW)

#define PERF_COUNTER_CONFIG_BITS	63
#define PERF_COUNTER_CONFIG_SHIFT	0
#define PERF_COUNTER_CONFIG_MASK	__PERF_COUNTER_MASK(CONFIG)

#define PERF_COUNTER_TYPE_BITS		7
#define PERF_COUNTER_TYPE_SHIFT		56
#define PERF_COUNTER_TYPE_MASK		__PERF_COUNTER_MASK(TYPE)

#define PERF_COUNTER_EVENT_BITS		56
#define PERF_COUNTER_EVENT_SHIFT	0
#define PERF_COUNTER_EVENT_MASK		__PERF_COUNTER_MASK(EVENT)

/*
 * Bits that can be set in hw_event.record_type to request information
 * in the overflow packets.
 */
enum perf_counter_record_format {
	PERF_RECORD_IP		= 1U << 0,
	PERF_RECORD_TID		= 1U << 1,
	PERF_RECORD_GROUP	= 1U << 2,
	PERF_RECORD_CALLCHAIN	= 1U << 3,
};

/*
 * Bits that can be set in hw_event.read_format to request that
 * reads on the counter should return the indicated quantities,
 * in increasing order of bit value, after the counter value.
 */
enum perf_counter_read_format {
	PERF_FORMAT_TOTAL_TIME_ENABLED	=  1,
	PERF_FORMAT_TOTAL_TIME_RUNNING	=  2,
};

/*
 * Hardware event to monitor via a performance monitoring counter:
 */
struct perf_counter_hw_event {
	/*
	 * The MSB of the config word signifies if the rest contains cpu
	 * specific (raw) counter configuration data, if unset, the next
	 * 7 bits are an event type and the rest of the bits are the event
	 * identifier.
	 */
	__u64			config;

	__u64			irq_period;
	__u32			record_type;
	__u32			read_format;

	__u64			disabled       :  1, /* off by default        */
				nmi	       :  1, /* NMI sampling          */
				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      */
				exclude_idle   :  1, /* don't count when idle */
				mmap           :  1, /* include mmap data     */
				munmap         :  1, /* include munmap data   */

				__reserved_1   : 53;

	__u32			extra_config_len;
	__u32			wakeup_events;	/* wakeup every n events */

	__u64			__reserved_2;
	__u64			__reserved_3;
};

/*
 * Ioctls that can be done on a perf counter fd:
 */
#define PERF_COUNTER_IOC_ENABLE		_IO('$', 0)
#define PERF_COUNTER_IOC_DISABLE	_IO('$', 1)

/*
 * 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 */

	/*
	 * Bits needed to read the hw counters in user-space.
	 *
	 *   u32 seq;
	 *   s64 count;
	 *
	 *   do {
	 *     seq = pc->lock;
	 *
	 *     barrier()
	 *     if (pc->index) {
	 *       count = pmc_read(pc->index - 1);
	 *       count += pc->offset;
	 *     } else
	 *       goto regular_read;
	 *
	 *     barrier();
	 *   } while (pc->lock != seq);
	 *
	 * NOTE: for obvious reason this only works on self-monitoring
	 *       processes.
	 */
	__u32	lock;			/* seqlock for synchronization */
	__u32	index;			/* hardware counter identifier */
	__s64	offset;			/* add to hardware counter value */

	/*
	 * Control data for the mmap() data buffer.
	 *
	 * User-space reading this value should issue an rmb(), on SMP capable
	 * platforms, after reading this value -- see perf_counter_wakeup().
	 */
	__u32   data_head;		/* head in the data section */
};

struct perf_event_header {
	__u32	type;
	__u32	size;
};

enum perf_event_type {

	PERF_EVENT_MMAP			= 1,
	PERF_EVENT_MUNMAP		= 2,

	/*
	 * Half the event type space is reserved for the counter overflow
	 * bitfields, as found in hw_event.record_type.
	 *
	 * These events will have types of the form:
	 *   PERF_EVENT_COUNTER_OVERFLOW { | __PERF_EVENT_* } *
	 */
	PERF_EVENT_COUNTER_OVERFLOW	= 1UL << 31,
	__PERF_EVENT_IP			= PERF_RECORD_IP,
	__PERF_EVENT_TID		= PERF_RECORD_TID,
	__PERF_EVENT_GROUP		= PERF_RECORD_GROUP,
	__PERF_EVENT_CALLCHAIN		= PERF_RECORD_CALLCHAIN,
};

#ifdef __KERNEL__
/*
 * Kernel-internal data types and definitions:
 */

#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>
#include <linux/hrtimer.h>
#include <linux/fs.h>
#include <asm/atomic.h>

struct task_struct;

static inline u64 perf_event_raw(struct perf_counter_hw_event *hw_event)
{
	return hw_event->config & PERF_COUNTER_RAW_MASK;
}

static inline u64 perf_event_config(struct perf_counter_hw_event *hw_event)
{
	return hw_event->config & PERF_COUNTER_CONFIG_MASK;
}

static inline u64 perf_event_type(struct perf_counter_hw_event *hw_event)
{
	return (hw_event->config & PERF_COUNTER_TYPE_MASK) >>
		PERF_COUNTER_TYPE_SHIFT;
}

static inline u64 perf_event_id(struct perf_counter_hw_event *hw_event)
{
	return hw_event->config & PERF_COUNTER_EVENT_MASK;
}

/**
 * struct hw_perf_counter - performance counter hardware details:
 */
struct hw_perf_counter {
#ifdef CONFIG_PERF_COUNTERS
	union {
		struct { /* hardware */
			u64				config;
			unsigned long			config_base;
			unsigned long			counter_base;
			int				nmi;
			unsigned int			idx;
		};
		union { /* software */
			atomic64_t			count;
			struct hrtimer			hrtimer;
		};
	};
	atomic64_t			prev_count;
	u64				irq_period;
	atomic64_t			period_left;
#endif
};

struct perf_counter;

/**
 * struct hw_perf_counter_ops - performance counter hw ops
 */
struct hw_perf_counter_ops {
	int (*enable)			(struct perf_counter *counter);
	void (*disable)			(struct perf_counter *counter);
	void (*read)			(struct perf_counter *counter);
};

/**
 * enum perf_counter_active_state - the states of a counter
 */
enum perf_counter_active_state {
	PERF_COUNTER_STATE_ERROR	= -2,
	PERF_COUNTER_STATE_OFF		= -1,
	PERF_COUNTER_STATE_INACTIVE	=  0,
	PERF_COUNTER_STATE_ACTIVE	=  1,
};

struct file;

struct perf_mmap_data {
	struct rcu_head			rcu_head;
	int				nr_pages;
	atomic_t			wakeup;
	atomic_t			head;
	atomic_t			events;
	struct perf_counter_mmap_page   *user_page;
	void 				*data_pages[0];
};

struct perf_pending_entry {
	struct perf_pending_entry *next;
	void (*func)(struct perf_pending_entry *);
};

/**
 * struct perf_counter - performance counter kernel representation:
 */
struct perf_counter {
#ifdef CONFIG_PERF_COUNTERS
	struct list_head		list_entry;
	struct list_head		event_entry;
	struct list_head		sibling_list;
	int 				nr_siblings;
	struct perf_counter		*group_leader;
	const struct hw_perf_counter_ops *hw_ops;

	enum perf_counter_active_state	state;
	enum perf_counter_active_state	prev_state;
	atomic64_t			count;

	/*
	 * 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;

	struct perf_counter_hw_event	hw_event;
	struct hw_perf_counter		hw;

	struct perf_counter_context	*ctx;
	struct task_struct		*task;
	struct file			*filp;

	struct perf_counter		*parent;
	struct list_head		child_list;

	/*
	 * 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;

	/*
	 * Protect attach/detach and child_list:
	 */
	struct mutex			mutex;

	int				oncpu;
	int				cpu;

	/* mmap bits */
	struct mutex			mmap_mutex;
	atomic_t			mmap_count;
	struct perf_mmap_data		*data;

	/* poll related */
	wait_queue_head_t		waitq;
	struct fasync_struct		*fasync;
	/* optional: for NMIs */
	struct perf_pending_entry	pending;

	void (*destroy)(struct perf_counter *);
	struct rcu_head			rcu_head;
#endif
};

/**
 * struct perf_counter_context - counter context structure
 *
 * Used as a container for task counters and CPU counters as well:
 */
struct perf_counter_context {
#ifdef CONFIG_PERF_COUNTERS
	/*
	 * Protect the states of the counters in the list,
	 * nr_active, and the list:
	 */
	spinlock_t		lock;
	/*
	 * 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.
	 */
	struct mutex		mutex;

	struct list_head	counter_list;
	struct list_head	event_list;
	int			nr_counters;
	int			nr_active;
	int			is_active;
	struct task_struct	*task;

	/*
	 * time_now is the current time in nanoseconds since an arbitrary
	 * point in the past.  For per-task counters, this is based on the
	 * task clock, and for per-cpu counters it is based on the cpu clock.
	 * time_lost is an offset from the task/cpu clock, used to make it
	 * appear that time only passes while the context is scheduled in.
	 */
	u64			time_now;
	u64			time_lost;
#endif
};

/**
 * 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;
	int				exclusive;

	/*
	 * Recursion avoidance:
	 *
	 * task, softirq, irq, nmi context
	 */
	int			recursion[4];
};

/*
 * Set by architecture code:
 */
extern int perf_max_counters;

#ifdef CONFIG_PERF_COUNTERS
extern const struct hw_perf_counter_ops *
hw_perf_counter_init(struct perf_counter *counter);

extern void perf_counter_task_sched_in(struct task_struct *task, int cpu);
extern void perf_counter_task_sched_out(struct task_struct *task, int cpu);
extern void perf_counter_task_tick(struct task_struct *task, int cpu);
extern void perf_counter_init_task(struct task_struct *child);
extern void perf_counter_exit_task(struct task_struct *child);
extern void perf_counter_do_pending(void);
extern void perf_counter_print_debug(void);
extern void perf_counter_unthrottle(void);
extern u64 hw_perf_save_disable(void);
extern void hw_perf_restore(u64 ctrl);
extern int perf_counter_task_disable(void);
extern int perf_counter_task_enable(void);
extern int hw_perf_group_sched_in(struct perf_counter *group_leader,
	       struct perf_cpu_context *cpuctx,
	       struct perf_counter_context *ctx, int cpu);
extern void perf_counter_update_userpage(struct perf_counter *counter);

extern int perf_counter_overflow(struct perf_counter *counter,
				 int nmi, struct pt_regs *regs);
/*
 * Return 1 for a software counter, 0 for a hardware counter
 */
static inline int is_software_counter(struct perf_counter *counter)
{
	return !perf_event_raw(&counter->hw_event) &&
		perf_event_type(&counter->hw_event) != PERF_TYPE_HARDWARE;
}

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

extern void perf_counter_mmap(unsigned long addr, unsigned long len,
			      unsigned long pgoff, struct file *file);

extern void perf_counter_munmap(unsigned long addr, unsigned long len,
				unsigned long pgoff, struct file *file);

#define MAX_STACK_DEPTH		255

struct perf_callchain_entry {
	u16	nr, hv, kernel, user;
	u64	ip[MAX_STACK_DEPTH];
};

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

#else
static inline void
perf_counter_task_sched_in(struct task_struct *task, int cpu)		{ }
static inline void
perf_counter_task_sched_out(struct task_struct *task, int cpu)		{ }
static inline void
perf_counter_task_tick(struct task_struct *task, int cpu)		{ }
static inline void perf_counter_init_task(struct task_struct *child)	{ }
static inline void perf_counter_exit_task(struct task_struct *child)	{ }
static inline void perf_counter_do_pending(void)			{ }
static inline void perf_counter_print_debug(void)			{ }
static inline void perf_counter_unthrottle(void)			{ }
static inline void hw_perf_restore(u64 ctrl)				{ }
static inline u64 hw_perf_save_disable(void)		      { return 0; }
static inline int perf_counter_task_disable(void)	{ return -EINVAL; }
static inline int perf_counter_task_enable(void)	{ return -EINVAL; }

static inline void
perf_swcounter_event(u32 event, u64 nr, int nmi, struct pt_regs *regs)	{ }


static inline void
perf_counter_mmap(unsigned long addr, unsigned long len,
		  unsigned long pgoff, struct file *file)		{ }

static inline void
perf_counter_munmap(unsigned long addr, unsigned long len,
		    unsigned long pgoff, struct file *file) 		{ }

#endif

#endif /* __KERNEL__ */
#endif /* _LINUX_PERF_COUNTER_H */
Commit	Line	Data
0793a61d TG	1	/*
	2	* Performance counters:
	3	*
	4	* Copyright(C) 2008, Thomas Gleixner <tglx@linutronix.de>
	5	* Copyright(C) 2008, Red Hat, Inc., Ingo Molnar
	6	*
	7	* Data type definitions, declarations, prototypes.
	8	*
	9	* Started by: Thomas Gleixner and Ingo Molnar
	10	*
	11	* For licencing details see kernel-base/COPYING
	12	*/
	13	#ifndef _LINUX_PERF_COUNTER_H
	14	#define _LINUX_PERF_COUNTER_H
	15
f3dfd265 PM	16	#include <linux/types.h>
f3dfd265 PM	17	#include <linux/ioctl.h>
9aaa131a	18	#include <asm/byteorder.h>
0793a61d TG	19
0793a61d TG	20	/*
9f66a381 IM	21	* User-space ABI bits:
	22	*/
	23
	24	/*
b8e83514	25	* hw_event.type
0793a61d	26	*/
b8e83514 PZ	27	enum perf_event_types {
	28	PERF_TYPE_HARDWARE = 0,
	29	PERF_TYPE_SOFTWARE = 1,
	30	PERF_TYPE_TRACEPOINT = 2,
	31
0793a61d	32	/*
b8e83514	33	* available TYPE space, raw is the max value.
0793a61d	34	*/
9f66a381	35
b8e83514 PZ	36	PERF_TYPE_RAW = 128,
b8e83514 PZ	37	};
6c594c21	38
b8e83514 PZ	39	/*
	40	* Generalized performance counter event types, used by the hw_event.event_id
	41	* parameter of the sys_perf_counter_open() syscall:
	42	*/
	43	enum hw_event_ids {
9f66a381	44	/*
b8e83514	45	* Common hardware events, generalized by the kernel:
9f66a381	46	*/
b8e83514 PZ	47	PERF_COUNT_CPU_CYCLES = 0,
	48	PERF_COUNT_INSTRUCTIONS = 1,
	49	PERF_COUNT_CACHE_REFERENCES = 2,
	50	PERF_COUNT_CACHE_MISSES = 3,
	51	PERF_COUNT_BRANCH_INSTRUCTIONS = 4,
	52	PERF_COUNT_BRANCH_MISSES = 5,
	53	PERF_COUNT_BUS_CYCLES = 6,
	54
	55	PERF_HW_EVENTS_MAX = 7,
	56	};
e077df4f	57
b8e83514 PZ	58	/*
	59	* Special "software" counters provided by the kernel, even if the hardware
	60	* does not support performance counters. These counters measure various
	61	* physical and sw events of the kernel (and allow the profiling of them as
	62	* well):
	63	*/
	64	enum sw_event_ids {
	65	PERF_COUNT_CPU_CLOCK = 0,
	66	PERF_COUNT_TASK_CLOCK = 1,
	67	PERF_COUNT_PAGE_FAULTS = 2,
	68	PERF_COUNT_CONTEXT_SWITCHES = 3,
	69	PERF_COUNT_CPU_MIGRATIONS = 4,
	70	PERF_COUNT_PAGE_FAULTS_MIN = 5,
	71	PERF_COUNT_PAGE_FAULTS_MAJ = 6,
	72
	73	PERF_SW_EVENTS_MAX = 7,
0793a61d TG	74	};
0793a61d TG	75
f4a2deb4 PZ	76	#define __PERF_COUNTER_MASK(name) \
	77	(((1ULL << PERF_COUNTER_##name##_BITS) - 1) << \
	78	PERF_COUNTER_##name##_SHIFT)
	79
	80	#define PERF_COUNTER_RAW_BITS 1
	81	#define PERF_COUNTER_RAW_SHIFT 63
	82	#define PERF_COUNTER_RAW_MASK __PERF_COUNTER_MASK(RAW)
	83
	84	#define PERF_COUNTER_CONFIG_BITS 63
	85	#define PERF_COUNTER_CONFIG_SHIFT 0
	86	#define PERF_COUNTER_CONFIG_MASK __PERF_COUNTER_MASK(CONFIG)
	87
	88	#define PERF_COUNTER_TYPE_BITS 7
	89	#define PERF_COUNTER_TYPE_SHIFT 56
	90	#define PERF_COUNTER_TYPE_MASK __PERF_COUNTER_MASK(TYPE)
	91
	92	#define PERF_COUNTER_EVENT_BITS 56
	93	#define PERF_COUNTER_EVENT_SHIFT 0
	94	#define PERF_COUNTER_EVENT_MASK __PERF_COUNTER_MASK(EVENT)
	95
8a057d84 PZ	96	/*
	97	* Bits that can be set in hw_event.record_type to request information
	98	* in the overflow packets.
	99	*/
	100	enum perf_counter_record_format {
	101	PERF_RECORD_IP = 1U << 0,
	102	PERF_RECORD_TID = 1U << 1,
	103	PERF_RECORD_GROUP = 1U << 2,
	104	PERF_RECORD_CALLCHAIN = 1U << 3,
	105	};
	106
53cfbf59 PM	107	/*
	108	* Bits that can be set in hw_event.read_format to request that
	109	* reads on the counter should return the indicated quantities,
	110	* in increasing order of bit value, after the counter value.
	111	*/
	112	enum perf_counter_read_format {
	113	PERF_FORMAT_TOTAL_TIME_ENABLED = 1,
	114	PERF_FORMAT_TOTAL_TIME_RUNNING = 2,
	115	};
	116
9f66a381 IM	117	/*
	118	* Hardware event to monitor via a performance monitoring counter:
	119	*/
	120	struct perf_counter_hw_event {
f4a2deb4 PZ	121	/*
	122	* The MSB of the config word signifies if the rest contains cpu
	123	* specific (raw) counter configuration data, if unset, the next
	124	* 7 bits are an event type and the rest of the bits are the event
	125	* identifier.
	126	*/
	127	__u64 config;
9f66a381	128
f3dfd265	129	__u64 irq_period;
8a057d84 PZ	130	__u32 record_type;
8a057d84 PZ	131	__u32 read_format;
9f66a381	132
2743a5b0	133	__u64 disabled : 1, /* off by default */
0475f9ea	134	nmi : 1, /* NMI sampling */
0475f9ea PM	135	inherit : 1, /* children inherit it */
	136	pinned : 1, /* must always be on PMU */
	137	exclusive : 1, /* only group on PMU */
	138	exclude_user : 1, /* don't count user */
	139	exclude_kernel : 1, /* ditto kernel */
	140	exclude_hv : 1, /* ditto hypervisor */
2743a5b0	141	exclude_idle : 1, /* don't count when idle */
0a4a9391 PZ	142	mmap : 1, /* include mmap data */
0a4a9391 PZ	143	munmap : 1, /* include munmap data */
0475f9ea	144
8a057d84	145	__reserved_1 : 53;
2743a5b0 PM	146
2743a5b0 PM	147	__u32 extra_config_len;
c457810a	148	__u32 wakeup_events; /* wakeup every n events */
9f66a381	149
f3dfd265	150	__u64 __reserved_2;
2743a5b0	151	__u64 __reserved_3;
eab656ae TG	152	};
eab656ae TG	153
d859e29f PM	154	/*
	155	* Ioctls that can be done on a perf counter fd:
	156	*/
	157	#define PERF_COUNTER_IOC_ENABLE _IO('$', 0)
	158	#define PERF_COUNTER_IOC_DISABLE _IO('$', 1)
	159
37d81828 PM	160	/*
	161	* Structure of the page that can be mapped via mmap
	162	*/
	163	struct perf_counter_mmap_page {
	164	__u32 version; /* version number of this structure */
	165	__u32 compat_version; /* lowest version this is compat with */
38ff667b PZ	166
	167	/*
	168	* Bits needed to read the hw counters in user-space.
	169	*
92f22a38 PZ	170	* u32 seq;
92f22a38 PZ	171	* s64 count;
38ff667b	172	*
a2e87d06 PZ	173	* do {
a2e87d06 PZ	174	* seq = pc->lock;
38ff667b	175	*
a2e87d06 PZ	176	* barrier()
	177	* if (pc->index) {
	178	* count = pmc_read(pc->index - 1);
	179	* count += pc->offset;
	180	* } else
	181	* goto regular_read;
38ff667b	182	*
a2e87d06 PZ	183	* barrier();
a2e87d06 PZ	184	* } while (pc->lock != seq);
38ff667b	185	*
92f22a38 PZ	186	* NOTE: for obvious reason this only works on self-monitoring
92f22a38 PZ	187	* processes.
38ff667b	188	*/
37d81828 PM	189	__u32 lock; /* seqlock for synchronization */
	190	__u32 index; /* hardware counter identifier */
	191	__s64 offset; /* add to hardware counter value */
7b732a75	192
38ff667b PZ	193	/*
	194	* Control data for the mmap() data buffer.
	195	*
	196	* User-space reading this value should issue an rmb(), on SMP capable
	197	* platforms, after reading this value -- see perf_counter_wakeup().
	198	*/
7b732a75	199	__u32 data_head; /* head in the data section */
37d81828 PM	200	};
37d81828 PM	201
5c148194 PZ	202	struct perf_event_header {
	203	__u32 type;
	204	__u32 size;
	205	};
	206
	207	enum perf_event_type {
5ed00415	208
8a057d84 PZ	209	PERF_EVENT_MMAP = 1,
8a057d84 PZ	210	PERF_EVENT_MUNMAP = 2,
0a4a9391	211
8a057d84 PZ	212	/*
	213	* Half the event type space is reserved for the counter overflow
	214	* bitfields, as found in hw_event.record_type.
	215	*
	216	* These events will have types of the form:
	217	* PERF_EVENT_COUNTER_OVERFLOW { \| __PERF_EVENT_* } *
	218	*/
	219	PERF_EVENT_COUNTER_OVERFLOW = 1UL << 31,
	220	__PERF_EVENT_IP = PERF_RECORD_IP,
	221	__PERF_EVENT_TID = PERF_RECORD_TID,
	222	__PERF_EVENT_GROUP = PERF_RECORD_GROUP,
	223	__PERF_EVENT_CALLCHAIN = PERF_RECORD_CALLCHAIN,
5c148194 PZ	224	};
5c148194 PZ	225
f3dfd265	226	#ifdef __KERNEL__
9f66a381	227	/*
f3dfd265	228	* Kernel-internal data types and definitions:
9f66a381 IM	229	*/
9f66a381 IM	230
f3dfd265 PM	231	#ifdef CONFIG_PERF_COUNTERS
	232	# include <asm/perf_counter.h>
	233	#endif
	234
	235	#include <linux/list.h>
	236	#include <linux/mutex.h>
	237	#include <linux/rculist.h>
	238	#include <linux/rcupdate.h>
	239	#include <linux/spinlock.h>
d6d020e9	240	#include <linux/hrtimer.h>
3c446b3d	241	#include <linux/fs.h>
f3dfd265 PM	242	#include <asm/atomic.h>
	243
	244	struct task_struct;
	245
f4a2deb4 PZ	246	static inline u64 perf_event_raw(struct perf_counter_hw_event *hw_event)
	247	{
	248	return hw_event->config & PERF_COUNTER_RAW_MASK;
	249	}
	250
	251	static inline u64 perf_event_config(struct perf_counter_hw_event *hw_event)
	252	{
	253	return hw_event->config & PERF_COUNTER_CONFIG_MASK;
	254	}
	255
	256	static inline u64 perf_event_type(struct perf_counter_hw_event *hw_event)
	257	{
	258	return (hw_event->config & PERF_COUNTER_TYPE_MASK) >>
	259	PERF_COUNTER_TYPE_SHIFT;
	260	}
	261
	262	static inline u64 perf_event_id(struct perf_counter_hw_event *hw_event)
	263	{
	264	return hw_event->config & PERF_COUNTER_EVENT_MASK;
	265	}
	266
0793a61d	267	/**
9f66a381	268	* struct hw_perf_counter - performance counter hardware details:
0793a61d TG	269	*/
0793a61d TG	270	struct hw_perf_counter {
ee06094f	271	#ifdef CONFIG_PERF_COUNTERS
d6d020e9 PZ	272	union {
	273	struct { /* hardware */
	274	u64 config;
	275	unsigned long config_base;
	276	unsigned long counter_base;
	277	int nmi;
	278	unsigned int idx;
	279	};
	280	union { /* software */
	281	atomic64_t count;
	282	struct hrtimer hrtimer;
	283	};
	284	};
ee06094f	285	atomic64_t prev_count;
9f66a381	286	u64 irq_period;
ee06094f IM	287	atomic64_t period_left;
ee06094f IM	288	#endif
0793a61d TG	289	};
0793a61d TG	290
621a01ea IM	291	struct perf_counter;
	292
	293	/**
	294	* struct hw_perf_counter_ops - performance counter hw ops
	295	*/
	296	struct hw_perf_counter_ops {
95cdd2e7	297	int (enable) (struct perf_counter counter);
7671581f IM	298	void (disable) (struct perf_counter counter);
7671581f IM	299	void (read) (struct perf_counter counter);
621a01ea IM	300	};
621a01ea IM	301
6a930700 IM	302	/**
	303	* enum perf_counter_active_state - the states of a counter
	304	*/
	305	enum perf_counter_active_state {
3b6f9e5c	306	PERF_COUNTER_STATE_ERROR = -2,
6a930700 IM	307	PERF_COUNTER_STATE_OFF = -1,
	308	PERF_COUNTER_STATE_INACTIVE = 0,
	309	PERF_COUNTER_STATE_ACTIVE = 1,
	310	};
	311
9b51f66d IM	312	struct file;
9b51f66d IM	313
7b732a75 PZ	314	struct perf_mmap_data {
	315	struct rcu_head rcu_head;
	316	int nr_pages;
c7138f37	317	atomic_t wakeup;
7b732a75	318	atomic_t head;
c457810a	319	atomic_t events;
7b732a75 PZ	320	struct perf_counter_mmap_page *user_page;
	321	void *data_pages[0];
	322	};
	323
671dec5d PZ	324	struct perf_pending_entry {
	325	struct perf_pending_entry *next;
	326	void (func)(struct perf_pending_entry );
925d519a PZ	327	};
925d519a PZ	328
0793a61d TG	329	/**
	330	* struct perf_counter - performance counter kernel representation:
	331	*/
	332	struct perf_counter {
ee06094f	333	#ifdef CONFIG_PERF_COUNTERS
04289bb9	334	struct list_head list_entry;
592903cd	335	struct list_head event_entry;
04289bb9	336	struct list_head sibling_list;
5c148194	337	int nr_siblings;
04289bb9	338	struct perf_counter *group_leader;
5c92d124	339	const struct hw_perf_counter_ops *hw_ops;
04289bb9	340
6a930700	341	enum perf_counter_active_state state;
c07c99b6	342	enum perf_counter_active_state prev_state;
0793a61d	343	atomic64_t count;
ee06094f	344
53cfbf59 PM	345	/*
	346	* These are the total time in nanoseconds that the counter
	347	* has been enabled (i.e. eligible to run, and the task has
	348	* been scheduled in, if this is a per-task counter)
	349	* and running (scheduled onto the CPU), respectively.
	350	*
	351	* They are computed from tstamp_enabled, tstamp_running and
	352	* tstamp_stopped when the counter is in INACTIVE or ACTIVE state.
	353	*/
	354	u64 total_time_enabled;
	355	u64 total_time_running;
	356
	357	/*
	358	* These are timestamps used for computing total_time_enabled
	359	* and total_time_running when the counter is in INACTIVE or
	360	* ACTIVE state, measured in nanoseconds from an arbitrary point
	361	* in time.
	362	* tstamp_enabled: the notional time when the counter was enabled
	363	* tstamp_running: the notional time when the counter was scheduled on
	364	* tstamp_stopped: in INACTIVE state, the notional time when the
	365	* counter was scheduled off.
	366	*/
	367	u64 tstamp_enabled;
	368	u64 tstamp_running;
	369	u64 tstamp_stopped;
	370
9f66a381	371	struct perf_counter_hw_event hw_event;
0793a61d TG	372	struct hw_perf_counter hw;
	373
	374	struct perf_counter_context *ctx;
	375	struct task_struct *task;
9b51f66d	376	struct file *filp;
0793a61d	377
9b51f66d	378	struct perf_counter *parent;
d859e29f PM	379	struct list_head child_list;
d859e29f PM	380
53cfbf59 PM	381	/*
	382	* These accumulate total time (in nanoseconds) that children
	383	* counters have been enabled and running, respectively.
	384	*/
	385	atomic64_t child_total_time_enabled;
	386	atomic64_t child_total_time_running;
	387
0793a61d	388	/*
d859e29f	389	* Protect attach/detach and child_list:
0793a61d TG	390	*/
	391	struct mutex mutex;
	392
	393	int oncpu;
	394	int cpu;
	395
7b732a75 PZ	396	/* mmap bits */
	397	struct mutex mmap_mutex;
	398	atomic_t mmap_count;
	399	struct perf_mmap_data *data;
37d81828	400
7b732a75	401	/* poll related */
0793a61d	402	wait_queue_head_t waitq;
3c446b3d	403	struct fasync_struct *fasync;
0793a61d	404	/* optional: for NMIs */
671dec5d	405	struct perf_pending_entry pending;
592903cd	406
e077df4f	407	void (destroy)(struct perf_counter );
592903cd	408	struct rcu_head rcu_head;
ee06094f	409	#endif
0793a61d TG	410	};
	411
	412	/**
	413	* struct perf_counter_context - counter context structure
	414	*
	415	* Used as a container for task counters and CPU counters as well:
	416	*/
	417	struct perf_counter_context {
	418	#ifdef CONFIG_PERF_COUNTERS
	419	/*
d859e29f PM	420	* Protect the states of the counters in the list,
d859e29f PM	421	* nr_active, and the list:
0793a61d TG	422	*/
0793a61d TG	423	spinlock_t lock;
d859e29f PM	424	/*
	425	* Protect the list of counters. Locking either mutex or lock
	426	* is sufficient to ensure the list doesn't change; to change
	427	* the list you need to lock both the mutex and the spinlock.
	428	*/
	429	struct mutex mutex;
04289bb9 IM	430
04289bb9 IM	431	struct list_head counter_list;
592903cd	432	struct list_head event_list;
0793a61d TG	433	int nr_counters;
0793a61d TG	434	int nr_active;
d859e29f	435	int is_active;
0793a61d	436	struct task_struct *task;
53cfbf59 PM	437
	438	/*
	439	* time_now is the current time in nanoseconds since an arbitrary
	440	* point in the past. For per-task counters, this is based on the
	441	* task clock, and for per-cpu counters it is based on the cpu clock.
	442	* time_lost is an offset from the task/cpu clock, used to make it
	443	* appear that time only passes while the context is scheduled in.
	444	*/
	445	u64 time_now;
	446	u64 time_lost;
0793a61d TG	447	#endif
	448	};
	449
	450	/**
	451	* struct perf_counter_cpu_context - per cpu counter context structure
	452	*/
	453	struct perf_cpu_context {
	454	struct perf_counter_context ctx;
	455	struct perf_counter_context *task_ctx;
	456	int active_oncpu;
	457	int max_pertask;
3b6f9e5c	458	int exclusive;
96f6d444 PZ	459
	460	/*
	461	* Recursion avoidance:
	462	*
	463	* task, softirq, irq, nmi context
	464	*/
	465	int recursion[4];
0793a61d TG	466	};
	467
	468	/*
	469	* Set by architecture code:
	470	*/
	471	extern int perf_max_counters;
	472
	473	#ifdef CONFIG_PERF_COUNTERS
5c92d124	474	extern const struct hw_perf_counter_ops *
621a01ea IM	475	hw_perf_counter_init(struct perf_counter *counter);
621a01ea IM	476
0793a61d TG	477	extern void perf_counter_task_sched_in(struct task_struct *task, int cpu);
	478	extern void perf_counter_task_sched_out(struct task_struct *task, int cpu);
	479	extern void perf_counter_task_tick(struct task_struct *task, int cpu);
9b51f66d IM	480	extern void perf_counter_init_task(struct task_struct *child);
9b51f66d IM	481	extern void perf_counter_exit_task(struct task_struct *child);
925d519a	482	extern void perf_counter_do_pending(void);
0793a61d	483	extern void perf_counter_print_debug(void);
1b023a96	484	extern void perf_counter_unthrottle(void);
01b2838c IM	485	extern u64 hw_perf_save_disable(void);
01b2838c IM	486	extern void hw_perf_restore(u64 ctrl);
1d1c7ddb IM	487	extern int perf_counter_task_disable(void);
1d1c7ddb IM	488	extern int perf_counter_task_enable(void);
3cbed429 PM	489	extern int hw_perf_group_sched_in(struct perf_counter *group_leader,
	490	struct perf_cpu_context *cpuctx,
	491	struct perf_counter_context *ctx, int cpu);
37d81828	492	extern void perf_counter_update_userpage(struct perf_counter *counter);
5c92d124	493
f6c7d5fe PZ	494	extern int perf_counter_overflow(struct perf_counter *counter,
f6c7d5fe PZ	495	int nmi, struct pt_regs *regs);
3b6f9e5c PM	496	/*
	497	* Return 1 for a software counter, 0 for a hardware counter
	498	*/
	499	static inline int is_software_counter(struct perf_counter *counter)
	500	{
f4a2deb4 PZ	501	return !perf_event_raw(&counter->hw_event) &&
f4a2deb4 PZ	502	perf_event_type(&counter->hw_event) != PERF_TYPE_HARDWARE;
3b6f9e5c PM	503	}
3b6f9e5c PM	504
b8e83514	505	extern void perf_swcounter_event(u32, u64, int, struct pt_regs *);
15dbf27c	506
0a4a9391 PZ	507	extern void perf_counter_mmap(unsigned long addr, unsigned long len,
	508	unsigned long pgoff, struct file *file);
	509
	510	extern void perf_counter_munmap(unsigned long addr, unsigned long len,
	511	unsigned long pgoff, struct file *file);
	512
9c03d88e	513	#define MAX_STACK_DEPTH 255
394ee076 PZ	514
394ee076 PZ	515	struct perf_callchain_entry {
9c03d88e	516	u16 nr, hv, kernel, user;
394ee076 PZ	517	u64 ip[MAX_STACK_DEPTH];
	518	};
	519
	520	extern struct perf_callchain_entry perf_callchain(struct pt_regs regs);
	521
0793a61d TG	522	#else
	523	static inline void
	524	perf_counter_task_sched_in(struct task_struct *task, int cpu) { }
	525	static inline void
	526	perf_counter_task_sched_out(struct task_struct *task, int cpu) { }
	527	static inline void
	528	perf_counter_task_tick(struct task_struct *task, int cpu) { }
9b51f66d IM	529	static inline void perf_counter_init_task(struct task_struct *child) { }
9b51f66d IM	530	static inline void perf_counter_exit_task(struct task_struct *child) { }
925d519a	531	static inline void perf_counter_do_pending(void) { }
0793a61d	532	static inline void perf_counter_print_debug(void) { }
1b023a96	533	static inline void perf_counter_unthrottle(void) { }
15dbf27c	534	static inline void hw_perf_restore(u64 ctrl) { }
01b2838c	535	static inline u64 hw_perf_save_disable(void) { return 0; }
1d1c7ddb IM	536	static inline int perf_counter_task_disable(void) { return -EINVAL; }
1d1c7ddb IM	537	static inline int perf_counter_task_enable(void) { return -EINVAL; }
15dbf27c	538
925d519a PZ	539	static inline void
	540	perf_swcounter_event(u32 event, u64 nr, int nmi, struct pt_regs *regs) { }
	541
0a4a9391 PZ	542
	543	static inline void
	544	perf_counter_mmap(unsigned long addr, unsigned long len,
	545	unsigned long pgoff, struct file *file) { }
	546
	547	static inline void
	548	perf_counter_munmap(unsigned long addr, unsigned long len,
	549	unsigned long pgoff, struct file *file) { }
	550
0793a61d TG	551	#endif
0793a61d TG	552
f3dfd265	553	#endif /* __KERNEL__ */
0793a61d	554	#endif /* _LINUX_PERF_COUNTER_H */