[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>

/*
 * User-space ABI bits:
 */

/*
 * Generalized performance counter event types, used by the hw_event.type
 * parameter of the sys_perf_counter_open() syscall:
 */
enum hw_event_types {
	/*
	 * 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):
	 */
	PERF_COUNT_CPU_CLOCK		= -1,
	PERF_COUNT_TASK_CLOCK		= -2,
	PERF_COUNT_PAGE_FAULTS		= -3,
	PERF_COUNT_CONTEXT_SWITCHES	= -4,
	PERF_COUNT_CPU_MIGRATIONS	= -5,
	PERF_COUNT_PAGE_FAULTS_MIN	= -6,
	PERF_COUNT_PAGE_FAULTS_MAJ	= -7,

	PERF_SW_EVENTS_MIN		= -8,
};

/*
 * IRQ-notification data record type:
 */
enum perf_counter_record_type {
	PERF_RECORD_SIMPLE		=  0,
	PERF_RECORD_IRQ			=  1,
	PERF_RECORD_GROUP		=  2,
};

/*
 * Hardware event to monitor via a performance monitoring counter:
 */
struct perf_counter_hw_event {
	__s64			type;

	__u64			irq_period;
	__u64			record_type;
	__u64			read_format;

	__u64			disabled       :  1, /* off by default        */
				nmi	       :  1, /* NMI sampling          */
				raw	       :  1, /* raw event type        */
				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 */

				__reserved_1   : 54;

	__u32			extra_config_len;
	__u32			__reserved_4;

	__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)

#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 <asm/atomic.h>

struct task_struct;

/**
 * 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
};

/*
 * Hardcoded buffer length limit for now, for IRQ-fed events:
 */
#define PERF_DATA_BUFLEN		2048

/**
 * struct perf_data - performance counter IRQ data sampling ...
 */
struct perf_data {
	int				len;
	int				rd_idx;
	int				overrun;
	u8				data[PERF_DATA_BUFLEN];
};

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_counter - performance counter kernel representation:
 */
struct perf_counter {
#ifdef CONFIG_PERF_COUNTERS
	struct list_head		list_entry;
	struct list_head		sibling_list;
	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;

	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;

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

	int				oncpu;
	int				cpu;

	/* read() / irq related data */
	wait_queue_head_t		waitq;
	/* optional: for NMIs */
	int				wakeup_pending;
	struct perf_data		*irqdata;
	struct perf_data		*usrdata;
	struct perf_data		data[2];
#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;
	int			nr_counters;
	int			nr_active;
	int			is_active;
	struct task_struct	*task;
#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;
};

/*
 * 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_notify(struct pt_regs *regs);
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);

/*
 * Return 1 for a software counter, 0 for a hardware counter
 */
static inline int is_software_counter(struct perf_counter *counter)
{
	return !counter->hw_event.raw && counter->hw_event.type < 0;
}

extern void perf_swcounter_event(enum hw_event_types, u64, int, struct pt_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_notify(struct pt_regs *regs)		{ }
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(enum hw_event_types event, u64 nr,
					int nmi, struct pt_regs *regs)	{ }
#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>
0793a61d TG	18
0793a61d TG	19	/*
9f66a381 IM	20	* User-space ABI bits:
	21	*/
	22
	23	/*
	24	* Generalized performance counter event types, used by the hw_event.type
	25	* parameter of the sys_perf_counter_open() syscall:
0793a61d TG	26	*/
0793a61d TG	27	enum hw_event_types {
0793a61d	28	/*
9f66a381	29	* Common hardware events, generalized by the kernel:
0793a61d	30	*/
f650a672	31	PERF_COUNT_CPU_CYCLES = 0,
9f66a381 IM	32	PERF_COUNT_INSTRUCTIONS = 1,
	33	PERF_COUNT_CACHE_REFERENCES = 2,
	34	PERF_COUNT_CACHE_MISSES = 3,
	35	PERF_COUNT_BRANCH_INSTRUCTIONS = 4,
	36	PERF_COUNT_BRANCH_MISSES = 5,
f650a672	37	PERF_COUNT_BUS_CYCLES = 6,
9f66a381	38
f650a672	39	PERF_HW_EVENTS_MAX = 7,
6c594c21	40
9f66a381 IM	41	/*
	42	* Special "software" counters provided by the kernel, even if
	43	* the hardware does not support performance counters. These
	44	* counters measure various physical and sw events of the
	45	* kernel (and allow the profiling of them as well):
	46	*/
	47	PERF_COUNT_CPU_CLOCK = -1,
	48	PERF_COUNT_TASK_CLOCK = -2,
5d6a27d8 IM	49	PERF_COUNT_PAGE_FAULTS = -3,
5d6a27d8 IM	50	PERF_COUNT_CONTEXT_SWITCHES = -4,
6c594c21	51	PERF_COUNT_CPU_MIGRATIONS = -5,
ac17dc8e PZ	52	PERF_COUNT_PAGE_FAULTS_MIN = -6,
ac17dc8e PZ	53	PERF_COUNT_PAGE_FAULTS_MAJ = -7,
6c594c21	54
ac17dc8e	55	PERF_SW_EVENTS_MIN = -8,
0793a61d TG	56	};
	57
	58	/*
	59	* IRQ-notification data record type:
	60	*/
9f66a381 IM	61	enum perf_counter_record_type {
	62	PERF_RECORD_SIMPLE = 0,
	63	PERF_RECORD_IRQ = 1,
	64	PERF_RECORD_GROUP = 2,
0793a61d TG	65	};
0793a61d TG	66
9f66a381 IM	67	/*
	68	* Hardware event to monitor via a performance monitoring counter:
	69	*/
	70	struct perf_counter_hw_event {
f3dfd265	71	__s64 type;
9f66a381	72
f3dfd265	73	__u64 irq_period;
2743a5b0 PM	74	__u64 record_type;
2743a5b0 PM	75	__u64 read_format;
9f66a381	76
2743a5b0	77	__u64 disabled : 1, /* off by default */
0475f9ea PM	78	nmi : 1, /* NMI sampling */
	79	raw : 1, /* raw event type */
	80	inherit : 1, /* children inherit it */
	81	pinned : 1, /* must always be on PMU */
	82	exclusive : 1, /* only group on PMU */
	83	exclude_user : 1, /* don't count user */
	84	exclude_kernel : 1, /* ditto kernel */
	85	exclude_hv : 1, /* ditto hypervisor */
2743a5b0	86	exclude_idle : 1, /* don't count when idle */
0475f9ea	87
2485e518	88	__reserved_1 : 54;
2743a5b0 PM	89
	90	__u32 extra_config_len;
	91	__u32 __reserved_4;
9f66a381	92
f3dfd265	93	__u64 __reserved_2;
2743a5b0	94	__u64 __reserved_3;
eab656ae TG	95	};
eab656ae TG	96
d859e29f PM	97	/*
	98	* Ioctls that can be done on a perf counter fd:
	99	*/
	100	#define PERF_COUNTER_IOC_ENABLE _IO('$', 0)
	101	#define PERF_COUNTER_IOC_DISABLE _IO('$', 1)
	102
f3dfd265	103	#ifdef __KERNEL__
9f66a381	104	/*
f3dfd265	105	* Kernel-internal data types and definitions:
9f66a381 IM	106	*/
9f66a381 IM	107
f3dfd265 PM	108	#ifdef CONFIG_PERF_COUNTERS
	109	# include <asm/perf_counter.h>
	110	#endif
	111
	112	#include <linux/list.h>
	113	#include <linux/mutex.h>
	114	#include <linux/rculist.h>
	115	#include <linux/rcupdate.h>
	116	#include <linux/spinlock.h>
d6d020e9	117	#include <linux/hrtimer.h>
f3dfd265 PM	118	#include <asm/atomic.h>
	119
	120	struct task_struct;
	121
0793a61d	122	/**
9f66a381	123	* struct hw_perf_counter - performance counter hardware details:
0793a61d TG	124	*/
0793a61d TG	125	struct hw_perf_counter {
ee06094f	126	#ifdef CONFIG_PERF_COUNTERS
d6d020e9 PZ	127	union {
	128	struct { /* hardware */
	129	u64 config;
	130	unsigned long config_base;
	131	unsigned long counter_base;
	132	int nmi;
	133	unsigned int idx;
	134	};
	135	union { /* software */
	136	atomic64_t count;
	137	struct hrtimer hrtimer;
	138	};
	139	};
ee06094f	140	atomic64_t prev_count;
9f66a381	141	u64 irq_period;
ee06094f IM	142	atomic64_t period_left;
ee06094f IM	143	#endif
0793a61d TG	144	};
	145
	146	/*
	147	* Hardcoded buffer length limit for now, for IRQ-fed events:
	148	*/
9f66a381	149	#define PERF_DATA_BUFLEN 2048
0793a61d TG	150
	151	/**
	152	* struct perf_data - performance counter IRQ data sampling ...
	153	*/
	154	struct perf_data {
9f66a381 IM	155	int len;
	156	int rd_idx;
	157	int overrun;
	158	u8 data[PERF_DATA_BUFLEN];
0793a61d TG	159	};
0793a61d TG	160
621a01ea IM	161	struct perf_counter;
	162
	163	/**
	164	* struct hw_perf_counter_ops - performance counter hw ops
	165	*/
	166	struct hw_perf_counter_ops {
95cdd2e7	167	int (enable) (struct perf_counter counter);
7671581f IM	168	void (disable) (struct perf_counter counter);
7671581f IM	169	void (read) (struct perf_counter counter);
621a01ea IM	170	};
621a01ea IM	171
6a930700 IM	172	/**
	173	* enum perf_counter_active_state - the states of a counter
	174	*/
	175	enum perf_counter_active_state {
3b6f9e5c	176	PERF_COUNTER_STATE_ERROR = -2,
6a930700 IM	177	PERF_COUNTER_STATE_OFF = -1,
	178	PERF_COUNTER_STATE_INACTIVE = 0,
	179	PERF_COUNTER_STATE_ACTIVE = 1,
	180	};
	181
9b51f66d IM	182	struct file;
9b51f66d IM	183
0793a61d TG	184	/**
	185	* struct perf_counter - performance counter kernel representation:
	186	*/
	187	struct perf_counter {
ee06094f	188	#ifdef CONFIG_PERF_COUNTERS
04289bb9 IM	189	struct list_head list_entry;
	190	struct list_head sibling_list;
	191	struct perf_counter *group_leader;
5c92d124	192	const struct hw_perf_counter_ops *hw_ops;
04289bb9	193
6a930700	194	enum perf_counter_active_state state;
c07c99b6	195	enum perf_counter_active_state prev_state;
0793a61d	196	atomic64_t count;
ee06094f	197
9f66a381	198	struct perf_counter_hw_event hw_event;
0793a61d TG	199	struct hw_perf_counter hw;
	200
	201	struct perf_counter_context *ctx;
	202	struct task_struct *task;
9b51f66d	203	struct file *filp;
0793a61d	204
9b51f66d	205	struct perf_counter *parent;
d859e29f PM	206	struct list_head child_list;
d859e29f PM	207
0793a61d	208	/*
d859e29f	209	* Protect attach/detach and child_list:
0793a61d TG	210	*/
	211	struct mutex mutex;
	212
	213	int oncpu;
	214	int cpu;
	215
0793a61d TG	216	/* read() / irq related data */
	217	wait_queue_head_t waitq;
	218	/* optional: for NMIs */
	219	int wakeup_pending;
	220	struct perf_data *irqdata;
	221	struct perf_data *usrdata;
	222	struct perf_data data[2];
ee06094f	223	#endif
0793a61d TG	224	};
	225
	226	/**
	227	* struct perf_counter_context - counter context structure
	228	*
	229	* Used as a container for task counters and CPU counters as well:
	230	*/
	231	struct perf_counter_context {
	232	#ifdef CONFIG_PERF_COUNTERS
	233	/*
d859e29f PM	234	* Protect the states of the counters in the list,
d859e29f PM	235	* nr_active, and the list:
0793a61d TG	236	*/
0793a61d TG	237	spinlock_t lock;
d859e29f PM	238	/*
	239	* Protect the list of counters. Locking either mutex or lock
	240	* is sufficient to ensure the list doesn't change; to change
	241	* the list you need to lock both the mutex and the spinlock.
	242	*/
	243	struct mutex mutex;
04289bb9 IM	244
04289bb9 IM	245	struct list_head counter_list;
0793a61d TG	246	int nr_counters;
0793a61d TG	247	int nr_active;
d859e29f	248	int is_active;
0793a61d TG	249	struct task_struct *task;
	250	#endif
	251	};
	252
	253	/**
	254	* struct perf_counter_cpu_context - per cpu counter context structure
	255	*/
	256	struct perf_cpu_context {
	257	struct perf_counter_context ctx;
	258	struct perf_counter_context *task_ctx;
	259	int active_oncpu;
	260	int max_pertask;
3b6f9e5c	261	int exclusive;
0793a61d TG	262	};
	263
	264	/*
	265	* Set by architecture code:
	266	*/
	267	extern int perf_max_counters;
	268
	269	#ifdef CONFIG_PERF_COUNTERS
5c92d124	270	extern const struct hw_perf_counter_ops *
621a01ea IM	271	hw_perf_counter_init(struct perf_counter *counter);
621a01ea IM	272
0793a61d TG	273	extern void perf_counter_task_sched_in(struct task_struct *task, int cpu);
	274	extern void perf_counter_task_sched_out(struct task_struct *task, int cpu);
	275	extern void perf_counter_task_tick(struct task_struct *task, int cpu);
9b51f66d IM	276	extern void perf_counter_init_task(struct task_struct *child);
9b51f66d IM	277	extern void perf_counter_exit_task(struct task_struct *child);
0793a61d TG	278	extern void perf_counter_notify(struct pt_regs *regs);
0793a61d TG	279	extern void perf_counter_print_debug(void);
1b023a96	280	extern void perf_counter_unthrottle(void);
01b2838c IM	281	extern u64 hw_perf_save_disable(void);
01b2838c IM	282	extern void hw_perf_restore(u64 ctrl);
1d1c7ddb IM	283	extern int perf_counter_task_disable(void);
1d1c7ddb IM	284	extern int perf_counter_task_enable(void);
3cbed429 PM	285	extern int hw_perf_group_sched_in(struct perf_counter *group_leader,
	286	struct perf_cpu_context *cpuctx,
	287	struct perf_counter_context *ctx, int cpu);
5c92d124	288
3b6f9e5c PM	289	/*
	290	* Return 1 for a software counter, 0 for a hardware counter
	291	*/
	292	static inline int is_software_counter(struct perf_counter *counter)
	293	{
	294	return !counter->hw_event.raw && counter->hw_event.type < 0;
	295	}
	296
15dbf27c PZ	297	extern void perf_swcounter_event(enum hw_event_types, u64, int, struct pt_regs *);
15dbf27c PZ	298
0793a61d TG	299	#else
	300	static inline void
	301	perf_counter_task_sched_in(struct task_struct *task, int cpu) { }
	302	static inline void
	303	perf_counter_task_sched_out(struct task_struct *task, int cpu) { }
	304	static inline void
	305	perf_counter_task_tick(struct task_struct *task, int cpu) { }
9b51f66d IM	306	static inline void perf_counter_init_task(struct task_struct *child) { }
9b51f66d IM	307	static inline void perf_counter_exit_task(struct task_struct *child) { }
0793a61d TG	308	static inline void perf_counter_notify(struct pt_regs *regs) { }
0793a61d TG	309	static inline void perf_counter_print_debug(void) { }
1b023a96	310	static inline void perf_counter_unthrottle(void) { }
15dbf27c	311	static inline void hw_perf_restore(u64 ctrl) { }
01b2838c	312	static inline u64 hw_perf_save_disable(void) { return 0; }
1d1c7ddb IM	313	static inline int perf_counter_task_disable(void) { return -EINVAL; }
1d1c7ddb IM	314	static inline int perf_counter_task_enable(void) { return -EINVAL; }
15dbf27c PZ	315
	316	static inline void perf_swcounter_event(enum hw_event_types event, u64 nr,
	317	int nmi, struct pt_regs *regs) { }
0793a61d TG	318	#endif
0793a61d TG	319
f3dfd265	320	#endif /* __KERNEL__ */
0793a61d	321	#endif /* _LINUX_PERF_COUNTER_H */