perf_counter: Split the mmap control page in two parts
[safe/jmp/linux-2.6] / include / linux / perf_counter.h
index cc3a75a..489d5cb 100644 (file)
 /*
  *  Performance counters:
  *
- *   Copyright(C) 2008, Thomas Gleixner <tglx@linutronix.de>
- *   Copyright(C) 2008, Red Hat, Inc., Ingo Molnar
+ *    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
  *
  *  Data type definitions, declarations, prototypes.
  *
- *  Started by: Thomas Gleixner and Ingo Molnar
+ *    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 <asm/atomic.h>
-
-#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>
-
-struct task_struct;
+#include <linux/types.h>
+#include <linux/ioctl.h>
+#include <asm/byteorder.h>
 
 /*
  * User-space ABI bits:
  */
 
 /*
- * Generalized performance counter event types, used by the hw_event.type
- * parameter of the sys_perf_counter_open() syscall:
+ * attr.type
+ */
+enum perf_type_id {
+       PERF_TYPE_HARDWARE                      = 0,
+       PERF_TYPE_SOFTWARE                      = 1,
+       PERF_TYPE_TRACEPOINT                    = 2,
+       PERF_TYPE_HW_CACHE                      = 3,
+       PERF_TYPE_RAW                           = 4,
+
+       PERF_TYPE_MAX,                          /* non-ABI */
+};
+
+/*
+ * Generalized performance counter event types, used by the
+ * attr.event_id parameter of the sys_perf_counter_open()
+ * syscall:
  */
-enum hw_event_types {
+enum perf_hw_id {
        /*
         * 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_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,
+
+       PERF_COUNT_HW_MAX,                      /* non-ABI */
+};
+
+/*
+ * Generalized hardware cache counters:
+ *
+ *       { L1-D, L1-I, LLC, ITLB, DTLB, BPU } x
+ *       { read, write, prefetch } x
+ *       { accesses, misses }
+ */
+enum perf_hw_cache_id {
+       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 */
+};
+
+enum perf_hw_cache_op_id {
+       PERF_COUNT_HW_CACHE_OP_READ             = 0,
+       PERF_COUNT_HW_CACHE_OP_WRITE            = 1,
+       PERF_COUNT_HW_CACHE_OP_PREFETCH         = 2,
+
+       PERF_COUNT_HW_CACHE_OP_MAX,             /* non-ABI */
+};
+
+enum perf_hw_cache_op_result_id {
+       PERF_COUNT_HW_CACHE_RESULT_ACCESS       = 0,
+       PERF_COUNT_HW_CACHE_RESULT_MISS         = 1,
+
+       PERF_COUNT_HW_CACHE_RESULT_MAX,         /* non-ABI */
+};
+
+/*
+ * 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 perf_sw_ids {
+       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 */
+};
+
+/*
+ * Bits that can be set in attr.sample_type to request information
+ * in the overflow packets.
+ */
+enum perf_counter_sample_format {
+       PERF_SAMPLE_IP                          = 1U << 0,
+       PERF_SAMPLE_TID                         = 1U << 1,
+       PERF_SAMPLE_TIME                        = 1U << 2,
+       PERF_SAMPLE_ADDR                        = 1U << 3,
+       PERF_SAMPLE_GROUP                       = 1U << 4,
+       PERF_SAMPLE_CALLCHAIN                   = 1U << 5,
+       PERF_SAMPLE_ID                          = 1U << 6,
+       PERF_SAMPLE_CPU                         = 1U << 7,
+       PERF_SAMPLE_PERIOD                      = 1U << 8,
+
+       PERF_SAMPLE_MAX = 1U << 9,              /* non-ABI */
+};
+
+/*
+ * Bits that can be set in attr.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          = 1U << 0,
+       PERF_FORMAT_TOTAL_TIME_RUNNING          = 1U << 1,
+       PERF_FORMAT_ID                          = 1U << 2,
+
+       PERF_FORMAT_MAX = 1U << 3,              /* non-ABI */
+};
+
+#define PERF_ATTR_SIZE_VER0    64      /* sizeof first published struct */
+
+/*
+ * Hardware event to monitor via a performance monitoring counter:
+ */
+struct perf_counter_attr {
 
-       PERF_HW_EVENTS_MAX              =  7,
+       /*
+        * Major type: hardware/software/tracepoint/etc.
+        */
+       __u32                   type;
 
        /*
-        * 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):
+        * Size of the attr structure, for fwd/bwd compat.
         */
-       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,
+       __u32                   size;
 
-       PERF_SW_EVENTS_MIN              = -6,
+       /*
+        * Type specific configuration information.
+        */
+       __u64                   config;
+
+       union {
+               __u64           sample_period;
+               __u64           sample_freq;
+       };
+
+       __u64                   sample_type;
+       __u64                   read_format;
+
+       __u64                   disabled       :  1, /* off by default        */
+                               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     */
+                               comm           :  1, /* include comm data     */
+                               freq           :  1, /* use freq, not period  */
+
+                               __reserved_1   : 53;
+
+       __u32                   wakeup_events;  /* wakeup every n events */
+       __u32                   __reserved_2;
+
+       __u64                   __reserved_3;
 };
 
 /*
- * IRQ-notification data record type:
+ * Ioctls that can be done on a perf counter fd:
  */
-enum perf_counter_record_type {
-       PERF_RECORD_SIMPLE              =  0,
-       PERF_RECORD_IRQ                 =  1,
-       PERF_RECORD_GROUP               =  2,
+#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)
+
+enum perf_counter_ioc_flags {
+       PERF_IOC_FLAG_GROUP             = 1U << 0,
 };
 
 /*
- * Hardware event to monitor via a performance monitoring counter:
+ * Structure of the page that can be mapped via mmap
  */
-struct perf_counter_hw_event {
-       s64                     type;
+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 */
+
+               /*
+                * Hole for extension of the self monitor capabilities
+                */
+
+       __u64   __reserved[125];        /* align to 1k */
+
+       /*
+        * Control data for the mmap() data buffer.
+        *
+        * 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.
+        */
+       __u64   data_head;              /* head in the data section */
+       __u64   data_tail;              /* user-space written tail */
+};
+
+#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)
+#define PERF_EVENT_MISC_OVERFLOW               (1 << 2)
+
+struct perf_event_header {
+       __u32   type;
+       __u16   misc;
+       __u16   size;
+};
+
+enum perf_event_type {
+
+       /*
+        * The MMAP events record the PROT_EXEC mappings so that we can
+        * correlate userspace IPs to code. They have the following structure:
+        *
+        * struct {
+        *      struct perf_event_header        header;
+        *
+        *      u32                             pid, tid;
+        *      u64                             addr;
+        *      u64                             len;
+        *      u64                             pgoff;
+        *      char                            filename[];
+        * };
+        */
+       PERF_EVENT_MMAP                 = 1,
+
+       /*
+        * struct {
+        *      struct perf_event_header        header;
+        *      u64                             id;
+        *      u64                             lost;
+        * };
+        */
+       PERF_EVENT_LOST                 = 2,
+
+       /*
+        * struct {
+        *      struct perf_event_header        header;
+        *
+        *      u32                             pid, tid;
+        *      char                            comm[];
+        * };
+        */
+       PERF_EVENT_COMM                 = 3,
+
+       /*
+        * struct {
+        *      struct perf_event_header        header;
+        *      u64                             time;
+        *      u64                             id;
+        *      u64                             sample_period;
+        * };
+        */
+       PERF_EVENT_PERIOD               = 4,
+
+       /*
+        * struct {
+        *      struct perf_event_header        header;
+        *      u64                             time;
+        *      u64                             id;
+        * };
+        */
+       PERF_EVENT_THROTTLE             = 5,
+       PERF_EVENT_UNTHROTTLE           = 6,
 
-       u64                     irq_period;
-       u32                     record_type;
+       /*
+        * struct {
+        *      struct perf_event_header        header;
+        *      u32                             pid, ppid;
+        * };
+        */
+       PERF_EVENT_FORK                 = 7,
 
-       u32                     disabled     :  1, /* off by default      */
-                               nmi          :  1, /* NMI sampling        */
-                               raw          :  1, /* raw event type      */
-                               inherit      :  1, /* children inherit it */
-                               __reserved_1 : 28;
+       /*
+        * When header.misc & PERF_EVENT_MISC_OVERFLOW the event_type field
+        * will be PERF_SAMPLE_*
+        *
+        * struct {
+        *      struct perf_event_header        header;
+        *
+        *      { u64                   ip;       } && PERF_SAMPLE_IP
+        *      { u32                   pid, tid; } && PERF_SAMPLE_TID
+        *      { u64                   time;     } && PERF_SAMPLE_TIME
+        *      { u64                   addr;     } && PERF_SAMPLE_ADDR
+        *      { u64                   config;   } && PERF_SAMPLE_CONFIG
+        *      { u32                   cpu, res; } && PERF_SAMPLE_CPU
+        *
+        *      { u64                   nr;
+        *        { u64 id, val; }      cnt[nr];  } && PERF_SAMPLE_GROUP
+        *
+        *      { u64                   nr,
+        *        u64                   ips[nr];  } && PERF_SAMPLE_CALLCHAIN
+        * };
+        */
+};
+
+enum perf_callchain_context {
+       PERF_CONTEXT_HV                 = (__u64)-32,
+       PERF_CONTEXT_KERNEL             = (__u64)-128,
+       PERF_CONTEXT_USER               = (__u64)-512,
+
+       PERF_CONTEXT_GUEST              = (__u64)-2048,
+       PERF_CONTEXT_GUEST_KERNEL       = (__u64)-2176,
+       PERF_CONTEXT_GUEST_USER         = (__u64)-2560,
 
-       u64                     __reserved_2;
+       PERF_CONTEXT_MAX                = (__u64)-4095,
 };
 
+#ifdef __KERNEL__
 /*
- * Kernel-internal data types:
+ * 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 <linux/pid_namespace.h>
+#include <asm/atomic.h>
+
+#define PERF_MAX_STACK_DEPTH           255
+
+struct perf_callchain_entry {
+       __u64                           nr;
+       __u64                           ip[PERF_MAX_STACK_DEPTH];
+};
+
+struct task_struct;
+
 /**
  * struct hw_perf_counter - performance counter hardware details:
  */
 struct hw_perf_counter {
 #ifdef CONFIG_PERF_COUNTERS
-       u64                             config;
-       unsigned long                   config_base;
-       unsigned long                   counter_base;
-       int                             nmi;
-       unsigned int                    idx;
+       union {
+               struct { /* hardware */
+                       u64             config;
+                       unsigned long   config_base;
+                       unsigned long   counter_base;
+                       int             idx;
+               };
+               union { /* software */
+                       atomic64_t      count;
+                       struct hrtimer  hrtimer;
+               };
+       };
        atomic64_t                      prev_count;
-       u64                             irq_period;
+       u64                             sample_period;
+       u64                             last_period;
        atomic64_t                      period_left;
-#endif
-};
+       u64                             interrupts;
 
-/*
- * 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];
+       u64                             freq_count;
+       u64                             freq_interrupts;
+       u64                             freq_stamp;
+#endif
 };
 
 struct perf_counter;
 
 /**
- * struct hw_perf_counter_ops - performance counter hw ops
+ * struct pmu - generic performance monitoring unit
  */
-struct hw_perf_counter_ops {
+struct pmu {
        int (*enable)                   (struct perf_counter *counter);
        void (*disable)                 (struct perf_counter *counter);
        void (*read)                    (struct perf_counter *counter);
+       void (*unthrottle)              (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,
@@ -148,42 +448,120 @@ enum perf_counter_active_state {
 
 struct file;
 
+struct perf_mmap_data {
+       struct rcu_head                 rcu_head;
+       int                             nr_pages;       /* nr of data pages  */
+       int                             writable;       /* are we writable   */
+       int                             nr_locked;      /* nr pages mlocked  */
+
+       atomic_t                        poll;           /* POLL_ for wakeups */
+       atomic_t                        events;         /* event limit       */
+
+       atomic_long_t                   head;           /* write position    */
+       atomic_long_t                   done_head;      /* completed head    */
+
+       atomic_t                        lock;           /* concurrent writes */
+       atomic_t                        wakeup;         /* needs a wakeup    */
+       atomic_t                        lost;           /* nr records lost   */
+
+       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;
+       const struct pmu                *pmu;
 
        enum perf_counter_active_state  state;
        atomic64_t                      count;
 
-       struct perf_counter_hw_event    hw_event;
+       /*
+        * 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_attr        attr;
        struct hw_perf_counter          hw;
 
        struct perf_counter_context     *ctx;
-       struct task_struct              *task;
        struct file                     *filp;
 
-       struct perf_counter             *parent;
        /*
-        * Protect attach/detach:
+        * These accumulate total time (in nanoseconds) that children
+        * counters have been enabled and running, respectively.
         */
-       struct mutex                    mutex;
+       atomic64_t                      child_total_time_enabled;
+       atomic64_t                      child_total_time_running;
+
+       /*
+        * Protect attach/detach and child_list:
+        */
+       struct mutex                    child_mutex;
+       struct list_head                child_list;
+       struct perf_counter             *parent;
 
        int                             oncpu;
        int                             cpu;
 
-       /* read() / irq related data */
+       struct list_head                owner_entry;
+       struct task_struct              *owner;
+
+       /* mmap bits */
+       struct mutex                    mmap_mutex;
+       atomic_t                        mmap_count;
+       struct perf_mmap_data           *data;
+
+       /* poll related */
        wait_queue_head_t               waitq;
-       /* optional: for NMIs */
-       int                             wakeup_pending;
-       struct perf_data                *irqdata;
-       struct perf_data                *usrdata;
-       struct perf_data                data[2];
+       struct fasync_struct            *fasync;
+
+       /* delayed work for NMIs and such */
+       int                             pending_wakeup;
+       int                             pending_kill;
+       int                             pending_disable;
+       struct perf_pending_entry       pending;
+
+       atomic_t                        event_limit;
+
+       void (*destroy)(struct perf_counter *);
+       struct rcu_head                 rcu_head;
+
+       struct pid_namespace            *ns;
+       u64                             id;
 #endif
 };
 
@@ -193,17 +571,41 @@ struct perf_counter {
  * Used as a container for task counters and CPU counters as well:
  */
 struct perf_counter_context {
-#ifdef CONFIG_PERF_COUNTERS
        /*
-        * Protect the list of counters:
+        * Protect the states of the counters in the list,
+        * nr_active, and the list:
         */
-       spinlock_t              lock;
+       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;
-       struct task_struct      *task;
-#endif
+       struct list_head                counter_list;
+       struct list_head                event_list;
+       int                             nr_counters;
+       int                             nr_active;
+       int                             is_active;
+       atomic_t                        refcount;
+       struct task_struct              *task;
+
+       /*
+        * Context clock, runs when context enabled.
+        */
+       u64                             time;
+       u64                             timestamp;
+
+       /*
+        * These fields let us detect when two contexts have both
+        * been cloned (inherited) from a common ancestor.
+        */
+       struct perf_counter_context     *parent_ctx;
+       u64                             parent_gen;
+       u64                             generation;
+       int                             pin_count;
+       struct rcu_head                 rcu_head;
 };
 
 /**
@@ -214,44 +616,128 @@ struct perf_cpu_context {
        struct perf_counter_context     *task_ctx;
        int                             active_oncpu;
        int                             max_pertask;
+       int                             exclusive;
+
+       /*
+        * Recursion avoidance:
+        *
+        * task, softirq, irq, nmi context
+        */
+       int                             recursion[4];
 };
 
+#ifdef CONFIG_PERF_COUNTERS
+
 /*
  * 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 const struct pmu *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_sched_out(struct task_struct *task,
+                                       struct task_struct *next, 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 int 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_free_task(struct task_struct *task);
+extern void set_perf_counter_pending(void);
+extern void perf_counter_do_pending(void);
 extern void perf_counter_print_debug(void);
-extern u64 hw_perf_save_disable(void);
-extern void hw_perf_restore(u64 ctrl);
+extern void __perf_disable(void);
+extern bool __perf_enable(void);
+extern void perf_disable(void);
+extern void perf_enable(void);
 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);
+
+struct perf_sample_data {
+       struct pt_regs                  *regs;
+       u64                             addr;
+       u64                             period;
+};
+
+extern int perf_counter_overflow(struct perf_counter *counter, int nmi,
+                                struct perf_sample_data *data);
+
+/*
+ * Return 1 for a software counter, 0 for a hardware counter
+ */
+static inline int is_software_counter(struct perf_counter *counter)
+{
+       return (counter->attr.type != PERF_TYPE_RAW) &&
+               (counter->attr.type != PERF_TYPE_HARDWARE) &&
+               (counter->attr.type != PERF_TYPE_HW_CACHE);
+}
+
+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);
+}
+
+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);
+}
+
+extern void perf_counter_comm(struct task_struct *tsk);
+extern void perf_counter_fork(struct task_struct *tsk);
+
+extern struct perf_callchain_entry *perf_callchain(struct pt_regs *regs);
+
+extern int sysctl_perf_counter_paranoid;
+extern int sysctl_perf_counter_mlock;
+extern int sysctl_perf_counter_sample_rate;
+
+extern void perf_counter_init(void);
+
+#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
 
 #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)         { }
+perf_counter_task_sched_out(struct task_struct *task,
+                           struct task_struct *next, 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 int perf_counter_init_task(struct task_struct *child)    { return 0; }
 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_free_task(struct task_struct *task)    { }
+static inline void perf_counter_do_pending(void)                       { }
 static inline void perf_counter_print_debug(void)                      { }
-static inline void hw_perf_restore(u64 ctrl)                   { }
-static inline u64 hw_perf_save_disable(void)                 { return 0; }
+static inline void perf_disable(void)                                  { }
+static inline void perf_enable(void)                                   { }
 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, u64 addr)                    { }
+
+static inline void perf_counter_mmap(struct vm_area_struct *vma)       { }
+static inline void perf_counter_comm(struct task_struct *tsk)          { }
+static inline void perf_counter_fork(struct task_struct *tsk)          { }
+static inline void perf_counter_init(void)                             { }
 #endif
 
+#endif /* __KERNEL__ */
 #endif /* _LINUX_PERF_COUNTER_H */