perf-util.h

Go to the documentation of this file.

// -*-Mode: C++;-*- // technically C99

// * BeginRiceCopyright *****************************************************
//
// $HeadURL$
// $Id$
//
// --------------------------------------------------------------------------
// Part of HPCToolkit (hpctoolkit.org)
//
// Information about sources of support for research and development of
// HPCToolkit is at 'hpctoolkit.org' and in 'README.Acknowledgments'.
// --------------------------------------------------------------------------
//
// Copyright ((c)) 2002-2019, Rice University
// All rights reserved.
//
// Redistribution and use in source and binary forms, with or without
// modification, are permitted provided that the following conditions are
// met:
//
// * Redistributions of source code must retain the above copyright
//   notice, this list of conditions and the following disclaimer.
//
// * Redistributions in binary form must reproduce the above copyright
//   notice, this list of conditions and the following disclaimer in the
//   documentation and/or other materials provided with the distribution.
//
// * Neither the name of Rice University (RICE) nor the names of its
//   contributors may be used to endorse or promote products derived from
//   this software without specific prior written permission.
//
// This software is provided by RICE and contributors "as is" and any
// express or implied warranties, including, but not limited to, the
// implied warranties of merchantability and fitness for a particular
// purpose are disclaimed. In no event shall RICE or contributors be

#ifndef __PERF_UTIL_H__
#define __PERF_UTIL_H__

#include <sys/syscall.h> 

#include <unistd.h>

#include <linux/types.h>
#include <linux/perf_event.h>
#include <linux/version.h>

#include <lib/prof-lean/hpcrun-fmt.h>
#include <sample_event.h>

#include "perf_constants.h"

/******************************************************************************
 * macros
 *****************************************************************************/

#define KERNEL_SAMPLING_ENABLED (LINUX_VERSION_CODE >= KERNEL_VERSION(3,7,0))



// the number of maximum frames (call chains) 
// For kernel only call chain, I think 32 is a good number.
// If we include user call chains, it should be bigger than that.
#define MAX_CALLCHAIN_FRAMES 32

// --------------------------------------------------------------------
// threshold data structure (period or frequency
// --------------------------------------------------------------------

enum threshold_e { PERIOD, FREQUENCY };

struct event_threshold_s {
  long             threshold_num;
  enum threshold_e threshold_type;
};


/******************************************************************************
 * Data types
 *****************************************************************************/

// data from perf's mmap. See perf_event_open man page
typedef struct perf_mmap_data_s {
  struct perf_event_header header;
  u64    sample_id;  /* if PERF_SAMPLE_IDENTIFIER */
  u64    ip;         /* if PERF_SAMPLE_IP */
  u32    pid, tid;   /* if PERF_SAMPLE_TID */
  u64    time;       /* if PERF_SAMPLE_TIME */
  u64    addr;       /* if PERF_SAMPLE_ADDR */
  u64    id;         /* if PERF_SAMPLE_ID */
  u64    stream_id;  /* if PERF_SAMPLE_STREAM_ID */
  u32    cpu, res;   /* if PERF_SAMPLE_CPU */
  u64    period;     /* if PERF_SAMPLE_PERIOD */
                     /* if PERF_SAMPLE_READ */
  u64    nr;         /* if PERF_SAMPLE_CALLCHAIN */
  u64    ips[MAX_CALLCHAIN_FRAMES];       /* if PERF_SAMPLE_CALLCHAIN */
  u32    size;       /* if PERF_SAMPLE_RAW */
  char   *data;      /* if PERF_SAMPLE_RAW */
  /* if PERF_SAMPLE_BRANCH_STACK */
  
                     /* if PERF_SAMPLE_BRANCH_STACK */
  u64    abi;        /* if PERF_SAMPLE_REGS_USER */
  u64    *regs;
                     /* if PERF_SAMPLE_REGS_USER */
  u64    stack_size;             /* if PERF_SAMPLE_STACK_USER */
  char   *stack_data; /* if PERF_SAMPLE_STACK_USER */
  u64    stack_dyn_size;         /* if PERF_SAMPLE_STACK_USER &&
                                     size != 0 */
  u64    weight;     /* if PERF_SAMPLE_WEIGHT */
  u64    data_src;   /* if PERF_SAMPLE_DATA_SRC */
  u64    transaction;/* if PERF_SAMPLE_TRANSACTION */
  u64    intr_abi;        /* if PERF_SAMPLE_REGS_INTR */
  u64    *intr_regs;
                     /* if PERF_SAMPLE_REGS_INTR */

  // header information in the buffer
  u32   header_misc; /* information about the sample */
  u32   header_type; /* either sample record or other */

  // only for PERF_RECORD_SWITCH
  u64   context_switch_time;

} perf_mmap_data_t;

// forward declaratio for custom event
struct event_custom_s;

// --------------------------------------------------------------
// main data structure to store the information of an event.
// this structure is designed to be created once during the initialization.
// this code doesn't work if the number of events change dynamically.
//
// this data is designed to be stored in evlist.h's _ev_t.event_info
// --------------------------------------------------------------
typedef struct event_info_s {

  struct perf_event_attr attr; // the event attribute
  struct event_custom_s *metric_custom; // pointer to the predefined metric

} event_info_t;


typedef struct perf_event_mmap_page pe_mmap_t;


// --------------------------------------------------------------
// data perf event per thread per event
// this data is designed to be used within a thread
// --------------------------------------------------------------
typedef struct event_thread_s {

  pe_mmap_t    *mmap;  // mmap buffer
  int          fd;     // file descriptor of the event

} event_thread_t;




/******************************************************************************
 * Interfaces
 *****************************************************************************/

void
perf_util_init();

int
perf_util_attr_init(
  const char *event_name,
  struct perf_event_attr *attr,
  bool usePeriod, 
  u64 threshold,
  u64  sampletype
);

bool
perf_util_is_ksym_available();

int
perf_util_get_paranoid_level();

#if KERNEL_SAMPLING_ENABLED
cct_node_t *
perf_util_add_kernel_callchain( cct_node_t *leaf, void *data_aux);
#endif

void
perf_util_get_default_threshold(struct event_threshold_s *threshold);

int
perf_util_check_precise_ip_suffix(char *event);

#endif