pe_profiling/operf.cpp

/**
 * @file operf.cpp
 * Front-end (containing main) for handling a user request to run a profile
 * using the new Linux Performance Events Subsystem.
 *
 * @remark Copyright 2011 OProfile authors
 * @remark Read the file COPYING
 *
 * Created on: Dec 7, 2011
 * @author Maynard Johnson
 * (C) Copyright IBM Corp. 2011
 *
 * Modified by Maynard Johnson <maynardj@us.ibm.com>
 * (C) Copyright IBM Corporation 2012, 2013
 *
 */

#include "config.h"

#include <stdio.h>
#include <stdlib.h>
#include <getopt.h>
#include <dirent.h>
#include <exception>
#include <pwd.h>
#include <errno.h>
#include <sys/time.h>
#include <string.h>
#include <unistd.h>
#include <sys/stat.h>
#include <sys/types.h>
#include <fcntl.h>
#include <sys/wait.h>
#include <ftw.h>
#include <getopt.h>
#include <iostream>
#include "operf_utils.h"
#include "op_libiberty.h"
#include "string_manip.h"
#include "cverb.h"
#include "operf_counter.h"
#include "op_cpu_type.h"
#include "op_cpufreq.h"
#include "op_events.h"
#include "op_string.h"
#include "operf_kernel.h"
#include "child_reader.h"
#include "op_get_time.h"
#include "operf_stats.h"
#include "op_netburst.h"

using namespace std;

typedef enum END_CODE {
	ALL_OK = 0,
	APP_ABNORMAL_END =  1,
	PERF_RECORD_ERROR = 2,
	PERF_READ_ERROR   = 4,
	PERF_BOTH_ERROR   = 8
} end_code_t;

// Globals
char * app_name = NULL;
bool use_cpu_minus_one = false;
pid_t app_PID = -1;
uint64_t kernel_start, kernel_end;
operf_read operfRead;
op_cpu cpu_type;
double cpu_speed;
uint op_nr_events;
verbose vmisc("misc");
uid_t my_uid;
bool no_vmlinux;
int kptr_restrict;
char * start_time_human_readable;

#define DEFAULT_OPERF_OUTFILE "operf.data"
#define CALLGRAPH_MIN_COUNT_SCALE 15

static char full_pathname[PATH_MAX];
static char * app_name_SAVE = NULL;
static char ** app_args = NULL;
static 	pid_t jitconv_pid = -1;
static bool app_started;
static pid_t operf_record_pid;
static pid_t operf_read_pid;
static string samples_dir;
static bool startApp;
static string outputfile;
static char start_time_str[32];
static vector<operf_event_t> events;
static bool jit_conversion_running;
static void convert_sample_data(void);
static int sample_data_pipe[2];
bool ctl_c = false;
bool pipe_closed = false;


namespace operf_options {
bool system_wide;
bool append;
int pid;
bool callgraph;
int mmap_pages_mult;
string session_dir;
string vmlinux;
bool separate_cpu;
bool separate_thread;
bool post_conversion;
vector<string> evts;
}

static const char * valid_verbose_vals[] = { "debug", "record", "convert", "misc", "sfile", "arcs", "all"};
#define NUM_VERBOSE_OPTIONS (sizeof(valid_verbose_vals)/sizeof(char *))

struct option long_options [] =
{
 {"verbose", required_argument, NULL, 'V'},
 {"session-dir", required_argument, NULL, 'd'},
 {"vmlinux", required_argument, NULL, 'k'},
 {"callgraph", no_argument, NULL, 'g'},
 {"system-wide", no_argument, NULL, 's'},
 {"append", no_argument, NULL, 'a'},
 {"pid", required_argument, NULL, 'p'},
 {"events", required_argument, NULL, 'e'},
 {"separate-cpu", no_argument, NULL, 'c'},
 {"separate-thread", no_argument, NULL, 't'},
 {"lazy-conversion", no_argument, NULL, 'l'},
 {"help", no_argument, NULL, 'h'},
 {"version", no_argument, NULL, 'v'},
 {"usage", no_argument, NULL, 'u'},
 {NULL, 9, NULL, 0}
};

const char * short_options = "V:d:k:gsap:e:ctlhuv";

vector<string> verbose_string;

void __set_event_throttled(int index)
{
	if (index < 0) {
		cerr << "Unable to determine if throttling occurred for ";
		cerr << "event " << events[index].name << endl;
	} else {
		events[index].throttled = true;
	}
}

static void __print_usage_and_exit(const char * extra_msg)
{
	if (extra_msg)
		cerr << extra_msg << endl;
	cerr << "usage: operf [ options ] [ --system-wide | --pid <pid> | [ command [ args ] ] ]" << endl;
	cerr << "See operf man page for details." << endl;
	exit(EXIT_FAILURE);
}

// Signal handler for main (parent) process.
static void op_sig_stop(int val __attribute__((unused)))
{
	// Received a signal to quit, so we need to stop the
	// app being profiled.
	size_t dummy __attribute__ ((__unused__));
	ctl_c = true;
	if (cverb << vdebug)
		dummy = write(1, "in op_sig_stop\n", 15);
	if (startApp)
		kill(app_PID, SIGKILL);
}

// For child processes to manage a controlled stop after Ctl-C is done
static void _handle_sigint(int val __attribute__((unused)))
{
	size_t dummy __attribute__ ((__unused__));
	/* Each process (parent and each forked child) will have their own copy of
	 * the ctl_c variable, so this can be used by each process in managing their
	 * shutdown procedure.
	 */
	ctl_c = true;
	if (cverb << vdebug)
		dummy = write(1, "in _handle_sigint\n", 19);
	return;
}


void _set_basic_SIGINT_handler_for_child(void)
{
	struct sigaction act;
	sigset_t ss;

	sigfillset(&ss);
	sigprocmask(SIG_UNBLOCK, &ss, NULL);

	act.sa_handler = _handle_sigint;
	act.sa_flags = 0;
	sigemptyset(&act.sa_mask);
	sigaddset(&act.sa_mask, SIGINT);
	if (sigaction(SIGINT, &act, NULL)) {
		perror("operf: install of SIGINT handler failed: ");
		exit(EXIT_FAILURE);
	}
}

void set_signals_for_parent(void)
{
	struct sigaction act;
	sigset_t ss;

	sigfillset(&ss);
	sigprocmask(SIG_UNBLOCK, &ss, NULL);

	act.sa_handler = op_sig_stop;
	act.sa_flags = 0;
	sigemptyset(&act.sa_mask);
	sigaddset(&act.sa_mask, SIGINT);

	if (sigaction(SIGINT, &act, NULL)) {
		perror("operf: install of SIGINT handler failed: ");
		exit(EXIT_FAILURE);
	}
}

static int app_ready_pipe[2], start_app_pipe[2], operf_record_ready_pipe[2];

static string args_to_string(void)
{
	string ret;
	char * const * ptr = app_args + 1;
	while (*ptr != NULL) {
		ret.append(*ptr);
		ret += ' ';
		ptr++;
	}
	return ret;
}

void run_app(void)
{
	// ASSUMPTION: app_name is a fully-qualified pathname
	char * app_fname = rindex(app_name, '/') + 1;
	app_args[0] = app_fname;

	string arg_str = args_to_string();
	cverb << vdebug << "Exec args are: " << app_fname << " " << arg_str << endl;
	// Fake an exec to warm-up the resolver
	execvp("", app_args);
	// signal to the parent that we're ready to exec
	int startup = 1;
	if (write(app_ready_pipe[1], &startup, sizeof(startup)) < 0) {
		perror("Internal error on app_ready_pipe");
		_exit(EXIT_FAILURE);
	}

	// wait for parent to tell us to start
	int startme = 0;
	if (read(start_app_pipe[0], &startme, sizeof(startme)) == -1) {
		perror("Internal error in run_app on start_app_pipe");
		_exit(EXIT_FAILURE);
	}
	if (startme != 1)
		_exit(EXIT_SUCCESS);

	cverb << vdebug << "parent says start app " << app_name << endl;
	app_started = true;
	execvp(app_name, app_args);
	cerr <<  "Failed to exec " << app_fname << " " << arg_str << ": " << strerror(errno) << endl;
	/* We don't want any cleanup in the child */
	_exit(EXIT_FAILURE);

}

int start_profiling(void)
{
	// The only process that should return from this function is the process
	// which invoked it.  Any forked process must do _exit() rather than return().
	struct timeval tv;
	unsigned long long start_time = 0ULL;
	gettimeofday(&tv, NULL);
	start_time = 0ULL;
	start_time = tv.tv_sec;
	sprintf(start_time_str, "%llu", start_time);
	start_time_human_readable = op_get_time();
	startApp = ((app_PID != operf_options::pid) && (operf_options::system_wide == false));

	if (startApp) {
		if (pipe(app_ready_pipe) < 0 || pipe(start_app_pipe) < 0) {
			perror("Internal error: operf-record could not create pipe");
			_exit(EXIT_FAILURE);
		}
		app_PID = fork();
		if (app_PID < 0) {
			perror("Internal error: fork failed");
			_exit(EXIT_FAILURE);
		} else if (app_PID == 0) { // child process for exec'ing app
			if (!operf_options::post_conversion) {
				close(sample_data_pipe[0]);
				close(sample_data_pipe[1]);
			}
			run_app();
		}
	}

	// parent
	if (pipe(operf_record_ready_pipe) < 0) {
		perror("Internal error: could not create pipe");
		return -1;
	}
	operf_record_pid = fork();
	if (operf_record_pid < 0) {
		return -1;
	} else if (operf_record_pid == 0) { // operf-record process
		int ready = 0;
		int exit_code = EXIT_SUCCESS;
		_set_basic_SIGINT_handler_for_child();
		close(operf_record_ready_pipe[0]);
		if (!operf_options::post_conversion)
			close(sample_data_pipe[0]);
		/*
		 * Since an informative message will be displayed to the user if
		 * an error occurs, we don't want to blow chunks here; instead, we'll
		 * exit gracefully.  Clear out the operf.data file as an indication
		 * to the parent process that the profile data isn't valid.
		 */
		try {
			OP_perf_utils::vmlinux_info_t vi;
			int outfd;
			int flags = O_WRONLY | O_CREAT | O_TRUNC;
			vi.image_name = operf_options::vmlinux;
			vi.start = kernel_start;
			vi.end = kernel_end;
			if (operf_options::post_conversion) {
				outfd = open(outputfile.c_str(), flags, S_IRUSR|S_IWUSR);
				if (outfd < 0) {
					string errmsg = "Internal error: Could not create temporary output file. errno is ";
					errmsg += strerror(errno);
					throw runtime_error(errmsg);
				}
			} else {
				outfd = sample_data_pipe[1];
			}
			operf_record operfRecord(outfd, operf_options::system_wide, app_PID,
			                         (operf_options::pid == app_PID), events, vi,
			                         operf_options::callgraph,
			                         operf_options::separate_cpu, operf_options::post_conversion);
			if (operfRecord.get_valid() == false) {
				/* If valid is false, it means that one of the "known" errors has
				 * occurred:
				 *   - profiled process has already ended
				 *   - passed PID was invalid
				 *   - device or resource busy
				 *   - failure to mmap kernel profile data
				 */
				cerr << "operf record init failed" << endl;
				cerr << "usage: operf [ options ] [ --system-wide | --pid <pid> | [ command [ args ] ] ]" << endl;
				// Exit with SUCCESS to avoid the unnecessary "operf-record process ended
				// abnormally" message
				goto fail_out;
			}

			ready = 1;
			if (write(operf_record_ready_pipe[1], &ready, sizeof(ready)) < 0) {
				perror("Internal error on operf_record_ready_pipe");
				exit_code = EXIT_FAILURE;
				goto fail_out;
			}

			// start recording
			operfRecord.recordPerfData();
			cverb << vdebug << "Total bytes recorded from perf events: " << dec
					<< operfRecord.get_total_bytes_recorded() << endl;
		} catch (runtime_error re) {
			/* If the user does ctl-c, the operf-record process may get interrupted
			 * in a system call, causing problems with writes to the sample data pipe.
			 * So we'll ignore such errors unless the user requests debug info.
			 */
			if (!ctl_c || (cverb << vmisc)) {
				cerr << "Caught runtime_error: " << re.what() << endl;
				exit_code = EXIT_FAILURE;
			}
			goto fail_out;
		}
		// done
		_exit(exit_code);

fail_out:
		if (!ready){
			/* ready==0 means we've not yet told parent we're ready,
			 * but the parent is reading our pipe.  So we tell the
			 * parent we're not ready so it can continue.
			 */
			if (write(operf_record_ready_pipe[1], &ready, sizeof(ready)) < 0) {
				perror("Internal error on operf_record_ready_pipe");
			}
		}
		_exit(exit_code);
	} else {  // parent
		int recorder_ready = 0;
		int startup;
		close(operf_record_ready_pipe[1]);
		if (startApp) {
			if (read(app_ready_pipe[0], &startup, sizeof(startup)) == -1) {
				perror("Internal error on app_ready_pipe");
				return -1;
			} else if (startup != 1) {
				cerr << "app is not ready to start; exiting" << endl;
				return -1;
			}
		}

		if (read(operf_record_ready_pipe[0], &recorder_ready, sizeof(recorder_ready)) == -1) {
			perror("Internal error on operf_record_ready_pipe");
			return -1;
		} else if (recorder_ready != 1) {
			cverb << vdebug << "operf record process failure; exiting" << endl;
			if (startApp) {
				cverb << vdebug << "telling child to abort starting of app" << endl;
				startup = 0;
				if (write(start_app_pipe[1], &startup, sizeof(startup)) < 0) {
					perror("Internal error on start_app_pipe");
				}
			}
			return -1;
		}

		if (startApp) {
			// Tell app_PID to start the app
			cverb << vdebug << "telling child to start app" << endl;
			if (write(start_app_pipe[1], &startup, sizeof(startup)) < 0) {
				perror("Internal error on start_app_pipe");
				return -1;
			}
		}

	}
	if (!operf_options::system_wide)
		app_started = true;

	// parent returns
	return 0;
}

static end_code_t _kill_operf_read_pid(end_code_t rc)
{
	// Now stop the operf-read process
	int waitpid_status;
	struct timeval tv;
	long long start_time_sec;
	long long usec_timer;
	bool keep_trying = true;
	waitpid_status = 0;
	gettimeofday(&tv, NULL);
	start_time_sec = tv.tv_sec;
	usec_timer = tv.tv_usec;
	/* We'll initially try the waitpid with WNOHANG once every 100,000 usecs.
	 * If it hasn't ended within 5 seconds, we'll kill it and do one
	 * final wait.
	 */
	while (keep_trying) {
		int option = WNOHANG;
		int wait_rc;
		gettimeofday(&tv, NULL);
		if (tv.tv_sec > start_time_sec + 5) {
			keep_trying = false;
			option = 0;
			cerr << "now trying to kill convert pid..." << endl;

			if (kill(operf_read_pid, SIGUSR1) < 0) {
				perror("Attempt to stop operf-read process failed");
				rc = rc ? PERF_BOTH_ERROR : PERF_READ_ERROR;
				break;
			}
		} else {
			/* If we exceed the 100000 usec interval or if the tv_usec
			 * value has rolled over to restart at 0, then we reset
			 * the usec_timer to current tv_usec and try waitpid.
			 */
			if ((tv.tv_usec % 1000000) > (usec_timer + 100000)
					|| (tv.tv_usec < usec_timer))
				usec_timer = tv.tv_usec;
			else
				continue;
		}
		if ((wait_rc = waitpid(operf_read_pid, &waitpid_status, option)) < 0) {
			keep_trying = false;
			if (errno != ECHILD) {
				perror("waitpid for operf-read process failed");
				rc = rc ? PERF_BOTH_ERROR : PERF_READ_ERROR;
			}
		} else if (wait_rc) {
			if (WIFEXITED(waitpid_status)) {
				keep_trying = false;
				if (!WEXITSTATUS(waitpid_status)) {
					cverb << vdebug << "operf-read process returned OK" << endl;
				} else if (WIFEXITED(waitpid_status)) {
					/* If user did ctl-c, operf-read may get spurious errors, like
					 * broken pipe, etc.  We ignore these unless the user asks for
					 * debug output.
					 */
					if (!ctl_c || cverb << vdebug) {
						cerr <<  "operf-read process ended abnormally.  Status = "
						     << WEXITSTATUS(waitpid_status) << endl;
						rc = rc ? PERF_BOTH_ERROR : PERF_READ_ERROR;
					}
				}
			}  else if (WIFSIGNALED(waitpid_status)) {
				keep_trying = false;
				/* If user did ctl-c, operf-read may get spurious errors, like
				 * broken pipe, etc.  We ignore these unless the user asks for
				 * debug output.
				 */
				if (!ctl_c || cverb << vdebug) {
					cerr << "operf-read process killed by signal "
					     << WTERMSIG(waitpid_status) << endl;
					rc = PERF_RECORD_ERROR;
				}
			}
		}
	}
	return rc;
}

static end_code_t _kill_operf_record_pid(void)
{
	int waitpid_status = 0;
	end_code_t rc = ALL_OK;

	// stop operf-record process
	errno = 0;
	if (kill(operf_record_pid, SIGUSR1) < 0) {
		// If operf-record process is already ended, don't consider this an error.
		if (errno != ESRCH) {
			perror("Attempt to stop operf-record process failed");
			rc = PERF_RECORD_ERROR;
		}
	} else {
		if (waitpid(operf_record_pid, &waitpid_status, 0) < 0) {
			perror("waitpid for operf-record process failed");
			rc = PERF_RECORD_ERROR;
		} else {
			if (WIFEXITED(waitpid_status) && (!WEXITSTATUS(waitpid_status))) {
				cverb << vdebug << "operf-record process returned OK" << endl;
			} else if (WIFEXITED(waitpid_status)) {
				/* If user did ctl-c, operf-record may get spurious errors, like
				 * broken pipe, etc.  We ignore these unless the user asks for
				 * debug output.
				 */
				if (!ctl_c || cverb << vdebug) {
					cerr <<  "operf-record process ended abnormally: "
							<< WEXITSTATUS(waitpid_status) << endl;
					rc = PERF_RECORD_ERROR;
				}
			} else if (WIFSIGNALED(waitpid_status)) {
				if (!ctl_c || cverb << vdebug) {
					cerr << "operf-record process killed by signal "
					     << WTERMSIG(waitpid_status) << endl;
					rc = PERF_RECORD_ERROR;
				}
			}
		}
	}
	return rc;
}

static end_code_t _run(void)
{
	int waitpid_status = 0;
	end_code_t rc = ALL_OK;
	bool kill_record = true;

	// Fork processes with signals blocked.
	sigset_t ss;
	sigfillset(&ss);
	sigprocmask(SIG_BLOCK, &ss, NULL);

	/* By default (unless the user specifies --lazy-conversion), the operf-record process
	 * writes the sample data to a pipe, from which the operf-read process reads.
	 */
	if (!operf_options::post_conversion && pipe(sample_data_pipe) < 0) {
		perror("Internal error: operf-record could not create pipe");
		_exit(EXIT_FAILURE);
	}

	if (start_profiling() < 0) {
		return PERF_RECORD_ERROR;
	}
	// parent continues here
	if (startApp)
		cverb << vdebug << "app " << app_PID << " is running" << endl;

	/* If we're not doing system wide profiling and no app is started, then
	 * there's no profile data to convert. So if this condition is NOT true,
	 * then we'll do the convert.
	 * Note that if --lazy-conversion is passed, then operf_options::post_conversion
	 * will be set, and we will defer conversion until after the operf-record
	 * process is done.
	 */
	if (!operf_options::post_conversion) {
		if (!(!app_started && !operf_options::system_wide)) {
			cverb << vdebug << "Forking read pid" << endl;
			operf_read_pid = fork();
			if (operf_read_pid < 0) {
				perror("Internal error: fork failed");
				_exit(EXIT_FAILURE);
			} else if (operf_read_pid == 0) { // child process
				close(sample_data_pipe[1]);
				_set_basic_SIGINT_handler_for_child();
				convert_sample_data();
				_exit(EXIT_SUCCESS);
			}
			// parent
			close(sample_data_pipe[0]);
			close(sample_data_pipe[1]);
		}
	}

	set_signals_for_parent();
	if (startApp) {
		/* The user passed in a command or program name to start, so we'll need to do waitpid on that
		 * process.  However, while that user-requested process is running, it's possible we
		 * may get an error in the operf-record process.  If that happens, we want to know it right
		 * away so we can stop profiling and kill the user app.  Therefore, we must use WNOHANG
		 * on the waitpid call and bounce back and forth between the user app and the operf-record
		 * process, checking their status.  The profiled app may end normally, abnormally, or by way
		 * of ctrl-C.  The operf-record process should not end here, except abnormally.  The normal
		 * flow is:
		 *    1. profiled app ends or is stopped via ctrl-C
		 *    2. keep_trying is set to false, so we drop out of while loop and proceed to end of function
		 *    3. call _kill_operf_record_pid and _kill_operf_read_pid
		 */
		struct timeval tv;
		long long usec_timer;
		bool keep_trying = true;
		const char * app_process = "profiled app";
		const char * record_process = "operf-record process";
		waitpid_status = 0;
		gettimeofday(&tv, NULL);
		usec_timer = tv.tv_usec;
		cverb << vdebug << "going into waitpid on profiled app " << app_PID << endl;

		// We'll try the waitpid with WNOHANG once every 100,000 usecs.
		while (keep_trying) {
			pid_t the_pid = app_PID;
			int wait_rc;
			const char * the_process = app_process;
			gettimeofday(&tv, NULL);
			/* If we exceed the 100000 usec interval or if the tv_usec
			 * value has rolled over to restart at 0, then we reset
			 * the usec_timer to current tv_usec and try waitpid.
			 */
			if ((tv.tv_usec % 1000000) > (usec_timer + 100000)
					|| (tv.tv_usec < usec_timer))
				usec_timer = tv.tv_usec;
			else
				continue;

			bool trying_user_app = true;
again:
			if ((wait_rc = waitpid(the_pid, &waitpid_status, WNOHANG)) < 0) {
				keep_trying = false;
				if (errno == EINTR) {
					//  Ctrl-C will only kill the profiled app.  See the op_sig_stop signal handler.
					cverb << vdebug << "Caught ctrl-C.  Killed " << the_process << "." << endl;
				} else {
					cerr << "waitpid for " << the_process << " failed: " << strerror(errno) << endl;
					rc = trying_user_app ? APP_ABNORMAL_END : PERF_RECORD_ERROR;
				}
			} else if (wait_rc) {
				keep_trying = false;
				if (WIFEXITED(waitpid_status) && (!WEXITSTATUS(waitpid_status))) {
					cverb << vdebug << the_process << " ended normally." << endl;
				} else if (WIFEXITED(waitpid_status)) {
					cerr <<  the_process << " exited with the following status: "
					     << WEXITSTATUS(waitpid_status) << endl;
					rc = trying_user_app ? APP_ABNORMAL_END : PERF_RECORD_ERROR;
				}  else if (WIFSIGNALED(waitpid_status)) {
					if (WTERMSIG(waitpid_status) != SIGKILL) {
						cerr << the_process << " killed by signal "
						     << WTERMSIG(waitpid_status) << endl;
						rc = trying_user_app ? APP_ABNORMAL_END : PERF_RECORD_ERROR;
					}
				} else {
					keep_trying = true;
				}
			}
			if (trying_user_app && (rc == ALL_OK)) {
				trying_user_app = false;
				the_pid = operf_record_pid;
				the_process = record_process;
				goto again;
			} else if (rc != ALL_OK) {
				// If trying_user_app == true, implies profiled app ended; otherwise, operf-record process abended.
				if (!trying_user_app)
					kill_record = false;
			}
		}
	} else {
		// User passed in --pid or --system-wide
		cout << "operf: Press Ctl-c or 'kill -SIGINT " << getpid() << "' to stop profiling" << endl;
		cverb << vdebug << "going into waitpid on operf record process " << operf_record_pid << endl;
		if (waitpid(operf_record_pid, &waitpid_status, 0) < 0) {
			if (errno == EINTR) {
				cverb << vdebug << "Caught ctrl-C. Killing operf-record process . . ." << endl;
			} else {
				cerr << "waitpid errno is " << errno << endl;
				perror("waitpid for operf-record process failed");
				kill_record = false;
				rc = PERF_RECORD_ERROR;
			}
		} else {
			if (WIFEXITED(waitpid_status) && (!WEXITSTATUS(waitpid_status))) {
				cverb << vdebug << "waitpid for operf-record process returned OK" << endl;
			} else if (WIFEXITED(waitpid_status)) {
				kill_record = false;
				cerr <<  "operf-record process ended abnormally: "
				     << WEXITSTATUS(waitpid_status) << endl;
				rc = PERF_RECORD_ERROR;
			} else if (WIFSIGNALED(waitpid_status)) {
				kill_record = false;
				cerr << "operf-record process killed by signal "
				     << WTERMSIG(waitpid_status) << endl;
				rc = PERF_RECORD_ERROR;
			}
		}
	}
	if (kill_record) {
		if (operf_options::post_conversion)
			rc = _kill_operf_record_pid();
		else
			rc = _kill_operf_read_pid(_kill_operf_record_pid());
	} else {
		if (!operf_options::post_conversion)
			rc = _kill_operf_read_pid(rc);
	}

	return rc;
}

static void cleanup(void)
{
	free(app_name_SAVE);
	free(app_args);
	events.clear();
	verbose_string.clear();
	if (operf_options::post_conversion) {
		string cmd = "rm -f " + outputfile;
		if (system(cmd.c_str()) != 0)
			cerr << "Unable to remove " << outputfile << endl;
	}
}

static void _jitconv_complete(int val __attribute__((unused)))
{
	int child_status;
	pid_t the_pid = wait(&child_status);
	if (the_pid != jitconv_pid) {
		return;
	}
	jit_conversion_running = false;
	if (WIFEXITED(child_status) && (!WEXITSTATUS(child_status))) {
		cverb << vdebug << "JIT dump processing complete." << endl;
	} else {
		 if (WIFSIGNALED(child_status))
			 cerr << "child received signal " << WTERMSIG(child_status) << endl;
		 else
			 cerr << "JIT dump processing exited abnormally: "
			 << WEXITSTATUS(child_status) << endl;
	}
}

static void _set_signals_for_convert(void)
{
	struct sigaction act;
	sigset_t ss;

	sigfillset(&ss);
	sigprocmask(SIG_UNBLOCK, &ss, NULL);

	act.sa_handler = _jitconv_complete;
	act.sa_flags = 0;
	sigemptyset(&act.sa_mask);
	sigaddset(&act.sa_mask, SIGCHLD);
	if (sigaction(SIGCHLD, &act, NULL)) {
		perror("operf: install of SIGCHLD handler failed: ");
		exit(EXIT_FAILURE);
	}
}

static void _do_jitdump_convert()
{
	int arg_num;
	unsigned long long end_time = 0ULL;
	struct timeval tv;
	char end_time_str[32];
	char opjitconv_path[PATH_MAX + 1];
	char * exec_args[8];

	jitconv_pid = fork();
	switch (jitconv_pid) {
	case -1:
		perror("Error forking JIT dump process!");
		break;
	case 0: {
		const char * jitconv_pgm = "opjitconv";
		const char * debug_option = "-d";
		const char * non_root_user = "--non-root";
		const char * delete_jitdumps = "--delete-jitdumps";
		gettimeofday(&tv, NULL);
		end_time = tv.tv_sec;
		sprintf(end_time_str, "%llu", end_time);
		sprintf(opjitconv_path, "%s/%s", OP_BINDIR, jitconv_pgm);
		arg_num = 0;
		exec_args[arg_num++] = (char *)jitconv_pgm;
		if (cverb << vdebug)
			exec_args[arg_num++] = (char *)debug_option;
		if (my_uid != 0)
			exec_args[arg_num++] = (char *)non_root_user;
		exec_args[arg_num++] = (char *)delete_jitdumps;
		exec_args[arg_num++] = (char *)operf_options::session_dir.c_str();
		exec_args[arg_num++] = start_time_str;
		exec_args[arg_num++] = end_time_str;
		exec_args[arg_num] = (char *) NULL;
		execvp(opjitconv_path, exec_args);
		fprintf(stderr, "Failed to exec %s: %s\n",
		        exec_args[0], strerror(errno));
		/* We don't want any cleanup in the child */
		_exit(EXIT_FAILURE);
		break;
	}
	default: // parent
		jit_conversion_running = true;
		break;
	}

}

static int __delete_old_previous_sample_data(const char *fpath,
                                const struct stat *sb  __attribute__((unused)),
                                int tflag  __attribute__((unused)),
                                struct FTW *ftwbuf __attribute__((unused)))
{
	if (remove(fpath)) {
		perror("sample data removal error");
		return FTW_STOP;
	} else {
		return FTW_CONTINUE;
	}
}

/* Read perf_events sample data written by the operf-record process through
 * the sample_data_pipe or file (dependent on 'lazy-conversion' option)
 * and convert the perf format sample data to to oprofile format sample files.
 *
 * If not invoked with --lazy-conversion option, this function is executed by
 * the "operf-read" child process.  If user does a ctrl-C, the parent will
 * execute _kill_operf_read_pid which will try to allow the conversion process
 * to complete, waiting 5 seconds before it forcefully kills the operf-read
 * process via 'kill SIGUSR1'.
 *
 * But if --lazy-conversion option is used, then it's the parent process that's
 * running convert_sample_data.  If the user does a ctrl-C during this procedure,
 * the ctrl-C is handled via op_sig_stop which essentially does nothing to stop
 * the conversion procedure, which in general is fine.  On the very rare chance
 * that the procedure gets stuck (hung) somehow, the user will have to do a
 * 'kill -KILL'.
 */
static void convert_sample_data(void)
{
	int inputfd;
	string inputfname;
	int rc = EXIT_SUCCESS;
	int keep_waiting = 0;
	string current_sampledir = samples_dir + "/current/";
	string previous_sampledir = samples_dir + "/previous";
	string stats_dir = "";
	current_sampledir.copy(op_samples_current_dir, current_sampledir.length(), 0);

	if (!app_started && !operf_options::system_wide)
		return;

	if (!operf_options::append) {
                int flags = FTW_DEPTH | FTW_ACTIONRETVAL;
		errno = 0;
		if (nftw(previous_sampledir.c_str(), __delete_old_previous_sample_data, 32, flags) !=0 &&
				errno != ENOENT) {
			cerr << "Unable to remove old sample data at " << previous_sampledir << "." << endl;
			if (errno)
				cerr << strerror(errno) << endl;
			rc = EXIT_FAILURE;
			goto out;
		}
		if (rename(current_sampledir.c_str(), previous_sampledir.c_str()) < 0) {
			if (errno && (errno != ENOENT)) {
				cerr << "Unable to move old profile data to " << previous_sampledir << endl;
				cerr << strerror(errno) << endl;
				rc = EXIT_FAILURE;
				goto out;
			}
		}
	}
	rc = mkdir(current_sampledir.c_str(), S_IRWXU | S_IRWXG | S_IROTH | S_IXOTH);
	if (rc && (errno != EEXIST)) {
		cerr << "Error trying to create " << current_sampledir << " dir." << endl;
		perror("mkdir failed with");
		rc = EXIT_FAILURE;
		goto out;
	}

	if (operf_options::post_conversion) {
		inputfd = -1;
		inputfname = outputfile;
	} else {
		inputfd = sample_data_pipe[0];
		inputfname = "";
	}
	operfRead.init(inputfd, inputfname, current_sampledir, cpu_type, events, operf_options::system_wide);
	if ((rc = operfRead.readPerfHeader()) < 0) {
		if (rc != OP_PERF_HANDLED_ERROR)
			cerr << "Error: Cannot create read header info for sample data " << endl;
		rc = EXIT_FAILURE;
		goto out;
	}
	cverb << vdebug << "Successfully read header info for sample data " << endl;
	if (operfRead.is_valid()) {
		try {
			unsigned int num = operfRead.convertPerfData();
			cverb << vdebug << "operf_read: Total bytes received from operf_record process: " << dec << num << endl;
		} catch (runtime_error e) {
			cerr << "Caught runtime error from operf_read::convertPerfData" << endl;
			cerr << e.what() << endl;
			rc = EXIT_FAILURE;
			goto out;
		}
	}

	_set_signals_for_convert();
	cverb << vdebug << "Calling _do_jitdump_convert" << endl;
	_do_jitdump_convert();
	while (jit_conversion_running && (keep_waiting < 2)) {
		sleep(1);
		keep_waiting++;
	}
	if (jit_conversion_running) {
		kill(jitconv_pid, SIGKILL);
	}
out:
	if (!operf_options::post_conversion)
		_exit(rc);
}


static int find_app_file_in_dir(const struct dirent * d)
{
	if (!strcmp(d->d_name, app_name))
		return 1;
	else
		return 0;
}

static int get_PATH_based_pathname(char * path_holder, size_t n)
{
	int retval = -1;

	char * real_path = getenv("PATH");
	char * path = (char *) xstrdup(real_path);
	char * segment = strtok(path, ":");
	while (segment) {
		struct dirent ** namelist;
		int rc = scandir(segment, &namelist, find_app_file_in_dir, NULL);
		if (rc < 0) {
			if (errno != ENOENT) {
				cerr << strerror(errno) << endl;
				cerr << app_name << " cannot be found in your PATH." << endl;
				break;
			}
		} else if (rc == 1) {
			size_t applen = strlen(app_name);
			size_t dirlen = strlen(segment);

			if (applen + dirlen + 2 > n) {
				cerr << "Path segment " << segment
				     << " prepended to the passed app name is too long"
				     << endl;
				retval = -1;
				break;
			}

			if (!strcmp(segment, ".")) {
				if (getcwd(path_holder, PATH_MAX) == NULL) {
					retval = -1;
					cerr << "getcwd [3] failed when processing <cur-dir>/" << app_name << " found via PATH. Aborting."
							<< endl;
					break;
				}
			} else {
				strncpy(path_holder, segment, dirlen);
			}
			strcat(path_holder, "/");
			strncat(path_holder, app_name, applen);
			retval = 0;
			free(namelist[0]);
			free(namelist);

			break;
		}
		segment = strtok(NULL, ":");
	}
	free(path);
	return retval;
}
int validate_app_name(void)
{
	int rc = 0;
	struct stat filestat;
	size_t len = strlen(app_name);

	if (len > (size_t) (OP_APPNAME_LEN - 1)) {
		cerr << "app name longer than max allowed (" << OP_APPNAME_LEN
		     << " chars)\n";
		cerr << app_name << endl;
		rc = -1;
		goto out;
	}

	if (index(app_name, '/') == app_name) {
		// Full pathname of app was specified, starting with "/".
		strncpy(full_pathname, app_name, len);
	} else if ((app_name[0] == '.') && (app_name[1] == '/')) {
		// Passed app is in current directory; e.g., "./myApp"
		if (getcwd(full_pathname, PATH_MAX) == NULL) {
			rc = -1;
			cerr << "getcwd [1] failed when trying to find app name " << app_name << ". Aborting."
			     << endl;
			goto out;
		}
		strcat(full_pathname, "/");
		if ((strlen(full_pathname) + strlen(app_name + 2) + 1) > PATH_MAX) {
			rc = -1;
			cerr << "Length of current dir (" << full_pathname << ") and app name ("
			     << (app_name + 2) << ") exceeds max allowed (" << PATH_MAX << "). Aborting."
			     << endl;
			goto out;
		}
		strcat(full_pathname, (app_name + 2));
	} else if (index(app_name, '/')) {
		// Passed app is in a subdirectory of cur dir; e.g., "test-stuff/myApp"
		if (getcwd(full_pathname, PATH_MAX) == NULL) {
			rc = -1;
			cerr << "getcwd [2] failed when trying to find app name " << app_name << ". Aborting."
			     << endl;
			goto out;
		}
		strcat(full_pathname, "/");
		strcat(full_pathname, app_name);
	} else {
		// Passed app name, at this point, MUST be found in PATH
		rc = get_PATH_based_pathname(full_pathname, PATH_MAX);
	}

	if (rc) {
		cerr << "Problem finding app name " << app_name << ". Aborting."
		     << endl;
		goto out;
	}
	app_name_SAVE = app_name;
	app_name = full_pathname;
	if (stat(app_name, &filestat)) {
		char msg[OP_APPNAME_LEN + 50];
		snprintf(msg, OP_APPNAME_LEN + 50, "Non-existent app name \"%s\"",
		         app_name);
		perror(msg);
		rc = -1;
	}

	out: return rc;
}

static void _get_event_code(operf_event_t * event)
{
	FILE * fp;
	char oprof_event_code[9];
	string command;
	u64 base_code, config;
	char buf[20];
	if ((snprintf(buf, 20, "%lu", event->count)) < 0) {
		cerr << "Error parsing event count of " << event->count << endl;
		exit(EXIT_FAILURE);
	}

	base_code = config = 0ULL;

	command = OP_BINDIR;
	command += "ophelp ";
	command += event->name;

	fp = popen(command.c_str(), "r");
	if (fp == NULL) {
		cerr << "Unable to execute ophelp to get info for event "
		     << event->name << endl;
		exit(EXIT_FAILURE);
	}
	if (fgets(oprof_event_code, sizeof(oprof_event_code), fp) == NULL) {
		pclose(fp);
		cerr << "Unable to find info for event "
		     << event->name << endl;
		exit(EXIT_FAILURE);
	}

	pclose(fp);

	base_code = strtoull(oprof_event_code, (char **) NULL, 10);


#if defined(__i386__) || defined(__x86_64__)
	// Setup EventSelct[11:8] field for AMD
	char mask[12];
	const char * vendor_AMD = "AuthenticAMD";
	if (op_is_cpu_vendor((char *)vendor_AMD)) {
		config = base_code & 0xF00ULL;
		config = config << 32;
	}

	// Setup EventSelct[7:0] field
	config |= base_code & 0xFFULL;

	// Setup unitmask field
handle_named_um:
	if (event->um_name[0]) {
		command = OP_BINDIR;
		command += "ophelp ";
		command += "--extra-mask ";
		command += event->name;
		command += ":";
		command += buf;
		command += ":";
		command += event->um_name;
		fp = popen(command.c_str(), "r");
		if (fp == NULL) {
			cerr << "Unable to execute ophelp to get info for event "
			     << event->name << endl;
			exit(EXIT_FAILURE);
		}
		if (fgets(mask, sizeof(mask), fp) == NULL) {
			pclose(fp);
			cerr << "Unable to find unit mask info for " << event->um_name << " for event "
			     << event->name << endl;
			exit(EXIT_FAILURE);
		}
		pclose(fp);
		// FIXME:  The mask value here is the extra bits from the named unit mask.  It's not
		// ideal to put that value into the UM's mask, since that's what will show up in
		// opreport.  It would be better if we could somehow have the unit mask name that the
		// user passed to us show up in opreort.
		event->evt_um = strtoull(mask, (char **) NULL, 10);
		/* A value >= EXTRA_MIN_VAL returned by 'ophelp --extra-mask' is interpreted as a
		 * valid extra value; otherwise we interpret it as a simple unit mask value
		 * for a named unit mask with EXTRA_NONE.
		 */
		if (event->evt_um >= EXTRA_MIN_VAL)
			config |= event->evt_um;
		else
			config |= ((event->evt_um & 0xFFULL) << 8);
	} else if (!event->evt_um) {
		char * endptr;
		command.clear();
		command = OP_BINDIR;
		command += "ophelp ";
		command += "--unit-mask ";
		command += event->name;
		command += ":";
		command += buf;
		fp = popen(command.c_str(), "r");
		if (fp == NULL) {
			cerr << "Unable to execute ophelp to get unit mask for event "
			     << event->name << endl;
			exit(EXIT_FAILURE);
		}
		if (fgets(mask, sizeof(mask), fp) == NULL) {
			pclose(fp);
			cerr << "Unable to find unit mask info for event " << event->name << endl;
			exit(EXIT_FAILURE);
		}
		pclose(fp);
		event->evt_um = strtoull(mask, &endptr, 10);
		if ((endptr >= mask) &&
				(endptr <= (mask + strlen(mask) - 1))) {
			// Must be a default named unit mask
			strncpy(event->um_name, mask, OP_MAX_UM_NAME_LEN);
			goto handle_named_um;
		}
		config |= ((event->evt_um & 0xFFULL) << 8);
	} else {
		config |= ((event->evt_um & 0xFFULL) << 8);
	}
#else
	config = base_code;
#endif

	event->op_evt_code = base_code;
	if (cpu_type == CPU_P4 || cpu_type == CPU_P4_HT2) {
		if (op_netburst_get_perf_encoding(event->name, event->evt_um, 1, 1, &config)) {
			cerr << "Unable to get event encoding for " << event->name << endl;
			exit(EXIT_FAILURE);
		}
	}
	event->evt_code = config;
}

#if PPC64_ARCH
/* All ppc64 events (except CYCLES) have a _GRP<n> suffix.  This is
 * because the legacy opcontrol profiler can only profile events in
 * the same group (i.e., having the same _GRP<n> suffix).  But operf
 * can multiplex events, so we should allow the user to pass event
 * names without the _GRP<n> suffix.
 *
 * If event name is not CYCLES or does not have a _GRP<n> suffix,
 * we'll call ophelp and scan the list of events, searching for one
 * that matches up to the _GRP<n> suffix.  If we don't find a match,
 * then we'll exit with the expected error message for invalid event name.
 */
static string _handle_powerpc_event_spec(string event_spec)
{
	FILE * fp;
	char line[MAX_INPUT];
	size_t grp_pos;
	string evt, retval, err_msg;
	size_t evt_name_len;
	bool first_non_cyc_evt_found = false;
	bool event_found = false;
	char event_name[OP_MAX_EVT_NAME_LEN], event_spec_str[OP_MAX_EVT_NAME_LEN + 20], * count_str;
	string cmd = OP_BINDIR;
	cmd += "/ophelp";

	strncpy(event_spec_str, event_spec.c_str(), event_spec.length() + 1);

	strncpy(event_name, strtok(event_spec_str, ":"), OP_MAX_EVT_NAME_LEN);
	count_str = strtok(NULL, ":");
	if (!count_str) {
		err_msg = "Invalid count for event ";
		goto out;
	}

	if (!strcmp("CYCLES", event_name)) {
		event_found = true;
		goto out;
	}

	evt = event_name;
	// Need to make sure the event name truly has a _GRP<n> suffix.
	grp_pos = evt.rfind("_GRP");
	if ((grp_pos != string::npos) && ((evt = evt.substr(grp_pos, string::npos))).length() > 4) {
		char * end;
		strtoul(evt.substr(4, string::npos).c_str(), &end, 0);
		if (end && (*end == '\0')) {
		// Valid group number found after _GRP, so we can skip to the end.
			event_found = true;
			goto out;
		}
	}

	// If we get here, it implies the user passed a non-CYCLES event without a GRP suffix.
	// Lets try to find a valid suffix for it.
	fp = popen(cmd.c_str(), "r");
	if (fp == NULL) {
		cerr << "Unable to execute ophelp to get info for event "
		     << event_spec << endl;
		exit(EXIT_FAILURE);
	}
	evt_name_len = strlen(event_name);
	err_msg = "Cannot find event ";
	while (fgets(line, MAX_INPUT, fp)) {
		if (!first_non_cyc_evt_found) {
			if (!strncmp(line, "PM_", 3))
				first_non_cyc_evt_found = true;
			else
				continue;
		}
		if (line[0] == ' ' || line[0] == '\t')
			continue;
		if (!strncmp(line, event_name, evt_name_len)) {
			// Found a potential match.  Check if it's a perfect match.
			string save_event_name = event_name;
			size_t full_evt_len = index(line, ':') - line;
			memset(event_name, '\0', OP_MAX_EVT_NAME_LEN);
			strncpy(event_name, line, full_evt_len);
			string candidate = event_name;
			if (candidate.rfind("_GRP") == evt_name_len) {
				event_found = true;
				break;
			} else {
				memset(event_name, '\0', OP_MAX_EVT_NAME_LEN);
				strncpy(event_name, save_event_name.c_str(), evt_name_len);
			}
		}
	}
	pclose(fp);

out:
	if (!event_found) {
		cerr << err_msg << event_name << endl;
		cerr << "Error retrieving info for event "
				<< event_spec << endl;
		exit(EXIT_FAILURE);
	}
	retval = event_name;
	return retval + ":" + count_str;
}
#endif

static void _process_events_list(void)
{
	string cmd = OP_BINDIR;
	if (operf_options::evts.size() > OP_MAX_EVENTS) {
		cerr << "Number of events specified is greater than allowed maximum of "
		     << OP_MAX_EVENTS << "." << endl;
		exit(EXIT_FAILURE);
	}
	cmd += "/ophelp --check-events ";
	for (unsigned int i = 0; i <  operf_options::evts.size(); i++) {
		FILE * fp;
		string full_cmd = cmd;
		string event_spec = operf_options::evts[i];

#if PPC64_ARCH
		// Starting with CPU_PPC64_ARCH_V1, ppc64 events files are formatted like
		// other architectures, so no special handling is needed.
		if (cpu_type < CPU_PPC64_ARCH_V1)
			event_spec = _handle_powerpc_event_spec(event_spec);
#endif

		if (operf_options::callgraph) {
			full_cmd += " --callgraph=1 ";
		}
		full_cmd += event_spec;
		fp = popen(full_cmd.c_str(), "r");
		if (fp == NULL) {
			cerr << "Unable to execute ophelp to get info for event "
			     << event_spec << endl;
			exit(EXIT_FAILURE);
		}
		if (fgetc(fp) == EOF) {
			pclose(fp);
			cerr << "Error retrieving info for event "
			     << event_spec << endl;
			if (operf_options::callgraph)
				cerr << "Note: When doing callgraph profiling, the sample count must be"
				     << endl << "15 times the minimum count value for the event."  << endl;
			exit(EXIT_FAILURE);
		}
		pclose(fp);
		char * event_str = op_xstrndup(event_spec.c_str(), event_spec.length());
		operf_event_t event;
		strncpy(event.name, strtok(event_str, ":"), OP_MAX_EVT_NAME_LEN - 1);
		event.count = atoi(strtok(NULL, ":"));
		/* Name and count are required in the event spec in order for
		 * 'ophelp --check-events' to pass.  But since unit mask and domain
		 * control bits are optional, we need to ensure the result of strtok
		 * is valid.
		 */
		char * info;
#define	_OP_UM 1
#define	_OP_KERNEL 2
#define	_OP_USER 3
		int place =  _OP_UM;
		char * endptr = NULL;
		event.evt_um = 0ULL;
		event.no_kernel = 0;
		event.no_user = 0;
		event.throttled = false;
		memset(event.um_name, '\0', OP_MAX_UM_NAME_LEN);
		while ((info = strtok(NULL, ":"))) {
			switch (place) {
			case _OP_UM:
				event.evt_um = strtoul(info, &endptr, 0);
				// If any of the UM part is not a number, then we
				// consider the entire part a string.
				if (*endptr) {
					event.evt_um = 0;
					strncpy(event.um_name, info, OP_MAX_UM_NAME_LEN - 1);
				}
				break;
			case _OP_KERNEL:
				if (atoi(info) == 0)
					event.no_kernel = 1;
				break;
			case _OP_USER:
				if (atoi(info) == 0)
					event.no_user = 1;
				break;
			}
			place++;
		}
		free(event_str);
		_get_event_code(&event);
		events.push_back(event);
	}
#if PPC64_ARCH
	{
		/* For ppc64 architecture processors prior to the introduction of
		 * architected_events_v1, the oprofile event code needs to be converted
		 * to the appropriate event code to pass to the perf_event_open syscall.
		 * But as of the introduction of architected_events_v1, the events
		 * file contains the necessary event code information, so this conversion
		 * step is no longer needed.
		 */

		using namespace OP_perf_utils;
		if ((cpu_type < CPU_PPC64_ARCH_V1) && !op_convert_event_vals(&events)) {
			cerr << "Unable to convert all oprofile event values to perf_event values" << endl;
			exit(EXIT_FAILURE);
		}
	}
#endif
}

static void get_default_event(void)
{
	operf_event_t dft_evt;
	struct op_default_event_descr descr;
	vector<operf_event_t> tmp_events;


	op_default_event(cpu_type, &descr);
	if (descr.name[0] == '\0') {
		cerr << "Unable to find default event" << endl;
		exit(EXIT_FAILURE);
	}

	memset(&dft_evt, 0, sizeof(dft_evt));
	if (operf_options::callgraph) {
		struct op_event * _event;
		op_events(cpu_type);
		if ((_event = find_event_by_name(descr.name, 0, 0))) {
			dft_evt.count = _event->min_count * CALLGRAPH_MIN_COUNT_SCALE;
		} else {
			cerr << "Error getting event info for " << descr.name << endl;
			exit(EXIT_FAILURE);
		}
	} else {
		dft_evt.count = descr.count;
	}
	dft_evt.evt_um = descr.um;
	strncpy(dft_evt.name, descr.name, OP_MAX_EVT_NAME_LEN - 1);
	_get_event_code(&dft_evt);
	events.push_back(dft_evt);

#if PPC64_ARCH
	{
		/* This section of code is for architectures such as ppc[64] for which
		 * the oprofile event code needs to be converted to the appropriate event
		 * code to pass to the perf_event_open syscall.
		 */

		using namespace OP_perf_utils;
		if ((cpu_type < CPU_PPC64_ARCH_V1) && !op_convert_event_vals(&events)) {
			cerr << "Unable to convert all oprofile event values to perf_event values" << endl;
			exit(EXIT_FAILURE);
		}
	}
#endif
}

static void _process_session_dir(void)
{
	if (operf_options::session_dir.empty()) {
		char * cwd = NULL;
		int rc;
		cwd = (char *) xmalloc(PATH_MAX);
		// set default session dir
		cwd = getcwd(cwd, PATH_MAX);
		if (cwd == NULL) {
			perror("Error calling getcwd");
			exit(EXIT_FAILURE);
		}
		operf_options::session_dir = cwd;
		operf_options::session_dir +="/oprofile_data";
		samples_dir = operf_options::session_dir + "/samples";
		free(cwd);
		rc = mkdir(operf_options::session_dir.c_str(), S_IRWXU | S_IRWXG | S_IROTH | S_IXOTH);
		if (rc && (errno != EEXIST)) {
			cerr << "Error trying to create " << operf_options::session_dir << " dir." << endl;
			perror("mkdir failed with");
			exit(EXIT_FAILURE);
		}
		rc = mkdir(samples_dir.c_str(), S_IRWXU | S_IRWXG | S_IROTH | S_IXOTH);
		if (rc && (errno != EEXIST)) {
			cerr << "Error trying to create " << samples_dir << " dir." << endl;
			perror("mkdir failed with");
			exit(EXIT_FAILURE);
		}
	} else {
		struct stat filestat;
		int rc;
		if (stat(operf_options::session_dir.c_str(), &filestat)) {
			perror("stat operation on passed session-dir failed");
			exit(EXIT_FAILURE);
		}
		if (!S_ISDIR(filestat.st_mode)) {
			cerr << "Passed session-dir " << operf_options::session_dir
			     << " is not a directory" << endl;
			exit(EXIT_FAILURE);
		}
		samples_dir = operf_options::session_dir + "/samples";
		rc = mkdir(samples_dir.c_str(), S_IRWXU);
		if (rc && (errno != EEXIST)) {
			cerr << "Error trying to create " << samples_dir << " dir." << endl;
			perror("mkdir failed with");
			exit(EXIT_FAILURE);
		}
	}

	cverb << vdebug << "Using samples dir " << samples_dir << endl;
}

bool _get_vmlinux_address_info(vector<string> args, string cmp_val, string &str)
{
	bool found = false;
	child_reader reader("objdump", args);
	if (reader.error()) {
		cerr << "An error occurred while trying to get vmlinux address info:\n\n";
		cerr << reader.error_str() << endl;
		exit(EXIT_FAILURE);
	}

	while (reader.getline(str)) {
		if (str.find(cmp_val.c_str()) != string::npos) {
			found = true;
			break;
		}
	}
	// objdump always returns SUCCESS so we must rely on the stderr state
	// of objdump. If objdump error message is cryptic our own error
	// message will be probably also cryptic
	ostringstream std_err;
	ostringstream std_out;
	reader.get_data(std_out, std_err);
	if (std_err.str().length()) {
		cerr << "An error occurred while getting vmlinux address info:\n\n";
		cerr << std_err.str() << endl;
		// If we found the string we were looking for in objdump output,
		// treat this as non-fatal error.
		if (!found)
			exit(EXIT_FAILURE);
	}

	// force error code to be acquired
	reader.terminate_process();

	// required because if objdump stop by signal all above things suceeed
	// (signal error message are not output through stdout/stderr)
	if (reader.error()) {
		cerr << "An error occur during the execution of objdump to get vmlinux address info:\n\n";
		cerr << reader.error_str() << endl;
		if (!found)
			exit(EXIT_FAILURE);
	}
	return found;
}

string _process_vmlinux(string vmlinux_file)
{
	vector<string> args;
	char start[17], end[17];
	string str, start_end;
	bool found;
	int ret;

	no_vmlinux = false;
	args.push_back("-h");
	args.push_back(vmlinux_file);
	if ((found = _get_vmlinux_address_info(args, " .text", str))) {
		cverb << vmisc << str << endl;
		ret = sscanf(str.c_str(), " %*s %*s %*s %s", start);
	}
	if (!found || ret != 1){
		cerr << "Unable to obtain vmlinux start address." << endl;
		cerr << "The specified vmlinux file (" << vmlinux_file << ") "
		     << "does not seem to be valid." << endl;
		cerr << "Make sure you are using a non-compressed image file "
		     << "(e.g. vmlinux not vmlinuz)" << endl;
		exit(EXIT_FAILURE);
	}

	args.clear();
	args.push_back("-t");
	args.push_back(vmlinux_file);
	if ((found = _get_vmlinux_address_info(args, " _etext", str))) {
		cverb << vmisc << str << endl;
		ret = sscanf(str.c_str(), "%s", end);
	}
	if (!found || ret != 1){
		cerr << "Unable to obtain vmlinux end address." << endl;
		cerr << "The specified vmlinux file (" << vmlinux_file << ") "
		     << "does not seem to be valid." << endl;
		cerr << "Make sure you are using a non-compressed image file "
		     << "(e.g. vmlinux not vmlinuz)" << endl;
		exit(EXIT_FAILURE);
	}

	errno = 0;
	kernel_start = strtoull(start, NULL, 16);
	if (errno) {
		cerr << "Unable to convert vmlinux start address " << start
		     << " to a valid hex value. errno is " << strerror(errno) << endl;
		exit(EXIT_FAILURE);
	}
	errno = 0;
	kernel_end =  strtoull(end, NULL, 16);
	if (errno) {
		cerr << "Unable to convert vmlinux end address " << start
		     << " to a valid hex value. errno is " << strerror(errno) << endl;
		exit(EXIT_FAILURE);
	}

	start_end = start;
	start_end.append(",");
	start_end.append(end);
	return start_end;
}

static void _print_valid_verbose_options(void)
{
	cerr << "Valid verbosity options are: ";
	for (unsigned i = 0; i < (NUM_VERBOSE_OPTIONS - 1); i++)
		cerr << valid_verbose_vals[i] << ",";
	cerr << valid_verbose_vals[NUM_VERBOSE_OPTIONS - 1] << endl;
}

static bool _validate_verbose_args(char * verbosity)
{
	bool valid_verbosity = true;

	char * verbose_cand = strtok(verbosity, ",");
	do {
		unsigned i;
		for (i = 0; i < (NUM_VERBOSE_OPTIONS); i++) {
			if (!strcmp(verbose_cand, valid_verbose_vals[i])) {
				verbose_string.push_back(verbose_cand);
				break;
			}
		}
		if (i == NUM_VERBOSE_OPTIONS) {
			valid_verbosity = false;
			cerr << "Verbosity argument " << verbose_cand << " is not valid." << endl;
			_print_valid_verbose_options();
		}
	} while ((verbose_cand = strtok(NULL, ",")) && valid_verbosity);
	return valid_verbosity;
}

static int _process_operf_and_app_args(int argc, char * const argv[])
{
	bool keep_trying = true;
	int idx_of_non_options = 0;
	setenv("POSIXLY_CORRECT", "1", 0);
	while (keep_trying) {
		int option_idx = 0;
		int c = getopt_long(argc, argv, short_options, long_options, &option_idx);
		switch (c) {
		char * endptr;
		char * event;

		case -1:
			if (optind != argc) {
				idx_of_non_options = optind;
			}
			keep_trying = false;
			break;
		case '?':
			cerr << "non-option detected at optind " << optind << endl;
			keep_trying = false;
			idx_of_non_options = -1;
			break;
		case 'V':
			if (!_validate_verbose_args(optarg))
				__print_usage_and_exit("NULL");
			break;
		case 'd':
			operf_options::session_dir = optarg;
			break;
		case 'k':
			operf_options::vmlinux = optarg;
			break;
		case 'g':
			operf_options::callgraph = true;
			break;
		case 's':
			operf_options::system_wide = true;
			break;
		case 'a':
			operf_options::append = true;
			break;
		case 'p':
			operf_options::pid = strtol(optarg, &endptr, 10);
			if ((endptr >= optarg) && (endptr <= (optarg + strlen(optarg) - 1)))
				__print_usage_and_exit("operf: Invalid numeric value for --pid option.");
			break;
		case 'e':
			event = strtok(optarg, ",");
			do {
				operf_options::evts.push_back(event);
			} while ((event = strtok(NULL, ",")));
			break;
		case 'c':
			operf_options::separate_cpu = true;
			break;
		case 't':
			operf_options::separate_thread = true;
			break;
		case 'l':
			operf_options::post_conversion = true;
			break;
		case 'h':
			__print_usage_and_exit(NULL);
			break;
		case 'u':
			__print_usage_and_exit(NULL);
			break;
		case 'v':
			cout << argv[0] << ": " << PACKAGE << " " << VERSION << " compiled on " << __DATE__
			     << " " << __TIME__ << endl;
			exit(EXIT_SUCCESS);
			break;
		default:
			__print_usage_and_exit("unexpected end of arg parsing");
		}
	}
	return idx_of_non_options;
}

static void process_args(int argc, char * const argv[])
{
	int non_options_idx  = _process_operf_and_app_args(argc, argv);

	if (non_options_idx < 0) {
		__print_usage_and_exit(NULL);
	} else if ((non_options_idx) > 0) {
		if (operf_options::pid || operf_options::system_wide)
			__print_usage_and_exit(NULL);

		app_name = (char *) xmalloc(strlen(argv[non_options_idx]) + 1);
		strcpy(app_name, argv[non_options_idx]);
		// Note 1: app_args[0] is placeholder for app_fname (filled in later).
		// Note 2: app_args[<end>] is set to NULL (required by execvp)
		if (non_options_idx < (argc -1)) {
			app_args = (char **) xmalloc((sizeof *app_args) *
			                             (argc - non_options_idx + 1));
			for(int i = non_options_idx + 1; i < argc; i++) {
				app_args[i - non_options_idx] = argv[i];
			}
			app_args[argc - non_options_idx] = NULL;
		} else {
			app_args = (char **) xmalloc((sizeof *app_args) * 2);
			app_args[1] = NULL;
		}
		if (validate_app_name() < 0) {
			__print_usage_and_exit(NULL);
		}
	} else {  // non_options_idx == 0
		if (operf_options::pid) {
			if (operf_options::system_wide)
				__print_usage_and_exit(NULL);
			app_PID = operf_options::pid;
		} else if (operf_options::system_wide) {
			app_PID = -1;
		} else {
			__print_usage_and_exit(NULL);
		}
	}
	/*  At this point, we know which of the three kinds of profiles the user requested:
	 *    - profile app by name
	 *    - profile app by PID
	 *    - profile whole system
	 */

	if (!verbose::setup(verbose_string)) {
		cerr << "unknown --verbose= options\n";
		exit(EXIT_FAILURE);
	}

	_process_session_dir();
	if (operf_options::post_conversion)
		outputfile = samples_dir + "/" + DEFAULT_OPERF_OUTFILE;

	if (operf_options::evts.empty()) {
		// Use default event
		get_default_event();
	} else  {
		_process_events_list();
	}
	op_nr_events = events.size();

	if (operf_options::vmlinux.empty()) {
		no_vmlinux = true;
		operf_create_vmlinux(NULL, NULL);
	} else {
		string startEnd = _process_vmlinux(operf_options::vmlinux);
		operf_create_vmlinux(operf_options::vmlinux.c_str(), startEnd.c_str());
	}

	return;
}

static int _get_cpu_for_perf_events_cap(void)
{
	int retval;
	string err_msg;
	char cpus_online[257];
	FILE * online_cpus;
	DIR *dir = NULL;

	int total_cpus = sysconf(_SC_NPROCESSORS_ONLN);
	if (!total_cpus) {
		err_msg = "Internal Error (1): Number of online cpus cannot be determined.";
		retval = -1;
		goto error;
	}

	online_cpus = fopen("/sys/devices/system/cpu/online", "r");
	if (!online_cpus) {
		err_msg = "Internal Error (2): Number of online cpus cannot be determined.";
		retval = -1;
		goto error;
	}
	memset(cpus_online, 0, sizeof(cpus_online));

	if ( fgets(cpus_online, sizeof(cpus_online), online_cpus) == NULL) {
		fclose(online_cpus);
		err_msg = "Internal Error (3): Number of online cpus cannot be determined.";
		retval = -1;
		goto error;
	}

	if (!cpus_online[0]) {
		fclose(online_cpus);
		err_msg = "Internal Error (4): Number of online cpus cannot be determined.";
		retval = -1;
		goto error;

	}
	if (index(cpus_online, ',') || cpus_online[0] != '0') {
		// A comma in cpus_online implies a gap, which in turn implies that not all
		// CPUs are online.
		if ((dir = opendir("/sys/devices/system/cpu")) == NULL) {
			fclose(online_cpus);
			err_msg = "Internal Error (5): Number of online cpus cannot be determined.";
			retval = -1;
			goto error;
		} else {
			struct dirent *entry = NULL;
			retval = OP_perf_utils::op_get_next_online_cpu(dir, entry);
			closedir(dir);
		}
	} else {
		// All CPUs are available, so we just arbitrarily choose CPU 0.
		retval = 0;
	}
	fclose(online_cpus);
error:
	return retval;
}


static int _check_perf_events_cap(bool use_cpu_minus_one)
{
	/* If perf_events syscall is not implemented, the syscall below will fail
	 * with ENOSYS (38).  If implemented, but the processor type on which this
	 * program is running is not supported by perf_events, the syscall returns
	 * ENOENT (2).
	 */
	struct perf_event_attr attr;
	pid_t pid ;
	int cpu_to_try = use_cpu_minus_one ? -1 : _get_cpu_for_perf_events_cap();
	errno = 0;
        memset(&attr, 0, sizeof(attr));
        attr.size = sizeof(attr);
        attr.sample_type = PERF_SAMPLE_IP;

	pid = getpid();
	syscall(__NR_perf_event_open, &attr, pid, cpu_to_try, -1, 0);
	return errno;

}

static void _precheck_permissions_to_samplesdir(string sampledir, bool for_current)
{
	/* Pre-check to make sure we have permission to remove old sample data
	 * or to create new sample data in the specified sample data directory.
	 * If the user wants us to remove old data, we don't actually do it now,
	 * since the profile session may fail for some reason or the user may do ctl-c.
	 * We should exit without unnecessarily removing the old sample data as
	 * the user may expect it to still be there after an aborted run.
	 */
	string sampledir_testfile = sampledir + "/.xxxTeStFiLe";
	ofstream afile;
	errno = 0;
	afile.open(sampledir_testfile.c_str());
	if (!afile.is_open() && (errno != ENOENT)) {
		if (operf_options::append && for_current)
			cerr << "Unable to write to sample data directory at "
			     << sampledir << "." << endl;
		else
			cerr << "Unable to remove old sample data at "
			     << sampledir << "." << endl;
		if (errno)
			cerr << strerror(errno) << endl;
		cerr << "Try a manual removal of " << sampledir << endl;
		cleanup();
		exit(1);
	}
	afile.close();

}

static int _get_sys_value(const char * filename)
{
	char str[10];
	int _val = -999;
	FILE * fp = fopen(filename, "r");
	if (fp == NULL)
		return _val;
	if (fgets(str, 9, fp))
		sscanf(str, "%d", &_val);
	fclose(fp);
	return _val;
}


int main(int argc, char * const argv[])
{
	int rc;
	int perf_event_paranoid = _get_sys_value("/proc/sys/kernel/perf_event_paranoid");

	my_uid = geteuid();
	throttled = false;
	rc = _check_perf_events_cap(use_cpu_minus_one);
	if (rc == EACCES) {
		/* Early perf_events kernels required the cpu argument to perf_event_open
		 * to be '-1' when setting up to profile a single process if 1) the user is
		 * not root; and 2) perf_event_paranoid is > 0.  An EACCES error would be
		 * returned if passing '0' or greater for the cpu arg and the above criteria
		 * was not met.  Unfortunately, later kernels turned this requirement around
		 * such that the passed cpu arg must be '0' or greater when the user is not
		 * root.
		 *
		 * We don't really have a good way to check whether we're running on such an
		 * early kernel except to try the perf_event_open with different values to see
		 * what works.
		 */
		if (my_uid != 0 && perf_event_paranoid > 0) {
			use_cpu_minus_one = true;
			rc = _check_perf_events_cap(use_cpu_minus_one);
		}
	}
	if (rc == EBUSY) {
		cerr << "Performance monitor unit is busy.  Do 'opcontrol --deinit' and try again." << endl;
		exit(1);
	}
	if (rc == ENOSYS) {
		cerr << "Your kernel does not implement a required syscall"
		     << " for the operf program." << endl;
	} else if (rc == ENOENT) {
		cerr << "Your kernel's Performance Events Subsystem does not support"
		     << " your processor type." << endl;
	} else if (rc) {
		cerr << "Unexpected error running operf: " << strerror(rc) << endl;
	}
	if (rc) {
		cerr << "Please use the opcontrol command instead of operf." << endl;
		exit(1);
	}

	cpu_type = op_get_cpu_type();
	cpu_speed = op_cpu_frequency();
	process_args(argc, argv);

	if (operf_options::system_wide && ((my_uid != 0) && (perf_event_paranoid > 0))) {
		cerr << "To do system-wide profiling, either you must be root or" << endl;
		cerr << "/proc/sys/kernel/perf_event_paranoid must be set to 0 or -1." << endl;
		cleanup();
		exit(1);
	}

	if (cpu_type == CPU_NO_GOOD) {
		cerr << "Unable to ascertain cpu type.  Exiting." << endl;
		cleanup();
		exit(1);
	}

	if (my_uid != 0) {
		bool for_current = true;
		string current_sampledir = samples_dir + "/current";
		_precheck_permissions_to_samplesdir(current_sampledir, for_current);
		if (!operf_options::append) {
			string previous_sampledir = samples_dir + "/previous";
			for_current = false;
			_precheck_permissions_to_samplesdir(previous_sampledir, for_current);
		}
	}
	kptr_restrict = _get_sys_value("/proc/sys/kernel/kptr_restrict");
	end_code_t run_result;
	if ((run_result = _run())) {
		if (startApp && app_started && (run_result != APP_ABNORMAL_END)) {
			int rc;
			cverb << vdebug << "Killing profiled app . . ." << endl;
			rc = kill(app_PID, SIGKILL);
			if (rc) {
				if (errno == ESRCH)
					cverb << vdebug
					      << "Unable to kill profiled app because it has already ended"
					      << endl;
				else
					perror("Attempt to kill profiled app failed.");
			}
		}
		if ((run_result == PERF_RECORD_ERROR) || (run_result == PERF_BOTH_ERROR)) {
			cerr <<  "Error running profiler" << endl;
		} else if (run_result == PERF_READ_ERROR) {
			cerr << "Error converting operf sample data to oprofile sample format" << endl;
		} else {
			cerr << "WARNING: Profile results may be incomplete due to to abend of profiled app." << endl;
		}
	} else {
		cerr << endl << "Profiling done." << endl;
	}
	if (operf_options::post_conversion) {
		if (!(!app_started && !operf_options::system_wide))
			convert_sample_data();
	}
	cleanup();
	return run_result;;
}