blob: a84482a812472be4a2122c4313dacf1b8ef53a44 [file] [log] [blame]
/**
* @file parse_filename.cpp
* Split a sample filename into its constituent parts
*
* @remark Copyright 2003 OProfile authors
* @remark Read the file COPYING
*
* @author Philippe Elie
*/
#include <stdexcept>
#include <vector>
#include <string>
#include <iostream>
#include <sys/stat.h>
#include "parse_filename.h"
#include "file_manip.h"
#include "string_manip.h"
#include "locate_images.h"
using namespace std;
namespace {
// PP:3.19 event_name.count.unitmask.tgid.tid.cpu
parsed_filename parse_event_spec(string const & event_spec)
{
typedef vector<string> parts_type;
typedef parts_type::size_type size_type;
size_type const nr_parts = 6;
parts_type parts = separate_token(event_spec, '.');
if (parts.size() != nr_parts) {
throw invalid_argument("parse_event_spec(): bad event specification: " + event_spec);
}
for (size_type i = 0; i < nr_parts ; ++i) {
if (parts[i].empty()) {
throw invalid_argument("parse_event_spec(): bad event specification: " + event_spec);
}
}
parsed_filename result;
size_type i = 0;
result.event = parts[i++];
result.count = parts[i++];
result.unitmask = parts[i++];
result.tgid = parts[i++];
result.tid = parts[i++];
result.cpu = parts[i++];
return result;
}
/**
* @param component path component
*
* remove from path_component all directory left to {root}, {kern} or {anon}
*/
void remove_base_dir(vector<string> & path)
{
vector<string>::iterator it;
for (it = path.begin(); it != path.end(); ++it) {
if (*it == "{root}" || *it == "{kern}" || *it == "{anon}")
break;
}
path.erase(path.begin(), it);
}
/// Handle an anon region. Pretty print the details.
/// The second argument is the anon portion of the path which will
/// contain extra details such as the anon region name (unknown, vdso, heap etc.)
string const parse_anon(string const & str, string const & str2)
{
string name = str2;
// Get rid of "{anon:
name.erase(0, 6);
// Catch the case where we end up with an empty string. This should
// never happen, except where things have gone awfully bad with profile
// data collection, resulting in one or more bogus sample files.
if(0 == name.size())
throw invalid_argument("parse_anon() invalid name: " + str2 + "\n"
+ "This error indicates your sample data is suspect. It is "
+ "recommended you do a --reset and collect new profile data.");
// Get rid of the trailing '}'
name.erase(name.size() - 1, 1);
vector<string> parts = separate_token(str, '.');
if (parts.size() != 3)
throw invalid_argument("parse_anon() invalid name: " + str);
string ret = name +" (tgid:";
ret += parts[0] + " range:" + parts[1] + "-" + parts[2] + ")";
return ret;
}
} // anonymous namespace
/*
* valid filename are variations on:
*
* {kern}/name/event_spec
* {root}/path/to/bin/{dep}/{root}/path/to/bin/event_spec
* {root}/path/to/bin/{dep}/{anon:anon}/pid.start.end/event_spec
* {root}/path/to/bin/{dep}/{anon:[vdso]}/pid.start.end/event_spec
* {root}/path/to/bin/{dep}/{kern}/name/event_spec
* {root}/path/to/bin/{dep}/{root}/path/to/bin/{cg}/{root}/path/to/bin/event_spec
*
* where /name/ denote a unique path component
*/
parsed_filename parse_filename(string const & filename,
extra_images const & extra_found_images)
{
struct stat st;
string::size_type pos = filename.find_last_of('/');
if (pos == string::npos) {
throw invalid_argument("parse_filename() invalid filename: " +
filename);
}
string event_spec = filename.substr(pos + 1);
string filename_spec = filename.substr(0, pos);
parsed_filename result = parse_event_spec(event_spec);
result.filename = filename;
vector<string> path = separate_token(filename_spec, '/');
remove_base_dir(path);
// pp_interface PP:3.19 to PP:3.23 path must start either with {root}
// or {kern} and we must found at least 2 component, remove_base_dir()
// return an empty path if {root} or {kern} are not found
if (path.size() < 2) {
throw invalid_argument("parse_filename() invalid filename: " +
filename);
}
size_t i;
for (i = 1 ; i < path.size() ; ++i) {
if (path[i] == "{dep}")
break;
result.image += "/" + path[i];
}
if (i == path.size()) {
throw invalid_argument("parse_filename() invalid filename: " +
filename);
}
// skip "{dep}"
++i;
// PP:3.19 {dep}/ must be followed by {kern}/, {root}/ or {anon}/
if (path[i] != "{kern}" && path[i] != "{root}" &&
path[i].find("{anon", 0) != 0) {
throw invalid_argument("parse_filename() invalid filename: " +
filename);
}
bool anon = path[i].find("{anon:", 0) == 0;
// skip "{root}", "{kern}" or "{anon:.*}"
++i;
for (; i < path.size(); ++i) {
if (path[i] == "{cg}")
break;
if (anon) {
pos = filename_spec.rfind('.');
pos = filename_spec.rfind('.', pos-1);
if (pos == string::npos) {
throw invalid_argument("parse_filename() pid.addr.addr name expected: " +
filename_spec);
}
string jitdump = filename_spec.substr(0, pos) + ".jo";
// if a jitdump file exists, we point to this file
if (!stat(jitdump.c_str(), &st)) {
// later code assumes an optional prefix path
// is stripped from the lib_image.
result.lib_image =
extra_found_images.strip_path_prefix(jitdump);
result.jit_dumpfile_exists = true;
} else {
result.lib_image = parse_anon(path[i], path[i - 1]);
}
i++;
break;
} else {
result.lib_image += "/" + path[i];
}
}
if (i == path.size())
return result;
// skip "{cg}"
++i;
if (i == path.size() ||
(path[i] != "{kern}" && path[i] != "{root}" &&
path[i].find("{anon", 0) != 0)) {
throw invalid_argument("parse_filename() invalid filename: "
+ filename);
}
// skip "{root}", "{kern}" or "{anon}"
anon = (path[i].find("{anon", 0) == 0);
++i;
if (anon) {
result.cg_image = parse_anon(path[i], path[i - 1]);
i++;
} else {
for (; i < path.size(); ++i)
result.cg_image += "/" + path[i];
}
return result;
}
bool parsed_filename::profile_spec_equal(parsed_filename const & parsed)
{
return event == parsed.event &&
count == parsed.count &&
unitmask == parsed.unitmask &&
tgid == parsed.tgid &&
tid == parsed.tid &&
cpu == parsed.cpu;
}
ostream & operator<<(ostream & out, parsed_filename const & data)
{
out << data.filename << endl;
out << data.image << " " << data.lib_image << " "
<< data.event << " " << data.count << " "
<< data.unitmask << " " << data.tgid << " "
<< data.tid << " " << data.cpu << endl;
return out;
}