open_toontown_panda3d/direct/src/plugin/fileSpec.cxx

/**
 * PANDA 3D SOFTWARE
 * Copyright (c) Carnegie Mellon University.  All rights reserved.
 *
 * All use of this software is subject to the terms of the revised BSD
 * license.  You should have received a copy of this license along
 * with this source code in a file named "LICENSE."
 *
 * @file fileSpec.cxx
 * @author drose
 * @date 2009-06-29
 */

#include "fileSpec.h"
#include "wstring_encode.h"
#include "openssl/md5.h"

#include <fstream>
#include <fcntl.h>
#include <assert.h>

#include <sys/types.h>
#include <sys/stat.h>
#include <time.h>

#ifdef _WIN32
#include <sys/utime.h>
#include <direct.h>
#define utimbuf _utimbuf

#else
#include <utime.h>

#endif

/**
 *
 */
FileSpec::
FileSpec() {
  _size = 0;
  _timestamp = 0;
  memset(_hash, 0, hash_size);
  _got_hash = false;
  _actual_file = NULL;
}

/**
 *
 */
FileSpec::
FileSpec(const FileSpec &copy) :
  _filename(copy._filename),
  _size(copy._size),
  _timestamp(copy._timestamp),
  _got_hash(copy._got_hash)
{
  memcpy(_hash, copy._hash, hash_size);
  _actual_file = NULL;
}

/**
 *
 */
void FileSpec::
operator = (const FileSpec &copy) {
  _filename = copy._filename;
  _size = copy._size;
  _timestamp = copy._timestamp;
  memcpy(_hash, copy._hash, hash_size);
  _got_hash = copy._got_hash;
}

/**
 *
 */
FileSpec::
~FileSpec() {
  if (_actual_file != NULL) {
    delete _actual_file;
  }
}

/**
 * Reads the data from the indicated XML file.
 */
void FileSpec::
load_xml(TiXmlElement *xelement) {
  const char *filename = xelement->Attribute("filename");
  if (filename != NULL) {
    _filename = filename;
  }

  const char *size = xelement->Attribute("size");
  if (size != NULL) {
    char *endptr;
    _size = strtoul(size, &endptr, 10);
  }

  const char *timestamp = xelement->Attribute("timestamp");
  if (timestamp != NULL) {
    char *endptr;
    _timestamp = strtoul(timestamp, &endptr, 10);
  }

  _got_hash = false;
  const char *hash = xelement->Attribute("hash");
  if (hash != NULL && strlen(hash) == (hash_size * 2)) {
    // Decode the hex hash string.
    _got_hash = decode_hex(_hash, hash, hash_size);
  }
}

/**
 * Stores the data to the indicated XML file.
 */
void FileSpec::
store_xml(TiXmlElement *xelement) {
  if (!_filename.empty()) {
    xelement->SetAttribute("filename", _filename);
  }
  if (_size != 0) {
    xelement->SetAttribute("size", _size);
  }
  if (_timestamp != 0) {
    xelement->SetAttribute("timestamp", (int)_timestamp);
  }
  if (_got_hash) {
    char hash[hash_size * 2 + 1];
    encode_hex(hash, _hash, hash_size);
    hash[hash_size * 2] = '\0';
    xelement->SetAttribute("hash", hash);
  }
}

/**
 * Performs a quick test to ensure the file has not been modified.  This test
 * is vulnerable to people maliciously attempting to fool the program (by
 * setting datestamps etc.).
 *
 * Returns true if it is intact, false if it needs to be redownloaded.
 */
bool FileSpec::
quick_verify(const string &package_dir) {
  return quick_verify_pathname(get_pathname(package_dir));
}

/**
 * Works like quick_verify(), above, with an explicit pathname.  Useful for
 * verifying the copy of a file in a temporary location.
 */
bool FileSpec::
quick_verify_pathname(const string &pathname) {
  if (_actual_file != NULL) {
    delete _actual_file;
    _actual_file = NULL;
  }

  int result = 1;
#ifdef _WIN32
  struct _stat st;
  wstring pathname_w;
  if (string_to_wstring(pathname_w, pathname)) {
    result = _wstat(pathname_w.c_str(), &st);
  }
#else // _WIN32
  struct stat st;
  result = stat(pathname.c_str(), &st);
#endif  // _WIN32

  if (result != 0) {
    // cerr << "file not found: " << _filename << "\n";
    return false;
  }

  if (st.st_size != _size) {
    // If the size is wrong, the file fails.  cerr << "size wrong: " <<
    // _filename << "\n";
    return false;
  }

  if (st.st_mtime == _timestamp) {
    // If the size is right and the timestamp is right, the file passes.  cerr
    // << "file ok: " << _filename << "\n";
    return true;
  }

  // cerr << "modification time wrong: " << _filename << "\n";

  // If the size is right but the timestamp is wrong, the file soft-fails.  We
  // follow this up with a hash check.
  if (!priv_check_hash(pathname, &st)) {
    // Hard fail, the hash is wrong.  cerr << "hash check wrong: " <<
    // _filename << "\n";
    return false;
  }

  // cerr << "hash check ok: " << _filename << "\n";

  // The hash is OK after all.  Change the file's timestamp back to what we
  // expect it to be, so we can quick-verify it successfully next time.
  utimbuf utb;
  utb.actime = st.st_atime;
  utb.modtime = _timestamp;

#ifdef _WIN32
  _wutime(pathname_w.c_str(), &utb);
#else // _WIN32
  utime(pathname.c_str(), &utb);
#endif  // _WIN32

  return true;
}

/**
 * Performs a more thorough test to ensure the file has not been modified.
 * This test is less vulnerable to malicious attacks, since it reads and
 * verifies the entire file.
 *
 * Returns true if it is intact, false if it needs to be redownloaded.
 */
bool FileSpec::
full_verify(const string &package_dir) {
  if (_actual_file != NULL) {
    delete _actual_file;
    _actual_file = NULL;
  }

  string pathname = get_pathname(package_dir);
  int result = 1;
#ifdef _WIN32
  struct _stat st;
  wstring pathname_w;
  if (string_to_wstring(pathname_w, pathname)) {
    result = _wstat(pathname_w.c_str(), &st);
  }
#else // _WIN32
  struct stat st;
  result = stat(pathname.c_str(), &st);
#endif  // _WIN32

  if (result != 0) {
    // cerr << "file not found: " << _filename << "\n";
    return false;
  }

  if (st.st_size != _size) {
    // If the size is wrong, the file fails.  cerr << "size wrong: " <<
    // _filename << "\n";
    return false;
  }

  if (!priv_check_hash(pathname, &st)) {
    // Hard fail, the hash is wrong.  cerr << "hash check wrong: " <<
    // _filename << "\n";
    return false;
  }

  // cerr << "hash check ok: " << _filename << "\n";

  // The hash is OK.  If the timestamp is wrong, change it back to what we
  // expect it to be, so we can quick-verify it successfully next time.

  if (st.st_mtime != _timestamp) {
    utimbuf utb;
    utb.actime = st.st_atime;
    utb.modtime = _timestamp;
#ifdef _WIN32
    _wutime(pathname_w.c_str(), &utb);
#else // _WIN32
    utime(pathname.c_str(), &utb);
#endif  // _WIN32
  }

  return true;
}

/**
 * Returns a FileSpec that represents the actual data read on disk.  This will
 * read the disk to determine the data if necessary.
 */
const FileSpec *FileSpec::
force_get_actual_file(const string &pathname) {
  if (_actual_file == NULL) {
#ifdef _WIN32
    struct _stat st;
    wstring pathname_w;
    if (string_to_wstring(pathname_w, pathname)) {
      _wstat(pathname_w.c_str(), &st);
    }
#else // _WIN32
    struct stat st;
    stat(pathname.c_str(), &st);
#endif  // _WIN32

    priv_check_hash(pathname, &st);
  }

  return _actual_file;
}

/**
 * Returns true if the file has the expected md5 hash, false otherwise.
 */
bool FileSpec::
check_hash(const string &pathname) const {
  FileSpec other;
  if (!other.read_hash(pathname)) {
    return false;
  }

  return (memcmp(_hash, other._hash, hash_size) == 0);
}

/**
 * Computes the hash from the indicated pathname and stores it within the
 * FileSpec.
 */
bool FileSpec::
read_hash(const string &pathname) {
  memset(_hash, 0, hash_size);
  _got_hash = false;

  ifstream stream;
#ifdef _WIN32
  wstring pathname_w;
  if (string_to_wstring(pathname_w, pathname)) {
    stream.open(pathname_w.c_str(), ios::in | ios::binary);
  }
#else // _WIN32
  stream.open(pathname.c_str(), ios::in | ios::binary);
#endif  // _WIN32

  if (!stream) {
    // cerr << "unable to read " << pathname << "\n";
    return false;
  }

  MD5_CTX ctx;
  MD5_Init(&ctx);

  static const int buffer_size = 4096;
  char buffer[buffer_size];

  stream.read(buffer, buffer_size);
  size_t count = stream.gcount();
  while (count != 0) {
    MD5_Update(&ctx, buffer, count);
    stream.read(buffer, buffer_size);
    count = stream.gcount();
  }

  MD5_Final(_hash, &ctx);
  _got_hash = true;

  return true;
}

/**
 * Reads the hash from the next 16 bytes on the indicated istream, in the same
 * unusual order observed by Panda's HashVal::read_stream() method.
 */
bool FileSpec::
read_hash_stream(istream &in) {
  for (int i = 0; i < hash_size; i += 4) {
    unsigned int a = in.get();
    unsigned int b = in.get();
    unsigned int c = in.get();
    unsigned int d = in.get();
    _hash[i + 0] = d;
    _hash[i + 1] = c;
    _hash[i + 2] = b;
    _hash[i + 3] = a;
  }

  return !in.fail();
}

/**
 * Returns < 0 if this hash sorts before the other hash, > 0 if it sorts
 * after, 0 if they are the same.
 */
int FileSpec::
compare_hash(const FileSpec &other) const {
  return memcmp(_hash, other._hash, hash_size);
}

/**
 * Describes the data in the FileSpec.
 */
void FileSpec::
write(ostream &out) const {
  out << "filename: " << _filename << ", " << _size << " bytes, "
      << asctime(localtime(&_timestamp));
  // asctime includes a newline.
  out << "hash: ";
  stream_hex(out, _hash, hash_size);
  out << "\n";
}

/**
 * Writes just the hash code.
 */
void FileSpec::
output_hash(ostream &out) const {
  stream_hex(out, _hash, hash_size);
}

/**
 * Returns true if the file has the expected md5 hash, false otherwise.
 * Updates _actual_file with the data read from disk, including the hash, for
 * future reference.
 *
 * The parameter stp is a pointer to a stat structure.  It's declared as a
 * void * to get around issues with the nonstandard declaration of this
 * structure in Windows.
 */
bool FileSpec::
priv_check_hash(const string &pathname, void *stp) {
  const struct stat &st = *(const struct stat *)stp;
  assert(_actual_file == NULL);
  _actual_file = new FileSpec;
  _actual_file->_filename = pathname;
  _actual_file->_size = st.st_size;
  _actual_file->_timestamp = st.st_mtime;

  if (!_actual_file->read_hash(pathname)) {
    return false;
  }

  return (memcmp(_hash, _actual_file->_hash, hash_size) == 0);
}

/**
 * Decodes the hex string in source into the character array in dest.  dest
 * must have has least size bytes; source must have size * 2 bytes.
 *
 * Returns true on success, false if there was a non-hex digit in the string.
 */
bool FileSpec::
decode_hex(unsigned char *dest, const char *source, size_t size) {
  for (size_t i = 0; i < size; ++i) {
    int high = decode_hexdigit(source[i * 2]);
    int low = decode_hexdigit(source[i * 2 + 1]);
    if (high < 0 || low < 0) {
      return false;
    }
    dest[i] = (high << 4) | low;
  }

  return true;
}

/**
 * Encodes a character array into a hex string for output.  dest must have at
 * least size * 2 bytes; source must have size bytes.  The result is not null-
 * terminated.
 */
void FileSpec::
encode_hex(char *dest, const unsigned char *source, size_t size) {
  for (size_t i = 0; i < size; ++i) {
    int high = (source[i] >> 4) & 0xf;
    int low = source[i] & 0xf;
    dest[2 * i] = encode_hexdigit(high);
    dest[2 * i + 1] = encode_hexdigit(low);
  }
}

/**
 * Writes the indicated buffer as a string of hex characters to the given
 * ostream.
 */
void FileSpec::
stream_hex(ostream &out, const unsigned char *source, size_t size) {
  for (size_t i = 0; i < size; ++i) {
    int high = (source[i] >> 4) & 0xf;
    int low = source[i] & 0xf;
    out.put(encode_hexdigit(high));
    out.put(encode_hexdigit(low));
  }
}