sparse.c

gnu tar 源码包。 tar 软件是 Unix 系统下的一个打包软件

  if (rc && file.optab->dump_region)
    {
      tar_sparse_dump_header (&file);

      if (fd >= 0)
	{
	  size_t i;

	  mv_begin (file.stat_info);
	  for (i = 0; rc && i < file.stat_info->sparse_map_avail; i++)
	    rc = tar_sparse_dump_region (&file, i);
	  mv_end ();
	}
    }

  pad_archive (file.stat_info->archive_file_size - file.dumped_size);
  return (tar_sparse_done (&file) && rc) ? dump_status_ok : dump_status_short;
}

bool
sparse_member_p (struct tar_stat_info *st)
{
  struct tar_sparse_file file;

  if (!tar_sparse_init (&file))
    return false;
  file.stat_info = st;
  return tar_sparse_member_p (&file);
}

bool
sparse_fixup_header (struct tar_stat_info *st)
{
  struct tar_sparse_file file;

  if (!tar_sparse_init (&file))
    return false;
  file.stat_info = st;
  return tar_sparse_fixup_header (&file);
}

enum dump_status
sparse_extract_file (int fd, struct tar_stat_info *st, off_t *size)
{
  bool rc = true;
  struct tar_sparse_file file;
  size_t i;

  if (!tar_sparse_init (&file))
    return dump_status_not_implemented;

  file.stat_info = st;
  file.fd = fd;
  file.seekable = lseek (fd, 0, SEEK_SET) == 0;
  file.offset = 0;

  rc = tar_sparse_decode_header (&file);
  for (i = 0; rc && i < file.stat_info->sparse_map_avail; i++)
    rc = tar_sparse_extract_region (&file, i);
  *size = file.stat_info->archive_file_size - file.dumped_size;
  return (tar_sparse_done (&file) && rc) ? dump_status_ok : dump_status_short;
}

enum dump_status
sparse_skip_file (struct tar_stat_info *st)
{
  bool rc = true;
  struct tar_sparse_file file;

  if (!tar_sparse_init (&file))
    return dump_status_not_implemented;

  file.stat_info = st;
  file.fd = -1;

  rc = tar_sparse_decode_header (&file);
  skip_file (file.stat_info->archive_file_size - file.dumped_size);
  return (tar_sparse_done (&file) && rc) ? dump_status_ok : dump_status_short;
}


static bool
check_sparse_region (struct tar_sparse_file *file, off_t beg, off_t end)
{
  if (!lseek_or_error (file, beg))
    return false;

  while (beg < end)
    {
      size_t bytes_read;
      size_t rdsize = BLOCKSIZE < end - beg ? BLOCKSIZE : end - beg;
      char diff_buffer[BLOCKSIZE];

      bytes_read = safe_read (file->fd, diff_buffer, rdsize);
      if (bytes_read == SAFE_READ_ERROR)
	{
          read_diag_details (file->stat_info->orig_file_name,
	                     beg,
			     rdsize);
	  return false;
	}
      if (!zero_block_p (diff_buffer, bytes_read))
	{
	  char begbuf[INT_BUFSIZE_BOUND (off_t)];
 	  report_difference (file->stat_info,
			     _("File fragment at %s is not a hole"),
			     offtostr (beg, begbuf));
	  return false;
	}

      beg += bytes_read;
    }
  return true;
}

static bool
check_data_region (struct tar_sparse_file *file, size_t i)
{
  size_t size_left;

  if (!lseek_or_error (file, file->stat_info->sparse_map[i].offset))
    return false;
  size_left = file->stat_info->sparse_map[i].numbytes;
  mv_size_left (file->stat_info->archive_file_size - file->dumped_size);
      
  while (size_left > 0)
    {
      size_t bytes_read;
      size_t rdsize = (size_left > BLOCKSIZE) ? BLOCKSIZE : size_left;
      char diff_buffer[BLOCKSIZE];

      union block *blk = find_next_block ();
      if (!blk)
	{
	  ERROR ((0, 0, _("Unexpected EOF in archive")));
	  return false;
	}
      set_next_block_after (blk);
      bytes_read = safe_read (file->fd, diff_buffer, rdsize);
      if (bytes_read == SAFE_READ_ERROR)
	{
          read_diag_details (file->stat_info->orig_file_name,
			     (file->stat_info->sparse_map[i].offset
			      + file->stat_info->sparse_map[i].numbytes
			      - size_left),
			     rdsize);
	  return false;
	}
      file->dumped_size += bytes_read;
      size_left -= bytes_read;
      mv_size_left (file->stat_info->archive_file_size - file->dumped_size);
      if (memcmp (blk->buffer, diff_buffer, rdsize))
	{
	  report_difference (file->stat_info, _("Contents differ"));
	  return false;
	}
    }
  return true;
}

bool
sparse_diff_file (int fd, struct tar_stat_info *st)
{
  bool rc = true;
  struct tar_sparse_file file;
  size_t i;
  off_t offset = 0;

  if (!tar_sparse_init (&file))
    return dump_status_not_implemented;

  file.stat_info = st;
  file.fd = fd;
  file.seekable = true; /* File *must* be seekable for compare to work */
  
  rc = tar_sparse_decode_header (&file);
  mv_begin (st);
  for (i = 0; rc && i < file.stat_info->sparse_map_avail; i++)
    {
      rc = check_sparse_region (&file,
				offset, file.stat_info->sparse_map[i].offset)
	    && check_data_region (&file, i);
      offset = file.stat_info->sparse_map[i].offset
	        + file.stat_info->sparse_map[i].numbytes;
    }

  if (!rc)
    skip_file (file.stat_info->archive_file_size - file.dumped_size);
  mv_end ();
  
  tar_sparse_done (&file);
  return rc;
}


/* Old GNU Format. The sparse file information is stored in the
   oldgnu_header in the following manner:

   The header is marked with type 'S'. Its `size' field contains
   the cumulative size of all non-empty blocks of the file. The
   actual file size is stored in `realsize' member of oldgnu_header.

   The map of the file is stored in a list of `struct sparse'.
   Each struct contains offset to the block of data and its
   size (both as octal numbers). The first file header contains
   at most 4 such structs (SPARSES_IN_OLDGNU_HEADER). If the map
   contains more structs, then the field `isextended' of the main
   header is set to 1 (binary) and the `struct sparse_header'
   header follows, containing at most 21 following structs
   (SPARSES_IN_SPARSE_HEADER). If more structs follow, `isextended'
   field of the extended header is set and next  next extension header
   follows, etc... */

enum oldgnu_add_status
  {
    add_ok,
    add_finish,
    add_fail
  };

static bool
oldgnu_sparse_member_p (struct tar_sparse_file *file __attribute__ ((unused)))
{
  return current_header->header.typeflag == GNUTYPE_SPARSE;
}

/* Add a sparse item to the sparse file and its obstack */
static enum oldgnu_add_status
oldgnu_add_sparse (struct tar_sparse_file *file, struct sparse *s)
{
  struct sp_array sp;

  if (s->numbytes[0] == '\0')
    return add_finish;
  sp.offset = OFF_FROM_HEADER (s->offset);
  sp.numbytes = SIZE_FROM_HEADER (s->numbytes);
  if (sp.offset < 0
      || file->stat_info->stat.st_size < sp.offset + sp.numbytes
      || file->stat_info->archive_file_size < 0)
    return add_fail;

  sparse_add_map (file->stat_info, &sp);
  return add_ok;
}

static bool
oldgnu_fixup_header (struct tar_sparse_file *file)
{
  /* NOTE! st_size was initialized from the header
     which actually contains archived size. The following fixes it */
  file->stat_info->archive_file_size = file->stat_info->stat.st_size;
  file->stat_info->stat.st_size =
    OFF_FROM_HEADER (current_header->oldgnu_header.realsize);
  return true;
}

/* Convert old GNU format sparse data to internal representation */
static bool
oldgnu_get_sparse_info (struct tar_sparse_file *file)
{
  size_t i;
  union block *h = current_header;
  int ext_p;
  enum oldgnu_add_status rc;

  file->stat_info->sparse_map_avail = 0;
  for (i = 0; i < SPARSES_IN_OLDGNU_HEADER; i++)
    {
      rc = oldgnu_add_sparse (file, &h->oldgnu_header.sp[i]);
      if (rc != add_ok)
	break;
    }

  for (ext_p = h->oldgnu_header.isextended;
       rc == add_ok && ext_p; ext_p = h->sparse_header.isextended)
    {
      h = find_next_block ();
      if (!h)
	{
	  ERROR ((0, 0, _("Unexpected EOF in archive")));
	  return false;
	}
      set_next_block_after (h);
      for (i = 0; i < SPARSES_IN_SPARSE_HEADER && rc == add_ok; i++)
	rc = oldgnu_add_sparse (file, &h->sparse_header.sp[i]);
    }

  if (rc == add_fail)
    {
      ERROR ((0, 0, _("%s: invalid sparse archive member"),
	      file->stat_info->orig_file_name));
      return false;
    }
  return true;
}

static void
oldgnu_store_sparse_info (struct tar_sparse_file *file, size_t *pindex,
			  struct sparse *sp, size_t sparse_size)
{
  for (; *pindex < file->stat_info->sparse_map_avail
	 && sparse_size > 0; sparse_size--, sp++, ++*pindex)
    {
      OFF_TO_CHARS (file->stat_info->sparse_map[*pindex].offset,
		    sp->offset);
      SIZE_TO_CHARS (file->stat_info->sparse_map[*pindex].numbytes,
		     sp->numbytes);
    }
}

static bool
oldgnu_dump_header (struct tar_sparse_file *file)
{
  off_t block_ordinal = current_block_ordinal ();
  union block *blk;
  size_t i;

  blk = start_header (file->stat_info);
  blk->header.typeflag = GNUTYPE_SPARSE;
  if (file->stat_info->sparse_map_avail > SPARSES_IN_OLDGNU_HEADER)
    blk->oldgnu_header.isextended = 1;

  /* Store the real file size */
  OFF_TO_CHARS (file->stat_info->stat.st_size, blk->oldgnu_header.realsize);
  /* Store the effective (shrunken) file size */
  OFF_TO_CHARS (file->stat_info->archive_file_size, blk->header.size);

  i = 0;
  oldgnu_store_sparse_info (file, &i,
			    blk->oldgnu_header.sp,
			    SPARSES_IN_OLDGNU_HEADER);
  blk->oldgnu_header.isextended = i < file->stat_info->sparse_map_avail;
  finish_header (file->stat_info, blk, block_ordinal);

  while (i < file->stat_info->sparse_map_avail)
    {
      blk = find_next_block ();
      memset (blk->buffer, 0, BLOCKSIZE);
      oldgnu_store_sparse_info (file, &i,
				blk->sparse_header.sp,
				SPARSES_IN_SPARSE_HEADER);
      if (i < file->stat_info->sparse_map_avail)
	blk->sparse_header.isextended = 1;
      set_next_block_after (blk);
    }
  return true;
}

static struct tar_sparse_optab const oldgnu_optab = {
  NULL,  /* No init function */
  NULL,  /* No done function */
  oldgnu_sparse_member_p,
  oldgnu_dump_header,
  oldgnu_fixup_header,
  oldgnu_get_sparse_info,
  NULL,  /* No scan_block function */
  sparse_dump_region,
  sparse_extract_region,
};


/* Star */

static bool
star_sparse_member_p (struct tar_sparse_file *file __attribute__ ((unused)))
{
  return current_header->header.typeflag == GNUTYPE_SPARSE;
}

static bool
star_fixup_header (struct tar_sparse_file *file)
{
  /* NOTE! st_size was initialized from the header
     which actually contains archived size. The following fixes it */
  file->stat_info->archive_file_size = file->stat_info->stat.st_size;
  file->stat_info->stat.st_size =
            OFF_FROM_HEADER (current_header->star_in_header.realsize);
  return true;
}

/* Convert STAR format sparse data to internal representation */
static bool
star_get_sparse_info (struct tar_sparse_file *file)
{
  size_t i;
  union block *h = current_header;
  int ext_p;
  enum oldgnu_add_status rc = add_ok;

  file->stat_info->sparse_map_avail = 0;

  if (h->star_in_header.prefix[0] == '\0'