Merging upstream version 0.8.

Signed-off-by: Daniel Baumann <daniel.baumann@progress-linux.org>

1 files changed, 267 insertions, 39 deletions

diff --git a/create.cc b/create.cc
index c5bed25..ba7d10a 100644
--- a/create.cc
+++ b/create.cc
@@ -28,6 +28,10 @@
 #include <stdint.h>
 #include <unistd.h>
 #include <sys/stat.h>
+#include <sys/types.h>
+#if defined(__GNU_LIBRARY__)
+#include <sys/sysmacros.h>		// for major, minor
+#endif
 #include <ftw.h>
 #include <grp.h>
 #include <pwd.h>
@@ -37,6 +41,9 @@
 #include "lzip.h"
 #include "tarlz.h"
 
+
+const CRC32C crc32c;
+
 int cl_owner = -1;		// global vars needed by add_member
 int cl_group = -1;
 int cl_solid = 0;		// 1 = dsolid, 2 = asolid, 3 = solid
@@ -44,6 +51,7 @@ int cl_solid = 0;		// 1 = dsolid, 2 = asolid, 3 = solid
 namespace {
 
 LZ_Encoder * encoder = 0;	// local vars needed by add_member
+const char * archive_namep = 0;
 int outfd = -1;
 int gretval = 0;
 
@@ -55,31 +63,67 @@ int seek_read( const int fd, uint8_t * const buf, const int size,
   return 0;
   }
 
-// Check archive type, remove EOF blocks, and leave outfd file pos at EOF
-bool check_appendable()
+// infd and outfd can refer to the same file if copying to a lower file
+// position or if source and destination blocks don't overlap.
+// max_size < 0 means no size limit.
+bool copy_file( const int infd, const int outfd, const long long max_size = -1 )
+  {
+  const int buffer_size = 65536;
+  // remaining number of bytes to copy
+  long long rest = ( ( max_size >= 0 ) ? max_size : buffer_size );
+  long long copied_size = 0;
+  uint8_t * const buffer = new uint8_t[buffer_size];
+  bool error = false;
+
+  while( rest > 0 )
+    {
+    const int size = std::min( (long long)buffer_size, rest );
+    if( max_size >= 0 ) rest -= size;
+    const int rd = readblock( infd, buffer, size );
+    if( rd != size && errno )
+      { show_error( "Error reading input file", errno ); error = true; break; }
+    if( rd > 0 )
+      {
+      const int wr = writeblock( outfd, buffer, rd );
+      if( wr != rd )
+        { show_error( "Error writing output file", errno );
+          error = true; break; }
+      copied_size += rd;
+      }
+    if( rd < size ) break;				// EOF
+    }
+  delete[] buffer;
+  return ( !error && ( max_size < 0 || copied_size == max_size ) );
+  }
+
+
+/* Check archive type. If success, leave fd file pos at 0.
+   If remove_eof, leave fd file pos at beginning of the EOF blocks. */
+bool check_appendable( const int fd, const bool remove_eof )
   {
   struct stat st;
-  if( fstat( outfd, &st ) != 0 || !S_ISREG( st.st_mode ) ) return false;
-  uint8_t buf[header_size];
-  int rd = readblock( outfd, buf, header_size );
+  if( fstat( fd, &st ) != 0 || !S_ISREG( st.st_mode ) ) return false;
+  if( lseek( fd, 0, SEEK_SET ) != 0 ) return false;
+  enum { bufsize = header_size + ( header_size / 8 ) };
+  uint8_t buf[bufsize];
+  int rd = readblock( fd, buf, bufsize );
   if( rd == 0 && errno == 0 ) return true;	// append to empty archive
-  if( rd < min_member_size || ( rd != header_size && errno ) ) return false;
-  const Lzip_header * const p = (Lzip_header *)buf;	// shut up gcc
+  if( rd < min_member_size || ( rd != bufsize && errno ) ) return false;
+  const Lzip_header * const p = (const Lzip_header *)buf;	// shut up gcc
   if( !p->verify_magic() ) return false;
   LZ_Decoder * decoder = LZ_decompress_open();	// decompress first header
   if( !decoder || LZ_decompress_errno( decoder ) != LZ_ok ||
       LZ_decompress_write( decoder, buf, rd ) != rd ||
-      ( rd = LZ_decompress_read( decoder, buf, header_size ) ) <
-      magic_o + magic_l )
+      ( rd = LZ_decompress_read( decoder, buf, header_size ) ) != header_size )
     { LZ_decompress_close( decoder ); return false; }
   LZ_decompress_close( decoder );
   const bool maybe_eof = ( buf[0] == 0 );
   if( !verify_ustar_chksum( buf ) && !maybe_eof ) return false;
-  const long long end = lseek( outfd, 0, SEEK_END );
+  const long long end = lseek( fd, 0, SEEK_END );
   if( end < min_member_size ) return false;
 
   Lzip_trailer trailer;
-  if( seek_read( outfd, trailer.data, Lzip_trailer::size,
+  if( seek_read( fd, trailer.data, Lzip_trailer::size,
                  end - Lzip_trailer::size ) != Lzip_trailer::size )
     return false;
   const long long member_size = trailer.member_size();
@@ -87,9 +131,8 @@ bool check_appendable()
       ( maybe_eof && member_size != end ) ) return false;
 
   Lzip_header header;
-  if( seek_read( outfd, header.data, Lzip_header::size,
-                 end - member_size ) != Lzip_header::size )
-    return false;
+  if( seek_read( fd, header.data, Lzip_header::size,
+                 end - member_size ) != Lzip_header::size ) return false;
   if( !header.verify_magic() || !isvalid_ds( header.dictionary_size() ) )
     return false;
 
@@ -102,12 +145,33 @@ bool check_appendable()
   crc ^= 0xFFFFFFFFU;
   if( crc != data_crc ) return false;
 
-  if( lseek( outfd, end - member_size, SEEK_SET ) != end - member_size ||
-      ftruncate( outfd, end - member_size ) != 0 ) return false;
-  return true;
+  const long long pos = remove_eof ? end - member_size : 0;
+  return ( lseek( fd, pos, SEEK_SET ) == pos );
   }
 
 
+class File_is_archive
+  {
+  dev_t archive_dev;
+  ino_t archive_ino;
+  bool initialized;
+public:
+  File_is_archive() : initialized( false ) {}
+  bool init()
+    {
+    struct stat st;
+    if( fstat( outfd, &st ) != 0 ) return false;
+    if( S_ISREG( st.st_mode ) )
+      { archive_dev = st.st_dev; archive_ino = st.st_ino; initialized = true; }
+    return true;
+    }
+  bool operator()( const struct stat & st ) const
+    {
+    return initialized && archive_dev == st.st_dev && archive_ino == st.st_ino;
+    }
+  } file_is_archive;
+
+
 bool archive_write( const uint8_t * const buf, const int size )
   {
   if( !encoder )					// uncompressed
@@ -121,9 +185,10 @@ bool archive_write( const uint8_t * const buf, const int size )
     const int wr = LZ_compress_write( encoder, buf + sz, size - sz );
     if( wr < 0 ) internal_error( "library error (LZ_compress_write)." );
     sz += wr;
+    if( sz >= size && size > 0 ) break;		// minimize dictionary size
     const int rd = LZ_compress_read( encoder, obuf, obuf_size );
     if( rd < 0 ) internal_error( "library error (LZ_compress_read)." );
-    if( rd == 0 && sz == size ) break;
+    if( rd == 0 && sz >= size ) break;
     if( writeblock( outfd, obuf, rd ) != rd ) return false;
     }
   if( LZ_compress_finished( encoder ) == 1 &&
@@ -133,11 +198,98 @@ bool archive_write( const uint8_t * const buf, const int size )
   }
 
 
+void init_tar_header( Tar_header header )	// set magic and version
+  {
+  std::memset( header, 0, header_size );
+  std::memcpy( header + magic_o, ustar_magic, magic_l - 1 );
+  header[version_o] = header[version_o+1] = '0';
+  }
+
+
+unsigned char xdigit( const unsigned value )
+  {
+  if( value <= 9 ) return '0' + value;
+  if( value <= 15 ) return 'A' + value - 10;
+  return 0;
+  }
+
+void print_hex( char * const buf, int size, unsigned long long num )
+  {
+  while( --size >= 0 ) { buf[size] = xdigit( num & 0x0F ); num >>= 4; }
+  }
+
 void print_octal( char * const buf, int size, unsigned long long num )
   {
   while( --size >= 0 ) { buf[size] = '0' + ( num % 8 ); num /= 8; }
   }
 
+unsigned decimal_digits( unsigned long long value )
+  {
+  unsigned digits = 1;
+  while( value >= 10 ) { value /= 10; ++digits; }
+  return digits;
+  }
+
+unsigned long long record_size( const unsigned keyword_size,
+                                const unsigned long long value_size )
+  {
+  // size = ' ' + keyword + '=' + value + '\n'
+  const unsigned long long size = 1 + keyword_size + 1 + value_size + 1;
+  const unsigned d1 = decimal_digits( size );
+  return decimal_digits( d1 + size ) + size;
+  }
+
+bool write_extended( const Extended & extended )
+  {
+  const int path_rec = extended.path.size() ?
+                       record_size( 4, extended.path.size() ) : 0;
+  const int lpath_rec = extended.linkpath.size() ?
+                        record_size( 8, extended.linkpath.size() ) : 0;
+  const int size_rec = ( extended.size > 0 ) ?
+                       record_size( 4, decimal_digits( extended.size ) ) : 0;
+  const unsigned long long edsize = path_rec + lpath_rec + size_rec + 22;
+  const unsigned long long bufsize = round_up( edsize );
+  if( edsize >= 1ULL << 33 ) return false;	// too much extended data
+  if( bufsize == 0 ) return edsize == 0;	// overflow or no extended data
+  char * const buf = new char[bufsize+1];	// extended records buffer
+  unsigned long long pos = path_rec;		// goto can't cross this
+  if( path_rec && snprintf( buf, path_rec + 1, "%d path=%s\n",
+                            path_rec, extended.path.c_str() ) != path_rec )
+    goto error;
+  if( lpath_rec && snprintf( buf + pos, lpath_rec + 1, "%d linkpath=%s\n",
+                     lpath_rec, extended.linkpath.c_str() ) != lpath_rec )
+    goto error;
+  pos += lpath_rec;
+  if( size_rec && snprintf( buf + pos, size_rec + 1, "%d size=%llu\n",
+                            size_rec, extended.size ) != size_rec )
+    goto error;
+  pos += size_rec;
+  if( snprintf( buf + pos, 23, "22 GNU.crc32=00000000\n" ) != 22 ) goto error;
+  pos += 22;
+  if( pos != edsize ) goto error;
+  print_hex( buf + edsize - 9, 8,
+             crc32c.windowed_crc( (const uint8_t *)buf, edsize - 9, edsize ) );
+  std::memset( buf + edsize, 0, bufsize - edsize );	// wipe padding
+  Tar_header header;				// extended header
+  init_tar_header( header );
+  header[typeflag_o] = tf_extended;		// fill only required fields
+  print_octal( header + size_o, size_l - 1, edsize );
+  print_octal( header + chksum_o, chksum_l - 1,
+               ustar_chksum( (const uint8_t *)header ) );
+  if( !archive_write( (const uint8_t *)header, header_size ) ) goto error;
+  for( pos = 0; pos < bufsize; )	// write extended records to archive
+    {
+    int size = std::min( bufsize - pos, 1ULL << 20 );
+    if( !archive_write( (const uint8_t *)buf + pos, size ) ) goto error;
+    pos += size;
+    }
+  delete[] buf;
+  return true;
+error:
+  delete[] buf;
+  return false;
+  }
+
 
 const char * remove_leading_dotdot( const char * const filename )
   {
@@ -164,24 +316,31 @@ const char * remove_leading_dotdot( const char * const filename )
   }
 
 
-bool split_name( const char * const filename, Tar_header header )
+// Return true if filename fits in the ustar header.
+bool store_name( const char * const filename, Extended & extended,
+                 Tar_header header )
   {
   const char * const stored_name = remove_leading_dotdot( filename );
   const int len = std::strlen( stored_name );
   enum { max_len = prefix_l + 1 + name_l };	// prefix + '/' + name
+
+  // first try storing filename in the ustar header
   if( len <= name_l )				// stored_name fits in name
     { std::memcpy( header + name_o, stored_name, len ); return true; }
   if( len <= max_len )				// find shortest prefix
     for( int i = len - name_l - 1; i < len && i <= prefix_l; ++i )
-      if( stored_name[i] == '/' )
+      if( stored_name[i] == '/' )		// stored_name can be split
         {
         std::memcpy( header + name_o, stored_name + i + 1, len - i - 1 );
         std::memcpy( header + prefix_o, stored_name, i );
         return true;
         }
+  // store filename in extended record, leave name zeroed in ustar header
+  extended.path = stored_name;
   return false;
   }
 
+
 int add_member( const char * const filename, const struct stat *,
                 const int flag, struct FTW * )
   {
@@ -189,11 +348,13 @@ int add_member( const char * const filename, const struct stat *,
   if( lstat( filename, &st ) != 0 )
     { show_file_error( filename, "Can't stat input file", errno );
       gretval = 1; return 0; }
+  if( file_is_archive( st ) )
+    { show_file_error( archive_namep, "File is the archive; not dumped." );
+      return 0; }
+  Extended extended;		// metadata for extended records
   Tar_header header;
-  std::memset( header, 0, header_size );
-  if( !split_name( filename, header ) )
-    { show_file_error( filename, "File name is too long." );
-      gretval = 2; return 0; }
+  init_tar_header( header );
+  store_name( filename, extended, header );
 
   const mode_t mode = st.st_mode;
   print_octal( header + mode_o, mode_l - 1,
@@ -201,10 +362,17 @@ int add_member( const char * const filename, const struct stat *,
                         S_IRWXU | S_IRWXG | S_IRWXO ) );
   const uid_t uid = ( cl_owner >= 0 ) ? (uid_t)cl_owner : st.st_uid;
   const gid_t gid = ( cl_group >= 0 ) ? (gid_t)cl_group : st.st_gid;
+  if( uid >= 2 << 20 || gid >= 2 << 20 )
+    { show_file_error( filename, "uid or gid is larger than 2_097_151." );
+      gretval = 1; return 0; }
   print_octal( header + uid_o, uid_l - 1, uid );
   print_octal( header + gid_o, gid_l - 1, gid );
+  const long long mtime = st.st_mtime;			// shut up gcc
+  if( mtime < 0 || mtime >= 1LL << 33 )
+    { show_file_error( filename, "mtime is out of ustar range [0, 8_589_934_591]." );
+      gretval = 1; return 0; }
+  print_octal( header + mtime_o, mtime_l - 1, mtime );
   unsigned long long file_size = 0;
-  print_octal( header + mtime_o, mtime_l - 1, st.st_mtime );
   Typeflag typeflag;
   if( S_ISREG( mode ) ) { typeflag = tf_regular; file_size = st.st_size; }
   else if( S_ISDIR( mode ) )
@@ -217,16 +385,26 @@ int add_member( const char * const filename, const struct stat *,
   else if( S_ISLNK( mode ) )
     {
     typeflag = tf_symlink;
-    if( st.st_size > linkname_l ||
-        readlink( filename, header + linkname_o, linkname_l ) != st.st_size )
+    long len;
+    if( st.st_size <= linkname_l )
+      len = readlink( filename, header + linkname_o, linkname_l );
+    else
       {
-      show_file_error( filename, "Link destination name is too long." );
-      gretval = 2; return 0;
+      char * const buf = new char[st.st_size+1];
+      len = readlink( filename, buf, st.st_size );
+      if( len == st.st_size ) { buf[len] = 0; extended.linkpath = buf; }
+      delete[] buf;
       }
+    if( len != st.st_size )
+      { show_file_error( filename, "Error reading link", (len < 0) ? errno : 0 );
+        gretval = 1; return 0; }
     }
   else if( S_ISCHR( mode ) || S_ISBLK( mode ) )
     {
     typeflag = S_ISCHR( mode ) ? tf_chardev : tf_blockdev;
+    if( major( st.st_dev ) >= 2 << 20 || minor( st.st_dev ) >= 2 << 20 )
+      { show_file_error( filename, "devmajor or devminor is larger than 2_097_151." );
+        gretval = 1; return 0; }
     print_octal( header + devmajor_o, devmajor_l - 1, major( st.st_dev ) );
     print_octal( header + devminor_o, devminor_l - 1, minor( st.st_dev ) );
     }
@@ -234,22 +412,23 @@ int add_member( const char * const filename, const struct stat *,
   else { show_file_error( filename, "Unknown file type." );
          gretval = 2; return 0; }
   header[typeflag_o] = typeflag;
-  std::memcpy( header + magic_o, ustar_magic, magic_l - 1 );
-  header[version_o] = header[version_o+1] = '0';
   const struct passwd * const pw = getpwuid( uid );
   if( pw && pw->pw_name )
     std::strncpy( header + uname_o, pw->pw_name, uname_l - 1 );
   const struct group * const gr = getgrgid( gid );
   if( gr && gr->gr_name )
     std::strncpy( header + gname_o, gr->gr_name, gname_l - 1 );
-  print_octal( header + size_o, size_l - 1, file_size );
+  if( file_size >= 1ULL << 33 ) extended.size = file_size;
+  else print_octal( header + size_o, size_l - 1, file_size );
   print_octal( header + chksum_o, chksum_l - 1,
                ustar_chksum( (const uint8_t *)header ) );
 
   const int infd = file_size ? open_instream( filename ) : -1;
   if( file_size && infd < 0 ) { gretval = 1; return 0; }
+  if( !extended.empty() && !write_extended( extended ) )
+    { show_error( "Error writing extended header", errno ); return 1; }
   if( !archive_write( (const uint8_t *)header, header_size ) )
-    { show_error( "Error writing archive header", errno ); return 1; }
+    { show_error( "Error writing ustar header", errno ); return 1; }
   if( file_size )
     {
     enum { bufsize = 32 * header_size };
@@ -304,6 +483,49 @@ bool verify_ustar_chksum( const uint8_t * const buf )
     ustar_chksum( buf ) == strtoul( (const char *)buf + chksum_o, 0, 8 ) ); }
 
 
+int concatenate( const std::string & archive_name, const Arg_parser & parser,
+                 const int filenames )
+  {
+  if( !filenames )
+    { if( verbosity >= 1 ) show_error( "Nothing to concatenate." ); return 0; }
+  if( archive_name.empty() )
+    { show_error( "'--concatenate' is incompatible with '-f -'.", 0, true );
+      return 1; }
+  if( ( outfd = open_outstream( archive_name, false ) ) < 0 ) return 1;
+  if( !file_is_archive.init() )
+    { show_file_error( archive_name.c_str(), "Can't stat", errno ); return 1; }
+
+  int retval = 0;
+  for( int i = 0; i < parser.arguments(); ++i )		// copy archives
+    {
+    if( parser.code( i ) ) continue;			// skip options
+    const char * const filename = parser.argument( i ).c_str();
+    const int infd = open_instream( filename );
+    if( infd < 0 )
+      { show_file_error( filename, "Can't open input file", errno );
+        retval = 1; break; }
+    if( !check_appendable( infd, false ) )
+      { show_file_error( filename, "Not an appendable tar.lz archive." );
+        close( infd ); retval = 2; break; }
+    struct stat st;
+    if( fstat( infd, &st ) == 0 && file_is_archive( st ) )
+      { show_file_error( filename, "File is the archive; not concatenated." );
+        close( infd ); continue; }
+    if( !check_appendable( outfd, true ) )
+      { show_error( "This does not look like an appendable tar.lz archive." );
+        close( infd ); retval = 2; break; }
+    if( !copy_file( infd, outfd ) || close( infd ) != 0 )
+      { show_file_error( filename, "Error copying archive", errno );
+        retval = 1; break; }
+    if( verbosity >= 1 ) std::fprintf( stderr, "%s\n", filename );
+    }
+
+  if( close( outfd ) != 0 && !retval )
+    { show_error( "Error closing archive", errno ); retval = 1; }
+  return retval;
+  }
+
+
 int encode( const std::string & archive_name, const Arg_parser & parser,
             const int filenames, const int level, const bool append )
   {
@@ -345,11 +567,15 @@ int encode( const std::string & archive_name, const Arg_parser & parser,
       { show_error( "'--append' is incompatible with '--uncompressed'.", 0, true );
         return 1; }
     if( ( outfd = open_outstream( archive_name, false ) ) < 0 ) return 1;
-    if( !check_appendable() )
+    if( !check_appendable( outfd, true ) )
       { show_error( "This does not look like an appendable tar.lz archive." );
         return 2; }
     }
 
+  archive_namep = archive_name.size() ? archive_name.c_str() : "(stdout)";
+  if( !file_is_archive.init() )
+    { show_file_error( archive_namep, "Can't stat", errno ); return 1; }
+
   if( compressed )
     {
     encoder = LZ_compress_open( option_mapping[level].dictionary_size,
@@ -365,7 +591,6 @@ int encode( const std::string & archive_name, const Arg_parser & parser,
     }
 
   int retval = 0;
-  std::string deslashed;		// arg without trailing slashes
   for( int i = 0; i < parser.arguments(); ++i )		// write members
     {
     const int code = parser.code( i );
@@ -375,6 +600,7 @@ int encode( const std::string & archive_name, const Arg_parser & parser,
       { show_file_error( filename, "Error changing working directory", errno );
         retval = 1; break; }
     if( code ) continue;				// skip options
+    std::string deslashed;		// arg without trailing slashes
     unsigned len = arg.size();
     while( len > 1 && arg[len-1] == '/' ) --len;
     if( len < arg.size() )
@@ -391,16 +617,18 @@ int encode( const std::string & archive_name, const Arg_parser & parser,
 
   if( !retval )			// write End-Of-Archive records
     {
-    uint8_t buf[header_size];
-    std::memset( buf, 0, header_size );
+    enum { bufsize = 2 * header_size };
+    uint8_t buf[bufsize];
+    std::memset( buf, 0, bufsize );
     if( encoder && cl_solid == 2 && !archive_write( 0, 0 ) )	// flush encoder
       { show_error( "Error flushing encoder", errno ); retval = 1; }
-    else if( !archive_write( buf, header_size ) ||
-             !archive_write( buf, header_size ) ||
+    else if( !archive_write( buf, bufsize ) ||
              ( encoder && !archive_write( 0, 0 ) ) )	// flush encoder
       { show_error( "Error writing end-of-archive blocks", errno );
         retval = 1; }
     }
+  if( encoder && LZ_compress_close( encoder ) < 0 )
+    { show_error( "LZ_compress_close failed." ); retval = 1; }
   if( close( outfd ) != 0 && !retval )
     { show_error( "Error closing archive", errno ); retval = 1; }
   if( retval && archive_name.size() && !append )