SEC-SMACK-CP

This project provides 2 programs:

• an extractor, smack-extract, that extract the smack labels and produce a file recording them

• a restorer, smack-restore, that, using the file recording smack labels, restores it.

Extracting extended attributes

The program sec-xattr-extract:

sec-xattr-extract [-d] [-m pattern] OUT-FILE ROOT-DIR

Extract in OUT-FILE the extended attributes of files at ROOT-DIR directory.

By default extract any extended attribute. But if an extended regular expression is given in pattern using option -m, only these patterns are extracted.

The option -ddumps out the extracted attributes.

Restoring extended attributes

The program sec-xattr-restore:

sec-xattr-rectore [-d] IN-FILE ROOT-DIR [program [arg ...]]

Set the extended attributes extracted in IN-FILE to files at ROOT-DIR.

The option '-d' is a dump out dry run of the process.

When program is given, on success, the restorer executes it, calling it with its optional arguments.

Building extended attributes

The program sec-xattr-build:

sec-xattr-build OUT-FILE

It reads from its standard input lines having 3 fields separated by spaces or tabs. Each line has 3 items separated by spaces or tabs:

• filename,
• attribute-name
• attribute-value

The lines are read as bytes and bytes out of the range 33 .. 126, except for for item separators (32 or 9) and line separators (10), must be escaped using octal '\xxx' (ex: space -> \040).

Format of the file recording the labels

The file contains 3 sections: ID CODE STRINGS

The program sec-xattr-debug can dump it.

Section ID

The ID section is 16 bytes long (alignment on 4 bytes boundary must be preserved)

For the version 1 the ID is the string "sec-xattr-cp 1\n\n"

0000000   s   e   c   -   x   a   t   t   r   -   c   p       1  \n  \n
         73  65  63  2d  78  61  74  74  72  2d  63  70  20  31  0a  0a

Section CODE

The section code is made of aligned 32 bits unsigned integers, named code, recorded in little endian order.

Each of these code is made of 2 lower bits recording the operation and 30 upper bits being an offset to a data argument.

  3 3 2
  1 0 9  ..  ...      2 1 0
 +-+-+-+ . . . . . . +-+-+-+
 | o f f s e t         | op|
 +-+-+-+ . . . . . . +-+-+-+

Operations are SUB (0), FILE (1), ATTR (2), SET(3).

The data argument of the operation is the data located at the given offset, computed by adding 4 + OFFSET to the adress of the code of the operation.

The below C fragment shows the process of retrieving the operation OP and the pointer to the DATA for a code pointed by PCODE

		CODE = *PCODE++;
		CODE = le32toh(CODE);
		DATA = &((char*)PCODE)[CODE >> 2];
        OP = CODE & 3;

The data is of to types: STRINGZ or BUFFER.

STRINGZ are zero terminated UTF8 encoded strings. It is in use for operation SUB, FILE and ATTR. That string can be retrieved using the below C code:

        STR = DATA;

BUFFER is a byte array whose length is encoded in 16 bits little endian at the start of the data. It is in use for SET operation. Be aware that the 16-bits integer may be not aligned on the 2 bytes boundary. The LENGTH and the CONTENT can be retrieved using the below C code:

        LENGTH = DATA[0] + 256 * DATA[1];
        CONTENT = &DATA[2];

Operation SUB

This is for entering or leaving a directory.

This operation has 2 cases. When the offset is 0 it means to pop the last entered directory or to terminate the process.

Otherwise, the string records the directory to enter.

Operation FILE

Set the file to be changed.

Operation ATTR

Set the attribute to be changed.

Operation SET

Set the current file with the current attribute with the value described at the offset. The value is recorded with its length given in 2 bytes little endian followed by the value without trailing zero.

Section STRING

Strings are zero terminated.