Re: glob in Safe compartment allows shell access

[p5sagit/p5-mst-13.2.git] / ext / Opcode / Opcode.pm

package Opcode;

require 5.002;

use vars qw($VERSION @ISA @EXPORT_OK);

$VERSION = "1.01";

use strict;
use Carp;
use Exporter ();
use DynaLoader ();
@ISA = qw(Exporter DynaLoader);

BEGIN {
    @EXPORT_OK = qw(
        opset ops_to_opset
        opset_to_ops opset_to_hex invert_opset
        empty_opset full_opset
        opdesc opcodes opmask define_optag
        opmask_add verify_opset opdump
    );
}

use subs @EXPORT_OK;

bootstrap Opcode $VERSION;

_init_optags();


*ops_to_opset = \&opset;        # alias for old name


sub opset_to_hex ($) {
    return "(invalid opset)" unless verify_opset($_[0]);
    unpack("h*",$_[0]);
}

sub opdump (;$) {
        my $pat = shift;
    # handy utility: perl -MOpcode=opdump -e 'opdump File'
    foreach(opset_to_ops(full_opset)) {
        my $op = sprintf "  %12s  %s\n", $_, opdesc($_);
                next if defined $pat and $op !~ m/$pat/i;
                print $op;
    }
}



sub _init_optags {
    my(%all, %seen);
    @all{opset_to_ops(full_opset)} = (); # keys only

    local($/) = "\n=cut"; # skip to optags definition section
    <DATA>;
    $/ = "\n=";         # now read in 'pod section' chunks
    while(<DATA>) {
        next unless m/^item\s+(:\w+)/;
        my $tag = $1;

        # Split into lines, keep only indented lines
        my @lines = grep { m/^\s/    } split(/\n/);
        foreach (@lines) { s/--.*//  } # delete comments
        my @ops   = map  { split ' ' } @lines; # get op words

        foreach(@ops) {
            warn "$tag - $_ already tagged in $seen{$_}\n" if $seen{$_};
            $seen{$_} = $tag;
            delete $all{$_};
        }
        # opset will croak on invalid names
        define_optag($tag, opset(@ops));
    }
    close(DATA);
    warn "Untagged opnames: ".join(' ',keys %all)."\n" if %all;
}


1;

__DATA__

=head1 NAME

Opcode - Disable named opcodes when compiling perl code

=head1 SYNOPSIS

  use Opcode;


=head1 DESCRIPTION

Perl code is always compiled into an internal format before execution.

Evaluating perl code (e.g. via "eval" or "do 'file'") causes
the code to be compiled into an internal format and then,
provided there was no error in the compilation, executed.
The internal format is based on many distinct I<opcodes>.

By default no opmask is in effect and any code can be compiled.

The Opcode module allow you to define an I<operator mask> to be in
effect when perl I<next> compiles any code.  Attempting to compile code
which contains a masked opcode will cause the compilation to fail
with an error. The code will not be executed.

=head1 NOTE

The Opcode module is not usually used directly. See the ops pragma and
Safe modules for more typical uses.

=head1 WARNING

The authors make B<no warranty>, implied or otherwise, about the
suitability of this software for safety or security purposes.

The authors shall not in any case be liable for special, incidental,
consequential, indirect or other similar damages arising from the use
of this software.

Your mileage will vary. If in any doubt B<do not use it>.


=head1 Operator Names and Operator Lists

The canonical list of operator names is the contents of the array
op_name defined and initialised in file F<opcode.h> of the Perl
source distribution (and installed into the perl library).

Each operator has both a terse name (its opname) and a more verbose or
recognisable descriptive name. The opdesc function can be used to
return a list of descriptions for a list of operators.

Many of the functions and methods listed below take a list of
operators as parameters. Most operator lists can be made up of several
types of element. Each element can be one of

=over 8

=item an operator name (opname)

Operator names are typically small lowercase words like enterloop,
leaveloop, last, next, redo etc. Sometimes they are rather cryptic
like gv2cv, i_ncmp and ftsvtx.

=item an operator tag name (optag)

Operator tags can be used to refer to groups (or sets) of operators.
Tag names always being with a colon. The Opcode module defines several
optags and the user can define others using the define_optag function.

=item a negated opname or optag

An opname or optag can be prefixed with an exclamation mark, e.g., !mkdir.
Negating an opname or optag means remove the corresponding ops from the
accumulated set of ops at that point.

=item an operator set (opset)

An I<opset> as a binary string of approximately 43 bytes which holds a
set or zero or more operators.

The opset and opset_to_ops functions can be used to convert from
a list of operators to an opset and I<vice versa>.

Wherever a list of operators can be given you can use one or more opsets.
See also Manipulating Opsets below.

=back


=head1 Opcode Functions

The Opcode package contains functions for manipulating operator names
tags and sets. All are available for export by the package.

=over 8

=item opcodes

In a scalar context opcodes returns the number of opcodes in this
version of perl (around 340 for perl5.002).

In a list context it returns a list of all the operator names.
(Not yet implemented, use @names = opset_to_ops(full_opset).)

=item opset (OP, ...)

Returns an opset containing the listed operators.

=item opset_to_ops (OPSET)

Returns a list of operator names corresponding to those operators in
the set.

=item opset_to_hex (OPSET)

Returns a string representation of an opset. Can be handy for debugging.

=item full_opset

Returns an opset which includes all operators.

=item empty_opset

Returns an opset which contains no operators.

=item invert_opset (OPSET)

Returns an opset which is the inverse set of the one supplied.

=item verify_opset (OPSET, ...)

Returns true if the supplied opset looks like a valid opset (is the
right length etc) otherwise it returns false. If an optional second
parameter is true then verify_opset will croak on an invalid opset
instead of returning false.

Most of the other Opcode functions call verify_opset automatically
and will croak if given an invalid opset.

=item define_optag (OPTAG, OPSET)

Define OPTAG as a symbolic name for OPSET. Optag names always start
with a colon C<:>.

The optag name used must not be defined already (define_optag will
croak if it is already defined). Optag names are global to the perl
process and optag definitions cannot be altered or deleted once
defined.

It is strongly recommended that applications using Opcode should use a
leading capital letter on their tag names since lowercase names are
reserved for use by the Opcode module. If using Opcode within a module
you should prefix your tags names with the name of your module to
ensure uniqueness and thus avoid clashes with other modules.

=item opmask_add (OPSET)

Adds the supplied opset to the current opmask. Note that there is
currently I<no> mechanism for unmasking ops once they have been masked.
This is intentional.

=item opmask

Returns an opset corresponding to the current opmask.

=item opdesc (OP, ...)

This takes a list of operator names and returns the corresponding list
of operator descriptions.

=item opdump (PAT)

Dumps to STDOUT a two column list of op names and op descriptions.
If an optional pattern is given then only lines which match the
(case insensitive) pattern will be output.

It's designed to be used as a handy command line utility:

        perl -MOpcode=opdump -e opdump
        perl -MOpcode=opdump -e 'opdump Eval'

=back

=head1 Manipulating Opsets

Opsets may be manipulated using the perl bit vector operators & (and), | (or),
^ (xor) and ~ (negate/invert).

However you should never rely on the numerical position of any opcode
within the opset. In other words both sides of a bit vector operator
should be opsets returned from Opcode functions.

Also, since the number of opcodes in your current version of perl might
not be an exact multiple of eight, there may be unused bits in the last
byte of an upset. This should not cause any problems (Opcode functions
ignore those extra bits) but it does mean that using the ~ operator
will typically not produce the same 'physical' opset 'string' as the
invert_opset function.


=head1 TO DO (maybe)

    $bool = opset_eq($opset1, $opset2)  true if opsets are logically eqiv

    $yes = opset_can($opset, @ops)      true if $opset has all @ops set

    @diff = opset_diff($opset1, $opset2) => ('foo', '!bar', ...)

=cut

# the =cut above is used by _init_optags() to get here quickly

=head1 Predefined Opcode Tags

=over 5

=item :base_core

    null stub scalar pushmark wantarray const defined undef

    rv2sv sassign

    rv2av aassign aelem aelemfast aslice av2arylen

    rv2hv helem hslice each values keys exists delete

    preinc i_preinc predec i_predec postinc i_postinc postdec i_postdec
    int hex oct abs pow multiply i_multiply divide i_divide
    modulo i_modulo add i_add subtract i_subtract

    left_shift right_shift bit_and bit_xor bit_or negate i_negate
    not complement

    lt i_lt gt i_gt le i_le ge i_ge eq i_eq ne i_ne ncmp i_ncmp
    slt sgt sle sge seq sne scmp

    substr vec stringify study pos length index rindex ord chr

    ucfirst lcfirst uc lc quotemeta trans chop schop chomp schomp

    match split

    list lslice splice push pop shift unshift reverse

    cond_expr flip flop andassign orassign and or xor

    warn die lineseq nextstate unstack scope enter leave

    rv2cv anoncode prototype

    entersub leavesub return method -- XXX loops via recursion?

    leaveeval -- needed for Safe to operate, is safe without entereval

=item :base_mem

These memory related ops are not included in :base_core because they
can easily be used to implement a resource attack (e.g., consume all
available memory).

    concat repeat join range

    anonlist anonhash

Note that despite the existance of this optag a memory resource attack
may still be possible using only :base_core ops.

Disabling these ops is a I<very> heavy handed way to attempt to prevent
a memory resource attack. It's probable that a specific memory limit
mechanism will be added to perl in the near future.

=item :base_loop

These loop ops are not included in :base_core because they can easily be
used to implement a resource attack (e.g., consume all available CPU time).

    grepstart grepwhile
    mapstart mapwhile
    enteriter iter
    enterloop leaveloop
    last next redo
    goto

=item :base_io

These ops enable I<filehandle> (rather than filename) based input and
output. These are safe on the assumption that only pre-existing
filehandles are available for use.  To create new filehandles other ops
such as open would need to be enabled.

    readline rcatline getc read

    formline enterwrite leavewrite

    print sysread syswrite send recv eof tell seek

    readdir telldir seekdir rewinddir

=item :base_orig

These are a hotchpotch of opcodes still waiting to be considered

    gvsv gv gelem

    padsv padav padhv padany

    rv2gv refgen srefgen ref

    bless -- could be used to change ownership of objects (reblessing)

    pushre regcmaybe regcomp subst substcont

    sprintf prtf -- can core dump

    crypt

    tie untie

    dbmopen dbmclose
    sselect select
    pipe_op sockpair

    getppid getpgrp setpgrp getpriority setpriority localtime gmtime

    entertry leavetry -- can be used to 'hide' fatal errors

=item :base_math

These ops are not included in :base_core because of the risk of them being
used to generate floating point exceptions (which would have to be caught
using a $SIG{FPE} handler).

    atan2 sin cos exp log sqrt

These ops are not included in :base_core because they have an effect
beyond the scope of the compartment.

    rand srand

=item :default

A handy tag name for a I<reasonable> default set of ops.  (The current ops
allowed are unstable while development continues. It will change.)

    :base_core :base_mem :base_loop :base_io :base_orig

If safety matters to you (and why else would you be using the Opcode module?)
then you should not rely on the definition of this, or indeed any other, optag!


=item :filesys_read

    stat lstat readlink

    ftatime ftblk ftchr ftctime ftdir fteexec fteowned fteread
    ftewrite ftfile ftis ftlink ftmtime ftpipe ftrexec ftrowned
    ftrread ftsgid ftsize ftsock ftsuid fttty ftzero ftrwrite ftsvtx

    fttext ftbinary

    fileno

=item :sys_db

    ghbyname ghbyaddr ghostent shostent ehostent      -- hosts
    gnbyname gnbyaddr gnetent snetent enetent         -- networks
    gpbyname gpbynumber gprotoent sprotoent eprotoent -- protocols
    gsbyname gsbyport gservent sservent eservent      -- services

    gpwnam gpwuid gpwent spwent epwent getlogin       -- users
    ggrnam ggrgid ggrent sgrent egrent                -- groups

=item :browse

A handy tag name for a I<reasonable> default set of ops beyond the
:default optag.  Like :default (and indeed all the other optags) its
current definition is unstable while development continues. It will change.

The :browse tag represents the next step beyond :default. It it a
superset of the :default ops and adds :filesys_read the :sys_db.
The intent being that scripts can access more (possibly sensitive)
information about your system but not be able to change it.

    :default :filesys_read :sys_db

=item :filesys_open

    sysopen open close
    umask binmode

    open_dir closedir -- other dir ops are in :base_io

=item :filesys_write

    link unlink rename symlink truncate

    mkdir rmdir

    utime chmod chown

    fcntl -- not strictly filesys related, but possibly as dangerous?

=item :subprocess

    backtick system

    fork

    wait waitpid

    glob -- access to Cshell via <`rm *`>

=item :ownprocess

    exec exit kill

    time tms -- could be used for timing attacks (paranoid?)

=item :others

This tag holds groups of assorted specialist opcodes that don't warrant
having optags defined for them.

SystemV Interprocess Communications:

    msgctl msgget msgrcv msgsnd

    semctl semget semop

    shmctl shmget shmread shmwrite

=item :still_to_be_decided

    chdir
    flock ioctl

    socket getpeername ssockopt
    bind connect listen accept shutdown gsockopt getsockname

    sleep alarm -- changes global timer state and signal handling
    sort -- assorted problems including core dumps
    tied -- can be used to access object implementing a tie
    pack unpack -- can be used to create/use memory pointers

    entereval -- can be used to hide code from initial compile
    require dofile 

    caller -- get info about calling environment and args

    reset

    dbstate -- perl -d version of nextstate(ment) opcode

=item :dangerous

This tag is simply a bucket for opcodes that are unlikely to be used via
a tag name but need to be tagged for completness and documentation.

    syscall dump chroot


=back

=head1 SEE ALSO

ops(3) -- perl pragma interface to Opcode module.

Safe(3) -- Opcode and namespace limited execution compartments

=head1 AUTHORS

Originally designed and implemented by Malcolm Beattie,
mbeattie@sable.ox.ac.uk as part of Safe version 1.

Split out from Safe module version 1, named opcode tags and other
changes added by Tim Bunce <Tim.Bunce@ig.co.uk>.

=cut