This is a live mirror of the Perl 5 development currently hosted at https://github.com/perl/perl5
Add Opcode extension
[perl5.git] / ext / Opcode / Opcode.pm
CommitLineData
6badd1a5
PP
1package Opcode;
2
3require 5.002;
4
5use vars qw($VERSION @ISA @EXPORT_OK);
6
7$VERSION = "1.01";
8
9use strict;
10use Carp;
11use Exporter ();
12use DynaLoader ();
13@ISA = qw(Exporter DynaLoader);
14
15BEGIN {
16 @EXPORT_OK = qw(
17 opset ops_to_opset
18 opset_to_ops opset_to_hex invert_opset
19 empty_opset full_opset
20 opdesc opcodes opmask define_optag
21 opmask_add verify_opset opdump
22 );
23}
24
25use subs @EXPORT_OK;
26
27bootstrap Opcode $VERSION;
28
29_init_optags();
30
31
32*ops_to_opset = \&opset; # alias for old name
33
34
35sub opset_to_hex ($) {
36 return "(invalid opset)" unless verify_opset($_[0]);
37 unpack("h*",$_[0]);
38}
39
40sub opdump (;$) {
41 my $pat = shift;
42 # handy utility: perl -MOpcode=opdump -e 'opdump File'
43 foreach(opset_to_ops(full_opset)) {
44 my $op = sprintf " %12s %s\n", $_, opdesc($_);
45 next if defined $pat and $op !~ m/$pat/i;
46 print $op;
47 }
48}
49
50
51
52sub _init_optags {
53 my(%all, %seen);
54 @all{opset_to_ops(full_opset)} = (); # keys only
55
56 local($/) = "\n=cut"; # skip to optags definition section
57 <DATA>;
58 $/ = "\n="; # now read in 'pod section' chunks
59 while(<DATA>) {
60 next unless m/^item\s+(:\w+)/;
61 my $tag = $1;
62
63 # Split into lines, keep only indented lines
64 my @lines = grep { m/^\s/ } split(/\n/);
65 foreach (@lines) { s/--.*// } # delete comments
66 my @ops = map { split ' ' } @lines; # get op words
67
68 foreach(@ops) {
69 warn "$tag - $_ already tagged in $seen{$_}\n" if $seen{$_};
70 $seen{$_} = $tag;
71 delete $all{$_};
72 }
73 # opset will croak on invalid names
74 define_optag($tag, opset(@ops));
75 }
76 close(DATA);
77 warn "Untagged opnames: ".join(' ',keys %all)."\n" if %all;
78}
79
80
811;
82
83__DATA__
84
85=head1 NAME
86
87Opcode - Disable named opcodes when compiling perl code
88
89=head1 SYNOPSIS
90
91 use Opcode;
92
93
94=head1 DESCRIPTION
95
96Perl code is always compiled into an internal format before execution.
97
98Evaluating perl code (e.g. via "eval" or "do 'file'") causes
99the code to be compiled into an internal format and then,
100provided there was no error in the compilation, executed.
101The internal format is based on many distinct I<opcodes>.
102
103By default no opmask is in effect and any code can be compiled.
104
105The Opcode module allow you to define an I<operator mask> to be in
106effect when perl I<next> compiles any code. Attempting to compile code
107which contains a masked opcode will cause the compilation to fail
108with an error. The code will not be executed.
109
110=head1 NOTE
111
112The Opcode module is not usually used directly. See the ops pragma and
113Safe modules for more typical uses.
114
115=head1 WARNING
116
117The authors make B<no warranty>, implied or otherwise, about the
118suitability of this software for safety or security purposes.
119
120The authors shall not in any case be liable for special, incidental,
121consequential, indirect or other similar damages arising from the use
122of this software.
123
124Your mileage will vary. If in any doubt B<do not use it>.
125
126
127=head1 Operator Names and Operator Lists
128
129The canonical list of operator names is the contents of the array
130op_name defined and initialised in file F<opcode.h> of the Perl
131source distribution (and installed into the perl library).
132
133Each operator has both a terse name (its opname) and a more verbose or
134recognisable descriptive name. The opdesc function can be used to
135return a list of descriptions for a list of operators.
136
137Many of the functions and methods listed below take a list of
138operators as parameters. Most operator lists can be made up of several
139types of element. Each element can be one of
140
141=over 8
142
143=item an operator name (opname)
144
145Operator names are typically small lowercase words like enterloop,
146leaveloop, last, next, redo etc. Sometimes they are rather cryptic
147like gv2cv, i_ncmp and ftsvtx.
148
149=item an operator tag name (optag)
150
151Operator tags can be used to refer to groups (or sets) of operators.
152Tag names always being with a colon. The Opcode module defines several
153optags and the user can define others using the define_optag function.
154
155=item a negated opname or optag
156
157An opname or optag can be prefixed with an exclamation mark, e.g., !mkdir.
158Negating an opname or optag means remove the corresponding ops from the
159accumulated set of ops at that point.
160
161=item an operator set (opset)
162
163An I<opset> as a binary string of approximately 43 bytes which holds a
164set or zero or more operators.
165
166The opset and opset_to_ops functions can be used to convert from
167a list of operators to an opset and I<vice versa>.
168
169Wherever a list of operators can be given you can use one or more opsets.
170See also Manipulating Opsets below.
171
172=back
173
174
175=head1 Opcode Functions
176
177The Opcode package contains functions for manipulating operator names
178tags and sets. All are available for export by the package.
179
180=over 8
181
182=item opcodes
183
184In a scalar context opcodes returns the number of opcodes in this
185version of perl (around 340 for perl5.002).
186
187In a list context it returns a list of all the operator names.
188(Not yet implemented, use @names = opset_to_ops(full_opset).)
189
190=item opset (OP, ...)
191
192Returns an opset containing the listed operators.
193
194=item opset_to_ops (OPSET)
195
196Returns a list of operator names corresponding to those operators in
197the set.
198
199=item opset_to_hex (OPSET)
200
201Returns a string representation of an opset. Can be handy for debugging.
202
203=item full_opset
204
205Returns an opset which includes all operators.
206
207=item empty_opset
208
209Returns an opset which contains no operators.
210
211=item invert_opset (OPSET)
212
213Returns an opset which is the inverse set of the one supplied.
214
215=item verify_opset (OPSET, ...)
216
217Returns true if the supplied opset looks like a valid opset (is the
218right length etc) otherwise it returns false. If an optional second
219parameter is true then verify_opset will croak on an invalid opset
220instead of returning false.
221
222Most of the other Opcode functions call verify_opset automatically
223and will croak if given an invalid opset.
224
225=item define_optag (OPTAG, OPSET)
226
227Define OPTAG as a symbolic name for OPSET. Optag names always start
228with a colon C<:>.
229
230The optag name used must not be defined already (define_optag will
231croak if it is already defined). Optag names are global to the perl
232process and optag definitions cannot be altered or deleted once
233defined.
234
235It is strongly recommended that applications using Opcode should use a
236leading capital letter on their tag names since lowercase names are
237reserved for use by the Opcode module. If using Opcode within a module
238you should prefix your tags names with the name of your module to
239ensure uniqueness and thus avoid clashes with other modules.
240
241=item opmask_add (OPSET)
242
243Adds the supplied opset to the current opmask. Note that there is
244currently I<no> mechanism for unmasking ops once they have been masked.
245This is intentional.
246
247=item opmask
248
249Returns an opset corresponding to the current opmask.
250
251=item opdesc (OP, ...)
252
253This takes a list of operator names and returns the corresponding list
254of operator descriptions.
255
256=item opdump (PAT)
257
258Dumps to STDOUT a two column list of op names and op descriptions.
259If an optional pattern is given then only lines which match the
260(case insensitive) pattern will be output.
261
262It's designed to be used as a handy command line utility:
263
264 perl -MOpcode=opdump -e opdump
265 perl -MOpcode=opdump -e 'opdump Eval'
266
267=back
268
269=head1 Manipulating Opsets
270
271Opsets may be manipulated using the perl bit vector operators & (and), | (or),
272^ (xor) and ~ (negate/invert).
273
274However you should never rely on the numerical position of any opcode
275within the opset. In other words both sides of a bit vector operator
276should be opsets returned from Opcode functions.
277
278Also, since the number of opcodes in your current version of perl might
279not be an exact multiple of eight, there may be unused bits in the last
280byte of an upset. This should not cause any problems (Opcode functions
281ignore those extra bits) but it does mean that using the ~ operator
282will typically not produce the same 'physical' opset 'string' as the
283invert_opset function.
284
285
286=head1 TO DO (maybe)
287
288 $bool = opset_eq($opset1, $opset2) true if opsets are logically eqiv
289
290 $yes = opset_can($opset, @ops) true if $opset has all @ops set
291
292 @diff = opset_diff($opset1, $opset2) => ('foo', '!bar', ...)
293
294=cut
295
296# the =cut above is used by _init_optags() to get here quickly
297
298=head1 Predefined Opcode Tags
299
300=over 5
301
302=item :base_core
303
304 null stub scalar pushmark wantarray const defined undef
305
306 rv2sv sassign
307
308 rv2av aassign aelem aelemfast aslice av2arylen
309
310 rv2hv helem hslice each values keys exists delete
311
312 preinc i_preinc predec i_predec postinc i_postinc postdec i_postdec
313 int hex oct abs pow multiply i_multiply divide i_divide
314 modulo i_modulo add i_add subtract i_subtract
315
316 left_shift right_shift bit_and bit_xor bit_or negate i_negate
317 not complement
318
319 lt i_lt gt i_gt le i_le ge i_ge eq i_eq ne i_ne ncmp i_ncmp
320 slt sgt sle sge seq sne scmp
321
322 substr vec stringify study pos length index rindex ord chr
323
324 ucfirst lcfirst uc lc quotemeta trans chop schop chomp schomp
325
326 match split
327
328 list lslice splice push pop shift unshift reverse
329
330 cond_expr flip flop andassign orassign and or xor
331
332 warn die lineseq nextstate unstack scope enter leave
333
334 rv2cv anoncode prototype
335
336 entersub leavesub return method -- XXX loops via recursion?
337
338 leaveeval -- needed for Safe to operate, is safe without entereval
339
340=item :base_mem
341
342These memory related ops are not included in :base_core because they
343can easily be used to implement a resource attack (e.g., consume all
344available memory).
345
346 concat repeat join range
347
348 anonlist anonhash
349
350Note that despite the existance of this optag a memory resource attack
351may still be possible using only :base_core ops.
352
353Disabling these ops is a I<very> heavy handed way to attempt to prevent
354a memory resource attack. It's probable that a specific memory limit
355mechanism will be added to perl in the near future.
356
357=item :base_loop
358
359These loop ops are not included in :base_core because they can easily be
360used to implement a resource attack (e.g., consume all available CPU time).
361
362 grepstart grepwhile
363 mapstart mapwhile
364 enteriter iter
365 enterloop leaveloop
366 last next redo
367 goto
368
369=item :base_io
370
371These ops enable I<filehandle> (rather than filename) based input and
372output. These are safe on the assumption that only pre-existing
373filehandles are available for use. To create new filehandles other ops
374such as open would need to be enabled.
375
376 readline rcatline getc read
377
378 formline enterwrite leavewrite
379
380 print sysread syswrite send recv eof tell seek
381
382 readdir telldir seekdir rewinddir
383
384=item :base_orig
385
386These are a hotchpotch of opcodes still waiting to be considered
387
388 gvsv gv gelem
389
390 padsv padav padhv padany
391
392 rv2gv refgen srefgen ref
393
394 bless -- could be used to change ownership of objects (reblessing)
395
396 glob
397
398 pushre regcmaybe regcomp subst substcont
399
400 sprintf prtf -- can core dump
401
402 crypt
403
404 tie untie
405
406 dbmopen dbmclose
407 sselect select
408 pipe_op sockpair
409
410 getppid getpgrp setpgrp getpriority setpriority localtime gmtime
411
412 entertry leavetry -- can be used to 'hide' fatal errors
413
414=item :base_math
415
416These ops are not included in :base_core because of the risk of them being
417used to generate floating point exceptions (which would have to be caught
418using a $SIG{FPE} handler).
419
420 atan2 sin cos exp log sqrt
421
422These ops are not included in :base_core because they have an effect
423beyond the scope of the compartment.
424
425 rand srand
426
427=item :default
428
429A handy tag name for a I<reasonable> default set of ops. (The current ops
430allowed are unstable while development continues. It will change.)
431
432 :base_core :base_mem :base_loop :base_io :base_orig
433
434If safety matters to you (and why else would you be using the Opcode module?)
435then you should not rely on the definition of this, or indeed any other, optag!
436
437
438=item :filesys_read
439
440 stat lstat readlink
441
442 ftatime ftblk ftchr ftctime ftdir fteexec fteowned fteread
443 ftewrite ftfile ftis ftlink ftmtime ftpipe ftrexec ftrowned
444 ftrread ftsgid ftsize ftsock ftsuid fttty ftzero ftrwrite ftsvtx
445
446 fttext ftbinary
447
448 fileno
449
450=item :sys_db
451
452 ghbyname ghbyaddr ghostent shostent ehostent -- hosts
453 gnbyname gnbyaddr gnetent snetent enetent -- networks
454 gpbyname gpbynumber gprotoent sprotoent eprotoent -- protocols
455 gsbyname gsbyport gservent sservent eservent -- services
456
457 gpwnam gpwuid gpwent spwent epwent getlogin -- users
458 ggrnam ggrgid ggrent sgrent egrent -- groups
459
460=item :browse
461
462A handy tag name for a I<reasonable> default set of ops beyond the
463:default optag. Like :default (and indeed all the other optags) its
464current definition is unstable while development continues. It will change.
465
466The :browse tag represents the next step beyond :default. It it a
467superset of the :default ops and adds :filesys_read the :sys_db.
468The intent being that scripts can access more (possibly sensitive)
469information about your system but not be able to change it.
470
471 :default :filesys_read :sys_db
472
473=item :filesys_open
474
475 sysopen open close
476 umask binmode
477
478 open_dir closedir -- other dir ops are in :base_io
479
480=item :filesys_write
481
482 link unlink rename symlink truncate
483
484 mkdir rmdir
485
486 utime chmod chown
487
488 fcntl -- not strictly filesys related, but possibly as dangerous?
489
490=item :subprocess
491
492 backtick system
493
494 fork
495
496 wait waitpid
497
498=item :ownprocess
499
500 exec exit kill
501
502 time tms -- could be used for timing attacks (paranoid?)
503
504=item :others
505
506This tag holds groups of assorted specialist opcodes that don't warrant
507having optags defined for them.
508
509SystemV Interprocess Communications:
510
511 msgctl msgget msgrcv msgsnd
512
513 semctl semget semop
514
515 shmctl shmget shmread shmwrite
516
517=item :still_to_be_decided
518
519 chdir
520 flock ioctl
521
522 socket getpeername ssockopt
523 bind connect listen accept shutdown gsockopt getsockname
524
525 sleep alarm -- changes global timer state and signal handling
526 sort -- assorted problems including core dumps
527 tied -- can be used to access object implementing a tie
528 pack unpack -- can be used to create/use memory pointers
529
530 entereval -- can be used to hide code from initial compile
531 require dofile
532
533 caller -- get info about calling environment and args
534
535 reset
536
537 dbstate -- perl -d version of nextstate(ment) opcode
538
539=item :dangerous
540
541This tag is simply a bucket for opcodes that are unlikely to be used via
542a tag name but need to be tagged for completness and documentation.
543
544 syscall dump chroot
545
546
547=back
548
549=head1 SEE ALSO
550
551ops(3) -- perl pragma interface to Opcode module.
552
553Safe(3) -- Opcode and namespace limited execution compartments
554
555=head1 AUTHORS
556
557Originally designed and implemented by Malcolm Beattie,
558mbeattie@sable.ox.ac.uk as part of Safe version 1.
559
560Split out from Safe module version 1, named opcode tags and other
561changes added by Tim Bunce <Tim.Bunce@ig.co.uk>.
562
563=cut
564