ViewVC Help
View File | Revision Log | Show Annotations | Download File
/cvs/CBOR-XS/XS.xs
(Generate patch)

Comparing CBOR-XS/XS.xs (file contents):
Revision 1.45 by root, Sun Dec 14 05:48:39 2014 UTC vs.
Revision 1.68 by root, Mon Nov 30 18:30:29 2020 UTC

6#include <string.h> 6#include <string.h>
7#include <stdlib.h> 7#include <stdlib.h>
8#include <stdio.h> 8#include <stdio.h>
9#include <limits.h> 9#include <limits.h>
10#include <float.h> 10#include <float.h>
11#include <inttypes.h>
11 12
12#define ECB_NO_THREADS 1 13#define ECB_NO_THREADS 1
13#include "ecb.h" 14#include "ecb.h"
14 15
15// compatibility with perl <5.18 16// compatibility with perl <5.18
20# define HvNAMELEN(hv) HvNAMELEN_get (hv) 21# define HvNAMELEN(hv) HvNAMELEN_get (hv)
21#endif 22#endif
22#ifndef HvNAMEUTF8 23#ifndef HvNAMEUTF8
23# define HvNAMEUTF8(hv) 0 24# define HvNAMEUTF8(hv) 0
24#endif 25#endif
26#ifndef SvREFCNT_inc_NN
27# define SvREFCNT_inc_NN(sv) SvREFCNT_inc (sv)
28#endif
25#ifndef SvREFCNT_dec_NN 29#ifndef SvREFCNT_dec_NN
26# define SvREFCNT_dec_NN(sv) SvREFCNT_dec (sv) 30# define SvREFCNT_dec_NN(sv) SvREFCNT_dec (sv)
27#endif 31#endif
28 32
29// known major and minor types 33// known major and minor types
95 CBOR_TAG_MIME = 36, // mime message rfc2045, utf-8 99 CBOR_TAG_MIME = 36, // mime message rfc2045, utf-8
96 100
97 CBOR_TAG_MAGIC = 55799, // self-describe cbor 101 CBOR_TAG_MAGIC = 55799, // self-describe cbor
98}; 102};
99 103
104// known forced types, also hardcoded in CBOR.pm
105enum
106{
107 AS_CBOR = 0,
108 AS_INT = 1,
109 AS_BYTES = 2,
110 AS_TEXT = 3,
111 AS_FLOAT16 = 4,
112 AS_FLOAT32 = 5,
113 AS_FLOAT64 = 6,
114 // possibly future enhancements: (generic) float, (generic) string
115};
116
100#define F_SHRINK 0x00000001UL 117#define F_SHRINK 0x00000001UL
101#define F_ALLOW_UNKNOWN 0x00000002UL 118#define F_ALLOW_UNKNOWN 0x00000002UL
102#define F_ALLOW_SHARING 0x00000004UL 119#define F_ALLOW_SHARING 0x00000004UL
103#define F_ALLOW_CYCLES 0x00000008UL 120#define F_ALLOW_CYCLES 0x00000008UL
121#define F_FORBID_OBJECTS 0x00000010UL
104#define F_PACK_STRINGS 0x00000010UL 122#define F_PACK_STRINGS 0x00000020UL
123#define F_TEXT_KEYS 0x00000040UL
124#define F_TEXT_STRINGS 0x00000080UL
105#define F_VALIDATE_UTF8 0x00000020UL 125#define F_VALIDATE_UTF8 0x00000100UL
106 126
107#define INIT_SIZE 32 // initial scalar size to be allocated 127#define INIT_SIZE 32 // initial scalar size to be allocated
108 128
109#define SB do { 129#define SB do {
110#define SE } while (0) 130#define SE } while (0)
183 203
184// minimum length of a string to be registered for stringref 204// minimum length of a string to be registered for stringref
185ecb_inline int 205ecb_inline int
186minimum_string_length (UV idx) 206minimum_string_length (UV idx)
187{ 207{
188 return idx > 23 208 return idx <= 23 ? 3
189 ? idx > 0xffU 209 : idx <= 0xffU ? 4
190 ? idx > 0xffffU 210 : idx <= 0xffffU ? 5
191 ? idx > 0xffffffffU 211 : idx <= 0xffffffffU ? 7
192 ? 11 212 : 11;
193 : 7
194 : 5
195 : 4
196 : 3;
197} 213}
198 214
199///////////////////////////////////////////////////////////////////////////// 215/////////////////////////////////////////////////////////////////////////////
200// encoder 216// encoder
201 217
214} enc_t; 230} enc_t;
215 231
216ecb_inline void 232ecb_inline void
217need (enc_t *enc, STRLEN len) 233need (enc_t *enc, STRLEN len)
218{ 234{
219 if (ecb_expect_false (enc->cur + len >= enc->end)) 235 if (ecb_expect_false ((uintptr_t)(enc->end - enc->cur) < len))
220 { 236 {
221 STRLEN cur = enc->cur - (char *)SvPVX (enc->sv); 237 STRLEN cur = enc->cur - (char *)SvPVX (enc->sv);
222 SvGROW (enc->sv, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1); 238 SvGROW (enc->sv, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1);
223 enc->cur = SvPVX (enc->sv) + cur; 239 enc->cur = SvPVX (enc->sv) + cur;
224 enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1; 240 enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1;
230{ 246{
231 need (enc, 1); 247 need (enc, 1);
232 *enc->cur++ = ch; 248 *enc->cur++ = ch;
233} 249}
234 250
251// used for tags, intregers, element counts and so on
235static void 252static void
236encode_uint (enc_t *enc, int major, UV len) 253encode_uint (enc_t *enc, int major, UV len)
237{ 254{
238 need (enc, 9); 255 need (enc, 9);
239 256
270 *enc->cur++ = len >> 8; 287 *enc->cur++ = len >> 8;
271 *enc->cur++ = len; 288 *enc->cur++ = len;
272 } 289 }
273} 290}
274 291
292// encodes a perl value into a CBOR integer
293ecb_inline void
294encode_int (enc_t *enc, SV *sv)
295{
296 if (SvIsUV (sv))
297 encode_uint (enc, MAJOR_POS_INT, SvUVX (sv));
298 else if (SvIVX (sv) >= 0)
299 encode_uint (enc, MAJOR_POS_INT, SvIVX (sv));
300 else
301 encode_uint (enc, MAJOR_NEG_INT, -(SvIVX (sv) + 1));
302}
303
275ecb_inline void 304ecb_inline void
276encode_tag (enc_t *enc, UV tag) 305encode_tag (enc_t *enc, UV tag)
277{ 306{
278 encode_uint (enc, MAJOR_TAG, tag); 307 encode_uint (enc, MAJOR_TAG, tag);
279} 308}
280 309
310// exceptional (hopefully) slow path for byte strings that need to be utf8-encoded
311ecb_noinline static void
312encode_str_utf8 (enc_t *enc, int utf8, char *str, STRLEN len)
313{
314 STRLEN ulen = len;
315 U8 *p, *pend = (U8 *)str + len;
316
317 for (p = (U8 *)str; p < pend; ++p)
318 ulen += *p >> 7; // count set high bits
319
320 encode_uint (enc, MAJOR_TEXT, ulen);
321
322 need (enc, ulen);
323 for (p = (U8 *)str; p < pend; ++p)
324 if (*p < 0x80)
325 *enc->cur++ = *p;
326 else
327 {
328 *enc->cur++ = 0xc0 + (*p >> 6);
329 *enc->cur++ = 0x80 + (*p & 63);
330 }
331}
332
281ecb_inline void 333ecb_inline void
282encode_str (enc_t *enc, int utf8, char *str, STRLEN len) 334encode_str (enc_t *enc, int upgrade_utf8, int utf8, char *str, STRLEN len)
283{ 335{
336 if (ecb_expect_false (upgrade_utf8))
337 if (!utf8)
338 {
339 encode_str_utf8 (enc, utf8, str, len);
340 return;
341 }
342
284 encode_uint (enc, utf8 ? MAJOR_TEXT : MAJOR_BYTES, len); 343 encode_uint (enc, utf8 ? MAJOR_TEXT : MAJOR_BYTES, len);
285 need (enc, len); 344 need (enc, len);
286 memcpy (enc->cur, str, len); 345 memcpy (enc->cur, str, len);
287 enc->cur += len; 346 enc->cur += len;
288} 347}
289 348
290static void 349ecb_inline void
291encode_strref (enc_t *enc, int utf8, char *str, STRLEN len) 350encode_strref (enc_t *enc, int upgrade_utf8, int utf8, char *str, STRLEN len)
292{ 351{
293 if (ecb_expect_false (enc->cbor.flags & F_PACK_STRINGS)) 352 if (ecb_expect_false (enc->cbor.flags & F_PACK_STRINGS))
294 { 353 {
295 SV **svp = hv_fetch (enc->stringref[!!utf8], str, len, 1); 354 SV **svp = hv_fetch (enc->stringref[!!utf8], str, len, 1);
296 355
307 sv_setuv (*svp, enc->stringref_idx); 366 sv_setuv (*svp, enc->stringref_idx);
308 ++enc->stringref_idx; 367 ++enc->stringref_idx;
309 } 368 }
310 } 369 }
311 370
312 encode_str (enc, utf8, str, len); 371 encode_str (enc, upgrade_utf8, utf8, str, len);
372}
373
374ecb_inline void
375encode_float16 (enc_t *enc, NV nv)
376{
377 need (enc, 1+2);
378
379 *enc->cur++ = MAJOR_MISC | MISC_FLOAT16;
380
381 uint16_t fp = ecb_float_to_binary16 (nv);
382
383 if (!ecb_big_endian ())
384 fp = ecb_bswap16 (fp);
385
386 memcpy (enc->cur, &fp, 2);
387 enc->cur += 2;
388}
389
390ecb_inline void
391encode_float32 (enc_t *enc, NV nv)
392{
393 need (enc, 1+4);
394
395 *enc->cur++ = MAJOR_MISC | MISC_FLOAT32;
396
397 uint32_t fp = ecb_float_to_binary32 (nv);
398
399 if (!ecb_big_endian ())
400 fp = ecb_bswap32 (fp);
401
402 memcpy (enc->cur, &fp, 4);
403 enc->cur += 4;
404}
405
406ecb_inline void
407encode_float64 (enc_t *enc, NV nv)
408{
409 need (enc, 1+8);
410
411 *enc->cur++ = MAJOR_MISC | MISC_FLOAT64;
412
413 uint64_t fp = ecb_double_to_binary64 (nv);
414
415 if (!ecb_big_endian ())
416 fp = ecb_bswap64 (fp);
417
418 memcpy (enc->cur, &fp, 8);
419 enc->cur += 8;
420}
421
422ecb_inline void
423encode_forced (enc_t *enc, UV type, SV *sv)
424{
425 switch (type)
426 {
427 case AS_CBOR:
428 {
429 STRLEN len;
430 char *str = SvPVbyte (sv, len);
431
432 need (enc, len);
433 memcpy (enc->cur, str, len);
434 enc->cur += len;
435 }
436 break;
437
438 case AS_BYTES:
439 {
440 STRLEN len;
441 char *str = SvPVbyte (sv, len);
442 encode_strref (enc, 0, 0, str, len);
443 }
444 break;
445
446 case AS_TEXT:
447 {
448 STRLEN len;
449 char *str = SvPVutf8 (sv, len);
450 encode_strref (enc, 1, 1, str, len);
451 }
452 break;
453
454 case AS_INT: encode_int (enc, sv); break;
455 case AS_FLOAT16: encode_float16 (enc, SvNV (sv)); break;
456 case AS_FLOAT32: encode_float32 (enc, SvNV (sv)); break;
457 case AS_FLOAT64: encode_float64 (enc, SvNV (sv)); break;
458
459 default:
460 croak ("encountered malformed CBOR::XS::Tagged object");
461 }
313} 462}
314 463
315static void encode_sv (enc_t *enc, SV *sv); 464static void encode_sv (enc_t *enc, SV *sv);
316 465
317static void 466static void
324 473
325 ++enc->depth; 474 ++enc->depth;
326 475
327 encode_uint (enc, MAJOR_ARRAY, len + 1); 476 encode_uint (enc, MAJOR_ARRAY, len + 1);
328 477
329 if (SvMAGICAL (av)) 478 if (ecb_expect_false (SvMAGICAL (av)))
330 for (i = 0; i <= len; ++i) 479 for (i = 0; i <= len; ++i)
331 { 480 {
332 SV **svp = av_fetch (av, i, 0); 481 SV **svp = av_fetch (av, i, 0);
333 encode_sv (enc, svp ? *svp : &PL_sv_undef); 482 encode_sv (enc, svp ? *svp : &PL_sv_undef);
334 } 483 }
335 else 484 else
336 for (i = 0; i <= len; ++i) 485 for (i = 0; i <= len; ++i)
337 encode_sv (enc, AvARRAY (av)[i]); 486 {
487 SV *sv = AvARRAY (av)[i];
488 encode_sv (enc, sv ? sv : &PL_sv_undef);
489 }
338 490
339 --enc->depth; 491 --enc->depth;
340} 492}
341 493
342static void 494static void
350 ++enc->depth; 502 ++enc->depth;
351 503
352 int pairs = hv_iterinit (hv); 504 int pairs = hv_iterinit (hv);
353 int mg = SvMAGICAL (hv); 505 int mg = SvMAGICAL (hv);
354 506
355 if (mg) 507 if (ecb_expect_false (mg))
356 encode_ch (enc, MAJOR_MAP | MINOR_INDEF); 508 encode_ch (enc, MAJOR_MAP | MINOR_INDEF);
357 else 509 else
358 encode_uint (enc, MAJOR_MAP, pairs); 510 encode_uint (enc, MAJOR_MAP, pairs);
359 511
360 while ((he = hv_iternext (hv))) 512 while ((he = hv_iternext (hv)))
361 { 513 {
362 if (HeKLEN (he) == HEf_SVKEY) 514 if (HeKLEN (he) == HEf_SVKEY)
363 encode_sv (enc, HeSVKEY (he)); 515 encode_sv (enc, HeSVKEY (he));
364 else 516 else
365 encode_strref (enc, HeKUTF8 (he), HeKEY (he), HeKLEN (he)); 517 encode_strref (enc, enc->cbor.flags & (F_TEXT_KEYS | F_TEXT_STRINGS), HeKUTF8 (he), HeKEY (he), HeKLEN (he));
366 518
367 encode_sv (enc, ecb_expect_false (mg) ? hv_iterval (hv, he) : HeVAL (he)); 519 encode_sv (enc, ecb_expect_false (mg) ? hv_iterval (hv, he) : HeVAL (he));
368 } 520 }
369 521
370 if (mg) 522 if (ecb_expect_false (mg))
371 encode_ch (enc, MAJOR_MISC | MINOR_INDEF); 523 encode_ch (enc, MAJOR_MISC | MINOR_INDEF);
372 524
373 --enc->depth; 525 --enc->depth;
374} 526}
375 527
408 else if (stash == tagged_stash) 560 else if (stash == tagged_stash)
409 { 561 {
410 if (svt != SVt_PVAV) 562 if (svt != SVt_PVAV)
411 croak ("encountered CBOR::XS::Tagged object that isn't an array"); 563 croak ("encountered CBOR::XS::Tagged object that isn't an array");
412 564
565 switch (av_len ((AV *)sv))
566 {
567 case 2-1:
568 // actually a tagged value
413 encode_uint (enc, MAJOR_TAG, SvUV (*av_fetch ((AV *)sv, 0, 1))); 569 encode_uint (enc, MAJOR_TAG, SvUV (*av_fetch ((AV *)sv, 0, 1)));
414 encode_sv (enc, *av_fetch ((AV *)sv, 1, 1)); 570 encode_sv (enc, *av_fetch ((AV *)sv, 1, 1));
571 break;
572
573 case 3-1:
574 // a forced type [value, type, undef]
575 encode_forced (enc, SvUV (*av_fetch ((AV *)sv, 1, 1)), *av_fetch ((AV *)sv, 0, 1));
576 break;
577
578 default:
579 croak ("encountered malformed CBOR::XS::Tagged object");
580 }
415 581
416 return; 582 return;
417 } 583 }
418 } 584 }
419 585
420 if (ecb_expect_false (SvREFCNT (sv) > 1) 586 if (ecb_expect_false (SvREFCNT (sv) > 1)
421 && ecb_expect_false (enc->cbor.flags & F_ALLOW_SHARING)) 587 && ecb_expect_false (enc->cbor.flags & F_ALLOW_SHARING))
422 { 588 {
423 if (!enc->shareable) 589 if (ecb_expect_false (!enc->shareable))
424 enc->shareable = (HV *)sv_2mortal ((SV *)newHV ()); 590 enc->shareable = (HV *)sv_2mortal ((SV *)newHV ());
425 591
426 SV **svp = hv_fetch (enc->shareable, (char *)&sv, sizeof (sv), 1); 592 SV **svp = hv_fetch (enc->shareable, (char *)&sv, sizeof (sv), 1);
427 593
428 if (SvOK (*svp)) 594 if (SvOK (*svp))
442 if (ecb_expect_false (SvOBJECT (sv))) 608 if (ecb_expect_false (SvOBJECT (sv)))
443 { 609 {
444 HV *stash = SvSTASH (sv); 610 HV *stash = SvSTASH (sv);
445 GV *method; 611 GV *method;
446 612
613 if (enc->cbor.flags & F_FORBID_OBJECTS)
614 croak ("encountered object '%s', but forbid_objects is enabled",
615 SvPV_nolen (sv_2mortal (newRV_inc (sv))));
447 if ((method = gv_fetchmethod_autoload (stash, "TO_CBOR", 0))) 616 else if ((method = gv_fetchmethod_autoload (stash, "TO_CBOR", 0)))
448 { 617 {
449 dSP; 618 dSP;
450 619
451 ENTER; SAVETMPS; PUSHMARK (SP); 620 ENTER; SAVETMPS;
621 PUSHMARK (SP);
452 // we re-bless the reference to get overload and other niceties right 622 // we re-bless the reference to get overload and other niceties right
453 XPUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), stash)); 623 XPUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), stash));
454 624
455 PUTBACK; 625 PUTBACK;
456 // G_SCALAR ensures that return value is 1 626 // G_SCALAR ensures that return value is 1
469 } 639 }
470 else if ((method = gv_fetchmethod_autoload (stash, "FREEZE", 0)) != 0) 640 else if ((method = gv_fetchmethod_autoload (stash, "FREEZE", 0)) != 0)
471 { 641 {
472 dSP; 642 dSP;
473 643
474 ENTER; SAVETMPS; PUSHMARK (SP); 644 ENTER; SAVETMPS;
645 PUSHMARK (SP);
475 EXTEND (SP, 2); 646 EXTEND (SP, 2);
476 // we re-bless the reference to get overload and other niceties right 647 // we re-bless the reference to get overload and other niceties right
477 PUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), stash)); 648 PUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), stash));
478 PUSHs (sv_cbor); 649 PUSHs (sv_cbor);
479 650
485 if (count == 1 && SvROK (TOPs) && SvRV (TOPs) == sv) 656 if (count == 1 && SvROK (TOPs) && SvRV (TOPs) == sv)
486 croak ("%s::FREEZE(CBOR) method returned same object as was passed instead of a new one", HvNAME (stash)); 657 croak ("%s::FREEZE(CBOR) method returned same object as was passed instead of a new one", HvNAME (stash));
487 658
488 encode_tag (enc, CBOR_TAG_PERL_OBJECT); 659 encode_tag (enc, CBOR_TAG_PERL_OBJECT);
489 encode_uint (enc, MAJOR_ARRAY, count + 1); 660 encode_uint (enc, MAJOR_ARRAY, count + 1);
490 encode_strref (enc, HvNAMEUTF8 (stash), HvNAME (stash), HvNAMELEN (stash)); 661 encode_strref (enc, 0, HvNAMEUTF8 (stash), HvNAME (stash), HvNAMELEN (stash));
491 662
492 while (count) 663 {
664 int i;
665
666 for (i = 0; i < count; ++i)
493 encode_sv (enc, SP[1 - count--]); 667 encode_sv (enc, SP[i + 1 - count]);
668
669 SP -= count;
670 }
494 671
495 PUTBACK; 672 PUTBACK;
496 673
497 FREETMPS; LEAVE; 674 FREETMPS; LEAVE;
498 } 675 }
520 697
521 if (ecb_expect_false (nv == (NV)(U32)nv)) 698 if (ecb_expect_false (nv == (NV)(U32)nv))
522 encode_uint (enc, MAJOR_POS_INT, (U32)nv); 699 encode_uint (enc, MAJOR_POS_INT, (U32)nv);
523 //TODO: maybe I32? 700 //TODO: maybe I32?
524 else if (ecb_expect_false (nv == (float)nv)) 701 else if (ecb_expect_false (nv == (float)nv))
525 { 702 encode_float32 (enc, nv);
526 uint32_t fp = ecb_float_to_binary32 (nv);
527
528 *enc->cur++ = MAJOR_MISC | MISC_FLOAT32;
529
530 if (!ecb_big_endian ())
531 fp = ecb_bswap32 (fp);
532
533 memcpy (enc->cur, &fp, 4);
534 enc->cur += 4;
535 }
536 else 703 else
537 { 704 encode_float64 (enc, nv);
538 uint64_t fp = ecb_double_to_binary64 (nv);
539
540 *enc->cur++ = MAJOR_MISC | MISC_FLOAT64;
541
542 if (!ecb_big_endian ())
543 fp = ecb_bswap64 (fp);
544
545 memcpy (enc->cur, &fp, 8);
546 enc->cur += 8;
547 }
548} 705}
549 706
550static void 707static void
551encode_sv (enc_t *enc, SV *sv) 708encode_sv (enc_t *enc, SV *sv)
552{ 709{
554 711
555 if (SvPOKp (sv)) 712 if (SvPOKp (sv))
556 { 713 {
557 STRLEN len; 714 STRLEN len;
558 char *str = SvPV (sv, len); 715 char *str = SvPV (sv, len);
559 encode_strref (enc, SvUTF8 (sv), str, len); 716 encode_strref (enc, enc->cbor.flags & F_TEXT_STRINGS, SvUTF8 (sv), str, len);
560 } 717 }
561 else if (SvNOKp (sv)) 718 else if (SvNOKp (sv))
562 encode_nv (enc, sv); 719 encode_nv (enc, sv);
563 else if (SvIOKp (sv)) 720 else if (SvIOKp (sv))
564 { 721 encode_int (enc, sv);
565 if (SvIsUV (sv))
566 encode_uint (enc, MAJOR_POS_INT, SvUVX (sv));
567 else if (SvIVX (sv) >= 0)
568 encode_uint (enc, MAJOR_POS_INT, SvIVX (sv));
569 else
570 encode_uint (enc, MAJOR_NEG_INT, -(SvIVX (sv) + 1));
571 }
572 else if (SvROK (sv)) 722 else if (SvROK (sv))
573 encode_rv (enc, SvRV (sv)); 723 encode_rv (enc, SvRV (sv));
574 else if (!SvOK (sv)) 724 else if (!SvOK (sv))
575 encode_ch (enc, MAJOR_MISC | SIMPLE_NULL); 725 encode_ch (enc, MAJOR_MISC | SIMPLE_NULL);
576 else if (enc->cbor.flags & F_ALLOW_UNKNOWN) 726 else if (enc->cbor.flags & F_ALLOW_UNKNOWN)
581} 731}
582 732
583static SV * 733static SV *
584encode_cbor (SV *scalar, CBOR *cbor) 734encode_cbor (SV *scalar, CBOR *cbor)
585{ 735{
586 enc_t enc = { }; 736 enc_t enc = { 0 };
587 737
588 enc.cbor = *cbor; 738 enc.cbor = *cbor;
589 enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE)); 739 enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));
590 enc.cur = SvPVX (enc.sv); 740 enc.cur = SvPVX (enc.sv);
591 enc.end = SvEND (enc.sv); 741 enc.end = SvEND (enc.sv);
592 742
593 SvPOK_only (enc.sv); 743 SvPOK_only (enc.sv);
594 744
595 if (cbor->flags & F_PACK_STRINGS) 745 if (cbor->flags & F_PACK_STRINGS)
596 { 746 {
623 U32 depth; // recursion depth 773 U32 depth; // recursion depth
624 U32 maxdepth; // recursion depth limit 774 U32 maxdepth; // recursion depth limit
625 AV *shareable; 775 AV *shareable;
626 AV *stringref; 776 AV *stringref;
627 SV *decode_tagged; 777 SV *decode_tagged;
778 SV *err_sv; // optional sv for error, needs to be freed
628} dec_t; 779} dec_t;
629 780
630#define ERR(reason) SB if (!dec->err) dec->err = reason; goto fail; SE 781// set dec->err to ERRSV
782ecb_cold static void
783err_errsv (dec_t *dec)
784{
785 if (!dec->err)
786 {
787 dec->err_sv = newSVsv (ERRSV);
631 788
632#define WANT(len) if (ecb_expect_false (dec->cur + len > dec->end)) ERR ("unexpected end of CBOR data") 789 // chop off the trailing \n
790 SvCUR_set (dec->err_sv, SvCUR (dec->err_sv) - 1);
791 *SvEND (dec->err_sv) = 0;
633 792
793 dec->err = SvPVutf8_nolen (dec->err_sv);
794 }
795}
796
797// the following functions are used to reduce code size and help the compiler to optimise
798ecb_cold static void
799err_set (dec_t *dec, const char *reason)
800{
801 if (!dec->err)
802 dec->err = reason;
803}
804
805ecb_cold static void
806err_unexpected_end (dec_t *dec)
807{
808 err_set (dec, "unexpected end of CBOR data");
809}
810
811#define ERR_DO(do) SB do; goto fail; SE
812#define ERR(reason) ERR_DO (err_set (dec, reason))
813#define ERR_ERRSV ERR_DO (err_errsv (dec))
814
815#define WANT(len) if (ecb_expect_false ((uintptr_t)(dec->end - dec->cur) < (STRLEN)len)) ERR_DO (err_unexpected_end (dec))
816
634#define DEC_INC_DEPTH if (++dec->depth > dec->cbor.max_depth) ERR (ERR_NESTING_EXCEEDED) 817#define DEC_INC_DEPTH if (ecb_expect_false (++dec->depth > dec->cbor.max_depth)) ERR (ERR_NESTING_EXCEEDED)
635#define DEC_DEC_DEPTH --dec->depth 818#define DEC_DEC_DEPTH --dec->depth
636 819
637static UV 820static UV
638decode_uint (dec_t *dec) 821decode_uint (dec_t *dec)
639{ 822{
705 888
706 for (;;) 889 for (;;)
707 { 890 {
708 WANT (1); 891 WANT (1);
709 892
710 if (*dec->cur == (MAJOR_MISC | MINOR_INDEF)) 893 if (*dec->cur == (MAJOR_MISC | MINOR_INDEF) || dec->err)
711 { 894 {
712 ++dec->cur; 895 ++dec->cur;
713 break; 896 break;
714 } 897 }
715 898
716 av_push (av, decode_sv (dec)); 899 av_push (av, decode_sv (dec));
717 } 900 }
718 } 901 }
719 else 902 else
720 { 903 {
721 int i, len = decode_uint (dec); 904 UV i, len = decode_uint (dec);
722 905
723 WANT (len); // complexity check for av_fill - need at least one byte per value, do not allow supersize arrays 906 WANT (len); // complexity check for av_fill - need at least one byte per value, do not allow supersize arrays
724 av_fill (av, len - 1); 907 av_fill (av, len - 1);
725 908
726 for (i = 0; i < len; ++i) 909 for (i = 0; i < len; ++i)
729 912
730 DEC_DEC_DEPTH; 913 DEC_DEC_DEPTH;
731 return newRV_noinc ((SV *)av); 914 return newRV_noinc ((SV *)av);
732 915
733fail: 916fail:
734 SvREFCNT_dec (av); 917 SvREFCNT_dec_NN (av);
735 DEC_DEC_DEPTH; 918 DEC_DEC_DEPTH;
736 return &PL_sv_undef; 919 return &PL_sv_undef;
737} 920}
738 921
739static void 922static void
743 // byte or utf-8 strings as keys, but only when !stringref 926 // byte or utf-8 strings as keys, but only when !stringref
744 927
745 if (ecb_expect_true (!dec->stringref)) 928 if (ecb_expect_true (!dec->stringref))
746 if (ecb_expect_true ((U8)(*dec->cur - MAJOR_BYTES) <= LENGTH_EXT8)) 929 if (ecb_expect_true ((U8)(*dec->cur - MAJOR_BYTES) <= LENGTH_EXT8))
747 { 930 {
748 I32 len = decode_uint (dec); 931 STRLEN len = decode_uint (dec);
749 char *key = (char *)dec->cur; 932 char *key = (char *)dec->cur;
750 933
934 WANT (len);
751 dec->cur += len; 935 dec->cur += len;
752 936
753 hv_store (hv, key, len, decode_sv (dec), 0); 937 hv_store (hv, key, len, decode_sv (dec), 0);
754 938
755 return; 939 return;
756 } 940 }
757 else if (ecb_expect_true ((U8)(*dec->cur - MAJOR_TEXT) <= LENGTH_EXT8)) 941 else if (ecb_expect_true ((U8)(*dec->cur - MAJOR_TEXT) <= LENGTH_EXT8))
758 { 942 {
759 I32 len = decode_uint (dec); 943 STRLEN len = decode_uint (dec);
760 char *key = (char *)dec->cur; 944 char *key = (char *)dec->cur;
761 945
946 WANT (len);
762 dec->cur += len; 947 dec->cur += len;
763 948
764 if (ecb_expect_false (dec->cbor.flags & F_VALIDATE_UTF8)) 949 if (ecb_expect_false (dec->cbor.flags & F_VALIDATE_UTF8))
765 if (!is_utf8_string (key, len)) 950 if (!is_utf8_string (key, len))
766 ERR ("corrupted CBOR data (invalid UTF-8 in map key)"); 951 ERR ("corrupted CBOR data (invalid UTF-8 in map key)");
771 } 956 }
772 957
773 SV *k = decode_sv (dec); 958 SV *k = decode_sv (dec);
774 SV *v = decode_sv (dec); 959 SV *v = decode_sv (dec);
775 960
961 // we leak memory if uncaught exceptions are thrown by random magical
962 // methods, and this is hopefully the only place where it can happen,
963 // so if there is a chance of an exception, take the very slow path.
964 // since catching exceptions is "undocumented/internal/forbidden" by
965 // the new p5p powers, we need to call out to a perl function :/
966 if (ecb_expect_false (SvAMAGIC (k)))
967 {
968 dSP;
969
970 ENTER; SAVETMPS;
971 PUSHMARK (SP);
972 EXTEND (SP, 3);
973 PUSHs (sv_2mortal (newRV_inc ((SV *)hv)));
974 PUSHs (sv_2mortal (k));
975 PUSHs (sv_2mortal (v));
976
977 PUTBACK;
978 call_pv ("CBOR::XS::_hv_store", G_VOID | G_DISCARD | G_EVAL);
979 SPAGAIN;
980
981 FREETMPS; LEAVE;
982
983 if (SvTRUE (ERRSV))
984 ERR_ERRSV;
985
986 return;
987 }
988
776 hv_store_ent (hv, k, v, 0); 989 hv_store_ent (hv, k, v, 0);
777 SvREFCNT_dec (k); 990 SvREFCNT_dec_NN (k);
778 991
779fail: 992fail:
780 ; 993 ;
781} 994}
782 995
793 1006
794 for (;;) 1007 for (;;)
795 { 1008 {
796 WANT (1); 1009 WANT (1);
797 1010
798 if (*dec->cur == (MAJOR_MISC | MINOR_INDEF)) 1011 if (*dec->cur == (MAJOR_MISC | MINOR_INDEF) || dec->err)
799 { 1012 {
800 ++dec->cur; 1013 ++dec->cur;
801 break; 1014 break;
802 } 1015 }
803 1016
804 decode_he (dec, hv); 1017 decode_he (dec, hv);
805 } 1018 }
806 } 1019 }
807 else 1020 else
808 { 1021 {
809 int pairs = decode_uint (dec); 1022 UV pairs = decode_uint (dec);
1023
1024 WANT (pairs); // complexity check - need at least one byte per value, do not allow supersize hashes
810 1025
811 while (pairs--) 1026 while (pairs--)
812 decode_he (dec, hv); 1027 decode_he (dec, hv);
813 } 1028 }
814 1029
815 DEC_DEC_DEPTH; 1030 DEC_DEC_DEPTH;
816 return newRV_noinc ((SV *)hv); 1031 return newRV_noinc ((SV *)hv);
817 1032
818fail: 1033fail:
819 SvREFCNT_dec (hv); 1034 SvREFCNT_dec_NN (hv);
820 DEC_DEC_DEPTH; 1035 DEC_DEC_DEPTH;
821 return &PL_sv_undef; 1036 return &PL_sv_undef;
822} 1037}
823 1038
824static SV * 1039static SV *
825decode_str (dec_t *dec, int utf8) 1040decode_str (dec_t *dec, int utf8)
826{ 1041{
827 SV *sv = 0; 1042 SV *sv = 0;
828 1043
829 if ((*dec->cur & MINOR_MASK) == MINOR_INDEF) 1044 if (ecb_expect_false ((*dec->cur & MINOR_MASK) == MINOR_INDEF))
830 { 1045 {
831 // indefinite length strings 1046 // indefinite length strings
832 ++dec->cur; 1047 ++dec->cur;
833 1048
834 U8 major = *dec->cur & MAJOR_MISC; 1049 U8 major = *dec->cur & MAJOR_MISC;
902 sv = newRV_noinc (decode_sv (dec)); 1117 sv = newRV_noinc (decode_sv (dec));
903 break; 1118 break;
904 1119
905 case CBOR_TAG_STRINGREF_NAMESPACE: 1120 case CBOR_TAG_STRINGREF_NAMESPACE:
906 { 1121 {
1122 // do not use SAVETMPS/FREETMPS, as these will
1123 // erase mortalised caches, e.g. "shareable"
907 ENTER; SAVETMPS; 1124 ENTER;
908 1125
909 SAVESPTR (dec->stringref); 1126 SAVESPTR (dec->stringref);
910 dec->stringref = (AV *)sv_2mortal ((SV *)newAV ()); 1127 dec->stringref = (AV *)sv_2mortal ((SV *)newAV ());
911 1128
912 sv = decode_sv (dec); 1129 sv = decode_sv (dec);
913 1130
914 FREETMPS; LEAVE; 1131 LEAVE;
915 } 1132 }
916 break; 1133 break;
917 1134
918 case CBOR_TAG_STRINGREF: 1135 case CBOR_TAG_STRINGREF:
919 { 1136 {
920 if ((*dec->cur >> MAJOR_SHIFT) != (MAJOR_POS_INT >> MAJOR_SHIFT)) 1137 if ((*dec->cur >> MAJOR_SHIFT) != (MAJOR_POS_INT >> MAJOR_SHIFT))
921 ERR ("corrupted CBOR data (stringref index not an unsigned integer)"); 1138 ERR ("corrupted CBOR data (stringref index not an unsigned integer)");
922 1139
923 UV idx = decode_uint (dec); 1140 UV idx = decode_uint (dec);
924 1141
925 if (!dec->stringref || (int)idx > AvFILLp (dec->stringref)) 1142 if (!dec->stringref || idx >= (UV)(1 + AvFILLp (dec->stringref)))
926 ERR ("corrupted CBOR data (stringref index out of bounds or outside namespace)"); 1143 ERR ("corrupted CBOR data (stringref index out of bounds or outside namespace)");
927 1144
928 sv = newSVsv (AvARRAY (dec->stringref)[idx]); 1145 sv = newSVsv (AvARRAY (dec->stringref)[idx]);
929 } 1146 }
930 break; 1147 break;
958 if ((*dec->cur >> MAJOR_SHIFT) != (MAJOR_POS_INT >> MAJOR_SHIFT)) 1175 if ((*dec->cur >> MAJOR_SHIFT) != (MAJOR_POS_INT >> MAJOR_SHIFT))
959 ERR ("corrupted CBOR data (sharedref index not an unsigned integer)"); 1176 ERR ("corrupted CBOR data (sharedref index not an unsigned integer)");
960 1177
961 UV idx = decode_uint (dec); 1178 UV idx = decode_uint (dec);
962 1179
963 if (!dec->shareable || (int)idx > AvFILLp (dec->shareable)) 1180 if (!dec->shareable || idx >= (UV)(1 + AvFILLp (dec->shareable)))
964 ERR ("corrupted CBOR data (sharedref index out of bounds)"); 1181 ERR ("corrupted CBOR data (sharedref index out of bounds)");
965 1182
966 sv = SvREFCNT_inc_NN (AvARRAY (dec->shareable)[idx]); 1183 sv = SvREFCNT_inc_NN (AvARRAY (dec->shareable)[idx]);
967 1184
968 if (sv == &PL_sv_undef) 1185 if (sv == &PL_sv_undef)
970 } 1187 }
971 break; 1188 break;
972 1189
973 case CBOR_TAG_PERL_OBJECT: 1190 case CBOR_TAG_PERL_OBJECT:
974 { 1191 {
1192 if (dec->cbor.flags & F_FORBID_OBJECTS)
1193 goto filter;
1194
975 sv = decode_sv (dec); 1195 sv = decode_sv (dec);
976 1196
977 if (!SvROK (sv) || SvTYPE (SvRV (sv)) != SVt_PVAV) 1197 if (!SvROK (sv) || SvTYPE (SvRV (sv)) != SVt_PVAV)
978 ERR ("corrupted CBOR data (non-array perl object)"); 1198 ERR ("corrupted CBOR data (non-array perl object)");
979 1199
989 if (!method) 1209 if (!method)
990 ERR ("cannot decode perl-object (package does not have a THAW method)"); 1210 ERR ("cannot decode perl-object (package does not have a THAW method)");
991 1211
992 dSP; 1212 dSP;
993 1213
994 ENTER; SAVETMPS; PUSHMARK (SP); 1214 ENTER; SAVETMPS;
1215 PUSHMARK (SP);
995 EXTEND (SP, len + 1); 1216 EXTEND (SP, len + 1);
996 // we re-bless the reference to get overload and other niceties right 1217 // we re-bless the reference to get overload and other niceties right
997 PUSHs (*av_fetch (av, 0, 1)); 1218 PUSHs (*av_fetch (av, 0, 1));
998 PUSHs (sv_cbor); 1219 PUSHs (sv_cbor);
999 1220
1007 SPAGAIN; 1228 SPAGAIN;
1008 1229
1009 if (SvTRUE (ERRSV)) 1230 if (SvTRUE (ERRSV))
1010 { 1231 {
1011 FREETMPS; LEAVE; 1232 FREETMPS; LEAVE;
1012 ERR (SvPVutf8_nolen (sv_2mortal (SvREFCNT_inc (ERRSV)))); 1233 ERR_ERRSV;
1013 } 1234 }
1014 1235
1015 SvREFCNT_dec (sv); 1236 SvREFCNT_dec_NN (sv);
1016 sv = SvREFCNT_inc (POPs); 1237 sv = SvREFCNT_inc (POPs);
1017 1238
1018 PUTBACK; 1239 PUTBACK;
1019 1240
1020 FREETMPS; LEAVE; 1241 FREETMPS; LEAVE;
1021 } 1242 }
1022 break; 1243 break;
1023 1244
1024 default: 1245 default:
1246 filter:
1025 { 1247 {
1248 SV *tag_sv = newSVuv (tag);
1249
1026 sv = decode_sv (dec); 1250 sv = decode_sv (dec);
1027 1251
1028 dSP; 1252 dSP;
1029 ENTER; SAVETMPS; PUSHMARK (SP); 1253 ENTER; SAVETMPS;
1254 PUSHMARK (SP);
1030 EXTEND (SP, 2); 1255 EXTEND (SP, 2);
1031 PUSHs (newSVuv (tag)); 1256 PUSHs (tag_sv);
1032 PUSHs (sv); 1257 PUSHs (sv);
1033 1258
1034 PUTBACK; 1259 PUTBACK;
1035 int count = call_sv (dec->cbor.filter ? dec->cbor.filter : default_filter, G_ARRAY | G_EVAL); 1260 int count = call_sv (dec->cbor.filter ? dec->cbor.filter : default_filter, G_ARRAY | G_EVAL);
1036 SPAGAIN; 1261 SPAGAIN;
1037 1262
1038 if (SvTRUE (ERRSV)) 1263 if (SvTRUE (ERRSV))
1039 { 1264 {
1265 SvREFCNT_dec_NN (tag_sv);
1040 FREETMPS; LEAVE; 1266 FREETMPS; LEAVE;
1041 ERR (SvPVutf8_nolen (sv_2mortal (SvREFCNT_inc (ERRSV)))); 1267 ERR_ERRSV;
1042 } 1268 }
1043 1269
1044 if (count) 1270 if (count)
1045 { 1271 {
1272 SvREFCNT_dec_NN (tag_sv);
1046 SvREFCNT_dec (sv); 1273 SvREFCNT_dec_NN (sv);
1047 sv = SvREFCNT_inc (POPs); 1274 sv = SvREFCNT_inc_NN (TOPs);
1275 SP -= count;
1048 } 1276 }
1049 else 1277 else
1050 { 1278 {
1051 AV *av = newAV (); 1279 AV *av = newAV ();
1052 av_push (av, newSVuv (tag)); 1280 av_push (av, tag_sv);
1053 av_push (av, sv); 1281 av_push (av, sv);
1054 1282
1055 HV *tagged_stash = !CBOR_SLOW || cbor_tagged_stash 1283 HV *tagged_stash = !CBOR_SLOW || cbor_tagged_stash
1056 ? cbor_tagged_stash 1284 ? cbor_tagged_stash
1057 : gv_stashpv ("CBOR::XS::Tagged" , 1); 1285 : gv_stashpv ("CBOR::XS::Tagged" , 1);
1160} 1388}
1161 1389
1162static SV * 1390static SV *
1163decode_cbor (SV *string, CBOR *cbor, char **offset_return) 1391decode_cbor (SV *string, CBOR *cbor, char **offset_return)
1164{ 1392{
1165 dec_t dec = { }; 1393 dec_t dec = { 0 };
1166 SV *sv; 1394 SV *sv;
1167 STRLEN len; 1395 STRLEN len;
1168 char *data = SvPVbyte (string, len); 1396 char *data = SvPVbyte (string, len);
1169 1397
1170 if (len > cbor->max_size && cbor->max_size) 1398 if (len > cbor->max_size && cbor->max_size)
1186 1414
1187 if (dec.err) 1415 if (dec.err)
1188 { 1416 {
1189 if (dec.shareable) 1417 if (dec.shareable)
1190 { 1418 {
1191 // need to break cyclic links, which whould all be in shareable 1419 // need to break cyclic links, which would all be in shareable
1192 int i; 1420 int i;
1193 SV **svp; 1421 SV **svp;
1194 1422
1195 for (i = av_len (dec.shareable) + 1; i--; ) 1423 for (i = av_len (dec.shareable) + 1; i--; )
1196 if ((svp = av_fetch (dec.shareable, i, 0))) 1424 if ((svp = av_fetch (dec.shareable, i, 0)))
1197 sv_setsv (*svp, &PL_sv_undef); 1425 sv_setsv (*svp, &PL_sv_undef);
1198 } 1426 }
1199 1427
1200 SvREFCNT_dec (sv); 1428 SvREFCNT_dec_NN (sv);
1429
1430 if (dec.err_sv)
1431 sv_2mortal (dec.err_sv);
1432
1201 croak ("%s, at offset %d (octet 0x%02x)", dec.err, dec.cur - (U8 *)data, (int)(uint8_t)*dec.cur); 1433 croak ("%s, at offset %d (octet 0x%02x)", dec.err, dec.cur - (U8 *)data, (int)(uint8_t)*dec.cur);
1202 } 1434 }
1203 1435
1204 sv = sv_2mortal (sv); 1436 sv = sv_2mortal (sv);
1205 1437
1261 1493
1262 int major = *p >> MAJOR_SHIFT; 1494 int major = *p >> MAJOR_SHIFT;
1263 1495
1264 switch (major) 1496 switch (major)
1265 { 1497 {
1498 case MAJOR_TAG >> MAJOR_SHIFT:
1499 ++count; // tags merely prefix another value
1500 break;
1501
1266 case MAJOR_BYTES >> MAJOR_SHIFT: 1502 case MAJOR_BYTES >> MAJOR_SHIFT:
1267 case MAJOR_TEXT >> MAJOR_SHIFT: 1503 case MAJOR_TEXT >> MAJOR_SHIFT:
1268 case MAJOR_ARRAY >> MAJOR_SHIFT: 1504 case MAJOR_ARRAY >> MAJOR_SHIFT:
1269 case MAJOR_MAP >> MAJOR_SHIFT: 1505 case MAJOR_MAP >> MAJOR_SHIFT:
1270 { 1506 {
1345 1581
1346 default_filter = newSVpv ("CBOR::XS::default_filter", 0); 1582 default_filter = newSVpv ("CBOR::XS::default_filter", 0);
1347 1583
1348 sv_cbor = newSVpv ("CBOR", 0); 1584 sv_cbor = newSVpv ("CBOR", 0);
1349 SvREADONLY_on (sv_cbor); 1585 SvREADONLY_on (sv_cbor);
1586
1587 assert (("STRLEN must be an unsigned type", 0 <= (STRLEN)-1));
1350} 1588}
1351 1589
1352PROTOTYPES: DISABLE 1590PROTOTYPES: DISABLE
1353 1591
1354void CLONE (...) 1592void CLONE (...)
1374 ALIAS: 1612 ALIAS:
1375 shrink = F_SHRINK 1613 shrink = F_SHRINK
1376 allow_unknown = F_ALLOW_UNKNOWN 1614 allow_unknown = F_ALLOW_UNKNOWN
1377 allow_sharing = F_ALLOW_SHARING 1615 allow_sharing = F_ALLOW_SHARING
1378 allow_cycles = F_ALLOW_CYCLES 1616 allow_cycles = F_ALLOW_CYCLES
1617 forbid_objects = F_FORBID_OBJECTS
1379 pack_strings = F_PACK_STRINGS 1618 pack_strings = F_PACK_STRINGS
1619 text_keys = F_TEXT_KEYS
1620 text_strings = F_TEXT_STRINGS
1380 validate_utf8 = F_VALIDATE_UTF8 1621 validate_utf8 = F_VALIDATE_UTF8
1381 PPCODE: 1622 PPCODE:
1382{ 1623{
1383 if (enable) 1624 if (enable)
1384 self->flags |= ix; 1625 self->flags |= ix;
1392 ALIAS: 1633 ALIAS:
1393 get_shrink = F_SHRINK 1634 get_shrink = F_SHRINK
1394 get_allow_unknown = F_ALLOW_UNKNOWN 1635 get_allow_unknown = F_ALLOW_UNKNOWN
1395 get_allow_sharing = F_ALLOW_SHARING 1636 get_allow_sharing = F_ALLOW_SHARING
1396 get_allow_cycles = F_ALLOW_CYCLES 1637 get_allow_cycles = F_ALLOW_CYCLES
1638 get_forbid_objects = F_FORBID_OBJECTS
1397 get_pack_strings = F_PACK_STRINGS 1639 get_pack_strings = F_PACK_STRINGS
1640 get_text_keys = F_TEXT_KEYS
1641 get_text_strings = F_TEXT_STRINGS
1398 get_validate_utf8 = F_VALIDATE_UTF8 1642 get_validate_utf8 = F_VALIDATE_UTF8
1399 PPCODE: 1643 PPCODE:
1400 XPUSHs (boolSV (self->flags & ix)); 1644 XPUSHs (boolSV (self->flags & ix));
1401 1645
1402void max_depth (CBOR *self, U32 max_depth = 0x80000000UL) 1646void max_depth (CBOR *self, U32 max_depth = 0x80000000UL)
1532 cbor_init (&cbor); 1776 cbor_init (&cbor);
1533 PUTBACK; cborstr = decode_cbor (cborstr, &cbor, 0); SPAGAIN; 1777 PUTBACK; cborstr = decode_cbor (cborstr, &cbor, 0); SPAGAIN;
1534 XPUSHs (cborstr); 1778 XPUSHs (cborstr);
1535} 1779}
1536 1780
1781#ifdef __AFL_COMPILER
1782
1783void
1784afl_init ()
1785 CODE:
1786 __AFL_INIT ();
1787
1788int
1789afl_loop (unsigned int count = 10000)
1790 CODE:
1791 RETVAL = __AFL_LOOP (count);
1792 OUTPUT:
1793 RETVAL
1794
1795#endif
1796

Diff Legend

Removed lines
+ Added lines
< Changed lines
> Changed lines