2 * Copyright (c) 2014 Sebastian Freundt
5 * Redistribution and use in source and binary forms, with or without
6 * modification, are permitted provided that the following conditions
8 * 1. Redistributions of source code must retain the above copyright
9 * notice, this list of conditions and the following disclaimer.
10 * 2. Redistributions in binary form must reproduce the above copyright
11 * notice, this list of conditions and the following disclaimer in the
12 * documentation and/or other materials provided with the distribution.
14 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR(S) ``AS IS'' AND ANY EXPRESS OR
15 * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
16 * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
17 * IN NO EVENT SHALL THE AUTHOR(S) BE LIABLE FOR ANY DIRECT, INDIRECT,
18 * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
19 * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
20 * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
21 * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
22 * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF
23 * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
26 #include "archive_platform.h"
27 __FBSDID("$FreeBSD$");
30 * WARC is standardised by ISO TC46/SC4/WG12 and currently available as
32 * For the purposes of this file we used the final draft from:
33 * http://bibnum.bnf.fr/warc/WARC_ISO_28500_version1_latestdraft.pdf
36 * [ ] real-world warcs can contain resources at endpoints ending in /
37 * e.g. http://bibnum.bnf.fr/warc/
38 * if you're lucky their response contains a Content-Location: header
39 * pointing to a unix-compliant filename, in the example above it's
40 * Content-Location: http://bibnum.bnf.fr/warc/index.html
41 * however, that's not mandated and github for example doesn't follow
43 * We need a set of archive options to control what to do with
44 * entries like these, at the moment care is taken to skip them.
48 #ifdef HAVE_SYS_STAT_H
71 #include "archive_entry.h"
72 #include "archive_private.h"
73 #include "archive_read_private.h"
83 /* request, unsupported */
85 /* response, unsupported */
87 /* revisit, unsupported */
89 /* conversion, unsupported */
91 /* continutation, unsupported at the moment */
108 /* content length ahead */
110 /* and how much we've processed so far */
112 /* and how much we need to consume between calls */
117 /* previous version */
119 /* stringified format name */
120 struct archive_string sver;
123 static int _warc_bid(struct archive_read *a, int);
124 static int _warc_cleanup(struct archive_read *a);
125 static int _warc_read(struct archive_read*, const void**, size_t*, int64_t*);
126 static int _warc_skip(struct archive_read *a);
127 static int _warc_rdhdr(struct archive_read *a, struct archive_entry *e);
129 /* private routines */
130 static unsigned int _warc_rdver(const char buf[10], size_t bsz);
131 static unsigned int _warc_rdtyp(const char *buf, size_t bsz);
132 static warc_string_t _warc_rduri(const char *buf, size_t bsz);
133 static ssize_t _warc_rdlen(const char *buf, size_t bsz);
134 static time_t _warc_rdrtm(const char *buf, size_t bsz);
135 static time_t _warc_rdmtm(const char *buf, size_t bsz);
136 static const char *_warc_find_eoh(const char *buf, size_t bsz);
140 archive_read_support_format_warc(struct archive *_a)
142 struct archive_read *a = (struct archive_read *)_a;
146 archive_check_magic(_a, ARCHIVE_READ_MAGIC,
147 ARCHIVE_STATE_NEW, "archive_read_support_format_warc");
149 if ((w = calloc(1, sizeof(*w))) == NULL) {
150 archive_set_error(&a->archive, ENOMEM,
151 "Can't allocate warc data");
152 return (ARCHIVE_FATAL);
155 r = __archive_read_register_format(
157 _warc_bid, NULL, _warc_rdhdr, _warc_read,
158 _warc_skip, NULL, _warc_cleanup, NULL, NULL);
160 if (r != ARCHIVE_OK) {
168 _warc_cleanup(struct archive_read *a)
170 struct warc_s *w = a->format->data;
172 if (w->pool.len > 0U) {
175 archive_string_free(&w->sver);
177 a->format->data = NULL;
182 _warc_bid(struct archive_read *a, int best_bid)
188 (void)best_bid; /* UNUSED */
190 /* check first line of file, it should be a record already */
191 if ((hdr = __archive_read_ahead(a, 12U, &nrd)) == NULL) {
192 /* no idea what to do */
194 } else if (nrd < 12) {
195 /* nah, not for us, our magic cookie is at least 12 bytes */
199 /* otherwise snarf the record's version number */
200 ver = _warc_rdver(hdr, nrd);
201 if (ver == 0U || ver > 10000U) {
202 /* oh oh oh, best not to wager ... */
206 /* otherwise be confident */
211 _warc_rdhdr(struct archive_read *a, struct archive_entry *entry)
213 #define HDR_PROBE_LEN (12U)
214 struct warc_s *w = a->format->data;
219 /* for the file name, saves some strndup()'ing */
221 /* warc record type, not that we really use it a lot */
223 /* content-length+error monad */
225 /* record time is the WARC-Date time we reinterpret it as ctime */
227 /* mtime is the Last-Modified time which will be the entry's mtime */
231 /* just use read_ahead() they keep track of unconsumed
232 * bits and bobs for us; no need to put an extra shift in
233 * and reproduce that functionality here */
234 buf = __archive_read_ahead(a, HDR_PROBE_LEN, &nrd);
239 &a->archive, ARCHIVE_ERRNO_MISC,
240 "Bad record header");
241 return (ARCHIVE_FATAL);
242 } else if (buf == NULL) {
243 /* there should be room for at least WARC/bla\r\n
244 * must be EOF therefore */
245 return (ARCHIVE_EOF);
247 /* looks good so far, try and find the end of the header now */
248 eoh = _warc_find_eoh(buf, nrd);
250 /* still no good, the header end might be beyond the
251 * probe we've requested, but then again who'd cram
252 * so much stuff into the header *and* be 28500-compliant */
254 &a->archive, ARCHIVE_ERRNO_MISC,
255 "Bad record header");
256 return (ARCHIVE_FATAL);
257 } else if ((ver = _warc_rdver(buf, eoh - buf)) > 10000U) {
258 /* nawww, I wish they promised backward compatibility
259 * anyhoo, in their infinite wisdom the 28500 guys might
260 * come up with something we can't possibly handle so
261 * best end things here */
263 &a->archive, ARCHIVE_ERRNO_MISC,
264 "Unsupported record version");
265 return (ARCHIVE_FATAL);
266 } else if ((cntlen = _warc_rdlen(buf, eoh - buf)) < 0) {
267 /* nightmare! the specs say content-length is mandatory
268 * so I don't feel overly bad stopping the reader here */
271 "Bad content length");
272 return (ARCHIVE_FATAL);
273 } else if ((rtime = _warc_rdrtm(buf, eoh - buf)) == (time_t)-1) {
274 /* record time is mandatory as per WARC/1.0,
275 * so just barf here, fast and loud */
279 return (ARCHIVE_FATAL);
282 /* let the world know we're a WARC archive */
283 a->archive.archive_format = ARCHIVE_FORMAT_WARC;
284 if (ver != w->pver) {
285 /* stringify this entry's version */
286 archive_string_sprintf(&w->sver,
287 "WARC/%u.%u", ver / 10000, ver % 10000);
288 /* remember the version */
291 /* start off with the type */
292 ftyp = _warc_rdtyp(buf, eoh - buf);
293 /* and let future calls know about the content */
296 mtime = 0;/* Avoid compiling error on some platform. */
301 /* only try and read the filename in the cases that are
302 * guaranteed to have one */
303 fnam = _warc_rduri(buf, eoh - buf);
304 /* check the last character in the URI to avoid creating
305 * directory endpoints as files, see Todo above */
306 if (fnam.len == 0 || fnam.str[fnam.len - 1] == '/') {
307 /* break here for now */
312 /* bang to our string pool, so we save a
313 * malloc()+free() roundtrip */
314 if (fnam.len + 1U > w->pool.len) {
315 w->pool.len = ((fnam.len + 64U) / 64U) * 64U;
316 w->pool.str = realloc(w->pool.str, w->pool.len);
318 memcpy(w->pool.str, fnam.str, fnam.len);
319 w->pool.str[fnam.len] = '\0';
320 /* let no one else know about the pool, it's a secret, shhh */
321 fnam.str = w->pool.str;
323 /* snarf mtime or deduce from rtime
324 * this is a custom header added by our writer, it's quite
325 * hard to believe anyone else would go through with it
326 * (apart from being part of some http responses of course) */
327 if ((mtime = _warc_rdmtm(buf, eoh - buf)) == (time_t)-1) {
337 /* now eat some of those delicious buffer bits */
338 __archive_read_consume(a, eoh - buf);
344 /* populate entry object */
345 archive_entry_set_filetype(entry, AE_IFREG);
346 archive_entry_copy_pathname(entry, fnam.str);
347 archive_entry_set_size(entry, cntlen);
348 archive_entry_set_perm(entry, 0644);
349 /* rtime is the new ctime, mtime stays mtime */
350 archive_entry_set_ctime(entry, rtime, 0L);
351 archive_entry_set_mtime(entry, mtime, 0L);
356 /* consume the content and start over */
364 _warc_read(struct archive_read *a, const void **buf, size_t *bsz, int64_t *off)
366 struct warc_s *w = a->format->data;
370 if (w->cntoff >= w->cntlen) {
372 /* it's our lucky day, no work, we can leave early */
375 *off = w->cntoff + 4U/*for \r\n\r\n separator*/;
377 return (ARCHIVE_EOF);
380 rab = __archive_read_ahead(a, 1U, &nrd);
383 /* big catastrophe */
385 } else if (nrd == 0) {
387 } else if ((size_t)nrd > w->cntlen - w->cntoff) {
388 /* clamp to content-length */
389 nrd = w->cntlen - w->cntoff;
396 w->unconsumed = (size_t)nrd;
401 _warc_skip(struct archive_read *a)
403 struct warc_s *w = a->format->data;
405 __archive_read_consume(a, w->cntlen + 4U/*\r\n\r\n separator*/);
412 /* private routines */
414 deconst(const void *c)
416 return (char *)0x1 + (((const char *)c) - (const char *)0x1);
420 xmemmem(const char *hay, const size_t haysize,
421 const char *needle, const size_t needlesize)
423 const char *const eoh = hay + haysize;
424 const char *const eon = needle + needlesize;
432 /* trivial checks first
433 * a 0-sized needle is defined to be found anywhere in haystack
434 * then run strchr() to find a candidate in HAYSTACK (i.e. a portion
435 * that happens to begin with *NEEDLE) */
436 if (needlesize == 0UL) {
438 } else if ((hay = memchr(hay, *needle, haysize)) == NULL) {
443 /* First characters of haystack and needle are the same now. Both are
444 * guaranteed to be at least one character long. Now computes the sum
445 * of characters values of needle together with the sum of the first
446 * needle_len characters of haystack. */
447 for (hp = hay + 1U, np = needle + 1U, hsum = *hay, nsum = *hay, eqp = 1U;
448 hp < eoh && np < eon;
449 hsum ^= *hp, nsum ^= *np, eqp &= *hp == *np, hp++, np++);
451 /* HP now references the (NEEDLESIZE + 1)-th character. */
453 /* haystack is smaller than needle, :O */
460 /* now loop through the rest of haystack,
461 * updating the sum iteratively */
462 for (cand = hay; hp < eoh; hp++) {
466 /* Since the sum of the characters is already known to be
467 * equal at that point, it is enough to check just NEEDLESIZE - 1
468 * characters for equality,
469 * also CAND is by design < HP, so no need for range checks */
470 if (hsum == nsum && memcmp(cand, needle, needlesize - 1U) == 0) {
471 return deconst(cand);
478 strtoi_lim(const char *str, const char **ep, int llim, int ulim)
482 /* we keep track of the number of digits via rulim */
485 for (sp = str, rulim = ulim > 10 ? ulim : 10;
486 res * 10 <= ulim && rulim && *sp >= '0' && *sp <= '9';
493 } else if (res < llim || res > ulim) {
496 *ep = (const char*)sp;
501 time_from_tm(struct tm *t)
504 /* Use platform timegm() if available. */
506 #elif HAVE__MKGMTIME64
507 return (_mkgmtime64(t));
509 /* Else use direct calculation using POSIX assumptions. */
510 /* First, fix up tm_yday based on the year/month/day. */
511 if (mktime(t) == (time_t)-1)
513 /* Then we can compute timegm() from first principles. */
518 + (t->tm_year - 70) * 31536000
519 + ((t->tm_year - 69) / 4) * 86400
520 - ((t->tm_year - 1) / 100) * 86400
521 + ((t->tm_year + 299) / 400) * 86400);
526 xstrpisotime(const char *s, char **endptr)
528 /** like strptime() but strictly for ISO 8601 Zulu strings */
530 time_t res = (time_t)-1;
532 /* make sure tm is clean */
533 memset(&tm, 0, sizeof(tm));
535 /* as a courtesy to our callers, and since this is a non-standard
536 * routine, we skip leading whitespace */
537 while (isspace((unsigned char)*s))
541 if ((tm.tm_year = strtoi_lim(s, &s, 1583, 4095)) < 0 || *s++ != '-') {
545 if ((tm.tm_mon = strtoi_lim(s, &s, 1, 12)) < 0 || *s++ != '-') {
548 /* read day-of-month */
549 if ((tm.tm_mday = strtoi_lim(s, &s, 1, 31)) < 0 || *s++ != 'T') {
553 if ((tm.tm_hour = strtoi_lim(s, &s, 0, 23)) < 0 || *s++ != ':') {
557 if ((tm.tm_min = strtoi_lim(s, &s, 0, 59)) < 0 || *s++ != ':') {
561 if ((tm.tm_sec = strtoi_lim(s, &s, 0, 60)) < 0 || *s++ != 'Z') {
565 /* massage TM to fulfill some of POSIX' contraints */
569 /* now convert our custom tm struct to a unix stamp using UTC */
570 res = time_from_tm(&tm);
573 if (endptr != NULL) {
574 *endptr = deconst(s);
580 _warc_rdver(const char buf[10], size_t bsz)
582 static const char magic[] = "WARC/";
585 (void)bsz; /* UNUSED */
587 if (memcmp(buf, magic, sizeof(magic) - 1U) != 0) {
591 /* looks good so far, read the version number for a laugh */
592 buf += sizeof(magic) - 1U;
593 /* most common case gets a quick-check here */
594 if (memcmp(buf, "1.0\r\n", 5U) == 0) {
607 if (buf[1U] == '.') {
610 /* set up major version */
611 ver = (buf[0U] - '0') * 10000U;
612 /* minor version, anyone? */
613 ver += (strtol(buf + 2U, &on, 10)) * 100U;
614 /* don't parse anything else */
622 /* just make the version ridiculously high */
631 _warc_rdtyp(const char *buf, size_t bsz)
633 static const char _key[] = "\r\nWARC-Type:";
634 const char *const eob = buf + bsz;
637 if ((val = xmemmem(buf, bsz, _key, sizeof(_key) - 1U)) == NULL) {
641 /* overread whitespace */
642 val += sizeof(_key) - 1U;
643 while (val < eob && isspace((unsigned char)*val))
646 if (val + 8U > eob) {
648 } else if (memcmp(val, "resource", 8U) == 0) {
650 } else if (memcmp(val, "warcinfo", 8U) == 0) {
652 } else if (memcmp(val, "metadata", 8U) == 0) {
654 } else if (memcmp(val, "request", 7U) == 0) {
656 } else if (memcmp(val, "response", 8U) == 0) {
658 } else if (memcmp(val, "conversi", 8U) == 0) {
660 } else if (memcmp(val, "continua", 8U) == 0) {
667 _warc_rduri(const char *buf, size_t bsz)
669 static const char _key[] = "\r\nWARC-Target-URI:";
670 const char *const eob = buf + bsz;
674 warc_string_t res = {0U, NULL};
676 if ((val = xmemmem(buf, bsz, _key, sizeof(_key) - 1U)) == NULL) {
680 /* overread whitespace */
681 val += sizeof(_key) - 1U;
682 while (val < eob && isspace((unsigned char)*val))
685 /* overread URL designators */
686 if ((uri = xmemmem(val, eob - val, "://", 3U)) == NULL) {
687 /* not touching that! */
689 } else if ((eol = memchr(uri, '\n', eob - uri)) == NULL) {
690 /* no end of line? :O */
694 /* massage uri to point to after :// */
696 /* also massage eol to point to the first whitespace
697 * after the last non-whitespace character before
698 * the end of the line */
699 while (eol > uri && isspace((unsigned char)eol[-1]))
702 /* now then, inspect the URI */
703 if (memcmp(val, "file", 4U) == 0) {
704 /* perfect, nothing left to do here */
706 } else if (memcmp(val, "http", 4U) == 0 ||
707 memcmp(val, "ftp", 3U) == 0) {
708 /* overread domain, and the first / */
709 while (uri < eol && *uri++ != '/');
711 /* not sure what to do? best to bugger off */
720 _warc_rdlen(const char *buf, size_t bsz)
722 static const char _key[] = "\r\nContent-Length:";
727 if ((val = xmemmem(buf, bsz, _key, sizeof(_key) - 1U)) == NULL) {
732 /* strtol kindly overreads whitespace for us, so use that */
733 val += sizeof(_key) - 1U;
734 len = strtol(val, &on, 10);
735 if (on == NULL || !isspace((unsigned char)*on)) {
736 /* hm, can we trust that number? Best not. */
743 _warc_rdrtm(const char *buf, size_t bsz)
745 static const char _key[] = "\r\nWARC-Date:";
750 if ((val = xmemmem(buf, bsz, _key, sizeof(_key) - 1U)) == NULL) {
755 /* xstrpisotime() kindly overreads whitespace for us, so use that */
756 val += sizeof(_key) - 1U;
757 res = xstrpisotime(val, &on);
758 if (on == NULL || !isspace((unsigned char)*on)) {
759 /* hm, can we trust that number? Best not. */
766 _warc_rdmtm(const char *buf, size_t bsz)
768 static const char _key[] = "\r\nLast-Modified:";
773 if ((val = xmemmem(buf, bsz, _key, sizeof(_key) - 1U)) == NULL) {
778 /* xstrpisotime() kindly overreads whitespace for us, so use that */
779 val += sizeof(_key) - 1U;
780 res = xstrpisotime(val, &on);
781 if (on == NULL || !isspace((unsigned char)*on)) {
782 /* hm, can we trust that number? Best not. */
789 _warc_find_eoh(const char *buf, size_t bsz)
791 static const char _marker[] = "\r\n\r\n";
792 const char *hit = xmemmem(buf, bsz, _marker, sizeof(_marker) - 1U);
795 hit += sizeof(_marker) - 1U;
800 /* archive_read_support_format_warc.c ends here */