]> git.ipfire.org Git - thirdparty/squid.git/blob - src/HttpHeader.cc
Source Format Enforcement (#1046)
[thirdparty/squid.git] / src / HttpHeader.cc
1 /*
2 * Copyright (C) 1996-2022 The Squid Software Foundation and contributors
3 *
4 * Squid software is distributed under GPLv2+ license and includes
5 * contributions from numerous individuals and organizations.
6 * Please see the COPYING and CONTRIBUTORS files for details.
7 */
8
9 /* DEBUG: section 55 HTTP Header */
10
11 #include "squid.h"
12 #include "base/CharacterSet.h"
13 #include "base/EnumIterator.h"
14 #include "base/Raw.h"
15 #include "base64.h"
16 #include "globals.h"
17 #include "http/ContentLengthInterpreter.h"
18 #include "HttpHdrCc.h"
19 #include "HttpHdrContRange.h"
20 #include "HttpHdrScTarget.h" // also includes HttpHdrSc.h
21 #include "HttpHeader.h"
22 #include "HttpHeaderFieldInfo.h"
23 #include "HttpHeaderStat.h"
24 #include "HttpHeaderTools.h"
25 #include "MemBuf.h"
26 #include "mgr/Registration.h"
27 #include "mime_header.h"
28 #include "sbuf/StringConvert.h"
29 #include "SquidConfig.h"
30 #include "StatHist.h"
31 #include "Store.h"
32 #include "StrList.h"
33 #include "time/gadgets.h"
34 #include "TimeOrTag.h"
35 #include "util.h"
36
37 #include <algorithm>
38 #include <array>
39
40 /* XXX: the whole set of API managing the entries vector should be rethought
41 * after the parse4r-ng effort is complete.
42 */
43
44 /*
45 * On naming conventions:
46 *
47 * HTTP/1.1 defines message-header as
48 *
49 * message-header = field-name ":" [ field-value ] CRLF
50 * field-name = token
51 * field-value = *( field-content | LWS )
52 *
53 * HTTP/1.1 does not give a name name a group of all message-headers in a message.
54 * Squid 1.1 seems to refer to that group _plus_ start-line as "headers".
55 *
56 * HttpHeader is an object that represents all message-headers in a message.
57 * HttpHeader does not manage start-line.
58 *
59 * HttpHeader is implemented as a collection of header "entries".
60 * An entry is a (field_id, field_name, field_value) triplet.
61 */
62
63 /*
64 * local constants and vars
65 */
66
67 // statistics counters for headers. clients must not allow Http::HdrType::BAD_HDR to be counted
68 std::vector<HttpHeaderFieldStat> headerStatsTable(Http::HdrType::enumEnd_);
69
70 /* request-only headers. Used for cachemgr */
71 static HttpHeaderMask RequestHeadersMask; /* set run-time using RequestHeaders */
72
73 /* reply-only headers. Used for cachemgr */
74 static HttpHeaderMask ReplyHeadersMask; /* set run-time using ReplyHeaders */
75
76 /* header accounting */
77 // NP: keep in sync with enum http_hdr_owner_type
78 static std::array<HttpHeaderStat, hoEnd> HttpHeaderStats = {
79 HttpHeaderStat(/*hoNone*/ "all", NULL),
80 #if USE_HTCP
81 HttpHeaderStat(/*hoHtcpReply*/ "HTCP reply", &ReplyHeadersMask),
82 #endif
83 HttpHeaderStat(/*hoRequest*/ "request", &RequestHeadersMask),
84 HttpHeaderStat(/*hoReply*/ "reply", &ReplyHeadersMask)
85 #if USE_OPENSSL
86 , HttpHeaderStat(/*hoErrorDetail*/ "error detail templates", nullptr)
87 #endif
88 /* hoEnd */
89 };
90
91 static int HeaderEntryParsedCount = 0;
92
93 /*
94 * forward declarations and local routines
95 */
96
97 class StoreEntry;
98
99 // update parse statistics for header id; if error is true also account
100 // for errors and write to debug log what happened
101 static void httpHeaderNoteParsedEntry(Http::HdrType id, String const &value, bool error);
102 static void httpHeaderStatDump(const HttpHeaderStat * hs, StoreEntry * e);
103 /** store report about current header usage and other stats */
104 static void httpHeaderStoreReport(StoreEntry * e);
105
106 /*
107 * Module initialization routines
108 */
109
110 static void
111 httpHeaderRegisterWithCacheManager(void)
112 {
113 Mgr::RegisterAction("http_headers",
114 "HTTP Header Statistics",
115 httpHeaderStoreReport, 0, 1);
116 }
117
118 void
119 httpHeaderInitModule(void)
120 {
121 /* check that we have enough space for masks */
122 assert(8 * sizeof(HttpHeaderMask) >= Http::HdrType::enumEnd_);
123
124 // masks are needed for stats page still
125 for (auto h : WholeEnum<Http::HdrType>()) {
126 if (Http::HeaderLookupTable.lookup(h).request)
127 CBIT_SET(RequestHeadersMask,h);
128 if (Http::HeaderLookupTable.lookup(h).reply)
129 CBIT_SET(ReplyHeadersMask,h);
130 }
131
132 assert(HttpHeaderStats[0].label && "httpHeaderInitModule() called via main()");
133 assert(HttpHeaderStats[hoEnd-1].label && "HttpHeaderStats created with all elements");
134
135 /* init dependent modules */
136 httpHdrCcInitModule();
137 httpHdrScInitModule();
138
139 httpHeaderRegisterWithCacheManager();
140 }
141
142 /*
143 * HttpHeader Implementation
144 */
145
146 HttpHeader::HttpHeader() : owner (hoNone), len (0), conflictingContentLength_(false)
147 {
148 entries.reserve(32);
149 httpHeaderMaskInit(&mask, 0);
150 }
151
152 HttpHeader::HttpHeader(const http_hdr_owner_type anOwner): owner(anOwner), len(0), conflictingContentLength_(false)
153 {
154 assert(anOwner > hoNone && anOwner < hoEnd);
155 debugs(55, 7, "init-ing hdr: " << this << " owner: " << owner);
156 entries.reserve(32);
157 httpHeaderMaskInit(&mask, 0);
158 }
159
160 // XXX: Delete as unused, expensive, and violating copy semantics by skipping Warnings
161 HttpHeader::HttpHeader(const HttpHeader &other): owner(other.owner), len(other.len), conflictingContentLength_(false)
162 {
163 entries.reserve(other.entries.capacity());
164 httpHeaderMaskInit(&mask, 0);
165 update(&other); // will update the mask as well
166 }
167
168 HttpHeader::~HttpHeader()
169 {
170 clean();
171 }
172
173 // XXX: Delete as unused, expensive, and violating assignment semantics by skipping Warnings
174 HttpHeader &
175 HttpHeader::operator =(const HttpHeader &other)
176 {
177 if (this != &other) {
178 // we do not really care, but the caller probably does
179 assert(owner == other.owner);
180 clean();
181 update(&other); // will update the mask as well
182 len = other.len;
183 conflictingContentLength_ = other.conflictingContentLength_;
184 teUnsupported_ = other.teUnsupported_;
185 }
186 return *this;
187 }
188
189 void
190 HttpHeader::clean()
191 {
192
193 assert(owner > hoNone && owner < hoEnd);
194 debugs(55, 7, "cleaning hdr: " << this << " owner: " << owner);
195
196 if (owner <= hoReply) {
197 /*
198 * An unfortunate bug. The entries array is initialized
199 * such that count is set to zero. httpHeaderClean() seems to
200 * be called both when 'hdr' is created, and destroyed. Thus,
201 * we accumulate a large number of zero counts for 'hdr' before
202 * it is ever used. Can't think of a good way to fix it, except
203 * adding a state variable that indicates whether or not 'hdr'
204 * has been used. As a hack, just never count zero-sized header
205 * arrays.
206 */
207 if (!entries.empty())
208 HttpHeaderStats[owner].hdrUCountDistr.count(entries.size());
209
210 ++ HttpHeaderStats[owner].destroyedCount;
211
212 HttpHeaderStats[owner].busyDestroyedCount += entries.size() > 0;
213 } // if (owner <= hoReply)
214
215 for (HttpHeaderEntry *e : entries) {
216 if (e == nullptr)
217 continue;
218 if (!Http::any_valid_header(e->id)) {
219 debugs(55, DBG_CRITICAL, "ERROR: Squid BUG: invalid entry (" << e->id << "). Ignored.");
220 } else {
221 if (owner <= hoReply)
222 HttpHeaderStats[owner].fieldTypeDistr.count(e->id);
223 delete e;
224 }
225 }
226
227 entries.clear();
228 httpHeaderMaskInit(&mask, 0);
229 len = 0;
230 conflictingContentLength_ = false;
231 teUnsupported_ = false;
232 }
233
234 /* append entries (also see httpHeaderUpdate) */
235 void
236 HttpHeader::append(const HttpHeader * src)
237 {
238 assert(src);
239 assert(src != this);
240 debugs(55, 7, "appending hdr: " << this << " += " << src);
241
242 for (auto e : src->entries) {
243 if (e)
244 addEntry(e->clone());
245 }
246 }
247
248 bool
249 HttpHeader::needUpdate(HttpHeader const *fresh) const
250 {
251 // our 1xx Warnings must be removed
252 for (const auto e: entries) {
253 // TODO: Move into HttpHeaderEntry::is1xxWarning() before official commit.
254 if (e && e->id == Http::HdrType::WARNING && (e->getInt()/100 == 1))
255 return true;
256 }
257
258 for (const auto e: fresh->entries) {
259 if (!e || skipUpdateHeader(e->id))
260 continue;
261 String value;
262 if (!hasNamed(e->name, &value) ||
263 (value != fresh->getByName(e->name)))
264 return true;
265 }
266 return false;
267 }
268
269 void
270 HttpHeader::updateWarnings()
271 {
272 int count = 0;
273 HttpHeaderPos pos = HttpHeaderInitPos;
274
275 // RFC 7234, section 4.3.4: delete 1xx warnings and retain 2xx warnings
276 while (HttpHeaderEntry *e = getEntry(&pos)) {
277 if (e->id == Http::HdrType::WARNING && (e->getInt()/100 == 1) )
278 delAt(pos, count);
279 }
280 }
281
282 bool
283 HttpHeader::skipUpdateHeader(const Http::HdrType id) const
284 {
285 return
286 // RFC 7234, section 4.3.4: use header fields other than Warning
287 (id == Http::HdrType::WARNING) ||
288 // TODO: Consider updating Vary headers after comparing the magnitude of
289 // the required changes (and/or cache losses) with compliance gains.
290 (id == Http::HdrType::VARY);
291 }
292
293 void
294 HttpHeader::update(HttpHeader const *fresh)
295 {
296 assert(fresh);
297 assert(this != fresh);
298
299 updateWarnings();
300
301 const HttpHeaderEntry *e;
302 HttpHeaderPos pos = HttpHeaderInitPos;
303
304 while ((e = fresh->getEntry(&pos))) {
305 /* deny bad guys (ok to check for Http::HdrType::OTHER) here */
306
307 if (skipUpdateHeader(e->id))
308 continue;
309
310 if (e->id != Http::HdrType::OTHER)
311 delById(e->id);
312 else
313 delByName(e->name);
314 }
315
316 pos = HttpHeaderInitPos;
317 while ((e = fresh->getEntry(&pos))) {
318 /* deny bad guys (ok to check for Http::HdrType::OTHER) here */
319
320 if (skipUpdateHeader(e->id))
321 continue;
322
323 debugs(55, 7, "Updating header '" << Http::HeaderLookupTable.lookup(e->id).name << "' in cached entry");
324
325 addEntry(e->clone());
326 }
327 }
328
329 bool
330 HttpHeader::Isolate(const char **parse_start, size_t l, const char **blk_start, const char **blk_end)
331 {
332 /*
333 * parse_start points to the first line of HTTP message *headers*,
334 * not including the request or status lines
335 */
336 const size_t end = headersEnd(*parse_start, l);
337
338 if (end) {
339 *blk_start = *parse_start;
340 *blk_end = *parse_start + end - 1;
341 assert(**blk_end == '\n');
342 // Point blk_end to the first character after the last header field.
343 // In other words, blk_end should point to the CR?LF header terminator.
344 if (end > 1 && *(*blk_end - 1) == '\r')
345 --(*blk_end);
346 *parse_start += end;
347 }
348 return end;
349 }
350
351 int
352 HttpHeader::parse(const char *buf, size_t buf_len, bool atEnd, size_t &hdr_sz, Http::ContentLengthInterpreter &clen)
353 {
354 const char *parse_start = buf;
355 const char *blk_start, *blk_end;
356 hdr_sz = 0;
357
358 if (!Isolate(&parse_start, buf_len, &blk_start, &blk_end)) {
359 // XXX: do not parse non-isolated headers even if the connection is closed.
360 // Treat unterminated headers as "partial headers" framing errors.
361 if (!atEnd)
362 return 0;
363 blk_start = parse_start;
364 blk_end = blk_start + strlen(blk_start);
365 }
366
367 if (parse(blk_start, blk_end - blk_start, clen)) {
368 hdr_sz = parse_start - buf;
369 return 1;
370 }
371 return -1;
372 }
373
374 // XXX: callers treat this return as boolean.
375 // XXX: A better mechanism is needed to signal different types of error.
376 // lexicon, syntax, semantics, validation, access policy - are all (ab)using 'return 0'
377 int
378 HttpHeader::parse(const char *header_start, size_t hdrLen, Http::ContentLengthInterpreter &clen)
379 {
380 const char *field_ptr = header_start;
381 const char *header_end = header_start + hdrLen; // XXX: remove
382 int warnOnError = (Config.onoff.relaxed_header_parser <= 0 ? DBG_IMPORTANT : 2);
383
384 assert(header_start && header_end);
385 debugs(55, 7, "parsing hdr: (" << this << ")" << std::endl << getStringPrefix(header_start, hdrLen));
386 ++ HttpHeaderStats[owner].parsedCount;
387
388 char *nulpos;
389 if ((nulpos = (char*)memchr(header_start, '\0', hdrLen))) {
390 debugs(55, DBG_IMPORTANT, "WARNING: HTTP header contains NULL characters {" <<
391 getStringPrefix(header_start, nulpos-header_start) << "}\nNULL\n{" << getStringPrefix(nulpos+1, hdrLen-(nulpos-header_start)-1));
392 clean();
393 return 0;
394 }
395
396 /* common format headers are "<name>:[ws]<value>" lines delimited by <CRLF>.
397 * continuation lines start with a (single) space or tab */
398 while (field_ptr < header_end) {
399 const char *field_start = field_ptr;
400 const char *field_end;
401
402 const char *hasBareCr = nullptr;
403 size_t lines = 0;
404 do {
405 const char *this_line = field_ptr;
406 field_ptr = (const char *)memchr(field_ptr, '\n', header_end - field_ptr);
407 ++lines;
408
409 if (!field_ptr) {
410 // missing <LF>
411 clean();
412 return 0;
413 }
414
415 field_end = field_ptr;
416
417 ++field_ptr; /* Move to next line */
418
419 if (field_end > this_line && field_end[-1] == '\r') {
420 --field_end; /* Ignore CR LF */
421
422 if (owner == hoRequest && field_end > this_line) {
423 bool cr_only = true;
424 for (const char *p = this_line; p < field_end && cr_only; ++p) {
425 if (*p != '\r')
426 cr_only = false;
427 }
428 if (cr_only) {
429 debugs(55, DBG_IMPORTANT, "SECURITY WARNING: Rejecting HTTP request with a CR+ "
430 "header field to prevent request smuggling attacks: {" <<
431 getStringPrefix(header_start, hdrLen) << "}");
432 clean();
433 return 0;
434 }
435 }
436 }
437
438 /* Barf on stray CR characters */
439 if (memchr(this_line, '\r', field_end - this_line)) {
440 hasBareCr = "bare CR";
441 debugs(55, warnOnError, "WARNING: suspicious CR characters in HTTP header {" <<
442 getStringPrefix(field_start, field_end-field_start) << "}");
443
444 if (Config.onoff.relaxed_header_parser) {
445 char *p = (char *) this_line; /* XXX Warning! This destroys original header content and violates specifications somewhat */
446
447 while ((p = (char *)memchr(p, '\r', field_end - p)) != NULL) {
448 *p = ' ';
449 ++p;
450 }
451 } else {
452 clean();
453 return 0;
454 }
455 }
456
457 if (this_line + 1 == field_end && this_line > field_start) {
458 debugs(55, warnOnError, "WARNING: Blank continuation line in HTTP header {" <<
459 getStringPrefix(header_start, hdrLen) << "}");
460 clean();
461 return 0;
462 }
463 } while (field_ptr < header_end && (*field_ptr == ' ' || *field_ptr == '\t'));
464
465 if (field_start == field_end) {
466 if (field_ptr < header_end) {
467 debugs(55, warnOnError, "WARNING: unparsable HTTP header field near {" <<
468 getStringPrefix(field_start, hdrLen-(field_start-header_start)) << "}");
469 clean();
470 return 0;
471 }
472
473 break; /* terminating blank line */
474 }
475
476 const auto e = HttpHeaderEntry::parse(field_start, field_end, owner);
477 if (!e) {
478 debugs(55, warnOnError, "WARNING: unparsable HTTP header field {" <<
479 getStringPrefix(field_start, field_end-field_start) << "}");
480 debugs(55, warnOnError, " in {" << getStringPrefix(header_start, hdrLen) << "}");
481
482 clean();
483 return 0;
484 }
485
486 if (lines > 1 || hasBareCr) {
487 const auto framingHeader = (e->id == Http::HdrType::CONTENT_LENGTH || e->id == Http::HdrType::TRANSFER_ENCODING);
488 if (framingHeader) {
489 if (!hasBareCr) // already warned about bare CRs
490 debugs(55, warnOnError, "WARNING: obs-fold in framing-sensitive " << e->name << ": " << e->value);
491 delete e;
492 clean();
493 return 0;
494 }
495 }
496
497 if (e->id == Http::HdrType::CONTENT_LENGTH && !clen.checkField(e->value)) {
498 delete e;
499
500 if (Config.onoff.relaxed_header_parser)
501 continue; // clen has printed any necessary warnings
502
503 clean();
504 return 0;
505 }
506
507 addEntry(e);
508 }
509
510 if (clen.headerWideProblem) {
511 debugs(55, warnOnError, "WARNING: " << clen.headerWideProblem <<
512 " Content-Length field values in" <<
513 Raw("header", header_start, hdrLen));
514 }
515
516 String rawTe;
517 if (clen.prohibitedAndIgnored()) {
518 // prohibitedAndIgnored() includes trailer header blocks
519 // being parsed as a case to forbid/ignore these headers.
520
521 // RFC 7230 section 3.3.2: A server MUST NOT send a Content-Length
522 // header field in any response with a status code of 1xx (Informational)
523 // or 204 (No Content). And RFC 7230 3.3.3#1 tells recipients to ignore
524 // such Content-Lengths.
525 if (delById(Http::HdrType::CONTENT_LENGTH))
526 debugs(55, 3, "Content-Length is " << clen.prohibitedAndIgnored());
527
528 // The same RFC 7230 3.3.3#1-based logic applies to Transfer-Encoding
529 // banned by RFC 7230 section 3.3.1.
530 if (delById(Http::HdrType::TRANSFER_ENCODING))
531 debugs(55, 3, "Transfer-Encoding is " << clen.prohibitedAndIgnored());
532
533 } else if (getByIdIfPresent(Http::HdrType::TRANSFER_ENCODING, &rawTe)) {
534 // RFC 2616 section 4.4: ignore Content-Length with Transfer-Encoding
535 // RFC 7230 section 3.3.3 #3: Transfer-Encoding overwrites Content-Length
536 delById(Http::HdrType::CONTENT_LENGTH);
537 // and clen state becomes irrelevant
538
539 if (rawTe.caseCmp("chunked") == 0) {
540 ; // leave header present for chunked() method
541 } else if (rawTe.caseCmp("identity") == 0) { // deprecated. no coding
542 delById(Http::HdrType::TRANSFER_ENCODING);
543 } else {
544 // This also rejects multiple encodings until we support them properly.
545 debugs(55, warnOnError, "WARNING: unsupported Transfer-Encoding used by client: " << rawTe);
546 teUnsupported_ = true;
547 }
548
549 } else if (clen.sawBad) {
550 // ensure our callers do not accidentally see bad Content-Length values
551 delById(Http::HdrType::CONTENT_LENGTH);
552 conflictingContentLength_ = true; // TODO: Rename to badContentLength_.
553 } else if (clen.needsSanitizing) {
554 // RFC 7230 section 3.3.2: MUST either reject or ... [sanitize];
555 // ensure our callers see a clean Content-Length value or none at all
556 delById(Http::HdrType::CONTENT_LENGTH);
557 if (clen.sawGood) {
558 putInt64(Http::HdrType::CONTENT_LENGTH, clen.value);
559 debugs(55, 5, "sanitized Content-Length to be " << clen.value);
560 }
561 }
562
563 return 1; /* even if no fields where found, it is a valid header */
564 }
565
566 /* packs all the entries using supplied packer */
567 void
568 HttpHeader::packInto(Packable * p, bool mask_sensitive_info) const
569 {
570 HttpHeaderPos pos = HttpHeaderInitPos;
571 const HttpHeaderEntry *e;
572 assert(p);
573 debugs(55, 7, this << " into " << p <<
574 (mask_sensitive_info ? " while masking" : ""));
575 /* pack all entries one by one */
576 while ((e = getEntry(&pos))) {
577 if (!mask_sensitive_info) {
578 e->packInto(p);
579 continue;
580 }
581
582 bool maskThisEntry = false;
583 switch (e->id) {
584 case Http::HdrType::AUTHORIZATION:
585 case Http::HdrType::PROXY_AUTHORIZATION:
586 maskThisEntry = true;
587 break;
588
589 case Http::HdrType::FTP_ARGUMENTS:
590 if (const HttpHeaderEntry *cmd = findEntry(Http::HdrType::FTP_COMMAND))
591 maskThisEntry = (cmd->value == "PASS");
592 break;
593
594 default:
595 break;
596 }
597 if (maskThisEntry) {
598 p->append(e->name.rawContent(), e->name.length());
599 p->append(": ** NOT DISPLAYED **\r\n", 23);
600 } else {
601 e->packInto(p);
602 }
603
604 }
605 /* Pack in the "special" entries */
606
607 /* Cache-Control */
608 }
609
610 /* returns next valid entry */
611 HttpHeaderEntry *
612 HttpHeader::getEntry(HttpHeaderPos * pos) const
613 {
614 assert(pos);
615 assert(*pos >= HttpHeaderInitPos && *pos < static_cast<ssize_t>(entries.size()));
616
617 for (++(*pos); *pos < static_cast<ssize_t>(entries.size()); ++(*pos)) {
618 if (entries[*pos])
619 return static_cast<HttpHeaderEntry*>(entries[*pos]);
620 }
621
622 return NULL;
623 }
624
625 /*
626 * returns a pointer to a specified entry if any
627 * note that we return one entry so it does not make much sense to ask for
628 * "list" headers
629 */
630 HttpHeaderEntry *
631 HttpHeader::findEntry(Http::HdrType id) const
632 {
633 assert(any_registered_header(id));
634 assert(!Http::HeaderLookupTable.lookup(id).list);
635
636 /* check mask first */
637
638 if (!CBIT_TEST(mask, id))
639 return NULL;
640
641 /* looks like we must have it, do linear search */
642 for (auto e : entries) {
643 if (e && e->id == id)
644 return e;
645 }
646
647 /* hm.. we thought it was there, but it was not found */
648 assert(false);
649 return nullptr; /* not reached */
650 }
651
652 /*
653 * same as httpHeaderFindEntry
654 */
655 HttpHeaderEntry *
656 HttpHeader::findLastEntry(Http::HdrType id) const
657 {
658 assert(any_registered_header(id));
659 assert(!Http::HeaderLookupTable.lookup(id).list);
660
661 /* check mask first */
662 if (!CBIT_TEST(mask, id))
663 return NULL;
664
665 for (auto e = entries.rbegin(); e != entries.rend(); ++e) {
666 if (*e && (*e)->id == id)
667 return *e;
668 }
669
670 /* hm.. we thought it was there, but it was not found */
671 assert(false);
672 return nullptr; /* not reached */
673 }
674
675 int
676 HttpHeader::delByName(const SBuf &name)
677 {
678 int count = 0;
679 HttpHeaderPos pos = HttpHeaderInitPos;
680 httpHeaderMaskInit(&mask, 0); /* temporal inconsistency */
681 debugs(55, 9, "deleting '" << name << "' fields in hdr " << this);
682
683 while (const HttpHeaderEntry *e = getEntry(&pos)) {
684 if (!e->name.caseCmp(name))
685 delAt(pos, count);
686 else
687 CBIT_SET(mask, e->id);
688 }
689
690 return count;
691 }
692
693 /* deletes all entries with a given id, returns the #entries deleted */
694 int
695 HttpHeader::delById(Http::HdrType id)
696 {
697 debugs(55, 8, this << " del-by-id " << id);
698 assert(any_registered_header(id));
699
700 if (!CBIT_TEST(mask, id))
701 return 0;
702
703 int count = 0;
704
705 HttpHeaderPos pos = HttpHeaderInitPos;
706 while (HttpHeaderEntry *e = getEntry(&pos)) {
707 if (e->id == id)
708 delAt(pos, count); // deletes e
709 }
710
711 CBIT_CLR(mask, id);
712 assert(count);
713 return count;
714 }
715
716 /*
717 * deletes an entry at pos and leaves a gap; leaving a gap makes it
718 * possible to iterate(search) and delete fields at the same time
719 * NOTE: Does not update the header mask. Caller must follow up with
720 * a call to refreshMask() if headers_deleted was incremented.
721 */
722 void
723 HttpHeader::delAt(HttpHeaderPos pos, int &headers_deleted)
724 {
725 HttpHeaderEntry *e;
726 assert(pos >= HttpHeaderInitPos && pos < static_cast<ssize_t>(entries.size()));
727 e = static_cast<HttpHeaderEntry*>(entries[pos]);
728 entries[pos] = NULL;
729 /* decrement header length, allow for ": " and crlf */
730 len -= e->name.length() + 2 + e->value.size() + 2;
731 assert(len >= 0);
732 delete e;
733 ++headers_deleted;
734 }
735
736 /*
737 * Compacts the header storage
738 */
739 void
740 HttpHeader::compact()
741 {
742 // TODO: optimize removal, or possibly make it so that's not needed.
743 entries.erase( std::remove(entries.begin(), entries.end(), nullptr),
744 entries.end());
745 }
746
747 /*
748 * Refreshes the header mask. Required after delAt() calls.
749 */
750 void
751 HttpHeader::refreshMask()
752 {
753 httpHeaderMaskInit(&mask, 0);
754 debugs(55, 7, "refreshing the mask in hdr " << this);
755 for (auto e : entries) {
756 if (e)
757 CBIT_SET(mask, e->id);
758 }
759 }
760
761 /* appends an entry;
762 * does not call e->clone() so one should not reuse "*e"
763 */
764 void
765 HttpHeader::addEntry(HttpHeaderEntry * e)
766 {
767 assert(e);
768 assert(any_HdrType_enum_value(e->id));
769 assert(e->name.length());
770
771 debugs(55, 7, this << " adding entry: " << e->id << " at " << entries.size());
772
773 if (e->id != Http::HdrType::BAD_HDR) {
774 if (CBIT_TEST(mask, e->id)) {
775 ++ headerStatsTable[e->id].repCount;
776 } else {
777 CBIT_SET(mask, e->id);
778 }
779 }
780
781 entries.push_back(e);
782
783 /* increment header length, allow for ": " and crlf */
784 len += e->name.length() + 2 + e->value.size() + 2;
785 }
786
787 /* inserts an entry;
788 * does not call e->clone() so one should not reuse "*e"
789 */
790 void
791 HttpHeader::insertEntry(HttpHeaderEntry * e)
792 {
793 assert(e);
794 assert(any_valid_header(e->id));
795
796 debugs(55, 7, this << " adding entry: " << e->id << " at " << entries.size());
797
798 // Http::HdrType::BAD_HDR is filtered out by assert_any_valid_header
799 if (CBIT_TEST(mask, e->id)) {
800 ++ headerStatsTable[e->id].repCount;
801 } else {
802 CBIT_SET(mask, e->id);
803 }
804
805 entries.insert(entries.begin(),e);
806
807 /* increment header length, allow for ": " and crlf */
808 len += e->name.length() + 2 + e->value.size() + 2;
809 }
810
811 bool
812 HttpHeader::getList(Http::HdrType id, String *s) const
813 {
814 debugs(55, 9, this << " joining for id " << id);
815 /* only fields from ListHeaders array can be "listed" */
816 assert(Http::HeaderLookupTable.lookup(id).list);
817
818 if (!CBIT_TEST(mask, id))
819 return false;
820
821 for (auto e: entries) {
822 if (e && e->id == id)
823 strListAdd(s, e->value.termedBuf(), ',');
824 }
825
826 /*
827 * note: we might get an empty (size==0) string if there was an "empty"
828 * header. This results in an empty length String, which may have a NULL
829 * buffer.
830 */
831 /* temporary warning: remove it? (Is it useful for diagnostics ?) */
832 if (!s->size())
833 debugs(55, 3, "empty list header: " << Http::HeaderLookupTable.lookup(id).name << "(" << id << ")");
834 else
835 debugs(55, 6, this << ": joined for id " << id << ": " << s);
836
837 return true;
838 }
839
840 /* return a list of entries with the same id separated by ',' and ws */
841 String
842 HttpHeader::getList(Http::HdrType id) const
843 {
844 HttpHeaderEntry *e;
845 HttpHeaderPos pos = HttpHeaderInitPos;
846 debugs(55, 9, this << "joining for id " << id);
847 /* only fields from ListHeaders array can be "listed" */
848 assert(Http::HeaderLookupTable.lookup(id).list);
849
850 if (!CBIT_TEST(mask, id))
851 return String();
852
853 String s;
854
855 while ((e = getEntry(&pos))) {
856 if (e->id == id)
857 strListAdd(&s, e->value.termedBuf(), ',');
858 }
859
860 /*
861 * note: we might get an empty (size==0) string if there was an "empty"
862 * header. This results in an empty length String, which may have a NULL
863 * buffer.
864 */
865 /* temporary warning: remove it? (Is it useful for diagnostics ?) */
866 if (!s.size())
867 debugs(55, 3, "empty list header: " << Http::HeaderLookupTable.lookup(id).name << "(" << id << ")");
868 else
869 debugs(55, 6, this << ": joined for id " << id << ": " << s);
870
871 return s;
872 }
873
874 /* return a string or list of entries with the same id separated by ',' and ws */
875 String
876 HttpHeader::getStrOrList(Http::HdrType id) const
877 {
878 HttpHeaderEntry *e;
879
880 if (Http::HeaderLookupTable.lookup(id).list)
881 return getList(id);
882
883 if ((e = findEntry(id)))
884 return e->value;
885
886 return String();
887 }
888
889 /*
890 * Returns the value of the specified header and/or an undefined String.
891 */
892 String
893 HttpHeader::getByName(const char *name) const
894 {
895 String result;
896 // ignore presence: return undefined string if an empty header is present
897 (void)hasNamed(name, strlen(name), &result);
898 return result;
899 }
900
901 String
902 HttpHeader::getByName(const SBuf &name) const
903 {
904 String result;
905 // ignore presence: return undefined string if an empty header is present
906 (void)hasNamed(name, &result);
907 return result;
908 }
909
910 String
911 HttpHeader::getById(Http::HdrType id) const
912 {
913 String result;
914 (void)getByIdIfPresent(id, &result);
915 return result;
916 }
917
918 bool
919 HttpHeader::hasNamed(const SBuf &s, String *result) const
920 {
921 return hasNamed(s.rawContent(), s.length(), result);
922 }
923
924 bool
925 HttpHeader::getByIdIfPresent(Http::HdrType id, String *result) const
926 {
927 if (id == Http::HdrType::BAD_HDR)
928 return false;
929 if (!has(id))
930 return false;
931 if (result)
932 *result = getStrOrList(id);
933 return true;
934 }
935
936 bool
937 HttpHeader::hasNamed(const char *name, unsigned int namelen, String *result) const
938 {
939 Http::HdrType id;
940 HttpHeaderPos pos = HttpHeaderInitPos;
941 HttpHeaderEntry *e;
942
943 assert(name);
944
945 /* First try the quick path */
946 id = Http::HeaderLookupTable.lookup(name,namelen).id;
947
948 if (id != Http::HdrType::BAD_HDR) {
949 if (getByIdIfPresent(id, result))
950 return true;
951 }
952
953 /* Sorry, an unknown header name. Do linear search */
954 bool found = false;
955 while ((e = getEntry(&pos))) {
956 if (e->id == Http::HdrType::OTHER && e->name.length() == namelen && e->name.caseCmp(name, namelen) == 0) {
957 found = true;
958 if (!result)
959 break;
960 strListAdd(result, e->value.termedBuf(), ',');
961 }
962 }
963
964 return found;
965 }
966
967 /*
968 * Returns a the value of the specified list member, if any.
969 */
970 SBuf
971 HttpHeader::getByNameListMember(const char *name, const char *member, const char separator) const
972 {
973 assert(name);
974 const auto header = getByName(name);
975 return ::getListMember(header, member, separator);
976 }
977
978 /*
979 * returns a the value of the specified list member, if any.
980 */
981 SBuf
982 HttpHeader::getListMember(Http::HdrType id, const char *member, const char separator) const
983 {
984 assert(any_registered_header(id));
985 const auto header = getStrOrList(id);
986 return ::getListMember(header, member, separator);
987 }
988
989 /* test if a field is present */
990 int
991 HttpHeader::has(Http::HdrType id) const
992 {
993 assert(any_registered_header(id));
994 debugs(55, 9, this << " lookup for " << id);
995 return CBIT_TEST(mask, id);
996 }
997
998 void
999 HttpHeader::addVia(const AnyP::ProtocolVersion &ver, const HttpHeader *from)
1000 {
1001 // TODO: do not add Via header for messages where Squid itself
1002 // generated the message (i.e., Downloader or ESI) there should be no Via header added at all.
1003
1004 if (Config.onoff.via) {
1005 SBuf buf;
1006 // RFC 7230 section 5.7.1.: protocol-name is omitted when
1007 // the received protocol is HTTP.
1008 if (ver.protocol > AnyP::PROTO_NONE && ver.protocol < AnyP::PROTO_UNKNOWN &&
1009 ver.protocol != AnyP::PROTO_HTTP && ver.protocol != AnyP::PROTO_HTTPS)
1010 buf.appendf("%s/", AnyP::ProtocolType_str[ver.protocol]);
1011 buf.appendf("%d.%d %s", ver.major, ver.minor, ThisCache);
1012 const HttpHeader *hdr = from ? from : this;
1013 SBuf strVia = StringToSBuf(hdr->getList(Http::HdrType::VIA));
1014 if (!strVia.isEmpty())
1015 strVia.append(", ", 2);
1016 strVia.append(buf);
1017 // XXX: putStr() still suffers from String size limits
1018 Must(strVia.length() < String::SizeMaxXXX());
1019 delById(Http::HdrType::VIA);
1020 putStr(Http::HdrType::VIA, strVia.c_str());
1021 }
1022 }
1023
1024 void
1025 HttpHeader::putInt(Http::HdrType id, int number)
1026 {
1027 assert(any_registered_header(id));
1028 assert(Http::HeaderLookupTable.lookup(id).type == Http::HdrFieldType::ftInt); /* must be of an appropriate type */
1029 assert(number >= 0);
1030 addEntry(new HttpHeaderEntry(id, SBuf(), xitoa(number)));
1031 }
1032
1033 void
1034 HttpHeader::putInt64(Http::HdrType id, int64_t number)
1035 {
1036 assert(any_registered_header(id));
1037 assert(Http::HeaderLookupTable.lookup(id).type == Http::HdrFieldType::ftInt64); /* must be of an appropriate type */
1038 assert(number >= 0);
1039 addEntry(new HttpHeaderEntry(id, SBuf(), xint64toa(number)));
1040 }
1041
1042 void
1043 HttpHeader::putTime(Http::HdrType id, time_t htime)
1044 {
1045 assert(any_registered_header(id));
1046 assert(Http::HeaderLookupTable.lookup(id).type == Http::HdrFieldType::ftDate_1123); /* must be of an appropriate type */
1047 assert(htime >= 0);
1048 addEntry(new HttpHeaderEntry(id, SBuf(), Time::FormatRfc1123(htime)));
1049 }
1050
1051 void
1052 HttpHeader::putStr(Http::HdrType id, const char *str)
1053 {
1054 assert(any_registered_header(id));
1055 assert(Http::HeaderLookupTable.lookup(id).type == Http::HdrFieldType::ftStr); /* must be of an appropriate type */
1056 assert(str);
1057 addEntry(new HttpHeaderEntry(id, SBuf(), str));
1058 }
1059
1060 void
1061 HttpHeader::putAuth(const char *auth_scheme, const char *realm)
1062 {
1063 assert(auth_scheme && realm);
1064 httpHeaderPutStrf(this, Http::HdrType::WWW_AUTHENTICATE, "%s realm=\"%s\"", auth_scheme, realm);
1065 }
1066
1067 void
1068 HttpHeader::putCc(const HttpHdrCc * cc)
1069 {
1070 assert(cc);
1071 /* remove old directives if any */
1072 delById(Http::HdrType::CACHE_CONTROL);
1073 /* pack into mb */
1074 MemBuf mb;
1075 mb.init();
1076 cc->packInto(&mb);
1077 /* put */
1078 addEntry(new HttpHeaderEntry(Http::HdrType::CACHE_CONTROL, SBuf(), mb.buf));
1079 /* cleanup */
1080 mb.clean();
1081 }
1082
1083 void
1084 HttpHeader::putContRange(const HttpHdrContRange * cr)
1085 {
1086 assert(cr);
1087 /* remove old directives if any */
1088 delById(Http::HdrType::CONTENT_RANGE);
1089 /* pack into mb */
1090 MemBuf mb;
1091 mb.init();
1092 httpHdrContRangePackInto(cr, &mb);
1093 /* put */
1094 addEntry(new HttpHeaderEntry(Http::HdrType::CONTENT_RANGE, SBuf(), mb.buf));
1095 /* cleanup */
1096 mb.clean();
1097 }
1098
1099 void
1100 HttpHeader::putRange(const HttpHdrRange * range)
1101 {
1102 assert(range);
1103 /* remove old directives if any */
1104 delById(Http::HdrType::RANGE);
1105 /* pack into mb */
1106 MemBuf mb;
1107 mb.init();
1108 range->packInto(&mb);
1109 /* put */
1110 addEntry(new HttpHeaderEntry(Http::HdrType::RANGE, SBuf(), mb.buf));
1111 /* cleanup */
1112 mb.clean();
1113 }
1114
1115 void
1116 HttpHeader::putSc(HttpHdrSc *sc)
1117 {
1118 assert(sc);
1119 /* remove old directives if any */
1120 delById(Http::HdrType::SURROGATE_CONTROL);
1121 /* pack into mb */
1122 MemBuf mb;
1123 mb.init();
1124 sc->packInto(&mb);
1125 /* put */
1126 addEntry(new HttpHeaderEntry(Http::HdrType::SURROGATE_CONTROL, SBuf(), mb.buf));
1127 /* cleanup */
1128 mb.clean();
1129 }
1130
1131 void
1132 HttpHeader::putWarning(const int code, const char *const text)
1133 {
1134 char buf[512];
1135 snprintf(buf, sizeof(buf), "%i %s \"%s\"", code, visible_appname_string, text);
1136 putStr(Http::HdrType::WARNING, buf);
1137 }
1138
1139 /* add extension header (these fields are not parsed/analyzed/joined, etc.) */
1140 void
1141 HttpHeader::putExt(const char *name, const char *value)
1142 {
1143 assert(name && value);
1144 debugs(55, 8, this << " adds ext entry " << name << " : " << value);
1145 addEntry(new HttpHeaderEntry(Http::HdrType::OTHER, SBuf(name), value));
1146 }
1147
1148 int
1149 HttpHeader::getInt(Http::HdrType id) const
1150 {
1151 assert(any_registered_header(id));
1152 assert(Http::HeaderLookupTable.lookup(id).type == Http::HdrFieldType::ftInt); /* must be of an appropriate type */
1153 HttpHeaderEntry *e;
1154
1155 if ((e = findEntry(id)))
1156 return e->getInt();
1157
1158 return -1;
1159 }
1160
1161 int64_t
1162 HttpHeader::getInt64(Http::HdrType id) const
1163 {
1164 assert(any_registered_header(id));
1165 assert(Http::HeaderLookupTable.lookup(id).type == Http::HdrFieldType::ftInt64); /* must be of an appropriate type */
1166 HttpHeaderEntry *e;
1167
1168 if ((e = findEntry(id)))
1169 return e->getInt64();
1170
1171 return -1;
1172 }
1173
1174 time_t
1175 HttpHeader::getTime(Http::HdrType id) const
1176 {
1177 HttpHeaderEntry *e;
1178 time_t value = -1;
1179 assert(any_registered_header(id));
1180 assert(Http::HeaderLookupTable.lookup(id).type == Http::HdrFieldType::ftDate_1123); /* must be of an appropriate type */
1181
1182 if ((e = findEntry(id))) {
1183 value = Time::ParseRfc1123(e->value.termedBuf());
1184 httpHeaderNoteParsedEntry(e->id, e->value, value < 0);
1185 }
1186
1187 return value;
1188 }
1189
1190 /* sync with httpHeaderGetLastStr */
1191 const char *
1192 HttpHeader::getStr(Http::HdrType id) const
1193 {
1194 HttpHeaderEntry *e;
1195 assert(any_registered_header(id));
1196 assert(Http::HeaderLookupTable.lookup(id).type == Http::HdrFieldType::ftStr); /* must be of an appropriate type */
1197
1198 if ((e = findEntry(id))) {
1199 httpHeaderNoteParsedEntry(e->id, e->value, false); /* no errors are possible */
1200 return e->value.termedBuf();
1201 }
1202
1203 return NULL;
1204 }
1205
1206 /* unusual */
1207 const char *
1208 HttpHeader::getLastStr(Http::HdrType id) const
1209 {
1210 HttpHeaderEntry *e;
1211 assert(any_registered_header(id));
1212 assert(Http::HeaderLookupTable.lookup(id).type == Http::HdrFieldType::ftStr); /* must be of an appropriate type */
1213
1214 if ((e = findLastEntry(id))) {
1215 httpHeaderNoteParsedEntry(e->id, e->value, false); /* no errors are possible */
1216 return e->value.termedBuf();
1217 }
1218
1219 return NULL;
1220 }
1221
1222 HttpHdrCc *
1223 HttpHeader::getCc() const
1224 {
1225 if (!CBIT_TEST(mask, Http::HdrType::CACHE_CONTROL))
1226 return NULL;
1227
1228 String s;
1229 getList(Http::HdrType::CACHE_CONTROL, &s);
1230
1231 HttpHdrCc *cc=new HttpHdrCc();
1232
1233 if (!cc->parse(s)) {
1234 delete cc;
1235 cc = NULL;
1236 }
1237
1238 ++ HttpHeaderStats[owner].ccParsedCount;
1239
1240 if (cc)
1241 httpHdrCcUpdateStats(cc, &HttpHeaderStats[owner].ccTypeDistr);
1242
1243 httpHeaderNoteParsedEntry(Http::HdrType::CACHE_CONTROL, s, !cc);
1244
1245 return cc;
1246 }
1247
1248 HttpHdrRange *
1249 HttpHeader::getRange() const
1250 {
1251 HttpHdrRange *r = NULL;
1252 HttpHeaderEntry *e;
1253 /* some clients will send "Request-Range" _and_ *matching* "Range"
1254 * who knows, some clients might send Request-Range only;
1255 * this "if" should work correctly in both cases;
1256 * hopefully no clients send mismatched headers! */
1257
1258 if ((e = findEntry(Http::HdrType::RANGE)) ||
1259 (e = findEntry(Http::HdrType::REQUEST_RANGE))) {
1260 r = HttpHdrRange::ParseCreate(&e->value);
1261 httpHeaderNoteParsedEntry(e->id, e->value, !r);
1262 }
1263
1264 return r;
1265 }
1266
1267 HttpHdrSc *
1268 HttpHeader::getSc() const
1269 {
1270 if (!CBIT_TEST(mask, Http::HdrType::SURROGATE_CONTROL))
1271 return NULL;
1272
1273 String s;
1274
1275 (void) getList(Http::HdrType::SURROGATE_CONTROL, &s);
1276
1277 HttpHdrSc *sc = httpHdrScParseCreate(s);
1278
1279 ++ HttpHeaderStats[owner].ccParsedCount;
1280
1281 if (sc)
1282 sc->updateStats(&HttpHeaderStats[owner].scTypeDistr);
1283
1284 httpHeaderNoteParsedEntry(Http::HdrType::SURROGATE_CONTROL, s, !sc);
1285
1286 return sc;
1287 }
1288
1289 HttpHdrContRange *
1290 HttpHeader::getContRange() const
1291 {
1292 HttpHdrContRange *cr = NULL;
1293 HttpHeaderEntry *e;
1294
1295 if ((e = findEntry(Http::HdrType::CONTENT_RANGE))) {
1296 cr = httpHdrContRangeParseCreate(e->value.termedBuf());
1297 httpHeaderNoteParsedEntry(e->id, e->value, !cr);
1298 }
1299
1300 return cr;
1301 }
1302
1303 SBuf
1304 HttpHeader::getAuthToken(Http::HdrType id, const char *auth_scheme) const
1305 {
1306 const char *field;
1307 int l;
1308 assert(auth_scheme);
1309 field = getStr(id);
1310
1311 static const SBuf nil;
1312 if (!field) /* no authorization field */
1313 return nil;
1314
1315 l = strlen(auth_scheme);
1316
1317 if (!l || strncasecmp(field, auth_scheme, l)) /* wrong scheme */
1318 return nil;
1319
1320 field += l;
1321
1322 if (!xisspace(*field)) /* wrong scheme */
1323 return nil;
1324
1325 /* skip white space */
1326 for (; field && xisspace(*field); ++field);
1327
1328 if (!*field) /* no authorization cookie */
1329 return nil;
1330
1331 const auto fieldLen = strlen(field);
1332 SBuf result;
1333 char *decodedAuthToken = result.rawAppendStart(BASE64_DECODE_LENGTH(fieldLen));
1334 struct base64_decode_ctx ctx;
1335 base64_decode_init(&ctx);
1336 size_t decodedLen = 0;
1337 if (!base64_decode_update(&ctx, &decodedLen, reinterpret_cast<uint8_t*>(decodedAuthToken), fieldLen, field) ||
1338 !base64_decode_final(&ctx)) {
1339 return nil;
1340 }
1341 result.rawAppendFinish(decodedAuthToken, decodedLen);
1342 return result;
1343 }
1344
1345 ETag
1346 HttpHeader::getETag(Http::HdrType id) const
1347 {
1348 ETag etag = {NULL, -1};
1349 HttpHeaderEntry *e;
1350 assert(Http::HeaderLookupTable.lookup(id).type == Http::HdrFieldType::ftETag); /* must be of an appropriate type */
1351
1352 if ((e = findEntry(id)))
1353 etagParseInit(&etag, e->value.termedBuf());
1354
1355 return etag;
1356 }
1357
1358 TimeOrTag
1359 HttpHeader::getTimeOrTag(Http::HdrType id) const
1360 {
1361 TimeOrTag tot;
1362 HttpHeaderEntry *e;
1363 assert(Http::HeaderLookupTable.lookup(id).type == Http::HdrFieldType::ftDate_1123_or_ETag); /* must be of an appropriate type */
1364 memset(&tot, 0, sizeof(tot));
1365
1366 if ((e = findEntry(id))) {
1367 const char *str = e->value.termedBuf();
1368 /* try as an ETag */
1369
1370 if (etagParseInit(&tot.tag, str)) {
1371 tot.valid = tot.tag.str != NULL;
1372 tot.time = -1;
1373 } else {
1374 /* or maybe it is time? */
1375 tot.time = Time::ParseRfc1123(str);
1376 tot.valid = tot.time >= 0;
1377 tot.tag.str = NULL;
1378 }
1379 }
1380
1381 assert(tot.time < 0 || !tot.tag.str); /* paranoid */
1382 return tot;
1383 }
1384
1385 /*
1386 * HttpHeaderEntry
1387 */
1388
1389 HttpHeaderEntry::HttpHeaderEntry(Http::HdrType anId, const SBuf &aName, const char *aValue)
1390 {
1391 assert(any_HdrType_enum_value(anId));
1392 id = anId;
1393
1394 if (id != Http::HdrType::OTHER)
1395 name = Http::HeaderLookupTable.lookup(id).name;
1396 else
1397 name = aName;
1398
1399 value = aValue;
1400
1401 if (id != Http::HdrType::BAD_HDR)
1402 ++ headerStatsTable[id].aliveCount;
1403
1404 debugs(55, 9, "created HttpHeaderEntry " << this << ": '" << name << " : " << value );
1405 }
1406
1407 HttpHeaderEntry::~HttpHeaderEntry()
1408 {
1409 debugs(55, 9, "destroying entry " << this << ": '" << name << ": " << value << "'");
1410
1411 if (id != Http::HdrType::BAD_HDR) {
1412 assert(headerStatsTable[id].aliveCount);
1413 -- headerStatsTable[id].aliveCount;
1414 id = Http::HdrType::BAD_HDR; // it already is BAD_HDR, no sense in resetting it
1415 }
1416
1417 }
1418
1419 /* parses and inits header entry, returns true/false */
1420 HttpHeaderEntry *
1421 HttpHeaderEntry::parse(const char *field_start, const char *field_end, const http_hdr_owner_type msgType)
1422 {
1423 /* note: name_start == field_start */
1424 const char *name_end = (const char *)memchr(field_start, ':', field_end - field_start);
1425 int name_len = name_end ? name_end - field_start :0;
1426 const char *value_start = field_start + name_len + 1; /* skip ':' */
1427 /* note: value_end == field_end */
1428
1429 ++ HeaderEntryParsedCount;
1430
1431 /* do we have a valid field name within this field? */
1432
1433 if (!name_len || name_end > field_end)
1434 return NULL;
1435
1436 if (name_len > 65534) {
1437 /* String must be LESS THAN 64K and it adds a terminating NULL */
1438 // TODO: update this to show proper name_len in Raw markup, but not print all that
1439 debugs(55, 2, "ignoring huge header field (" << Raw("field_start", field_start, 100) << "...)");
1440 return NULL;
1441 }
1442
1443 /*
1444 * RFC 7230 section 3.2.4:
1445 * "No whitespace is allowed between the header field-name and colon.
1446 * ...
1447 * A server MUST reject any received request message that contains
1448 * whitespace between a header field-name and colon with a response code
1449 * of 400 (Bad Request). A proxy MUST remove any such whitespace from a
1450 * response message before forwarding the message downstream."
1451 */
1452 if (xisspace(field_start[name_len - 1])) {
1453
1454 if (msgType == hoRequest)
1455 return nullptr;
1456
1457 // for now, also let relaxed parser remove this BWS from any non-HTTP messages
1458 const bool stripWhitespace = (msgType == hoReply) ||
1459 Config.onoff.relaxed_header_parser;
1460 if (!stripWhitespace)
1461 return nullptr; // reject if we cannot strip
1462
1463 debugs(55, Config.onoff.relaxed_header_parser <= 0 ? 1 : 2,
1464 "WARNING: Whitespace after header name in '" << getStringPrefix(field_start, field_end-field_start) << "'");
1465
1466 while (name_len > 0 && xisspace(field_start[name_len - 1]))
1467 --name_len;
1468
1469 if (!name_len) {
1470 debugs(55, 2, "found header with only whitespace for name");
1471 return NULL;
1472 }
1473 }
1474
1475 /* RFC 7230 section 3.2:
1476 *
1477 * header-field = field-name ":" OWS field-value OWS
1478 * field-name = token
1479 * token = 1*TCHAR
1480 */
1481 for (const char *pos = field_start; pos < (field_start+name_len); ++pos) {
1482 if (!CharacterSet::TCHAR[*pos]) {
1483 debugs(55, 2, "found header with invalid characters in " <<
1484 Raw("field-name", field_start, min(name_len,100)) << "...");
1485 return nullptr;
1486 }
1487 }
1488
1489 /* now we know we can parse it */
1490
1491 debugs(55, 9, "parsing HttpHeaderEntry: near '" << getStringPrefix(field_start, field_end-field_start) << "'");
1492
1493 /* is it a "known" field? */
1494 Http::HdrType id = Http::HeaderLookupTable.lookup(field_start,name_len).id;
1495 debugs(55, 9, "got hdr-id=" << id);
1496
1497 SBuf theName;
1498
1499 String value;
1500
1501 if (id == Http::HdrType::BAD_HDR)
1502 id = Http::HdrType::OTHER;
1503
1504 /* set field name */
1505 if (id == Http::HdrType::OTHER)
1506 theName.append(field_start, name_len);
1507 else
1508 theName = Http::HeaderLookupTable.lookup(id).name;
1509
1510 /* trim field value */
1511 while (value_start < field_end && xisspace(*value_start))
1512 ++value_start;
1513
1514 while (value_start < field_end && xisspace(field_end[-1]))
1515 --field_end;
1516
1517 if (field_end - value_start > 65534) {
1518 /* String must be LESS THAN 64K and it adds a terminating NULL */
1519 debugs(55, 2, "WARNING: found '" << theName << "' header of " << (field_end - value_start) << " bytes");
1520 return NULL;
1521 }
1522
1523 /* set field value */
1524 value.assign(value_start, field_end - value_start);
1525
1526 if (id != Http::HdrType::BAD_HDR)
1527 ++ headerStatsTable[id].seenCount;
1528
1529 debugs(55, 9, "parsed HttpHeaderEntry: '" << theName << ": " << value << "'");
1530
1531 return new HttpHeaderEntry(id, theName, value.termedBuf());
1532 }
1533
1534 HttpHeaderEntry *
1535 HttpHeaderEntry::clone() const
1536 {
1537 return new HttpHeaderEntry(id, name, value.termedBuf());
1538 }
1539
1540 void
1541 HttpHeaderEntry::packInto(Packable * p) const
1542 {
1543 assert(p);
1544 p->append(name.rawContent(), name.length());
1545 p->append(": ", 2);
1546 p->append(value.rawBuf(), value.size());
1547 p->append("\r\n", 2);
1548 }
1549
1550 int
1551 HttpHeaderEntry::getInt() const
1552 {
1553 int val = -1;
1554 int ok = httpHeaderParseInt(value.termedBuf(), &val);
1555 httpHeaderNoteParsedEntry(id, value, ok == 0);
1556 /* XXX: Should we check ok - ie
1557 * return ok ? -1 : value;
1558 */
1559 return val;
1560 }
1561
1562 int64_t
1563 HttpHeaderEntry::getInt64() const
1564 {
1565 int64_t val = -1;
1566 const bool ok = httpHeaderParseOffset(value.termedBuf(), &val);
1567 httpHeaderNoteParsedEntry(id, value, !ok);
1568 return val; // remains -1 if !ok (XXX: bad method API)
1569 }
1570
1571 static void
1572 httpHeaderNoteParsedEntry(Http::HdrType id, String const &context, bool error)
1573 {
1574 if (id != Http::HdrType::BAD_HDR)
1575 ++ headerStatsTable[id].parsCount;
1576
1577 if (error) {
1578 if (id != Http::HdrType::BAD_HDR)
1579 ++ headerStatsTable[id].errCount;
1580 debugs(55, 2, "cannot parse hdr field: '" << Http::HeaderLookupTable.lookup(id).name << ": " << context << "'");
1581 }
1582 }
1583
1584 /*
1585 * Reports
1586 */
1587
1588 /* tmp variable used to pass stat info to dumpers */
1589 extern const HttpHeaderStat *dump_stat; /* argh! */
1590 const HttpHeaderStat *dump_stat = NULL;
1591
1592 static void
1593 httpHeaderFieldStatDumper(StoreEntry * sentry, int, double val, double, int count)
1594 {
1595 const int id = static_cast<int>(val);
1596 const bool valid_id = Http::any_valid_header(static_cast<Http::HdrType>(id));
1597 const char *name = valid_id ? Http::HeaderLookupTable.lookup(static_cast<Http::HdrType>(id)).name : "INVALID";
1598 int visible = count > 0;
1599 /* for entries with zero count, list only those that belong to current type of message */
1600
1601 if (!visible && valid_id && dump_stat->owner_mask)
1602 visible = CBIT_TEST(*dump_stat->owner_mask, id);
1603
1604 if (visible)
1605 storeAppendPrintf(sentry, "%2d\t %-20s\t %5d\t %6.2f\n",
1606 id, name, count, xdiv(count, dump_stat->busyDestroyedCount));
1607 }
1608
1609 static void
1610 httpHeaderFldsPerHdrDumper(StoreEntry * sentry, int idx, double val, double, int count)
1611 {
1612 if (count)
1613 storeAppendPrintf(sentry, "%2d\t %5d\t %5d\t %6.2f\n",
1614 idx, (int) val, count,
1615 xpercent(count, dump_stat->destroyedCount));
1616 }
1617
1618 static void
1619 httpHeaderStatDump(const HttpHeaderStat * hs, StoreEntry * e)
1620 {
1621 assert(hs);
1622 assert(e);
1623
1624 if (!hs->owner_mask)
1625 return; // these HttpHeaderStat objects were not meant to be dumped here
1626
1627 dump_stat = hs;
1628 storeAppendPrintf(e, "\nHeader Stats: %s\n", hs->label);
1629 storeAppendPrintf(e, "\nField type distribution\n");
1630 storeAppendPrintf(e, "%2s\t %-20s\t %5s\t %6s\n",
1631 "id", "name", "count", "#/header");
1632 hs->fieldTypeDistr.dump(e, httpHeaderFieldStatDumper);
1633 storeAppendPrintf(e, "\nCache-control directives distribution\n");
1634 storeAppendPrintf(e, "%2s\t %-20s\t %5s\t %6s\n",
1635 "id", "name", "count", "#/cc_field");
1636 hs->ccTypeDistr.dump(e, httpHdrCcStatDumper);
1637 storeAppendPrintf(e, "\nSurrogate-control directives distribution\n");
1638 storeAppendPrintf(e, "%2s\t %-20s\t %5s\t %6s\n",
1639 "id", "name", "count", "#/sc_field");
1640 hs->scTypeDistr.dump(e, httpHdrScStatDumper);
1641 storeAppendPrintf(e, "\nNumber of fields per header distribution\n");
1642 storeAppendPrintf(e, "%2s\t %-5s\t %5s\t %6s\n",
1643 "id", "#flds", "count", "%total");
1644 hs->hdrUCountDistr.dump(e, httpHeaderFldsPerHdrDumper);
1645 storeAppendPrintf(e, "\n");
1646 dump_stat = NULL;
1647 }
1648
1649 void
1650 httpHeaderStoreReport(StoreEntry * e)
1651 {
1652 assert(e);
1653
1654 HttpHeaderStats[0].parsedCount =
1655 HttpHeaderStats[hoRequest].parsedCount + HttpHeaderStats[hoReply].parsedCount;
1656 HttpHeaderStats[0].ccParsedCount =
1657 HttpHeaderStats[hoRequest].ccParsedCount + HttpHeaderStats[hoReply].ccParsedCount;
1658 HttpHeaderStats[0].destroyedCount =
1659 HttpHeaderStats[hoRequest].destroyedCount + HttpHeaderStats[hoReply].destroyedCount;
1660 HttpHeaderStats[0].busyDestroyedCount =
1661 HttpHeaderStats[hoRequest].busyDestroyedCount + HttpHeaderStats[hoReply].busyDestroyedCount;
1662
1663 for (const auto &stats: HttpHeaderStats)
1664 httpHeaderStatDump(&stats, e);
1665
1666 /* field stats for all messages */
1667 storeAppendPrintf(e, "\nHttp Fields Stats (replies and requests)\n");
1668
1669 storeAppendPrintf(e, "%2s\t %-25s\t %5s\t %6s\t %6s\n",
1670 "id", "name", "#alive", "%err", "%repeat");
1671
1672 // scan heaaderTable and output
1673 for (auto h : WholeEnum<Http::HdrType>()) {
1674 auto stats = headerStatsTable[h];
1675 storeAppendPrintf(e, "%2d\t %-25s\t %5d\t %6.3f\t %6.3f\n",
1676 Http::HeaderLookupTable.lookup(h).id,
1677 Http::HeaderLookupTable.lookup(h).name,
1678 stats.aliveCount,
1679 xpercent(stats.errCount, stats.parsCount),
1680 xpercent(stats.repCount, stats.seenCount));
1681 }
1682
1683 storeAppendPrintf(e, "Headers Parsed: %d + %d = %d\n",
1684 HttpHeaderStats[hoRequest].parsedCount,
1685 HttpHeaderStats[hoReply].parsedCount,
1686 HttpHeaderStats[0].parsedCount);
1687 storeAppendPrintf(e, "Hdr Fields Parsed: %d\n", HeaderEntryParsedCount);
1688 }
1689
1690 int
1691 HttpHeader::hasListMember(Http::HdrType id, const char *member, const char separator) const
1692 {
1693 int result = 0;
1694 const char *pos = NULL;
1695 const char *item;
1696 int ilen;
1697 int mlen = strlen(member);
1698
1699 assert(any_registered_header(id));
1700
1701 String header (getStrOrList(id));
1702
1703 while (strListGetItem(&header, separator, &item, &ilen, &pos)) {
1704 if (strncasecmp(item, member, mlen) == 0
1705 && (item[mlen] == '=' || item[mlen] == separator || item[mlen] == ';' || item[mlen] == '\0')) {
1706 result = 1;
1707 break;
1708 }
1709 }
1710
1711 return result;
1712 }
1713
1714 int
1715 HttpHeader::hasByNameListMember(const char *name, const char *member, const char separator) const
1716 {
1717 int result = 0;
1718 const char *pos = NULL;
1719 const char *item;
1720 int ilen;
1721 int mlen = strlen(member);
1722
1723 assert(name);
1724
1725 String header (getByName(name));
1726
1727 while (strListGetItem(&header, separator, &item, &ilen, &pos)) {
1728 if (strncasecmp(item, member, mlen) == 0
1729 && (item[mlen] == '=' || item[mlen] == separator || item[mlen] == ';' || item[mlen] == '\0')) {
1730 result = 1;
1731 break;
1732 }
1733 }
1734
1735 return result;
1736 }
1737
1738 void
1739 HttpHeader::removeHopByHopEntries()
1740 {
1741 removeConnectionHeaderEntries();
1742
1743 const HttpHeaderEntry *e;
1744 HttpHeaderPos pos = HttpHeaderInitPos;
1745 int headers_deleted = 0;
1746 while ((e = getEntry(&pos))) {
1747 Http::HdrType id = e->id;
1748 if (Http::HeaderLookupTable.lookup(id).hopbyhop) {
1749 delAt(pos, headers_deleted);
1750 CBIT_CLR(mask, id);
1751 }
1752 }
1753 }
1754
1755 void
1756 HttpHeader::removeConnectionHeaderEntries()
1757 {
1758 if (has(Http::HdrType::CONNECTION)) {
1759 /* anything that matches Connection list member will be deleted */
1760 String strConnection;
1761
1762 (void) getList(Http::HdrType::CONNECTION, &strConnection);
1763 const HttpHeaderEntry *e;
1764 HttpHeaderPos pos = HttpHeaderInitPos;
1765 /*
1766 * think: on-average-best nesting of the two loops (hdrEntry
1767 * and strListItem) @?@
1768 */
1769 /*
1770 * maybe we should delete standard stuff ("keep-alive","close")
1771 * from strConnection first?
1772 */
1773
1774 int headers_deleted = 0;
1775 while ((e = getEntry(&pos))) {
1776 if (strListIsMember(&strConnection, e->name, ','))
1777 delAt(pos, headers_deleted);
1778 }
1779 if (headers_deleted)
1780 refreshMask();
1781 }
1782 }
1783