Annotation of parser3/src/main/untaint.C, revision 1.150
1.7 paf 1: /** @file
1.8 paf 2: Parser: String class part: untaint mechanizm.
3:
1.135 paf 4: Copyright(c) 2001-2005 ArtLebedev Group (http://www.artlebedev.com)
1.89 paf 5: Author: Alexandr Petrosian <paf@design.ru> (http://paf.design.ru)
1.103 paf 6: */
1.8 paf 7:
1.150 ! misha 8: static const char * const IDENT_UNTAINT_C="$Date: 2009-07-07 05:48:05 $";
1.117 paf 9:
1.1 paf 10:
11: #include "pa_string.h"
12: #include "pa_hash.h"
13: #include "pa_exception.h"
1.13 paf 14: #include "pa_table.h"
1.32 paf 15: #include "pa_globals.h"
1.58 parser 16: #include "pa_dictionary.h"
1.66 parser 17: #include "pa_common.h"
1.85 paf 18: #include "pa_charset.h"
1.117 paf 19: #include "pa_request_charsets.h"
20: #include "pa_sapi.h"
1.1 paf 21:
1.117 paf 22: extern "C" { // author forgot to do that
23: #include "ec.h"
24: }
1.91 paf 25:
1.117 paf 26: #include "pa_sql_connection.h"
1.91 paf 27:
1.117 paf 28: // defines
29:
30: #undef CORD_ec_append
31: // redefining to intercept flushes and implement whitespace optimization
32: // of all consequent white space chars leaving only first one
33: #define CORD_ec_append(x, c) \
34: { \
35: bool skip=false; \
36: if(optimize) switch(c) { \
1.131 paf 37: case ' ': case '\n': case '\t': \
1.117 paf 38: if(whitespace) \
39: skip=true; /*skipping subsequent*/ \
40: else \
41: whitespace=true; \
42: break; \
43: default: \
44: whitespace=false; \
45: break; \
46: } \
47: if(!skip) { \
48: if ((x)[0].ec_bufptr == (x)[0].ec_buf + CORD_BUFSZ) { \
49: CORD_ec_flush_buf(x); \
50: } \
51: *((x)[0].ec_bufptr)++ = (c); \
52: } \
53: }
54:
55:
1.140 misha 56: #define escape_fragment(action) \
1.119 paf 57: for(; fragment_length--; CORD_next(info->pos)) { \
58: char c=CORD_pos_fetch(info->pos); \
1.117 paf 59: action \
1.1 paf 60: }
1.133 paf 61: #define _default CORD_ec_append(info->result, c)
1.117 paf 62: #define encode(need_encode_func, prefix, otherwise) \
63: if(need_encode_func(c)) { \
64: static const char* hex="0123456789ABCDEF"; \
1.119 paf 65: CORD_ec_append(info->result, prefix); \
66: CORD_ec_append(info->result, hex[((unsigned char)c)/0x10]); \
67: CORD_ec_append(info->result, hex[((unsigned char)c)%0x10]); \
1.117 paf 68: } else \
1.119 paf 69: CORD_ec_append(info->result, otherwise);
1.126 paf 70: #define to_char(c) { CORD_ec_append(info->result, c); whitespace=false; }
71: #define to_string(s) { CORD_ec_append_cord(info->result, s); whitespace=false; }
1.4 paf 72:
1.9 paf 73: inline bool need_file_encode(unsigned char c){
1.108 paf 74: // russian letters and space ENABLED
75: // encoding only these...
76: return strchr(
1.143 misha 77: "*?'\"<>|"
1.108 paf 78: #ifndef WIN32
1.143 misha 79: ":\\"
1.31 paf 80: #endif
1.143 misha 81: , c)!=0;
1.9 paf 82: }
1.143 misha 83:
1.5 paf 84: inline bool need_uri_encode(unsigned char c){
1.143 misha 85: if((c>='0') &&(c<='9') ||(c>='A') &&(c<='Z') ||(c>='a') &&(c<='z'))
1.4 paf 86: return false;
87:
1.143 misha 88: return !strchr("_-./", c);
1.5 paf 89: }
1.143 misha 90:
1.36 paf 91: inline bool need_http_header_encode(unsigned char c){
1.143 misha 92: if(strchr(" , :", c))
1.5 paf 93: return false;
94:
95: return need_uri_encode(c);
1.4 paf 96: }
1.143 misha 97:
1.136 paf 98: inline bool need_regex_escape(unsigned char c){
1.142 misha 99: return strchr("\\^$.[]|()?*+{}-", c)!=0;
1.136 paf 100: }
1.1 paf 101:
1.150 ! misha 102: inline bool need_parser_code_escape(unsigned char c){
! 103: return strchr("^$;@()[]{}:#\"", c)!=0;
! 104: }
! 105:
1.1 paf 106: // String
107:
1.41 paf 108: /*
109: HTTP-header = field-name ":" [ field-value ] CRLF
110:
111: field-name = token
112: field-value = *( field-content | LWS )
113:
114: field-content = <the OCTETs making up the field-value
115: and consisting of either *TEXT or combinations
116: of token, tspecials, and quoted-string>
117:
118:
119:
120: word = token | quoted-string
121:
122: token = 1*<any CHAR except CTLs or tspecials>
123:
124:
125:
126: tspecials = "(" | ")" | "<" | ">" | "@"
127: | "," | ";" | ":" | "\" | <">
128: | "/" | "[" | "]" | "?" | "="
129: | "{" | "}" | SP | HT
130:
131: SP = <US-ASCII SP, space (32)>
132: HT = <US-ASCII HT, horizontal-tab (9)>
133:
134: LWS = [CRLF] 1*( SP | HT )
135: TEXT = <any OCTET except CTLs,
136: but including LWS>
137:
138: quoted-pair = "\" CHAR
139:
140: if(strchr("()<>@,;:\\\"/[]?={} \t", *ptr))
141: */
1.117 paf 142: inline bool need_quote_http_header(const char* ptr, size_t size) {
1.41 paf 143: for(; size--; ptr++)
1.42 paf 144: if(strchr(";\\\"= \t" /* excluded ()<>@, :/ ? []{} */, *ptr))
1.41 paf 145: return true;
146: return false;
147: }
148:
1.119 paf 149: #ifndef DOXYGEN
150: struct Append_fragment_info {
151: String::Language lang;
152: String::Languages* dest_languages;
153: size_t dest_body_plan_length;
154: };
155: #endif
156: int append_fragment_optimizing(char alang, size_t asize, Append_fragment_info* info) {
157: const String::Language lang=(String::Language)(unsigned char)alang;
158: // main idea here:
159: // tainted piece would get OPTIMIZED bit from 'lang'
160: // clean piece would be marked OPTIMIZED manually
161: // pieces with determined languages [not tainted|clean] would retain theirs langs
162: info->dest_languages->append(info->dest_body_plan_length,
1.120 paf 163: lang==String::L_TAINTED?
1.119 paf 164: info->lang
1.120 paf 165: :lang==String::L_CLEAN?
166: (String::Language)(String::L_CLEAN|String::L_OPTIMIZE_BIT) // ORing with OPTIMIZED flag
1.119 paf 167: :lang,
168: asize);
169: info->dest_body_plan_length+=asize;
170:
171: return 0; // 0=continue
172: }
173: int append_fragment_nonoptimizing(char alang, size_t asize, Append_fragment_info* info) {
174: const String::Language lang=(String::Language)(unsigned char)alang;
175: // The core idea: tainted pieces got marked with context's lang
176: info->dest_languages->append(info->dest_body_plan_length,
1.120 paf 177: lang==String::L_TAINTED?
1.119 paf 178: info->lang
179: :lang,
180: asize);
181: info->dest_body_plan_length+=asize;
182:
183: return 0; // 0=continue
184: }
1.117 paf 185:
1.91 paf 186: /**
1.117 paf 187: appends to other String,
1.91 paf 188: marking all tainted pieces of it with @a lang.
1.92 paf 189: or marking ALL pieces of it with a @a lang when @a forced to,
190: and propagating OPTIMIZE language bit.
1.117 paf 191: */
1.147 misha 192: String& String::append_to(String& dest, Language ilang, bool forced) const {
1.117 paf 193: if(is_empty())
194: return dest;
1.91 paf 195:
1.119 paf 196: // first: fragment infos
1.117 paf 197:
1.147 misha 198: if(ilang==L_PASS_APPENDED) // without language-change?
199: dest.langs.appendHelper(dest.body, langs, body);
1.117 paf 200: else if(forced) //forcing passed lang?
1.147 misha 201: dest.langs.appendHelper(dest.body, ilang, body);
202: else {
203: if(langs.opt.is_not_just_lang){
204: Append_fragment_info info={ilang, &dest.langs, dest.body.length()};
205: langs.for_each(body, ilang&L_OPTIMIZE_BIT?
206: append_fragment_optimizing
207: :append_fragment_nonoptimizing, &info);
208: } else {
209: Language lang=langs.opt.lang;
210: // see append_fragment_* for explanation
211: if(ilang&L_OPTIMIZE_BIT){
212: dest.langs.appendHelper(dest.body,
213: lang==String::L_TAINTED?
214: ilang
215: :lang==String::L_CLEAN?
216: (String::Language)(String::L_CLEAN|String::L_OPTIMIZE_BIT)
217: :lang,
218: body);
219: } else {
220: dest.langs.appendHelper(dest.body, lang==String::L_TAINTED ? ilang:lang, body);
221: }
222: }
1.91 paf 223: }
224:
1.119 paf 225: // next: letters
226: dest.body<<body;
227:
1.117 paf 228: ASSERT_STRING_INVARIANT(dest);
1.75 paf 229: return dest;
1.51 parser 230: }
231:
1.109 paf 232: /** http://www.ietf.org/rfc/rfc2047.txt
233: RFC
234: (3) As a replacement for a 'word' entity within a 'phrase', for example,
235: one that precedes an address in a From, To, or Cc header. The ABNF
236: definition for 'phrase' from RFC 822 thus becomes:
237:
238: phrase = 1*( encoded-word / word )
239:
240: In this case the set of characters that may be used in a "Q"-encoded
241: 'encoded-word' is restricted to: <upper and lower case ASCII
242: letters, decimal digits, "!", "*", "+", "-", "/", "=", and "_"
243: (underscore, ASCII 95.)>. An 'encoded-word' that appears within a
244: 'phrase' MUST be separated from any adjacent 'word', 'text' or
245: 'special' by 'linear-white-space'.
246: ...
247: (2) The 8-bit hexadecimal value 20 (e.g., ISO-8859-1 SPACE) may be
248: represented as "_" (underscore, ASCII 95.). (This character may
249: not pass through some internetwork mail gateways, but its use
250: will greatly enhance readability of "Q" encoded data with mail
251: readers that do not support this encoding.) Note that the "_"
252: always represents hexadecimal 20, even if the SPACE character
253: occupies a different code position in the character set in use.
254:
255: paf: obviously,
256: without "=", or one could not differ "=E0" and "russian letter a"
257: and without "_", or in would mean 0x20
258: */
1.117 paf 259: inline bool mail_header_char_valid_within_Qencoded(char c) {
1.109 paf 260: return c>='A' && c<='Z'
261: || c>='a' && c<='Z'
262: || c>='0' && c<='9'
263: || strchr("!*+-/", c);
264: }
1.118 paf 265: inline bool addr_spec_soon(const char *src) {
1.124 paf 266: for(char c; (c=*src); src++)
1.118 paf 267: if(c=='<')
268: return true;
269: else if(!(c==' ' || c=='\t'))
270: return false;
271: return false;
272: }
1.117 paf 273: /**
274: RFC
275: Upper case should be used for hexadecimal digits "A" through "F"
276: The 8-bit hexadecimal value 20 (e.g., ISO-8859-1 SPACE)
277: may be represented as "_"
278: */
279: inline bool mail_header_nonspace_char(char c) {
280: return c != 0x20;
281: }
1.125 paf 282:
1.117 paf 283: inline void ec_append(CORD_ec& result, bool& optimize, bool& whitespace, CORD_pos pos, size_t size) {
284: while(size--) {
285: CORD_ec_append(result, CORD_pos_fetch(pos));
286: CORD_next(pos);
287: }
288: }
289: inline void pa_CORD_pos_advance(CORD_pos pos, size_t n) {
290: while(true) {
1.125 paf 291: long avail=CORD_pos_chars_left(pos);
292: if(avail<=0) {
1.117 paf 293: CORD_next(pos);
294: if(!--n)
295: break;
1.125 paf 296: } else if((size_t)avail<n) {
1.117 paf 297: CORD_pos_advance(pos, avail);
298: n-=avail;
299: } else { // avail>=n
300: CORD_pos_advance(pos, n);
301: break;
302: }
303: }
304: }
305:
1.119 paf 306: #ifndef DOXYGEN
307: struct Cstr_to_string_body_block_info {
308: // input
309: String::Language lang;
310: SQL_Connection* connection;
311: const Request_charsets* charsets;
312: const String::Body* body;
313:
314: // output
315: CORD_ec result;
316:
317: // private
318: CORD_pos pos;
319: size_t fragment_begin;
320: bool whitespace;
1.138 misha 321: const char* exception;
1.119 paf 322: };
323: #endif
1.148 misha 324: int cstr_to_string_body_block(String::Language to_lang, size_t fragment_length, Cstr_to_string_body_block_info* info) {
1.119 paf 325: bool& whitespace=info->whitespace;
326: size_t fragment_end=info->fragment_begin+fragment_length;
1.148 misha 327: //fprintf(stderr, "%d, %d =%s=\n", to_lang, fragment_length, info->body->cstr());
1.119 paf 328:
1.120 paf 329: bool optimize=(to_lang & String::L_OPTIMIZE_BIT)!=0;
1.119 paf 330: if(!optimize)
331: whitespace=false;
332:
1.120 paf 333: switch(to_lang & ~String::L_OPTIMIZE_BIT) {
334: case String::L_CLEAN:
335: case String::L_TAINTED:
336: case String::L_AS_IS:
1.119 paf 337: // clean piece
338:
339: // tainted piece, but undefined untaint language
340: // for VString.as_double of tainted values
341: // for ^process{body} evaluation
342:
343: // tainted, untaint language: as-is
344: ec_append(info->result, optimize, whitespace, info->pos, fragment_length);
345: break;
1.120 paf 346: case String::L_FILE_SPEC:
1.119 paf 347: // tainted, untaint language: file [name]
1.140 misha 348: {
349: bool is1251=(info->charsets && info->charsets->source().NAME()=="WINDOWS-1251");
350: escape_fragment(
351: // Macintosh has problems with small Russian letter 'r'
352: if( is1251 && c=='\xF0' ) {
353: // fixing that letter for most common charset
354: to_char('p');
355: } else // fallback to default
356: encode(need_file_encode, '_', c);
357: );
358: }
1.119 paf 359: break;
1.144 misha 360: case String::L_FILE_POST:
361: {
362: escape_fragment(switch(c) {
363: case '\0': to_string("\\0"); break;
364: case '\\': to_string("\\\\"); break;
365: default: _default; break;
366: });
367: }
368: break;
1.120 paf 369: case String::L_URI:
1.119 paf 370: // tainted, untaint language: uri
371: {
372: const char *fragment_str=info->body->mid(info->fragment_begin, fragment_length).cstr();
373: // skip source [we use recoded version]
374: pa_CORD_pos_advance(info->pos, fragment_length);
375: String::C output(fragment_str, fragment_length);
376: if(info->charsets)
377: output=Charset::transcode(output,
378: info->charsets->source(),
379: info->charsets->client());
1.77 paf 380:
1.119 paf 381: char c;
1.124 paf 382: for(const char* src=output.str; (c=*src++); )
1.117 paf 383: encode(need_uri_encode, '%', c);
1.119 paf 384: }
385: break;
1.120 paf 386: case String::L_HTTP_HEADER:
1.119 paf 387: // tainted, untaint language: http-field-content-text
1.139 misha 388: // the same as L_URI BUT not transcoded into $response:charset before encoding
1.140 misha 389: escape_fragment(
1.119 paf 390: encode(need_uri_encode, '%', c);
391: );
392: break;
1.120 paf 393: case String::L_MAIL_HEADER:
1.119 paf 394: // tainted, untaint language: mail-header
395: // http://www.ietf.org/rfc/rfc2047.txt
396: if(info->charsets) {
397: size_t mail_size;
398: const char *mail_ptr=
399: info->body->mid(info->fragment_begin, mail_size=fragment_length).cstr();
400: // skip source [we use recoded version]
401: pa_CORD_pos_advance(info->pos, mail_size);
402:
403: const char* charset_name=info->charsets->mail().NAME().cstr();
404:
405: // Subject: Re: parser3: =?koi8-r?Q?=D3=C5=CD=C9=CE=C1=D2?=
406: bool to_quoted_printable=false;
407:
408: bool email=false;
409: uchar c;
1.124 paf 410: for(const char* src=mail_ptr; (c=(uchar)*src++); ) {
1.137 paf 411: if(c=='\r' || c=='\n')
412: c=' ';
1.130 paf 413: if(to_quoted_printable && (c==',' || c == '"' || addr_spec_soon(src-1/*position to 'c'*/))) {
1.119 paf 414: email=c=='<';
415: to_string("?=");
416: to_quoted_printable=false;
417: }
1.130 paf 418: //RFC + An 'encoded-word' MUST NOT appear in any portion of an 'addr-spec'.
1.119 paf 419: if(!email && (
420: !to_quoted_printable && (c & 0x80) // starting quote-printable-encoding on first 8bit char
421: || to_quoted_printable && !mail_header_char_valid_within_Qencoded(c)
422: )) {
423: if(!to_quoted_printable) {
424: to_string("=?");
425: to_string(charset_name);
426: to_string("?Q?");
427: to_quoted_printable=true;
1.105 paf 428: }
1.119 paf 429: encode(mail_header_nonspace_char, '=', '_');
430: } else
431: to_char(c);
432: if(c=='>')
433: email=false;
434: }
435: if(to_quoted_printable) // close
436: to_string("?=");
437:
438: } else
439: ec_append(info->result, optimize, whitespace, info->pos, fragment_length);
440: break;
1.120 paf 441: case String::L_SQL:
1.119 paf 442: // tainted, untaint language: sql
443: if(info->connection) {
444: const char *fragment_str=info->body->mid(info->fragment_begin, fragment_length).cstr();
445: // skip source [we use recoded version]
446: pa_CORD_pos_advance(info->pos, fragment_length);
447:
448: to_string(info->connection->quote(fragment_str, fragment_length));
1.138 misha 449: } else {
450: info->exception="untaint in SQL language failed - no connection specified";
451: info->fragment_begin=fragment_end;
452: return 1; // stop processing. can't throw exception here
453: }
1.119 paf 454: break;
1.120 paf 455: case String::L_JS:
1.140 misha 456: escape_fragment(switch(c) {
457: case '\n': to_string("\\n"); break;
1.119 paf 458: case '"': to_string("\\\""); break;
459: case '\'': to_string("\\'"); break;
460: case '\\': to_string("\\\\"); break;
461: case '\xFF': to_string("\\\xFF"); break;
1.146 misha 462: case '\r': to_string("\\r"); break;
1.133 paf 463: default: _default; break;
1.119 paf 464: });
465: break;
1.120 paf 466: case String::L_XML:
1.133 paf 467: // [2] Char ::= #x9 | #xA | #xD | [#x20-#xD7FF] | [#xE000-#xFFFD] | [#x10000-#x10FFFF]
1.140 misha 468: escape_fragment(switch(c) {
469: case '\x20':
470: case '\x9':
471: case '\xA':
1.133 paf 472: case '\xD': // this is usually removed on input
473: _default;
474: break;
1.119 paf 475: case '&': to_string("&"); break;
476: case '>': to_string(">"); break;
477: case '<': to_string("<"); break;
478: case '"': to_string("""); break;
479: case '\'': to_string("'"); break;
1.133 paf 480: default:
481: if(((unsigned char)c)<0x20) {
482: // fixing it, so that libxml would not result
483: // in fatal error parsing text
484: // though it really violates standard.
485: // to indicate there were an error
486: // replace bad char not to it's code,
487: // which we can do,
488: // but rather to '!' to show that input were actually
489: // invalid.
490: // life: shows that MSIE can somehow garble form values
491: // so that they contain these chars.
492: to_char('!');
493: } else {
494: _default;
495: }
496: break;
1.119 paf 497: });
498: break;
1.120 paf 499: case String::L_HTML:
1.140 misha 500: escape_fragment(switch(c) {
1.119 paf 501: case '&': to_string("&"); break;
502: case '>': to_string(">"); break;
503: case '<': to_string("<"); break;
504: case '"': to_string("""); break;
1.133 paf 505: default: _default; break;
1.119 paf 506: });
507: break;
1.136 paf 508: case String::L_REGEX:
509: // tainted, untaint language: regex
1.140 misha 510: escape_fragment(
1.136 paf 511: if(need_regex_escape(c))
512: to_char('\\')
513: _default;
514: );
515: break;
1.140 misha 516: case String::L_HTTP_COOKIE:
517: // tainted, untaint language: cookie (3.3.0 and higher: %uXXXX in UTF-8)
518: {
519: const char *fragment_str=info->body->mid(info->fragment_begin, fragment_length).cstr();
520: // skip source [we use recoded version]
521: pa_CORD_pos_advance(info->pos, fragment_length);
522: String::C output(fragment_str, fragment_length);
523:
524: output=Charset::escape(output, info->charsets->source());
525: //throw Exception(0, 0, output);
526: to_string(output);
527:
528: }
529: break;
1.150 ! misha 530: case String::L_PARSER_CODE:
! 531: // for auto-untaint in process
! 532: escape_fragment(
! 533: if(need_parser_code_escape(c))
! 534: to_char('^');
! 535: _default;
! 536: );
! 537: break;
1.119 paf 538: default:
1.127 paf 539: SAPI::abort("unknown untaint language #%d",
1.119 paf 540: static_cast<int>(to_lang)); // should never
541: break; // never
1.117 paf 542: }
1.55 parser 543:
1.119 paf 544: info->fragment_begin=fragment_end;
545:
546: return 0; // 0=continue
547: }
548:
549:
1.148 misha 550: String::Body String::cstr_to_string_body_taint(Language lang, SQL_Connection* connection, const Request_charsets *charsets) const {
1.134 paf 551: if(is_empty())
552: return String::Body();
1.119 paf 553:
554: Cstr_to_string_body_block_info info;
555: // input
556: info.lang=lang;
557: info.connection=connection;
558: info.charsets=charsets;
559: info.body=&body;
560: // output
561: CORD_ec_init(info.result);
562: // private
563: body.set_pos(info.pos, 0);
564: info.fragment_begin=0;
1.138 misha 565: info.exception=0;
1.119 paf 566: info.whitespace=true;
567:
1.148 misha 568: cstr_to_string_body_block(lang, length(), &info);
569:
570: if(info.exception)
571: throw Exception(0,
572: 0,
573: info.exception);
574:
575: return String::Body(CORD_ec_to_cord(info.result));
576: }
577:
578: int cstr_to_string_body_block_untaint(char alang, size_t fragment_length, Cstr_to_string_body_block_info* info){
579: const String::Language lang=(String::Language)(unsigned char)alang;
580: // see append_fragment_* for explanation
581: if(info->lang&String::L_OPTIMIZE_BIT)
582: return cstr_to_string_body_block(
583: lang==String::L_TAINTED?
584: info->lang
585: :lang==String::L_CLEAN?
586: (String::Language)(String::L_CLEAN|String::L_OPTIMIZE_BIT)
587: :lang,
588: fragment_length, info);
589: else
590: return cstr_to_string_body_block(lang==String::L_TAINTED ? info->lang:lang, fragment_length, info);
591: }
592:
593: String::Body String::cstr_to_string_body_untaint(Language lang, SQL_Connection* connection, const Request_charsets *charsets) const {
594: if(is_empty())
595: return String::Body();
596:
597: Cstr_to_string_body_block_info info;
598: // input
599: info.lang=lang;
600: info.connection=connection;
601: info.charsets=charsets;
602: info.body=&body;
603: // output
604: CORD_ec_init(info.result);
605: // private
606: body.set_pos(info.pos, 0);
607: info.fragment_begin=0;
608: info.exception=0;
609: info.whitespace=true;
610:
611: langs.for_each(body, cstr_to_string_body_block_untaint, &info);
612: if(info.exception)
1.138 misha 613: throw Exception(0,
614: 0,
615: info.exception);
616:
1.119 paf 617: return String::Body(CORD_ec_to_cord(info.result));
1.1 paf 618: }
E-mail:
![[BACK]](/cvsweb/icons/back.gif){kind=icon}
Return to untaint.C CVS log ![[TXT]](/cvsweb/icons/text.gif){kind=icon}
![[DIR]](/cvsweb/icons/dir.gif){kind=icon}
Up to [parser3project] / parser3 / src / main