Annotation of parser3/src/main/untaint.C, revision 1.147
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:
1.147 ! misha 8: static const char * const IDENT_UNTAINT_C="$Date: 2009-04-18 22:49:47 $";
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:
102: // String
103:
1.41 paf 104: /*
105: HTTP-header = field-name ":" [ field-value ] CRLF
106:
107: field-name = token
108: field-value = *( field-content | LWS )
109:
110: field-content = <the OCTETs making up the field-value
111: and consisting of either *TEXT or combinations
112: of token, tspecials, and quoted-string>
113:
114:
115:
116: word = token | quoted-string
117:
118: token = 1*<any CHAR except CTLs or tspecials>
119:
120:
121:
122: tspecials = "(" | ")" | "<" | ">" | "@"
123: | "," | ";" | ":" | "\" | <">
124: | "/" | "[" | "]" | "?" | "="
125: | "{" | "}" | SP | HT
126:
127: SP = <US-ASCII SP, space (32)>
128: HT = <US-ASCII HT, horizontal-tab (9)>
129:
130: LWS = [CRLF] 1*( SP | HT )
131: TEXT = <any OCTET except CTLs,
132: but including LWS>
133:
134: quoted-pair = "\" CHAR
135:
136: if(strchr("()<>@,;:\\\"/[]?={} \t", *ptr))
137: */
1.117 paf 138: inline bool need_quote_http_header(const char* ptr, size_t size) {
1.41 paf 139: for(; size--; ptr++)
1.42 paf 140: if(strchr(";\\\"= \t" /* excluded ()<>@, :/ ? []{} */, *ptr))
1.41 paf 141: return true;
142: return false;
143: }
144:
1.119 paf 145: #ifndef DOXYGEN
146: struct Append_fragment_info {
147: String::Language lang;
148: String::Languages* dest_languages;
149: size_t dest_body_plan_length;
150: };
151: #endif
152: int append_fragment_optimizing(char alang, size_t asize, Append_fragment_info* info) {
153: const String::Language lang=(String::Language)(unsigned char)alang;
154: // main idea here:
155: // tainted piece would get OPTIMIZED bit from 'lang'
156: // clean piece would be marked OPTIMIZED manually
157: // pieces with determined languages [not tainted|clean] would retain theirs langs
158: info->dest_languages->append(info->dest_body_plan_length,
1.120 paf 159: lang==String::L_TAINTED?
1.119 paf 160: info->lang
1.120 paf 161: :lang==String::L_CLEAN?
162: (String::Language)(String::L_CLEAN|String::L_OPTIMIZE_BIT) // ORing with OPTIMIZED flag
1.119 paf 163: :lang,
164: asize);
165: info->dest_body_plan_length+=asize;
166:
167: return 0; // 0=continue
168: }
169: int append_fragment_nonoptimizing(char alang, size_t asize, Append_fragment_info* info) {
170: const String::Language lang=(String::Language)(unsigned char)alang;
171: // The core idea: tainted pieces got marked with context's lang
172: info->dest_languages->append(info->dest_body_plan_length,
1.120 paf 173: lang==String::L_TAINTED?
1.119 paf 174: info->lang
175: :lang,
176: asize);
177: info->dest_body_plan_length+=asize;
178:
179: return 0; // 0=continue
180: }
1.117 paf 181:
1.91 paf 182: /**
1.117 paf 183: appends to other String,
1.91 paf 184: marking all tainted pieces of it with @a lang.
1.92 paf 185: or marking ALL pieces of it with a @a lang when @a forced to,
186: and propagating OPTIMIZE language bit.
1.117 paf 187: */
1.147 ! misha 188: String& String::append_to(String& dest, Language ilang, bool forced) const {
1.117 paf 189: if(is_empty())
190: return dest;
1.91 paf 191:
1.119 paf 192: // first: fragment infos
1.117 paf 193:
1.147 ! misha 194: if(ilang==L_PASS_APPENDED) // without language-change?
! 195: dest.langs.appendHelper(dest.body, langs, body);
1.117 paf 196: else if(forced) //forcing passed lang?
1.147 ! misha 197: dest.langs.appendHelper(dest.body, ilang, body);
! 198: else {
! 199: if(langs.opt.is_not_just_lang){
! 200: Append_fragment_info info={ilang, &dest.langs, dest.body.length()};
! 201: langs.for_each(body, ilang&L_OPTIMIZE_BIT?
! 202: append_fragment_optimizing
! 203: :append_fragment_nonoptimizing, &info);
! 204: } else {
! 205: Language lang=langs.opt.lang;
! 206: // see append_fragment_* for explanation
! 207: if(ilang&L_OPTIMIZE_BIT){
! 208: dest.langs.appendHelper(dest.body,
! 209: lang==String::L_TAINTED?
! 210: ilang
! 211: :lang==String::L_CLEAN?
! 212: (String::Language)(String::L_CLEAN|String::L_OPTIMIZE_BIT)
! 213: :lang,
! 214: body);
! 215: } else {
! 216: dest.langs.appendHelper(dest.body, lang==String::L_TAINTED ? ilang:lang, body);
! 217: }
! 218: }
1.91 paf 219: }
220:
1.119 paf 221: // next: letters
222: dest.body<<body;
223:
1.117 paf 224: ASSERT_STRING_INVARIANT(dest);
1.75 paf 225: return dest;
1.51 parser 226: }
227:
1.109 paf 228: /** http://www.ietf.org/rfc/rfc2047.txt
229: RFC
230: (3) As a replacement for a 'word' entity within a 'phrase', for example,
231: one that precedes an address in a From, To, or Cc header. The ABNF
232: definition for 'phrase' from RFC 822 thus becomes:
233:
234: phrase = 1*( encoded-word / word )
235:
236: In this case the set of characters that may be used in a "Q"-encoded
237: 'encoded-word' is restricted to: <upper and lower case ASCII
238: letters, decimal digits, "!", "*", "+", "-", "/", "=", and "_"
239: (underscore, ASCII 95.)>. An 'encoded-word' that appears within a
240: 'phrase' MUST be separated from any adjacent 'word', 'text' or
241: 'special' by 'linear-white-space'.
242: ...
243: (2) The 8-bit hexadecimal value 20 (e.g., ISO-8859-1 SPACE) may be
244: represented as "_" (underscore, ASCII 95.). (This character may
245: not pass through some internetwork mail gateways, but its use
246: will greatly enhance readability of "Q" encoded data with mail
247: readers that do not support this encoding.) Note that the "_"
248: always represents hexadecimal 20, even if the SPACE character
249: occupies a different code position in the character set in use.
250:
251: paf: obviously,
252: without "=", or one could not differ "=E0" and "russian letter a"
253: and without "_", or in would mean 0x20
254: */
1.117 paf 255: inline bool mail_header_char_valid_within_Qencoded(char c) {
1.109 paf 256: return c>='A' && c<='Z'
257: || c>='a' && c<='Z'
258: || c>='0' && c<='9'
259: || strchr("!*+-/", c);
260: }
1.118 paf 261: inline bool addr_spec_soon(const char *src) {
1.124 paf 262: for(char c; (c=*src); src++)
1.118 paf 263: if(c=='<')
264: return true;
265: else if(!(c==' ' || c=='\t'))
266: return false;
267: return false;
268: }
1.117 paf 269: /**
270: RFC
271: Upper case should be used for hexadecimal digits "A" through "F"
272: The 8-bit hexadecimal value 20 (e.g., ISO-8859-1 SPACE)
273: may be represented as "_"
274: */
275: inline bool mail_header_nonspace_char(char c) {
276: return c != 0x20;
277: }
1.125 paf 278:
1.117 paf 279: inline void ec_append(CORD_ec& result, bool& optimize, bool& whitespace, CORD_pos pos, size_t size) {
280: while(size--) {
281: CORD_ec_append(result, CORD_pos_fetch(pos));
282: CORD_next(pos);
283: }
284: }
285: inline void pa_CORD_pos_advance(CORD_pos pos, size_t n) {
286: while(true) {
1.125 paf 287: long avail=CORD_pos_chars_left(pos);
288: if(avail<=0) {
1.117 paf 289: CORD_next(pos);
290: if(!--n)
291: break;
1.125 paf 292: } else if((size_t)avail<n) {
1.117 paf 293: CORD_pos_advance(pos, avail);
294: n-=avail;
295: } else { // avail>=n
296: CORD_pos_advance(pos, n);
297: break;
298: }
299: }
300: }
301:
1.119 paf 302: #ifndef DOXYGEN
303: struct Cstr_to_string_body_block_info {
304: // input
305: String::Language lang;
306: SQL_Connection* connection;
307: const Request_charsets* charsets;
308: const String::Body* body;
309:
310: // output
311: CORD_ec result;
312:
313: // private
314: CORD_pos pos;
315: size_t fragment_begin;
316: bool whitespace;
1.138 misha 317: const char* exception;
1.119 paf 318: };
319: #endif
320: int cstr_to_string_body_block(char alang, size_t fragment_length, Cstr_to_string_body_block_info* info) {
321: const String::Language fragment_lang=(String::Language)(unsigned char)alang;
322: bool& whitespace=info->whitespace;
323: size_t fragment_end=info->fragment_begin+fragment_length;
1.138 misha 324: //fprintf(stderr, "%d, %d =%s=\n", fragment_lang, fragment_length, info->body->cstr());
1.119 paf 325:
1.120 paf 326: String::Language to_lang=info->lang==String::L_UNSPECIFIED?fragment_lang:info->lang;
327: bool optimize=(to_lang & String::L_OPTIMIZE_BIT)!=0;
1.119 paf 328: if(!optimize)
329: whitespace=false;
330:
1.120 paf 331: switch(to_lang & ~String::L_OPTIMIZE_BIT) {
332: case String::L_CLEAN:
333: case String::L_TAINTED:
334: case String::L_AS_IS:
1.119 paf 335: // clean piece
336:
337: // tainted piece, but undefined untaint language
338: // for VString.as_double of tainted values
339: // for ^process{body} evaluation
340:
341: // tainted, untaint language: as-is
342: ec_append(info->result, optimize, whitespace, info->pos, fragment_length);
343: break;
1.120 paf 344: case String::L_FILE_SPEC:
1.119 paf 345: // tainted, untaint language: file [name]
1.140 misha 346: {
347: bool is1251=(info->charsets && info->charsets->source().NAME()=="WINDOWS-1251");
348: escape_fragment(
349: // Macintosh has problems with small Russian letter 'r'
350: if( is1251 && c=='\xF0' ) {
351: // fixing that letter for most common charset
352: to_char('p');
353: } else // fallback to default
354: encode(need_file_encode, '_', c);
355: );
356: }
1.119 paf 357: break;
1.144 misha 358: case String::L_FILE_POST:
359: {
360: escape_fragment(switch(c) {
361: case '\0': to_string("\\0"); break;
362: case '\\': to_string("\\\\"); break;
363: default: _default; break;
364: });
365: }
366: break;
1.120 paf 367: case String::L_URI:
1.119 paf 368: // tainted, untaint language: uri
369: {
370: const char *fragment_str=info->body->mid(info->fragment_begin, fragment_length).cstr();
371: // skip source [we use recoded version]
372: pa_CORD_pos_advance(info->pos, fragment_length);
373: String::C output(fragment_str, fragment_length);
374: if(info->charsets)
375: output=Charset::transcode(output,
376: info->charsets->source(),
377: info->charsets->client());
1.77 paf 378:
1.119 paf 379: char c;
1.124 paf 380: for(const char* src=output.str; (c=*src++); )
1.117 paf 381: encode(need_uri_encode, '%', c);
1.119 paf 382: }
383: break;
1.120 paf 384: case String::L_HTTP_HEADER:
1.119 paf 385: // tainted, untaint language: http-field-content-text
1.139 misha 386: // the same as L_URI BUT not transcoded into $response:charset before encoding
1.140 misha 387: escape_fragment(
1.119 paf 388: encode(need_uri_encode, '%', c);
389: );
390: break;
1.120 paf 391: case String::L_MAIL_HEADER:
1.119 paf 392: // tainted, untaint language: mail-header
393: // http://www.ietf.org/rfc/rfc2047.txt
394: if(info->charsets) {
395: size_t mail_size;
396: const char *mail_ptr=
397: info->body->mid(info->fragment_begin, mail_size=fragment_length).cstr();
398: // skip source [we use recoded version]
399: pa_CORD_pos_advance(info->pos, mail_size);
400:
401: const char* charset_name=info->charsets->mail().NAME().cstr();
402:
403: // Subject: Re: parser3: =?koi8-r?Q?=D3=C5=CD=C9=CE=C1=D2?=
404: bool to_quoted_printable=false;
405:
406: bool email=false;
407: uchar c;
1.124 paf 408: for(const char* src=mail_ptr; (c=(uchar)*src++); ) {
1.137 paf 409: if(c=='\r' || c=='\n')
410: c=' ';
1.130 paf 411: if(to_quoted_printable && (c==',' || c == '"' || addr_spec_soon(src-1/*position to 'c'*/))) {
1.119 paf 412: email=c=='<';
413: to_string("?=");
414: to_quoted_printable=false;
415: }
1.130 paf 416: //RFC + An 'encoded-word' MUST NOT appear in any portion of an 'addr-spec'.
1.119 paf 417: if(!email && (
418: !to_quoted_printable && (c & 0x80) // starting quote-printable-encoding on first 8bit char
419: || to_quoted_printable && !mail_header_char_valid_within_Qencoded(c)
420: )) {
421: if(!to_quoted_printable) {
422: to_string("=?");
423: to_string(charset_name);
424: to_string("?Q?");
425: to_quoted_printable=true;
1.105 paf 426: }
1.119 paf 427: encode(mail_header_nonspace_char, '=', '_');
428: } else
429: to_char(c);
430: if(c=='>')
431: email=false;
432: }
433: if(to_quoted_printable) // close
434: to_string("?=");
435:
436: } else
437: ec_append(info->result, optimize, whitespace, info->pos, fragment_length);
438: break;
1.120 paf 439: case String::L_SQL:
1.119 paf 440: // tainted, untaint language: sql
441: if(info->connection) {
442: const char *fragment_str=info->body->mid(info->fragment_begin, fragment_length).cstr();
443: // skip source [we use recoded version]
444: pa_CORD_pos_advance(info->pos, fragment_length);
445:
446: to_string(info->connection->quote(fragment_str, fragment_length));
1.138 misha 447: } else {
448: info->exception="untaint in SQL language failed - no connection specified";
449: info->fragment_begin=fragment_end;
450: return 1; // stop processing. can't throw exception here
451: }
1.119 paf 452: break;
1.120 paf 453: case String::L_JS:
1.140 misha 454: escape_fragment(switch(c) {
455: case '\n': to_string("\\n"); break;
1.119 paf 456: case '"': to_string("\\\""); break;
457: case '\'': to_string("\\'"); break;
458: case '\\': to_string("\\\\"); break;
459: case '\xFF': to_string("\\\xFF"); break;
1.146 misha 460: case '\r': to_string("\\r"); break;
1.133 paf 461: default: _default; break;
1.119 paf 462: });
463: break;
1.120 paf 464: case String::L_XML:
1.133 paf 465: // [2] Char ::= #x9 | #xA | #xD | [#x20-#xD7FF] | [#xE000-#xFFFD] | [#x10000-#x10FFFF]
1.140 misha 466: escape_fragment(switch(c) {
467: case '\x20':
468: case '\x9':
469: case '\xA':
1.133 paf 470: case '\xD': // this is usually removed on input
471: _default;
472: break;
1.119 paf 473: case '&': to_string("&"); break;
474: case '>': to_string(">"); break;
475: case '<': to_string("<"); break;
476: case '"': to_string("""); break;
477: case '\'': to_string("'"); break;
1.133 paf 478: default:
479: if(((unsigned char)c)<0x20) {
480: // fixing it, so that libxml would not result
481: // in fatal error parsing text
482: // though it really violates standard.
483: // to indicate there were an error
484: // replace bad char not to it's code,
485: // which we can do,
486: // but rather to '!' to show that input were actually
487: // invalid.
488: // life: shows that MSIE can somehow garble form values
489: // so that they contain these chars.
490: to_char('!');
491: } else {
492: _default;
493: }
494: break;
1.119 paf 495: });
496: break;
1.120 paf 497: case String::L_HTML:
1.140 misha 498: escape_fragment(switch(c) {
1.119 paf 499: case '&': to_string("&"); break;
500: case '>': to_string(">"); break;
501: case '<': to_string("<"); break;
502: case '"': to_string("""); break;
1.133 paf 503: default: _default; break;
1.119 paf 504: });
505: break;
1.136 paf 506: case String::L_REGEX:
507: // tainted, untaint language: regex
1.140 misha 508: escape_fragment(
1.136 paf 509: if(need_regex_escape(c))
510: to_char('\\')
511: _default;
512: );
513: break;
1.140 misha 514: case String::L_HTTP_COOKIE:
515: // tainted, untaint language: cookie (3.3.0 and higher: %uXXXX in UTF-8)
516: {
517: const char *fragment_str=info->body->mid(info->fragment_begin, fragment_length).cstr();
518: // skip source [we use recoded version]
519: pa_CORD_pos_advance(info->pos, fragment_length);
520: String::C output(fragment_str, fragment_length);
521:
522: output=Charset::escape(output, info->charsets->source());
523: //throw Exception(0, 0, output);
524: to_string(output);
525:
526: }
527: break;
1.119 paf 528: default:
1.127 paf 529: SAPI::abort("unknown untaint language #%d",
1.119 paf 530: static_cast<int>(to_lang)); // should never
531: break; // never
1.117 paf 532: }
1.55 parser 533:
1.119 paf 534: info->fragment_begin=fragment_end;
535:
536: return 0; // 0=continue
537: }
538:
539:
540: String::Body String::cstr_to_string_body(Language lang,
541: SQL_Connection* connection,
542: const Request_charsets *charsets) const {
1.134 paf 543: if(is_empty())
544: return String::Body();
1.119 paf 545:
546: Cstr_to_string_body_block_info info;
547: // input
548: info.lang=lang;
549: info.connection=connection;
550: info.charsets=charsets;
551: info.body=&body;
552: // output
553: CORD_ec_init(info.result);
554: // private
555: body.set_pos(info.pos, 0);
556: info.fragment_begin=0;
1.138 misha 557: info.exception=0;
1.119 paf 558: info.whitespace=true;
559:
1.134 paf 560: langs.for_each(body, cstr_to_string_body_block, &info);
1.138 misha 561: if(info.exception){
562: throw Exception(0,
563: 0,
564: info.exception);
565: }
566:
1.119 paf 567: return String::Body(CORD_ec_to_cord(info.result));
1.1 paf 568: }
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