Annotation of parser3/src/main/untaint.C, revision 1.143
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:
1.143 ! misha 8: static const char * const IDENT_UNTAINT_C="$Date: 2008-08-14 09:26:11 $";
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: */
188: String& String::append_to(String& dest, Language lang, bool forced) const {
189: if(is_empty())
190: return dest;
1.91 paf 191:
1.119 paf 192: // first: fragment infos
1.117 paf 193:
194: if(lang==L_PASS_APPENDED) // without language-change?
1.119 paf 195: dest.langs.append(dest.body, body.length(), langs);
1.117 paf 196: else if(forced) //forcing passed lang?
1.119 paf 197: dest.langs.append(dest.body, lang, length());
198: else {
1.122 paf 199: Append_fragment_info info={lang, &dest.langs, dest.body.length()};
1.119 paf 200: langs.for_each(body, lang&L_OPTIMIZE_BIT?
201: append_fragment_optimizing
202: :append_fragment_nonoptimizing, &info);
1.91 paf 203: }
204:
1.119 paf 205: // next: letters
206: dest.body<<body;
207:
1.117 paf 208: ASSERT_STRING_INVARIANT(dest);
1.75 paf 209: return dest;
1.51 parser 210: }
211:
1.109 paf 212: /** http://www.ietf.org/rfc/rfc2047.txt
213: RFC
214: (3) As a replacement for a 'word' entity within a 'phrase', for example,
215: one that precedes an address in a From, To, or Cc header. The ABNF
216: definition for 'phrase' from RFC 822 thus becomes:
217:
218: phrase = 1*( encoded-word / word )
219:
220: In this case the set of characters that may be used in a "Q"-encoded
221: 'encoded-word' is restricted to: <upper and lower case ASCII
222: letters, decimal digits, "!", "*", "+", "-", "/", "=", and "_"
223: (underscore, ASCII 95.)>. An 'encoded-word' that appears within a
224: 'phrase' MUST be separated from any adjacent 'word', 'text' or
225: 'special' by 'linear-white-space'.
226: ...
227: (2) The 8-bit hexadecimal value 20 (e.g., ISO-8859-1 SPACE) may be
228: represented as "_" (underscore, ASCII 95.). (This character may
229: not pass through some internetwork mail gateways, but its use
230: will greatly enhance readability of "Q" encoded data with mail
231: readers that do not support this encoding.) Note that the "_"
232: always represents hexadecimal 20, even if the SPACE character
233: occupies a different code position in the character set in use.
234:
235: paf: obviously,
236: without "=", or one could not differ "=E0" and "russian letter a"
237: and without "_", or in would mean 0x20
238: */
1.117 paf 239: inline bool mail_header_char_valid_within_Qencoded(char c) {
1.109 paf 240: return c>='A' && c<='Z'
241: || c>='a' && c<='Z'
242: || c>='0' && c<='9'
243: || strchr("!*+-/", c);
244: }
1.118 paf 245: inline bool addr_spec_soon(const char *src) {
1.124 paf 246: for(char c; (c=*src); src++)
1.118 paf 247: if(c=='<')
248: return true;
249: else if(!(c==' ' || c=='\t'))
250: return false;
251: return false;
252: }
1.117 paf 253: /**
254: RFC
255: Upper case should be used for hexadecimal digits "A" through "F"
256: The 8-bit hexadecimal value 20 (e.g., ISO-8859-1 SPACE)
257: may be represented as "_"
258: */
259: inline bool mail_header_nonspace_char(char c) {
260: return c != 0x20;
261: }
1.125 paf 262:
1.117 paf 263: inline void ec_append(CORD_ec& result, bool& optimize, bool& whitespace, CORD_pos pos, size_t size) {
264: while(size--) {
265: CORD_ec_append(result, CORD_pos_fetch(pos));
266: CORD_next(pos);
267: }
268: }
269: inline void pa_CORD_pos_advance(CORD_pos pos, size_t n) {
270: while(true) {
1.125 paf 271: long avail=CORD_pos_chars_left(pos);
272: if(avail<=0) {
1.117 paf 273: CORD_next(pos);
274: if(!--n)
275: break;
1.125 paf 276: } else if((size_t)avail<n) {
1.117 paf 277: CORD_pos_advance(pos, avail);
278: n-=avail;
279: } else { // avail>=n
280: CORD_pos_advance(pos, n);
281: break;
282: }
283: }
284: }
285:
1.119 paf 286: #ifndef DOXYGEN
287: struct Cstr_to_string_body_block_info {
288: // input
289: String::Language lang;
290: SQL_Connection* connection;
291: const Request_charsets* charsets;
292: const String::Body* body;
293:
294: // output
295: CORD_ec result;
296:
297: // private
298: CORD_pos pos;
299: size_t fragment_begin;
300: bool whitespace;
1.138 misha 301: const char* exception;
1.119 paf 302: };
303: #endif
304: int cstr_to_string_body_block(char alang, size_t fragment_length, Cstr_to_string_body_block_info* info) {
305: const String::Language fragment_lang=(String::Language)(unsigned char)alang;
306: bool& whitespace=info->whitespace;
307: size_t fragment_end=info->fragment_begin+fragment_length;
1.138 misha 308: //fprintf(stderr, "%d, %d =%s=\n", fragment_lang, fragment_length, info->body->cstr());
1.119 paf 309:
1.120 paf 310: String::Language to_lang=info->lang==String::L_UNSPECIFIED?fragment_lang:info->lang;
311: bool optimize=(to_lang & String::L_OPTIMIZE_BIT)!=0;
1.119 paf 312: if(!optimize)
313: whitespace=false;
314:
1.120 paf 315: switch(to_lang & ~String::L_OPTIMIZE_BIT) {
316: case String::L_CLEAN:
317: case String::L_TAINTED:
318: case String::L_AS_IS:
1.119 paf 319: // clean piece
320:
321: // tainted piece, but undefined untaint language
322: // for VString.as_double of tainted values
323: // for ^process{body} evaluation
324:
325: // tainted, untaint language: as-is
326: ec_append(info->result, optimize, whitespace, info->pos, fragment_length);
327: break;
1.120 paf 328: case String::L_FILE_SPEC:
1.119 paf 329: // tainted, untaint language: file [name]
1.140 misha 330: {
331: bool is1251=(info->charsets && info->charsets->source().NAME()=="WINDOWS-1251");
332: escape_fragment(
333: // Macintosh has problems with small Russian letter 'r'
334: if( is1251 && c=='\xF0' ) {
335: // fixing that letter for most common charset
336: to_char('p');
337: } else // fallback to default
338: encode(need_file_encode, '_', c);
339: );
340: }
1.119 paf 341: break;
1.120 paf 342: case String::L_URI:
1.119 paf 343: // tainted, untaint language: uri
344: {
345: const char *fragment_str=info->body->mid(info->fragment_begin, fragment_length).cstr();
346: // skip source [we use recoded version]
347: pa_CORD_pos_advance(info->pos, fragment_length);
348: String::C output(fragment_str, fragment_length);
349: if(info->charsets)
350: output=Charset::transcode(output,
351: info->charsets->source(),
352: info->charsets->client());
1.77 paf 353:
1.119 paf 354: char c;
1.124 paf 355: for(const char* src=output.str; (c=*src++); )
1.117 paf 356: encode(need_uri_encode, '%', c);
1.119 paf 357: }
358: break;
1.120 paf 359: case String::L_HTTP_HEADER:
1.119 paf 360: // tainted, untaint language: http-field-content-text
1.139 misha 361: // the same as L_URI BUT not transcoded into $response:charset before encoding
1.140 misha 362: escape_fragment(
1.119 paf 363: encode(need_uri_encode, '%', c);
364: );
365: break;
1.120 paf 366: case String::L_MAIL_HEADER:
1.119 paf 367: // tainted, untaint language: mail-header
368: // http://www.ietf.org/rfc/rfc2047.txt
369: if(info->charsets) {
370: size_t mail_size;
371: const char *mail_ptr=
372: info->body->mid(info->fragment_begin, mail_size=fragment_length).cstr();
373: // skip source [we use recoded version]
374: pa_CORD_pos_advance(info->pos, mail_size);
375:
376: const char* charset_name=info->charsets->mail().NAME().cstr();
377:
378: // Subject: Re: parser3: =?koi8-r?Q?=D3=C5=CD=C9=CE=C1=D2?=
379: bool to_quoted_printable=false;
380:
381: bool email=false;
382: uchar c;
1.124 paf 383: for(const char* src=mail_ptr; (c=(uchar)*src++); ) {
1.137 paf 384: if(c=='\r' || c=='\n')
385: c=' ';
1.130 paf 386: if(to_quoted_printable && (c==',' || c == '"' || addr_spec_soon(src-1/*position to 'c'*/))) {
1.119 paf 387: email=c=='<';
388: to_string("?=");
389: to_quoted_printable=false;
390: }
1.130 paf 391: //RFC + An 'encoded-word' MUST NOT appear in any portion of an 'addr-spec'.
1.119 paf 392: if(!email && (
393: !to_quoted_printable && (c & 0x80) // starting quote-printable-encoding on first 8bit char
394: || to_quoted_printable && !mail_header_char_valid_within_Qencoded(c)
395: )) {
396: if(!to_quoted_printable) {
397: to_string("=?");
398: to_string(charset_name);
399: to_string("?Q?");
400: to_quoted_printable=true;
1.105 paf 401: }
1.119 paf 402: encode(mail_header_nonspace_char, '=', '_');
403: } else
404: to_char(c);
405: if(c=='>')
406: email=false;
407: }
408: if(to_quoted_printable) // close
409: to_string("?=");
410:
411: } else
412: ec_append(info->result, optimize, whitespace, info->pos, fragment_length);
413: break;
1.120 paf 414: case String::L_SQL:
1.119 paf 415: // tainted, untaint language: sql
416: if(info->connection) {
417: const char *fragment_str=info->body->mid(info->fragment_begin, fragment_length).cstr();
418: // skip source [we use recoded version]
419: pa_CORD_pos_advance(info->pos, fragment_length);
420:
421: to_string(info->connection->quote(fragment_str, fragment_length));
1.138 misha 422: } else {
423: info->exception="untaint in SQL language failed - no connection specified";
424: info->fragment_begin=fragment_end;
425: return 1; // stop processing. can't throw exception here
426: }
1.119 paf 427: break;
1.120 paf 428: case String::L_JS:
1.140 misha 429: escape_fragment(switch(c) {
430: case '\n': to_string("\\n"); break;
1.119 paf 431: case '"': to_string("\\\""); break;
432: case '\'': to_string("\\'"); break;
433: case '\\': to_string("\\\\"); break;
434: case '\xFF': to_string("\\\xFF"); break;
1.133 paf 435: default: _default; break;
1.119 paf 436: });
437: break;
1.120 paf 438: case String::L_XML:
1.133 paf 439: // [2] Char ::= #x9 | #xA | #xD | [#x20-#xD7FF] | [#xE000-#xFFFD] | [#x10000-#x10FFFF]
1.140 misha 440: escape_fragment(switch(c) {
441: case '\x20':
442: case '\x9':
443: case '\xA':
1.133 paf 444: case '\xD': // this is usually removed on input
445: _default;
446: break;
1.119 paf 447: case '&': to_string("&"); break;
448: case '>': to_string(">"); break;
449: case '<': to_string("<"); break;
450: case '"': to_string("""); break;
451: case '\'': to_string("'"); break;
1.133 paf 452: default:
453: if(((unsigned char)c)<0x20) {
454: // fixing it, so that libxml would not result
455: // in fatal error parsing text
456: // though it really violates standard.
457: // to indicate there were an error
458: // replace bad char not to it's code,
459: // which we can do,
460: // but rather to '!' to show that input were actually
461: // invalid.
462: // life: shows that MSIE can somehow garble form values
463: // so that they contain these chars.
464: to_char('!');
465: } else {
466: _default;
467: }
468: break;
1.119 paf 469: });
470: break;
1.120 paf 471: case String::L_HTML:
1.140 misha 472: escape_fragment(switch(c) {
1.119 paf 473: case '&': to_string("&"); break;
474: case '>': to_string(">"); break;
475: case '<': to_string("<"); break;
476: case '"': to_string("""); break;
1.133 paf 477: default: _default; break;
1.119 paf 478: });
479: break;
1.136 paf 480: case String::L_REGEX:
481: // tainted, untaint language: regex
1.140 misha 482: escape_fragment(
1.136 paf 483: if(need_regex_escape(c))
484: to_char('\\')
485: _default;
486: );
487: break;
1.140 misha 488: case String::L_HTTP_COOKIE:
489: // tainted, untaint language: cookie (3.3.0 and higher: %uXXXX in UTF-8)
490: {
491: const char *fragment_str=info->body->mid(info->fragment_begin, fragment_length).cstr();
492: // skip source [we use recoded version]
493: pa_CORD_pos_advance(info->pos, fragment_length);
494: String::C output(fragment_str, fragment_length);
495:
496: output=Charset::escape(output, info->charsets->source());
497: //throw Exception(0, 0, output);
498: to_string(output);
499:
500: }
501: break;
1.119 paf 502: default:
1.127 paf 503: SAPI::abort("unknown untaint language #%d",
1.119 paf 504: static_cast<int>(to_lang)); // should never
505: break; // never
1.117 paf 506: }
1.55 parser 507:
1.119 paf 508: info->fragment_begin=fragment_end;
509:
510: return 0; // 0=continue
511: }
512:
513:
514: String::Body String::cstr_to_string_body(Language lang,
515: SQL_Connection* connection,
516: const Request_charsets *charsets) const {
1.134 paf 517: if(is_empty())
518: return String::Body();
1.119 paf 519:
520: Cstr_to_string_body_block_info info;
521: // input
522: info.lang=lang;
523: info.connection=connection;
524: info.charsets=charsets;
525: info.body=&body;
526: // output
527: CORD_ec_init(info.result);
528: // private
529: body.set_pos(info.pos, 0);
530: info.fragment_begin=0;
1.138 misha 531: info.exception=0;
1.119 paf 532: info.whitespace=true;
533:
1.134 paf 534: langs.for_each(body, cstr_to_string_body_block, &info);
1.138 misha 535: if(info.exception){
536: throw Exception(0,
537: 0,
538: info.exception);
539: }
540:
1.119 paf 541: return String::Body(CORD_ec_to_cord(info.result));
1.1 paf 542: }
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