Annotation of parser3/src/main/untaint.C, revision 1.136
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:
1.136 ! paf 8: static const char * const IDENT_UNTAINT_C="$Date: 2005/08/09 08:14:53 $";
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:
1.18 paf 56: #define escape(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(
77: "*?'\"<>|"
78: #ifndef WIN32
1.116 paf 79: ":\\"
1.31 paf 80: #endif
1.108 paf 81: , c)!=0;
1.9 paf 82: }
1.5 paf 83: inline bool need_uri_encode(unsigned char c){
1.13 paf 84: if((c>='0') &&(c<='9') ||(c>='A') &&(c<='Z') ||(c>='a') &&(c<='z'))
1.4 paf 85: return false;
86:
1.5 paf 87: return !strchr("_-./", c);
88: }
1.36 paf 89: inline bool need_http_header_encode(unsigned char c){
1.18 paf 90: if(strchr(" , :", c))
1.5 paf 91: return false;
92:
93: return need_uri_encode(c);
1.4 paf 94: }
1.136 ! paf 95: inline bool need_regex_escape(unsigned char c){
! 96: return strchr("\\^$.[]|()?*+{}", c)!=0;
! 97: }
1.1 paf 98:
99: // String
100:
1.41 paf 101: /*
102: HTTP-header = field-name ":" [ field-value ] CRLF
103:
104: field-name = token
105: field-value = *( field-content | LWS )
106:
107: field-content = <the OCTETs making up the field-value
108: and consisting of either *TEXT or combinations
109: of token, tspecials, and quoted-string>
110:
111:
112:
113: word = token | quoted-string
114:
115: token = 1*<any CHAR except CTLs or tspecials>
116:
117:
118:
119: tspecials = "(" | ")" | "<" | ">" | "@"
120: | "," | ";" | ":" | "\" | <">
121: | "/" | "[" | "]" | "?" | "="
122: | "{" | "}" | SP | HT
123:
124: SP = <US-ASCII SP, space (32)>
125: HT = <US-ASCII HT, horizontal-tab (9)>
126:
127: LWS = [CRLF] 1*( SP | HT )
128: TEXT = <any OCTET except CTLs,
129: but including LWS>
130:
131: quoted-pair = "\" CHAR
132:
133: if(strchr("()<>@,;:\\\"/[]?={} \t", *ptr))
134: */
1.117 paf 135: inline bool need_quote_http_header(const char* ptr, size_t size) {
1.41 paf 136: for(; size--; ptr++)
1.42 paf 137: if(strchr(";\\\"= \t" /* excluded ()<>@, :/ ? []{} */, *ptr))
1.41 paf 138: return true;
139: return false;
140: }
141:
1.119 paf 142: #ifndef DOXYGEN
143: struct Append_fragment_info {
144: String::Language lang;
145: String::Languages* dest_languages;
146: size_t dest_body_plan_length;
147: };
148: #endif
149: int append_fragment_optimizing(char alang, size_t asize, Append_fragment_info* info) {
150: const String::Language lang=(String::Language)(unsigned char)alang;
151: // main idea here:
152: // tainted piece would get OPTIMIZED bit from 'lang'
153: // clean piece would be marked OPTIMIZED manually
154: // pieces with determined languages [not tainted|clean] would retain theirs langs
155: info->dest_languages->append(info->dest_body_plan_length,
1.120 paf 156: lang==String::L_TAINTED?
1.119 paf 157: info->lang
1.120 paf 158: :lang==String::L_CLEAN?
159: (String::Language)(String::L_CLEAN|String::L_OPTIMIZE_BIT) // ORing with OPTIMIZED flag
1.119 paf 160: :lang,
161: asize);
162: info->dest_body_plan_length+=asize;
163:
164: return 0; // 0=continue
165: }
166: int append_fragment_nonoptimizing(char alang, size_t asize, Append_fragment_info* info) {
167: const String::Language lang=(String::Language)(unsigned char)alang;
168: // The core idea: tainted pieces got marked with context's lang
169: info->dest_languages->append(info->dest_body_plan_length,
1.120 paf 170: lang==String::L_TAINTED?
1.119 paf 171: info->lang
172: :lang,
173: asize);
174: info->dest_body_plan_length+=asize;
175:
176: return 0; // 0=continue
177: }
1.117 paf 178:
1.91 paf 179: /**
1.117 paf 180: appends to other String,
1.91 paf 181: marking all tainted pieces of it with @a lang.
1.92 paf 182: or marking ALL pieces of it with a @a lang when @a forced to,
183: and propagating OPTIMIZE language bit.
1.117 paf 184: */
185: String& String::append_to(String& dest, Language lang, bool forced) const {
186: if(is_empty())
187: return dest;
1.91 paf 188:
1.119 paf 189: // first: fragment infos
1.117 paf 190:
191: if(lang==L_PASS_APPENDED) // without language-change?
1.119 paf 192: dest.langs.append(dest.body, body.length(), langs);
1.117 paf 193: else if(forced) //forcing passed lang?
1.119 paf 194: dest.langs.append(dest.body, lang, length());
195: else {
1.122 paf 196: Append_fragment_info info={lang, &dest.langs, dest.body.length()};
1.119 paf 197: langs.for_each(body, lang&L_OPTIMIZE_BIT?
198: append_fragment_optimizing
199: :append_fragment_nonoptimizing, &info);
1.91 paf 200: }
201:
1.119 paf 202: // next: letters
203: dest.body<<body;
204:
1.117 paf 205: ASSERT_STRING_INVARIANT(dest);
1.75 paf 206: return dest;
1.51 parser 207: }
208:
1.109 paf 209: /** http://www.ietf.org/rfc/rfc2047.txt
210: RFC
211: (3) As a replacement for a 'word' entity within a 'phrase', for example,
212: one that precedes an address in a From, To, or Cc header. The ABNF
213: definition for 'phrase' from RFC 822 thus becomes:
214:
215: phrase = 1*( encoded-word / word )
216:
217: In this case the set of characters that may be used in a "Q"-encoded
218: 'encoded-word' is restricted to: <upper and lower case ASCII
219: letters, decimal digits, "!", "*", "+", "-", "/", "=", and "_"
220: (underscore, ASCII 95.)>. An 'encoded-word' that appears within a
221: 'phrase' MUST be separated from any adjacent 'word', 'text' or
222: 'special' by 'linear-white-space'.
223: ...
224: (2) The 8-bit hexadecimal value 20 (e.g., ISO-8859-1 SPACE) may be
225: represented as "_" (underscore, ASCII 95.). (This character may
226: not pass through some internetwork mail gateways, but its use
227: will greatly enhance readability of "Q" encoded data with mail
228: readers that do not support this encoding.) Note that the "_"
229: always represents hexadecimal 20, even if the SPACE character
230: occupies a different code position in the character set in use.
231:
232: paf: obviously,
233: without "=", or one could not differ "=E0" and "russian letter a"
234: and without "_", or in would mean 0x20
235: */
1.117 paf 236: inline bool mail_header_char_valid_within_Qencoded(char c) {
1.109 paf 237: return c>='A' && c<='Z'
238: || c>='a' && c<='Z'
239: || c>='0' && c<='9'
240: || strchr("!*+-/", c);
241: }
1.118 paf 242: inline bool addr_spec_soon(const char *src) {
1.124 paf 243: for(char c; (c=*src); src++)
1.118 paf 244: if(c=='<')
245: return true;
246: else if(!(c==' ' || c=='\t'))
247: return false;
248: return false;
249: }
1.117 paf 250: /**
251: RFC
252: Upper case should be used for hexadecimal digits "A" through "F"
253: The 8-bit hexadecimal value 20 (e.g., ISO-8859-1 SPACE)
254: may be represented as "_"
255: */
256: inline bool mail_header_nonspace_char(char c) {
257: return c != 0x20;
258: }
1.125 paf 259:
1.117 paf 260: inline void ec_append(CORD_ec& result, bool& optimize, bool& whitespace, CORD_pos pos, size_t size) {
261: while(size--) {
262: CORD_ec_append(result, CORD_pos_fetch(pos));
263: CORD_next(pos);
264: }
265: }
266: inline void pa_CORD_pos_advance(CORD_pos pos, size_t n) {
267: while(true) {
1.125 paf 268: long avail=CORD_pos_chars_left(pos);
269: if(avail<=0) {
1.117 paf 270: CORD_next(pos);
271: if(!--n)
272: break;
1.125 paf 273: } else if((size_t)avail<n) {
1.117 paf 274: CORD_pos_advance(pos, avail);
275: n-=avail;
276: } else { // avail>=n
277: CORD_pos_advance(pos, n);
278: break;
279: }
280: }
281: }
282:
1.119 paf 283: #ifndef DOXYGEN
284: struct Cstr_to_string_body_block_info {
285: // input
286: String::Language lang;
287: SQL_Connection* connection;
288: const Request_charsets* charsets;
289: const String::Body* body;
290:
291: // output
292: CORD_ec result;
293:
294: // private
295: CORD_pos pos;
296: size_t fragment_begin;
297: bool whitespace;
298: };
299: #endif
300: int cstr_to_string_body_block(char alang, size_t fragment_length, Cstr_to_string_body_block_info* info) {
301: const String::Language fragment_lang=(String::Language)(unsigned char)alang;
302: bool& whitespace=info->whitespace;
303: size_t fragment_end=info->fragment_begin+fragment_length;
304: //fprintf(stderr, "%d, %d\n", fragment.lang, fragment.length);
1.117 paf 305:
1.119 paf 306:
1.120 paf 307: String::Language to_lang=info->lang==String::L_UNSPECIFIED?fragment_lang:info->lang;
308: bool optimize=(to_lang & String::L_OPTIMIZE_BIT)!=0;
1.119 paf 309: if(!optimize)
310: whitespace=false;
311:
1.120 paf 312: switch(to_lang & ~String::L_OPTIMIZE_BIT) {
313: case String::L_CLEAN:
314: case String::L_TAINTED:
315: case String::L_AS_IS:
1.119 paf 316: // clean piece
317:
318: // tainted piece, but undefined untaint language
319: // for VString.as_double of tainted values
320: // for ^process{body} evaluation
321:
322: // tainted, untaint language: as-is
323: ec_append(info->result, optimize, whitespace, info->pos, fragment_length);
324: break;
1.120 paf 325: case String::L_FILE_SPEC:
1.119 paf 326: // tainted, untaint language: file [name]
327: escape(
328: // Macintosh has problems with small Russian letter 'r'
329: if( c=='\xF0' && info->charsets && info->charsets->source().NAME()=="WINDOWS-1251" ) {
330: // fixing that letter for most common charset
331: to_char('p');
332: } else // fallback to default
333: encode(need_file_encode, '_', c);
334: );
335: break;
1.120 paf 336: case String::L_URI:
1.119 paf 337: // tainted, untaint language: uri
338: {
339: const char *fragment_str=info->body->mid(info->fragment_begin, fragment_length).cstr();
340: // skip source [we use recoded version]
341: pa_CORD_pos_advance(info->pos, fragment_length);
342: String::C output(fragment_str, fragment_length);
343: if(info->charsets)
344: output=Charset::transcode(output,
345: info->charsets->source(),
346: info->charsets->client());
1.77 paf 347:
1.119 paf 348: char c;
1.124 paf 349: for(const char* src=output.str; (c=*src++); )
1.117 paf 350: encode(need_uri_encode, '%', c);
1.119 paf 351: }
352: break;
1.120 paf 353: case String::L_HTTP_HEADER:
1.119 paf 354: // tainted, untaint language: http-field-content-text
355: escape(
356: encode(need_uri_encode, '%', c);
357: );
358: break;
1.120 paf 359: case String::L_MAIL_HEADER:
1.119 paf 360: // tainted, untaint language: mail-header
361: // http://www.ietf.org/rfc/rfc2047.txt
362: if(info->charsets) {
363: size_t mail_size;
364: const char *mail_ptr=
365: info->body->mid(info->fragment_begin, mail_size=fragment_length).cstr();
366: // skip source [we use recoded version]
367: pa_CORD_pos_advance(info->pos, mail_size);
368:
369: const char* charset_name=info->charsets->mail().NAME().cstr();
370:
371: // Subject: Re: parser3: =?koi8-r?Q?=D3=C5=CD=C9=CE=C1=D2?=
372: bool to_quoted_printable=false;
373:
374: bool email=false;
375: uchar c;
1.124 paf 376: for(const char* src=mail_ptr; (c=(uchar)*src++); ) {
1.130 paf 377: if(to_quoted_printable && (c==',' || c == '"' || addr_spec_soon(src-1/*position to 'c'*/))) {
1.119 paf 378: email=c=='<';
379: to_string("?=");
380: to_quoted_printable=false;
381: }
1.130 paf 382: //RFC + An 'encoded-word' MUST NOT appear in any portion of an 'addr-spec'.
1.119 paf 383: if(!email && (
384: !to_quoted_printable && (c & 0x80) // starting quote-printable-encoding on first 8bit char
385: || to_quoted_printable && !mail_header_char_valid_within_Qencoded(c)
386: )) {
387: if(!to_quoted_printable) {
388: to_string("=?");
389: to_string(charset_name);
390: to_string("?Q?");
391: to_quoted_printable=true;
1.105 paf 392: }
1.119 paf 393: encode(mail_header_nonspace_char, '=', '_');
394: } else
395: to_char(c);
396: if(c=='>')
397: email=false;
398: }
399: if(to_quoted_printable) // close
400: to_string("?=");
401:
402: } else
403: ec_append(info->result, optimize, whitespace, info->pos, fragment_length);
404: break;
1.120 paf 405: case String::L_SQL:
1.119 paf 406: // tainted, untaint language: sql
407: if(info->connection) {
408: const char *fragment_str=info->body->mid(info->fragment_begin, fragment_length).cstr();
409: // skip source [we use recoded version]
410: pa_CORD_pos_advance(info->pos, fragment_length);
411:
412: to_string(info->connection->quote(fragment_str, fragment_length));
413: } else
414: throw Exception(0,
415: 0,
416: "untaint in SQL language failed - no connection specified");
417: break;
1.120 paf 418: case String::L_JS:
1.119 paf 419: escape(switch(c) {
420: case '"': to_string("\\\""); break;
421: case '\'': to_string("\\'"); break;
422: case '\n': to_string("\\n"); break;
423: case '\\': to_string("\\\\"); break;
424: case '\xFF': to_string("\\\xFF"); break;
1.133 paf 425: default: _default; break;
1.119 paf 426: });
427: break;
1.120 paf 428: case String::L_XML:
1.133 paf 429: // [2] Char ::= #x9 | #xA | #xD | [#x20-#xD7FF] | [#xE000-#xFFFD] | [#x10000-#x10FFFF]
1.119 paf 430: escape(switch(c) {
1.133 paf 431: case '\x9':
432: case '\xA':
433: case '\xD': // this is usually removed on input
434: _default;
435: break;
1.119 paf 436: case '&': to_string("&"); break;
437: case '>': to_string(">"); break;
438: case '<': to_string("<"); break;
439: case '"': to_string("""); break;
440: case '\'': to_string("'"); break;
1.133 paf 441: default:
442: if(((unsigned char)c)<0x20) {
443: // fixing it, so that libxml would not result
444: // in fatal error parsing text
445: // though it really violates standard.
446: // to indicate there were an error
447: // replace bad char not to it's code,
448: // which we can do,
449: // but rather to '!' to show that input were actually
450: // invalid.
451: // life: shows that MSIE can somehow garble form values
452: // so that they contain these chars.
453: to_char('!');
454: } else {
455: _default;
456: }
457: break;
1.119 paf 458: });
459: break;
1.120 paf 460: case String::L_HTML:
1.119 paf 461: escape(switch(c) {
462: case '&': to_string("&"); break;
463: case '>': to_string(">"); break;
464: case '<': to_string("<"); break;
465: case '"': to_string("""); break;
1.133 paf 466: default: _default; break;
1.119 paf 467: });
468: break;
1.136 ! paf 469: case String::L_REGEX:
! 470: // tainted, untaint language: regex
! 471: escape(
! 472: if(need_regex_escape(c))
! 473: to_char('\\')
! 474: _default;
! 475: );
! 476: break;
1.119 paf 477: default:
1.127 paf 478: SAPI::abort("unknown untaint language #%d",
1.119 paf 479: static_cast<int>(to_lang)); // should never
480: break; // never
1.117 paf 481: }
1.55 parser 482:
1.119 paf 483: info->fragment_begin=fragment_end;
484:
485: return 0; // 0=continue
486: }
487:
488:
489: String::Body String::cstr_to_string_body(Language lang,
490: SQL_Connection* connection,
491: const Request_charsets *charsets) const {
1.134 paf 492: if(is_empty())
493: return String::Body();
1.119 paf 494:
495: Cstr_to_string_body_block_info info;
496: // input
497: info.lang=lang;
498: info.connection=connection;
499: info.charsets=charsets;
500: info.body=&body;
501: // output
502: CORD_ec_init(info.result);
503: // private
504: body.set_pos(info.pos, 0);
505: info.fragment_begin=0;
506: info.whitespace=true;
507:
1.134 paf 508: langs.for_each(body, cstr_to_string_body_block, &info);
1.119 paf 509: return String::Body(CORD_ec_to_cord(info.result));
1.1 paf 510: }
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