Annotation of parser3/src/main/untaint.C, revision 1.134
1.7 paf 1: /** @file
1.8 paf 2: Parser: String class part: untaint mechanizm.
3:
1.129 paf 4: Copyright(c) 2001-2004 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:
1.134 ! paf 8: static const char * const IDENT_UNTAINT_C="$Date: 2004/07/01 07:20:10 $";
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: }
1.1 paf 95:
96: // String
97:
1.41 paf 98: /*
99: HTTP-header = field-name ":" [ field-value ] CRLF
100:
101: field-name = token
102: field-value = *( field-content | LWS )
103:
104: field-content = <the OCTETs making up the field-value
105: and consisting of either *TEXT or combinations
106: of token, tspecials, and quoted-string>
107:
108:
109:
110: word = token | quoted-string
111:
112: token = 1*<any CHAR except CTLs or tspecials>
113:
114:
115:
116: tspecials = "(" | ")" | "<" | ">" | "@"
117: | "," | ";" | ":" | "\" | <">
118: | "/" | "[" | "]" | "?" | "="
119: | "{" | "}" | SP | HT
120:
121: SP = <US-ASCII SP, space (32)>
122: HT = <US-ASCII HT, horizontal-tab (9)>
123:
124: LWS = [CRLF] 1*( SP | HT )
125: TEXT = <any OCTET except CTLs,
126: but including LWS>
127:
128: quoted-pair = "\" CHAR
129:
130: if(strchr("()<>@,;:\\\"/[]?={} \t", *ptr))
131: */
1.117 paf 132: inline bool need_quote_http_header(const char* ptr, size_t size) {
1.41 paf 133: for(; size--; ptr++)
1.42 paf 134: if(strchr(";\\\"= \t" /* excluded ()<>@, :/ ? []{} */, *ptr))
1.41 paf 135: return true;
136: return false;
137: }
138:
1.119 paf 139: #ifndef DOXYGEN
140: struct Append_fragment_info {
141: String::Language lang;
142: String::Languages* dest_languages;
143: size_t dest_body_plan_length;
144: };
145: #endif
146: int append_fragment_optimizing(char alang, size_t asize, Append_fragment_info* info) {
147: const String::Language lang=(String::Language)(unsigned char)alang;
148: // main idea here:
149: // tainted piece would get OPTIMIZED bit from 'lang'
150: // clean piece would be marked OPTIMIZED manually
151: // pieces with determined languages [not tainted|clean] would retain theirs langs
152: info->dest_languages->append(info->dest_body_plan_length,
1.120 paf 153: lang==String::L_TAINTED?
1.119 paf 154: info->lang
1.120 paf 155: :lang==String::L_CLEAN?
156: (String::Language)(String::L_CLEAN|String::L_OPTIMIZE_BIT) // ORing with OPTIMIZED flag
1.119 paf 157: :lang,
158: asize);
159: info->dest_body_plan_length+=asize;
160:
161: return 0; // 0=continue
162: }
163: int append_fragment_nonoptimizing(char alang, size_t asize, Append_fragment_info* info) {
164: const String::Language lang=(String::Language)(unsigned char)alang;
165: // The core idea: tainted pieces got marked with context's lang
166: info->dest_languages->append(info->dest_body_plan_length,
1.120 paf 167: lang==String::L_TAINTED?
1.119 paf 168: info->lang
169: :lang,
170: asize);
171: info->dest_body_plan_length+=asize;
172:
173: return 0; // 0=continue
174: }
1.117 paf 175:
1.91 paf 176: /**
1.117 paf 177: appends to other String,
1.91 paf 178: marking all tainted pieces of it with @a lang.
1.92 paf 179: or marking ALL pieces of it with a @a lang when @a forced to,
180: and propagating OPTIMIZE language bit.
1.117 paf 181: */
182: String& String::append_to(String& dest, Language lang, bool forced) const {
183: if(is_empty())
184: return dest;
1.91 paf 185:
1.119 paf 186: // first: fragment infos
1.117 paf 187:
188: if(lang==L_PASS_APPENDED) // without language-change?
1.119 paf 189: dest.langs.append(dest.body, body.length(), langs);
1.117 paf 190: else if(forced) //forcing passed lang?
1.119 paf 191: dest.langs.append(dest.body, lang, length());
192: else {
1.122 paf 193: Append_fragment_info info={lang, &dest.langs, dest.body.length()};
1.119 paf 194: langs.for_each(body, lang&L_OPTIMIZE_BIT?
195: append_fragment_optimizing
196: :append_fragment_nonoptimizing, &info);
1.91 paf 197: }
198:
1.119 paf 199: // next: letters
200: dest.body<<body;
201:
1.117 paf 202: ASSERT_STRING_INVARIANT(dest);
1.75 paf 203: return dest;
1.51 parser 204: }
205:
1.109 paf 206: /** http://www.ietf.org/rfc/rfc2047.txt
207: RFC
208: (3) As a replacement for a 'word' entity within a 'phrase', for example,
209: one that precedes an address in a From, To, or Cc header. The ABNF
210: definition for 'phrase' from RFC 822 thus becomes:
211:
212: phrase = 1*( encoded-word / word )
213:
214: In this case the set of characters that may be used in a "Q"-encoded
215: 'encoded-word' is restricted to: <upper and lower case ASCII
216: letters, decimal digits, "!", "*", "+", "-", "/", "=", and "_"
217: (underscore, ASCII 95.)>. An 'encoded-word' that appears within a
218: 'phrase' MUST be separated from any adjacent 'word', 'text' or
219: 'special' by 'linear-white-space'.
220: ...
221: (2) The 8-bit hexadecimal value 20 (e.g., ISO-8859-1 SPACE) may be
222: represented as "_" (underscore, ASCII 95.). (This character may
223: not pass through some internetwork mail gateways, but its use
224: will greatly enhance readability of "Q" encoded data with mail
225: readers that do not support this encoding.) Note that the "_"
226: always represents hexadecimal 20, even if the SPACE character
227: occupies a different code position in the character set in use.
228:
229: paf: obviously,
230: without "=", or one could not differ "=E0" and "russian letter a"
231: and without "_", or in would mean 0x20
232: */
1.117 paf 233: inline bool mail_header_char_valid_within_Qencoded(char c) {
1.109 paf 234: return c>='A' && c<='Z'
235: || c>='a' && c<='Z'
236: || c>='0' && c<='9'
237: || strchr("!*+-/", c);
238: }
1.118 paf 239: inline bool addr_spec_soon(const char *src) {
1.124 paf 240: for(char c; (c=*src); src++)
1.118 paf 241: if(c=='<')
242: return true;
243: else if(!(c==' ' || c=='\t'))
244: return false;
245: return false;
246: }
1.117 paf 247: /**
248: RFC
249: Upper case should be used for hexadecimal digits "A" through "F"
250: The 8-bit hexadecimal value 20 (e.g., ISO-8859-1 SPACE)
251: may be represented as "_"
252: */
253: inline bool mail_header_nonspace_char(char c) {
254: return c != 0x20;
255: }
1.125 paf 256:
1.117 paf 257: inline void ec_append(CORD_ec& result, bool& optimize, bool& whitespace, CORD_pos pos, size_t size) {
258: while(size--) {
259: CORD_ec_append(result, CORD_pos_fetch(pos));
260: CORD_next(pos);
261: }
262: }
263: inline void pa_CORD_pos_advance(CORD_pos pos, size_t n) {
264: while(true) {
1.125 paf 265: long avail=CORD_pos_chars_left(pos);
266: if(avail<=0) {
1.117 paf 267: CORD_next(pos);
268: if(!--n)
269: break;
1.125 paf 270: } else if((size_t)avail<n) {
1.117 paf 271: CORD_pos_advance(pos, avail);
272: n-=avail;
273: } else { // avail>=n
274: CORD_pos_advance(pos, n);
275: break;
276: }
277: }
278: }
279:
1.119 paf 280: #ifndef DOXYGEN
281: struct Cstr_to_string_body_block_info {
282: // input
283: String::Language lang;
284: SQL_Connection* connection;
285: const Request_charsets* charsets;
286: const String::Body* body;
287:
288: // output
289: CORD_ec result;
290:
291: // private
292: CORD_pos pos;
293: size_t fragment_begin;
294: bool whitespace;
295: };
296: #endif
297: int cstr_to_string_body_block(char alang, size_t fragment_length, Cstr_to_string_body_block_info* info) {
298: const String::Language fragment_lang=(String::Language)(unsigned char)alang;
299: bool& whitespace=info->whitespace;
300: size_t fragment_end=info->fragment_begin+fragment_length;
301: //fprintf(stderr, "%d, %d\n", fragment.lang, fragment.length);
1.117 paf 302:
1.119 paf 303:
1.120 paf 304: String::Language to_lang=info->lang==String::L_UNSPECIFIED?fragment_lang:info->lang;
305: bool optimize=(to_lang & String::L_OPTIMIZE_BIT)!=0;
1.119 paf 306: if(!optimize)
307: whitespace=false;
308:
1.120 paf 309: switch(to_lang & ~String::L_OPTIMIZE_BIT) {
310: case String::L_CLEAN:
311: case String::L_TAINTED:
312: case String::L_AS_IS:
1.119 paf 313: // clean piece
314:
315: // tainted piece, but undefined untaint language
316: // for VString.as_double of tainted values
317: // for ^process{body} evaluation
318:
319: // tainted, untaint language: as-is
320: ec_append(info->result, optimize, whitespace, info->pos, fragment_length);
321: break;
1.120 paf 322: case String::L_FILE_SPEC:
1.119 paf 323: // tainted, untaint language: file [name]
324: escape(
325: // Macintosh has problems with small Russian letter 'r'
326: if( c=='\xF0' && info->charsets && info->charsets->source().NAME()=="WINDOWS-1251" ) {
327: // fixing that letter for most common charset
328: to_char('p');
329: } else // fallback to default
330: encode(need_file_encode, '_', c);
331: );
332: break;
1.120 paf 333: case String::L_URI:
1.119 paf 334: // tainted, untaint language: uri
335: {
336: const char *fragment_str=info->body->mid(info->fragment_begin, fragment_length).cstr();
337: // skip source [we use recoded version]
338: pa_CORD_pos_advance(info->pos, fragment_length);
339: String::C output(fragment_str, fragment_length);
340: if(info->charsets)
341: output=Charset::transcode(output,
342: info->charsets->source(),
343: info->charsets->client());
1.77 paf 344:
1.119 paf 345: char c;
1.124 paf 346: for(const char* src=output.str; (c=*src++); )
1.117 paf 347: encode(need_uri_encode, '%', c);
1.119 paf 348: }
349: break;
1.120 paf 350: case String::L_HTTP_HEADER:
1.119 paf 351: // tainted, untaint language: http-field-content-text
352: escape(
353: encode(need_uri_encode, '%', c);
354: );
355: break;
1.120 paf 356: case String::L_MAIL_HEADER:
1.119 paf 357: // tainted, untaint language: mail-header
358: // http://www.ietf.org/rfc/rfc2047.txt
359: if(info->charsets) {
360: size_t mail_size;
361: const char *mail_ptr=
362: info->body->mid(info->fragment_begin, mail_size=fragment_length).cstr();
363: // skip source [we use recoded version]
364: pa_CORD_pos_advance(info->pos, mail_size);
365:
366: const char* charset_name=info->charsets->mail().NAME().cstr();
367:
368: // Subject: Re: parser3: =?koi8-r?Q?=D3=C5=CD=C9=CE=C1=D2?=
369: bool to_quoted_printable=false;
370:
371: bool email=false;
372: uchar c;
1.124 paf 373: for(const char* src=mail_ptr; (c=(uchar)*src++); ) {
1.130 paf 374: if(to_quoted_printable && (c==',' || c == '"' || addr_spec_soon(src-1/*position to 'c'*/))) {
1.119 paf 375: email=c=='<';
376: to_string("?=");
377: to_quoted_printable=false;
378: }
1.130 paf 379: //RFC + An 'encoded-word' MUST NOT appear in any portion of an 'addr-spec'.
1.119 paf 380: if(!email && (
381: !to_quoted_printable && (c & 0x80) // starting quote-printable-encoding on first 8bit char
382: || to_quoted_printable && !mail_header_char_valid_within_Qencoded(c)
383: )) {
384: if(!to_quoted_printable) {
385: to_string("=?");
386: to_string(charset_name);
387: to_string("?Q?");
388: to_quoted_printable=true;
1.105 paf 389: }
1.119 paf 390: encode(mail_header_nonspace_char, '=', '_');
391: } else
392: to_char(c);
393: if(c=='>')
394: email=false;
395: }
396: if(to_quoted_printable) // close
397: to_string("?=");
398:
399: } else
400: ec_append(info->result, optimize, whitespace, info->pos, fragment_length);
401: break;
1.120 paf 402: case String::L_SQL:
1.119 paf 403: // tainted, untaint language: sql
404: if(info->connection) {
405: const char *fragment_str=info->body->mid(info->fragment_begin, fragment_length).cstr();
406: // skip source [we use recoded version]
407: pa_CORD_pos_advance(info->pos, fragment_length);
408:
409: to_string(info->connection->quote(fragment_str, fragment_length));
410: } else
411: throw Exception(0,
412: 0,
413: "untaint in SQL language failed - no connection specified");
414: break;
1.120 paf 415: case String::L_JS:
1.119 paf 416: escape(switch(c) {
417: case '"': to_string("\\\""); break;
418: case '\'': to_string("\\'"); break;
419: case '\n': to_string("\\n"); break;
420: case '\\': to_string("\\\\"); break;
421: case '\xFF': to_string("\\\xFF"); break;
1.133 paf 422: default: _default; break;
1.119 paf 423: });
424: break;
1.120 paf 425: case String::L_XML:
1.133 paf 426: // [2] Char ::= #x9 | #xA | #xD | [#x20-#xD7FF] | [#xE000-#xFFFD] | [#x10000-#x10FFFF]
1.119 paf 427: escape(switch(c) {
1.133 paf 428: case '\x9':
429: case '\xA':
430: case '\xD': // this is usually removed on input
431: _default;
432: break;
1.119 paf 433: case '&': to_string("&"); break;
434: case '>': to_string(">"); break;
435: case '<': to_string("<"); break;
436: case '"': to_string("""); break;
437: case '\'': to_string("'"); break;
1.133 paf 438: default:
439: if(((unsigned char)c)<0x20) {
440: // fixing it, so that libxml would not result
441: // in fatal error parsing text
442: // though it really violates standard.
443: // to indicate there were an error
444: // replace bad char not to it's code,
445: // which we can do,
446: // but rather to '!' to show that input were actually
447: // invalid.
448: // life: shows that MSIE can somehow garble form values
449: // so that they contain these chars.
450: to_char('!');
451: } else {
452: _default;
453: }
454: break;
1.119 paf 455: });
456: break;
1.120 paf 457: case String::L_HTML:
1.119 paf 458: escape(switch(c) {
459: case '&': to_string("&"); break;
460: case '>': to_string(">"); break;
461: case '<': to_string("<"); break;
462: case '"': to_string("""); break;
1.133 paf 463: default: _default; break;
1.119 paf 464: });
465: break;
466: default:
1.127 paf 467: SAPI::abort("unknown untaint language #%d",
1.119 paf 468: static_cast<int>(to_lang)); // should never
469: break; // never
1.117 paf 470: }
1.55 parser 471:
1.119 paf 472: info->fragment_begin=fragment_end;
473:
474: return 0; // 0=continue
475: }
476:
477:
478: String::Body String::cstr_to_string_body(Language lang,
479: SQL_Connection* connection,
480: const Request_charsets *charsets) const {
1.134 ! paf 481: if(is_empty())
! 482: return String::Body();
1.119 paf 483:
484: Cstr_to_string_body_block_info info;
485: // input
486: info.lang=lang;
487: info.connection=connection;
488: info.charsets=charsets;
489: info.body=&body;
490: // output
491: CORD_ec_init(info.result);
492: // private
493: body.set_pos(info.pos, 0);
494: info.fragment_begin=0;
495: info.whitespace=true;
496:
1.134 ! paf 497: langs.for_each(body, cstr_to_string_body_block, &info);
1.119 paf 498: return String::Body(CORD_ec_to_cord(info.result));
1.1 paf 499: }
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