Annotation of parser3/src/main/untaint.C, revision 1.123
1.7 paf 1: /** @file
1.8 paf 2: Parser: String class part: untaint mechanizm.
3:
1.117 paf 4: Copyright(c) 2001-2003 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:
1.123 ! paf 8: static const char * const IDENT_UNTAINT_C="$Date: 2003/09/26 15:45:24 $";
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: #define PA_SQL
27:
28: #ifdef PA_SQL
29: #include "pa_sql_connection.h"
1.91 paf 30: #endif
31:
1.117 paf 32: // defines
33:
34:
35: #undef CORD_ec_append
36: // redefining to intercept flushes and implement whitespace optimization
37: // of all consequent white space chars leaving only first one
38: #define CORD_ec_append(x, c) \
39: { \
40: bool skip=false; \
41: if(optimize) switch(c) { \
42: case ' ': case '\r': case '\n': case '\t': \
43: if(whitespace) \
44: skip=true; /*skipping subsequent*/ \
45: else \
46: whitespace=true; \
47: break; \
48: default: \
49: whitespace=false; \
50: break; \
51: } \
52: if(!skip) { \
53: if ((x)[0].ec_bufptr == (x)[0].ec_buf + CORD_BUFSZ) { \
54: CORD_ec_flush_buf(x); \
55: } \
56: *((x)[0].ec_bufptr)++ = (c); \
57: } \
58: }
59:
60:
1.18 paf 61: #define escape(action) \
1.119 paf 62: for(; fragment_length--; CORD_next(info->pos)) { \
63: char c=CORD_pos_fetch(info->pos); \
1.117 paf 64: action \
1.1 paf 65: }
1.119 paf 66: #define _default default: CORD_ec_append(info->result, c); break
1.117 paf 67: #define encode(need_encode_func, prefix, otherwise) \
68: if(need_encode_func(c)) { \
69: static const char* hex="0123456789ABCDEF"; \
1.119 paf 70: CORD_ec_append(info->result, prefix); \
71: CORD_ec_append(info->result, hex[((unsigned char)c)/0x10]); \
72: CORD_ec_append(info->result, hex[((unsigned char)c)%0x10]); \
1.117 paf 73: } else \
1.119 paf 74: CORD_ec_append(info->result, otherwise);
75: #define to_char(c) CORD_ec_append(info->result, c)
76: #define to_string(s) CORD_ec_append_cord(info->result, s)
1.4 paf 77:
1.9 paf 78: inline bool need_file_encode(unsigned char c){
1.108 paf 79: // russian letters and space ENABLED
80: // encoding only these...
81: return strchr(
82: "*?'\"<>|"
83: #ifndef WIN32
1.116 paf 84: ":\\"
1.31 paf 85: #endif
1.108 paf 86: , c)!=0;
1.9 paf 87: }
1.5 paf 88: inline bool need_uri_encode(unsigned char c){
1.13 paf 89: if((c>='0') &&(c<='9') ||(c>='A') &&(c<='Z') ||(c>='a') &&(c<='z'))
1.4 paf 90: return false;
91:
1.5 paf 92: return !strchr("_-./", c);
93: }
1.36 paf 94: inline bool need_http_header_encode(unsigned char c){
1.18 paf 95: if(strchr(" , :", c))
1.5 paf 96: return false;
97:
98: return need_uri_encode(c);
1.4 paf 99: }
1.1 paf 100:
1.56 parser 101: //
102:
1.117 paf 103: static const char* String_Untaint_lang_name[]={
1.56 parser 104: "U", ///< zero value handy for hash lookup @see untaint_lang_name2enum
105: "C", ///< clean
106: "T", ///< tainted, untaint language as assigned later
107: // untaint languages. assigned by ^untaint[lang]{...}
108: "P",
109: /**<
110: leave language built into string being appended.
111: just a flag, that value not stored
112: */
113: "A", ///< leave all characters intact
1.68 parser 114: "F", ///< file specification
115: "H", ///< ext in HTTP response header
1.56 parser 116: "M", ///< text in mail header
117: "URI", ///< text in uri
118: "T", ///< ^table:set body
119: "SQL", ///< ^table:sql body
120: "JS", ///< JavaScript code
1.68 parser 121: "XML", ///< ^dom:set xml
1.82 paf 122: "HTML" ///< HTML code (for editing)
1.56 parser 123: };
124:
125:
1.1 paf 126: // String
127:
1.41 paf 128: /*
129:
130: HTTP-header = field-name ":" [ field-value ] CRLF
131:
132: field-name = token
133: field-value = *( field-content | LWS )
134:
135: field-content = <the OCTETs making up the field-value
136: and consisting of either *TEXT or combinations
137: of token, tspecials, and quoted-string>
138:
139:
140:
141: word = token | quoted-string
142:
143: token = 1*<any CHAR except CTLs or tspecials>
144:
145:
146:
147: tspecials = "(" | ")" | "<" | ">" | "@"
148: | "," | ";" | ":" | "\" | <">
149: | "/" | "[" | "]" | "?" | "="
150: | "{" | "}" | SP | HT
151:
152: SP = <US-ASCII SP, space (32)>
153: HT = <US-ASCII HT, horizontal-tab (9)>
154:
155: LWS = [CRLF] 1*( SP | HT )
156: TEXT = <any OCTET except CTLs,
157: but including LWS>
158:
159: quoted-pair = "\" CHAR
160:
161: if(strchr("()<>@,;:\\\"/[]?={} \t", *ptr))
162: */
1.117 paf 163: inline bool need_quote_http_header(const char* ptr, size_t size) {
1.41 paf 164: for(; size--; ptr++)
1.42 paf 165: if(strchr(";\\\"= \t" /* excluded ()<>@, :/ ? []{} */, *ptr))
1.41 paf 166: return true;
167: return false;
168: }
169:
1.119 paf 170: #ifndef DOXYGEN
171: struct Append_fragment_info {
172: String::Language lang;
173: String::Languages* dest_languages;
174: size_t dest_body_plan_length;
175: };
176: #endif
177: int append_fragment_optimizing(char alang, size_t asize, Append_fragment_info* info) {
178: const String::Language lang=(String::Language)(unsigned char)alang;
179: // main idea here:
180: // tainted piece would get OPTIMIZED bit from 'lang'
181: // clean piece would be marked OPTIMIZED manually
182: // pieces with determined languages [not tainted|clean] would retain theirs langs
183: info->dest_languages->append(info->dest_body_plan_length,
1.120 paf 184: lang==String::L_TAINTED?
1.119 paf 185: info->lang
1.120 paf 186: :lang==String::L_CLEAN?
187: (String::Language)(String::L_CLEAN|String::L_OPTIMIZE_BIT) // ORing with OPTIMIZED flag
1.119 paf 188: :lang,
189: asize);
190: info->dest_body_plan_length+=asize;
191:
192: return 0; // 0=continue
193: }
194: int append_fragment_nonoptimizing(char alang, size_t asize, Append_fragment_info* info) {
195: const String::Language lang=(String::Language)(unsigned char)alang;
196: // The core idea: tainted pieces got marked with context's lang
197: info->dest_languages->append(info->dest_body_plan_length,
1.120 paf 198: lang==String::L_TAINTED?
1.119 paf 199: info->lang
200: :lang,
201: asize);
202: info->dest_body_plan_length+=asize;
203:
204: return 0; // 0=continue
205: }
1.117 paf 206:
1.91 paf 207: /**
1.117 paf 208: appends to other String,
1.91 paf 209: marking all tainted pieces of it with @a lang.
1.92 paf 210: or marking ALL pieces of it with a @a lang when @a forced to,
211: and propagating OPTIMIZE language bit.
1.117 paf 212: */
213: String& String::append_to(String& dest, Language lang, bool forced) const {
214: if(is_empty())
215: return dest;
1.91 paf 216:
1.119 paf 217: // first: fragment infos
1.117 paf 218:
219: if(lang==L_PASS_APPENDED) // without language-change?
1.119 paf 220: dest.langs.append(dest.body, body.length(), langs);
1.117 paf 221: else if(forced) //forcing passed lang?
1.119 paf 222: dest.langs.append(dest.body, lang, length());
223: else {
1.122 paf 224: Append_fragment_info info={lang, &dest.langs, dest.body.length()};
1.119 paf 225: langs.for_each(body, lang&L_OPTIMIZE_BIT?
226: append_fragment_optimizing
227: :append_fragment_nonoptimizing, &info);
1.91 paf 228: }
229:
1.119 paf 230: // next: letters
231: dest.body<<body;
232:
1.117 paf 233: ASSERT_STRING_INVARIANT(dest);
1.75 paf 234: return dest;
1.51 parser 235: }
236:
1.109 paf 237: /** http://www.ietf.org/rfc/rfc2047.txt
238: RFC
239: (3) As a replacement for a 'word' entity within a 'phrase', for example,
240: one that precedes an address in a From, To, or Cc header. The ABNF
241: definition for 'phrase' from RFC 822 thus becomes:
242:
243: phrase = 1*( encoded-word / word )
244:
245: In this case the set of characters that may be used in a "Q"-encoded
246: 'encoded-word' is restricted to: <upper and lower case ASCII
247: letters, decimal digits, "!", "*", "+", "-", "/", "=", and "_"
248: (underscore, ASCII 95.)>. An 'encoded-word' that appears within a
249: 'phrase' MUST be separated from any adjacent 'word', 'text' or
250: 'special' by 'linear-white-space'.
251: ...
252: (2) The 8-bit hexadecimal value 20 (e.g., ISO-8859-1 SPACE) may be
253: represented as "_" (underscore, ASCII 95.). (This character may
254: not pass through some internetwork mail gateways, but its use
255: will greatly enhance readability of "Q" encoded data with mail
256: readers that do not support this encoding.) Note that the "_"
257: always represents hexadecimal 20, even if the SPACE character
258: occupies a different code position in the character set in use.
259:
260: paf: obviously,
261: without "=", or one could not differ "=E0" and "russian letter a"
262: and without "_", or in would mean 0x20
263: */
1.117 paf 264: inline bool mail_header_char_valid_within_Qencoded(char c) {
1.109 paf 265: return c>='A' && c<='Z'
266: || c>='a' && c<='Z'
267: || c>='0' && c<='9'
268: || strchr("!*+-/", c);
269: }
1.118 paf 270: inline bool addr_spec_soon(const char *src) {
271: for(char c; c=*src; src++)
272: if(c=='<')
273: return true;
274: else if(!(c==' ' || c=='\t'))
275: return false;
276: return false;
277: }
1.117 paf 278: /**
279: RFC
280: Upper case should be used for hexadecimal digits "A" through "F"
281: The 8-bit hexadecimal value 20 (e.g., ISO-8859-1 SPACE)
282: may be represented as "_"
283: */
284: inline bool mail_header_nonspace_char(char c) {
285: return c != 0x20;
286: }
287: inline void ec_append(CORD_ec& result, bool& optimize, bool& whitespace, CORD_pos pos, size_t size) {
288: while(size--) {
289: CORD_ec_append(result, CORD_pos_fetch(pos));
290: CORD_next(pos);
291: }
292: }
293: inline void pa_CORD_pos_advance(CORD_pos pos, size_t n) {
294: while(true) {
295: size_t avail=CORD_pos_chars_left(pos);
296: if(avail==0) {
297: CORD_next(pos);
298: if(!--n)
299: break;
300: } else if(avail<n) {
301: CORD_pos_advance(pos, avail);
302: n-=avail;
303: } else { // avail>=n
304: CORD_pos_advance(pos, n);
305: break;
306: }
307: }
308: }
309:
1.119 paf 310: #ifndef DOXYGEN
311: struct Cstr_to_string_body_block_info {
312: // input
313: String::Language lang;
314: SQL_Connection* connection;
315: const Request_charsets* charsets;
316: const String::Body* body;
317:
318: // output
319: CORD_ec result;
320:
321: // private
322: CORD_pos pos;
323: size_t fragment_begin;
324: bool whitespace;
325: };
326: #endif
327: int cstr_to_string_body_block(char alang, size_t fragment_length, Cstr_to_string_body_block_info* info) {
328: const String::Language fragment_lang=(String::Language)(unsigned char)alang;
329: bool& whitespace=info->whitespace;
330: size_t fragment_end=info->fragment_begin+fragment_length;
331: //fprintf(stderr, "%d, %d\n", fragment.lang, fragment.length);
1.117 paf 332:
1.119 paf 333:
1.120 paf 334: String::Language to_lang=info->lang==String::L_UNSPECIFIED?fragment_lang:info->lang;
335: bool optimize=(to_lang & String::L_OPTIMIZE_BIT)!=0;
1.119 paf 336: if(!optimize)
337: whitespace=false;
338:
1.120 paf 339: switch(to_lang & ~String::L_OPTIMIZE_BIT) {
340: case String::L_CLEAN:
341: case String::L_TAINTED:
342: case String::L_AS_IS:
1.119 paf 343: // clean piece
344:
345: // tainted piece, but undefined untaint language
346: // for VString.as_double of tainted values
347: // for ^process{body} evaluation
348:
349: // tainted, untaint language: as-is
350: ec_append(info->result, optimize, whitespace, info->pos, fragment_length);
351: break;
1.120 paf 352: case String::L_FILE_SPEC:
1.119 paf 353: // tainted, untaint language: file [name]
354: escape(
355: // Macintosh has problems with small Russian letter 'r'
356: if( c=='\xF0' && info->charsets && info->charsets->source().NAME()=="WINDOWS-1251" ) {
357: // fixing that letter for most common charset
358: to_char('p');
359: } else // fallback to default
360: encode(need_file_encode, '_', c);
361: );
362: break;
1.120 paf 363: case String::L_URI:
1.119 paf 364: // tainted, untaint language: uri
365: {
366: const char *fragment_str=info->body->mid(info->fragment_begin, fragment_length).cstr();
367: // skip source [we use recoded version]
368: pa_CORD_pos_advance(info->pos, fragment_length);
369: String::C output(fragment_str, fragment_length);
370: if(info->charsets)
371: output=Charset::transcode(output,
372: info->charsets->source(),
373: info->charsets->client());
1.77 paf 374:
1.119 paf 375: char c;
376: for(const char* src=output.str; c=*src++; )
1.117 paf 377: encode(need_uri_encode, '%', c);
1.119 paf 378: }
379: break;
1.120 paf 380: case String::L_HTTP_HEADER:
1.119 paf 381: // tainted, untaint language: http-field-content-text
382: escape(
383: encode(need_uri_encode, '%', c);
384: );
385: break;
1.120 paf 386: case String::L_MAIL_HEADER:
1.119 paf 387: // tainted, untaint language: mail-header
388: // http://www.ietf.org/rfc/rfc2047.txt
389: if(info->charsets) {
390: size_t mail_size;
391: const char *mail_ptr=
392: info->body->mid(info->fragment_begin, mail_size=fragment_length).cstr();
393: // skip source [we use recoded version]
394: pa_CORD_pos_advance(info->pos, mail_size);
395:
396: const char* charset_name=info->charsets->mail().NAME().cstr();
397:
398: // Subject: Re: parser3: =?koi8-r?Q?=D3=C5=CD=C9=CE=C1=D2?=
399: bool to_quoted_printable=false;
400:
401: bool email=false;
402: uchar c;
403: for(const char* src=mail_ptr; c=(uchar)*src++; ) {
404: //RFC + An 'encoded-word' MUST NOT appear in any portion of an 'addr-spec'.
405: if(to_quoted_printable && (c==',' || addr_spec_soon(src) || c == '"')) {
406: email=c=='<';
407: to_string("?=");
408: to_quoted_printable=false;
409: }
410: if(!email && (
411: !to_quoted_printable && (c & 0x80) // starting quote-printable-encoding on first 8bit char
412: || to_quoted_printable && !mail_header_char_valid_within_Qencoded(c)
413: )) {
414: if(!to_quoted_printable) {
415: to_string("=?");
416: to_string(charset_name);
417: to_string("?Q?");
418: to_quoted_printable=true;
1.105 paf 419: }
1.119 paf 420: encode(mail_header_nonspace_char, '=', '_');
421: } else
422: to_char(c);
423: if(c=='>')
424: email=false;
425: }
426: if(to_quoted_printable) // close
427: to_string("?=");
428:
429: } else
430: ec_append(info->result, optimize, whitespace, info->pos, fragment_length);
431: break;
1.120 paf 432: case String::L_TABLE:
1.119 paf 433: // tainted, untaint language: table
434: escape(switch(c) {
435: case '\t': to_char(' '); break;
436: case '\n': to_char(' '); break;
437: _default;
438: });
439: break;
1.117 paf 440: #ifdef PA_SQL
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));
449: } else
450: throw Exception(0,
451: 0,
452: "untaint in SQL language failed - no connection specified");
453: break;
1.117 paf 454: #endif
1.120 paf 455: case String::L_JS:
1.119 paf 456: escape(switch(c) {
457: case '"': to_string("\\\""); break;
458: case '\'': to_string("\\'"); break;
459: case '\n': to_string("\\n"); break;
460: case '\\': to_string("\\\\"); break;
461: case '\xFF': to_string("\\\xFF"); break;
462: _default;
463: });
464: break;
1.120 paf 465: case String::L_XML:
1.119 paf 466: escape(switch(c) {
467: case '&': to_string("&"); break;
468: case '>': to_string(">"); break;
469: case '<': to_string("<"); break;
470: case '"': to_string("""); break;
471: case '\'': to_string("'"); break;
472: _default;
473: });
474: break;
1.120 paf 475: case String::L_HTML:
1.119 paf 476: escape(switch(c) {
477: case '&': to_string("&"); break;
478: case '>': to_string(">"); break;
479: case '<': to_string("<"); break;
480: case '"': to_string("""); break;
481: _default;
482: });
483: break;
484: default:
485: assert(!"should never");
486: SAPI::die("unknown untaint language #%d",
487: static_cast<int>(to_lang)); // should never
488: break; // never
1.117 paf 489: }
1.55 parser 490:
1.119 paf 491: info->fragment_begin=fragment_end;
492:
493: return 0; // 0=continue
494: }
495:
496:
497: String::Body String::cstr_to_string_body(Language lang,
498: SQL_Connection* connection,
499: const Request_charsets *charsets) const {
500:
501: Cstr_to_string_body_block_info info;
502: // input
503: info.lang=lang;
504: info.connection=connection;
505: info.charsets=charsets;
506: info.body=&body;
507: // output
508: CORD_ec_init(info.result);
509: // private
510: body.set_pos(info.pos, 0);
511: info.fragment_begin=0;
512: info.whitespace=true;
513:
514: if(!is_empty())
515: langs.for_each(body, cstr_to_string_body_block, &info);
516:
517: return String::Body(CORD_ec_to_cord(info.result));
1.1 paf 518: }
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