--- parser3/src/main/untaint.C 2003/08/18 08:27:41 1.118 +++ parser3/src/main/untaint.C 2009/01/12 07:17:20 1.143 @@ -1,11 +1,11 @@ /** @file Parser: String class part: untaint mechanizm. - Copyright(c) 2001-2003 ArtLebedev Group (http://www.artlebedev.com) + Copyright(c) 2001-2005 ArtLebedev Group (http://www.artlebedev.com) Author: Alexandr Petrosian (http://paf.design.ru) */ -static const char* IDENT_UNTAINT_C="$Date: 2003/08/18 08:27:41 $"; +static const char * const IDENT_UNTAINT_C="$Date: 2009/01/12 07:17:20 $"; #include "pa_string.h" @@ -23,15 +23,10 @@ extern "C" { // author forgot to do that #include "ec.h" } -#define PA_SQL - -#ifdef PA_SQL #include "pa_sql_connection.h" -#endif // defines - #undef CORD_ec_append // redefining to intercept flushes and implement whitespace optimization // of all consequent white space chars leaving only first one @@ -39,7 +34,7 @@ extern "C" { // author forgot to do that { \ bool skip=false; \ if(optimize) switch(c) { \ - case ' ': case '\r': case '\n': case '\t': \ + case ' ': case '\n': case '\t': \ if(whitespace) \ skip=true; /*skipping subsequent*/ \ else \ @@ -58,75 +53,55 @@ extern "C" { // author forgot to do that } -#define escape(action) \ - for(; fragment_length--; CORD_next(pos)) { \ - char c=CORD_pos_fetch(pos); \ +#define escape_fragment(action) \ + for(; fragment_length--; CORD_next(info->pos)) { \ + char c=CORD_pos_fetch(info->pos); \ action \ } -#define _default default: CORD_ec_append(result, c); break +#define _default CORD_ec_append(info->result, c) #define encode(need_encode_func, prefix, otherwise) \ if(need_encode_func(c)) { \ static const char* hex="0123456789ABCDEF"; \ - CORD_ec_append(result, prefix); \ - CORD_ec_append(result, hex[((unsigned char)c)/0x10]); \ - CORD_ec_append(result, hex[((unsigned char)c)%0x10]); \ + CORD_ec_append(info->result, prefix); \ + CORD_ec_append(info->result, hex[((unsigned char)c)/0x10]); \ + CORD_ec_append(info->result, hex[((unsigned char)c)%0x10]); \ } else \ - CORD_ec_append(result, otherwise); -#define to_char(c) CORD_ec_append(result, c) -#define to_string(s) CORD_ec_append_cord(result, s) + CORD_ec_append(info->result, otherwise); +#define to_char(c) { CORD_ec_append(info->result, c); whitespace=false; } +#define to_string(s) { CORD_ec_append_cord(info->result, s); whitespace=false; } inline bool need_file_encode(unsigned char c){ // russian letters and space ENABLED // encoding only these... return strchr( - "*?'\"<>|" + "*?'\"<>|" #ifndef WIN32 - ":\\" + ":\\" #endif - , c)!=0; + , c)!=0; } + inline bool need_uri_encode(unsigned char c){ - if((c>='0') &&(c<='9') ||(c>='A') &&(c<='Z') ||(c>='a') &&(c<='z')) + if((c>='0') &&(c<='9') ||(c>='A') &&(c<='Z') ||(c>='a') &&(c<='z')) return false; - return !strchr("_-./", c); + return !strchr("_-./", c); } + inline bool need_http_header_encode(unsigned char c){ - if(strchr(" , :", c)) + if(strchr(" , :", c)) return false; return need_uri_encode(c); } -// - -static const char* String_Untaint_lang_name[]={ - "U", ///< zero value handy for hash lookup @see untaint_lang_name2enum - "C", ///< clean - "T", ///< tainted, untaint language as assigned later - // untaint languages. assigned by ^untaint[lang]{...} - "P", - /**< - leave language built into string being appended. - just a flag, that value not stored - */ - "A", ///< leave all characters intact - "F", ///< file specification - "H", ///< ext in HTTP response header - "M", ///< text in mail header - "URI", ///< text in uri - "T", ///< ^table:set body - "SQL", ///< ^table:sql body - "JS", ///< JavaScript code - "XML", ///< ^dom:set xml - "HTML" ///< HTML code (for editing) -}; - +inline bool need_regex_escape(unsigned char c){ + return strchr("\\^$.[]|()?*+{}-", c)!=0; +} // String /* - HTTP-header = field-name ":" [ field-value ] CRLF field-name = token @@ -167,6 +142,42 @@ inline bool need_quote_http_header(const return false; } +#ifndef DOXYGEN +struct Append_fragment_info { + String::Language lang; + String::Languages* dest_languages; + size_t dest_body_plan_length; +}; +#endif +int append_fragment_optimizing(char alang, size_t asize, Append_fragment_info* info) { + const String::Language lang=(String::Language)(unsigned char)alang; + // main idea here: + // tainted piece would get OPTIMIZED bit from 'lang' + // clean piece would be marked OPTIMIZED manually + // pieces with determined languages [not tainted|clean] would retain theirs langs + info->dest_languages->append(info->dest_body_plan_length, + lang==String::L_TAINTED? + info->lang + :lang==String::L_CLEAN? + (String::Language)(String::L_CLEAN|String::L_OPTIMIZE_BIT) // ORing with OPTIMIZED flag + :lang, + asize); + info->dest_body_plan_length+=asize; + + return 0; // 0=continue +} +int append_fragment_nonoptimizing(char alang, size_t asize, Append_fragment_info* info) { + const String::Language lang=(String::Language)(unsigned char)alang; + // The core idea: tainted pieces got marked with context's lang + info->dest_languages->append(info->dest_body_plan_length, + lang==String::L_TAINTED? + info->lang + :lang, + asize); + info->dest_body_plan_length+=asize; + + return 0; // 0=continue +} /** appends to other String, @@ -178,37 +189,22 @@ String& String::append_to(String& dest, if(is_empty()) return dest; - // first: letters - dest.body< i(fragments); i.has_next(); ) { - const Fragment fragment=i.next(); - // main idea here: - // tainted piece would get OPTIMIZED bit from 'lang' - // clean piece would be marked OPTIMIZED manually - // pieces with determined languages [not tainted|clean] would retain theirs langs - dest.fragments+=Fragment(static_cast( - fragment.lang==L_TAINTED?lang:( - fragment.lang==L_CLEAN?L_CLEAN|L_OPTIMIZE_BIT: // ORing with OPTIMIZED flag - fragment.lang)), - fragment.length); - } - } else { // The core idea: tainted pieces got marked with context's lang - for(Array_iterator i(fragments); i.has_next(); ) { - const Fragment fragment=i.next(); - dest.fragments+=Fragment( - fragment.lang==L_TAINTED?lang:fragment.lang, - fragment.length); - } + dest.langs.append(dest.body, lang, length()); + else { + Append_fragment_info info={lang, &dest.langs, dest.body.length()}; + langs.for_each(body, lang&L_OPTIMIZE_BIT? + append_fragment_optimizing + :append_fragment_nonoptimizing, &info); } + // next: letters + dest.body<=n @@ -285,178 +282,261 @@ inline void pa_CORD_pos_advance(CORD_pos } } } -StringBody String::cstr_to_string_body(Language lang, - SQL_Connection* connection, - const Request_charsets *charsets) const { - CORD_ec result; CORD_ec_init(result); - CORD_pos pos; body.set_pos(pos, 0); - - bool whitespace=true; - size_t fragment_begin=0; - size_t fragment_end; - for(Array_iterator i(fragments); i.has_next(); fragment_begin=fragment_end) { - const Fragment fragment=i.next(); - size_t fragment_length=fragment.length; - fragment_end=fragment_begin+fragment_length; - //fprintf(stderr, "%d, %d\n", fragment.lang, fragment.length); - - Language to_lang=lang==L_UNSPECIFIED?fragment.lang:lang; - bool optimize=(to_lang & L_OPTIMIZE_BIT)!=0; - if(!optimize) - whitespace=false; - - switch(to_lang & ~L_OPTIMIZE_BIT) { - case L_CLEAN: - case L_TAINTED: - case L_AS_IS: - // clean piece - - // tainted piece, but undefined untaint language - // for VString.as_double of tainted values - // for ^process{body} evaluation - - // tainted, untaint language: as-is - ec_append(result, optimize, whitespace, pos, fragment_length); - break; - case L_FILE_SPEC: - // tainted, untaint language: file [name] - escape( +#ifndef DOXYGEN +struct Cstr_to_string_body_block_info { + // input + String::Language lang; + SQL_Connection* connection; + const Request_charsets* charsets; + const String::Body* body; + + // output + CORD_ec result; + + // private + CORD_pos pos; + size_t fragment_begin; + bool whitespace; + const char* exception; +}; +#endif +int cstr_to_string_body_block(char alang, size_t fragment_length, Cstr_to_string_body_block_info* info) { + const String::Language fragment_lang=(String::Language)(unsigned char)alang; + bool& whitespace=info->whitespace; + size_t fragment_end=info->fragment_begin+fragment_length; + //fprintf(stderr, "%d, %d =%s=\n", fragment_lang, fragment_length, info->body->cstr()); + + String::Language to_lang=info->lang==String::L_UNSPECIFIED?fragment_lang:info->lang; + bool optimize=(to_lang & String::L_OPTIMIZE_BIT)!=0; + if(!optimize) + whitespace=false; + + switch(to_lang & ~String::L_OPTIMIZE_BIT) { + case String::L_CLEAN: + case String::L_TAINTED: + case String::L_AS_IS: + // clean piece + + // tainted piece, but undefined untaint language + // for VString.as_double of tainted values + // for ^process{body} evaluation + + // tainted, untaint language: as-is + ec_append(info->result, optimize, whitespace, info->pos, fragment_length); + break; + case String::L_FILE_SPEC: + // tainted, untaint language: file [name] + { + bool is1251=(info->charsets && info->charsets->source().NAME()=="WINDOWS-1251"); + escape_fragment( // Macintosh has problems with small Russian letter 'r' - if( c=='\xF0' && charsets && charsets->source().NAME()=="WINDOWS-1251" ) { + if( is1251 && c=='\xF0' ) { // fixing that letter for most common charset to_char('p'); } else // fallback to default encode(need_file_encode, '_', c); ); - break; - case L_URI: - // tainted, untaint language: uri - { - const char *fragment_str=body.mid(fragment_begin, fragment_length).cstr(); - // skip source [we use recoded version] - pa_CORD_pos_advance(pos, fragment_length); - String::C output(fragment_str, fragment_length); - if(charsets) - output=Charset::transcode(output, - charsets->source(), - charsets->client()); - - char c; - for(const char* src=output.str; c=*src++; ) - encode(need_uri_encode, '%', c); - } - break; - case L_HTTP_HEADER: - // tainted, untaint language: http-field-content-text - escape( + } + break; + case String::L_URI: + // tainted, untaint language: uri + { + const char *fragment_str=info->body->mid(info->fragment_begin, fragment_length).cstr(); + // skip source [we use recoded version] + pa_CORD_pos_advance(info->pos, fragment_length); + String::C output(fragment_str, fragment_length); + if(info->charsets) + output=Charset::transcode(output, + info->charsets->source(), + info->charsets->client()); + + char c; + for(const char* src=output.str; (c=*src++); ) encode(need_uri_encode, '%', c); - ); - break; - case L_MAIL_HEADER: - // tainted, untaint language: mail-header - // http://www.ietf.org/rfc/rfc2047.txt - if(charsets) { - size_t mail_size; - const char *mail_ptr= - body.mid(fragment_begin, mail_size=fragment_length).cstr(); - // skip source [we use recoded version] - pa_CORD_pos_advance(pos, mail_size); - - const char* charset_name=charsets->mail().NAME().cstr(); - - // Subject: Re: parser3: =?koi8-r?Q?=D3=C5=CD=C9=CE=C1=D2?= - bool to_quoted_printable=false; - - bool email=false; - uchar c; - for(const char* src=mail_ptr; c=(uchar)*src++; ) { - //RFC + An 'encoded-word' MUST NOT appear in any portion of an 'addr-spec'. - if(to_quoted_printable && (c==',' || addr_spec_soon(src) || c == '"')) { - email=c=='<'; - to_string("?="); - to_quoted_printable=false; + } + break; + case String::L_HTTP_HEADER: + // tainted, untaint language: http-field-content-text + // the same as L_URI BUT not transcoded into $response:charset before encoding + escape_fragment( + encode(need_uri_encode, '%', c); + ); + break; + case String::L_MAIL_HEADER: + // tainted, untaint language: mail-header + // http://www.ietf.org/rfc/rfc2047.txt + if(info->charsets) { + size_t mail_size; + const char *mail_ptr= + info->body->mid(info->fragment_begin, mail_size=fragment_length).cstr(); + // skip source [we use recoded version] + pa_CORD_pos_advance(info->pos, mail_size); + + const char* charset_name=info->charsets->mail().NAME().cstr(); + + // Subject: Re: parser3: =?koi8-r?Q?=D3=C5=CD=C9=CE=C1=D2?= + bool to_quoted_printable=false; + + bool email=false; + uchar c; + for(const char* src=mail_ptr; (c=(uchar)*src++); ) { + if(c=='\r' || c=='\n') + c=' '; + if(to_quoted_printable && (c==',' || c == '"' || addr_spec_soon(src-1/*position to 'c'*/))) { + email=c=='<'; + to_string("?="); + to_quoted_printable=false; + } + //RFC + An 'encoded-word' MUST NOT appear in any portion of an 'addr-spec'. + if(!email && ( + !to_quoted_printable && (c & 0x80) // starting quote-printable-encoding on first 8bit char + || to_quoted_printable && !mail_header_char_valid_within_Qencoded(c) + )) { + if(!to_quoted_printable) { + to_string("=?"); + to_string(charset_name); + to_string("?Q?"); + to_quoted_printable=true; } - if(!email && ( - !to_quoted_printable && (c & 0x80) // starting quote-printable-encoding on first 8bit char - || to_quoted_printable && !mail_header_char_valid_within_Qencoded(c) - )) { - if(!to_quoted_printable) { - to_string("=?"); - to_string(charset_name); - to_string("?Q?"); - to_quoted_printable=true; - } - encode(mail_header_nonspace_char, '=', '_'); - } else - to_char(c); - if(c=='>') - email=false; + encode(mail_header_nonspace_char, '=', '_'); + } else + to_char(c); + if(c=='>') + email=false; + } + if(to_quoted_printable) // close + to_string("?="); + + } else + ec_append(info->result, optimize, whitespace, info->pos, fragment_length); + break; + case String::L_SQL: + // tainted, untaint language: sql + if(info->connection) { + const char *fragment_str=info->body->mid(info->fragment_begin, fragment_length).cstr(); + // skip source [we use recoded version] + pa_CORD_pos_advance(info->pos, fragment_length); + + to_string(info->connection->quote(fragment_str, fragment_length)); + } else { + info->exception="untaint in SQL language failed - no connection specified"; + info->fragment_begin=fragment_end; + return 1; // stop processing. can't throw exception here + } + break; + case String::L_JS: + escape_fragment(switch(c) { + case '\n': to_string("\\n"); break; + case '"': to_string("\\\""); break; + case '\'': to_string("\\'"); break; + case '\\': to_string("\\\\"); break; + case '\xFF': to_string("\\\xFF"); break; + default: _default; break; + }); + break; + case String::L_XML: + // [2] Char ::= #x9 | #xA | #xD | [#x20-#xD7FF] | [#xE000-#xFFFD] | [#x10000-#x10FFFF] + escape_fragment(switch(c) { + case '\x20': + case '\x9': + case '\xA': + case '\xD': // this is usually removed on input + _default; + break; + case '&': to_string("&"); break; + case '>': to_string(">"); break; + case '<': to_string("<"); break; + case '"': to_string("""); break; + case '\'': to_string("'"); break; + default: + if(((unsigned char)c)<0x20) { + // fixing it, so that libxml would not result + // in fatal error parsing text + // though it really violates standard. + // to indicate there were an error + // replace bad char not to it's code, + // which we can do, + // but rather to '!' to show that input were actually + // invalid. + // life: shows that MSIE can somehow garble form values + // so that they contain these chars. + to_char('!'); + } else { + _default; } - if(to_quoted_printable) // close - to_string("?="); + break; + }); + break; + case String::L_HTML: + escape_fragment(switch(c) { + case '&': to_string("&"); break; + case '>': to_string(">"); break; + case '<': to_string("<"); break; + case '"': to_string("""); break; + default: _default; break; + }); + break; + case String::L_REGEX: + // tainted, untaint language: regex + escape_fragment( + if(need_regex_escape(c)) + to_char('\\') + _default; + ); + break; + case String::L_HTTP_COOKIE: + // tainted, untaint language: cookie (3.3.0 and higher: %uXXXX in UTF-8) + { + const char *fragment_str=info->body->mid(info->fragment_begin, fragment_length).cstr(); + // skip source [we use recoded version] + pa_CORD_pos_advance(info->pos, fragment_length); + String::C output(fragment_str, fragment_length); - } else - ec_append(result, optimize, whitespace, pos, fragment_length); - break; - case L_TABLE: - // tainted, untaint language: table - escape(switch(c) { - case '\t': to_char(' '); break; - case '\n': to_char(' '); break; - _default; - }); - break; -#ifdef PA_SQL - case L_SQL: - // tainted, untaint language: sql - if(connection) { - const char *fragment_str=body.mid(fragment_begin, fragment_length).cstr(); - // skip source [we use recoded version] - pa_CORD_pos_advance(pos, fragment_length); - - to_string(connection->quote(fragment_str, fragment_length)); - } else - throw Exception(0, - 0, - "untaint in SQL language failed - no connection specified"); - break; -#endif - case L_JS: - escape(switch(c) { - case '"': to_string("\\\""); break; - case '\'': to_string("\\'"); break; - case '\n': to_string("\\n"); break; - case '\\': to_string("\\\\"); break; - case '\xFF': to_string("\\\xFF"); break; - _default; - }); - break; - case L_XML: - escape(switch(c) { - case '&': to_string("&"); break; - case '>': to_string(">"); break; - case '<': to_string("<"); break; - case '"': to_string("""); break; - case '\'': to_string("'"); break; - _default; - }); - break; - case L_HTML: - escape(switch(c) { - case '&': to_string("&"); break; - case '>': to_string(">"); break; - case '<': to_string("<"); break; - case '"': to_string("""); break; - _default; - }); - break; - default: - SAPI::die("unknown untaint language #%d", - static_cast(to_lang)); // should never - break; // never + output=Charset::escape(output, info->charsets->source()); + //throw Exception(0, 0, output); + to_string(output); + } + break; + default: + SAPI::abort("unknown untaint language #%d", + static_cast(to_lang)); // should never + break; // never + } + + info->fragment_begin=fragment_end; + + return 0; // 0=continue +} + + +String::Body String::cstr_to_string_body(Language lang, + SQL_Connection* connection, + const Request_charsets *charsets) const { + if(is_empty()) + return String::Body(); + + Cstr_to_string_body_block_info info; + // input + info.lang=lang; + info.connection=connection; + info.charsets=charsets; + info.body=&body; + // output + CORD_ec_init(info.result); + // private + body.set_pos(info.pos, 0); + info.fragment_begin=0; + info.exception=0; + info.whitespace=true; + + langs.for_each(body, cstr_to_string_body_block, &info); + if(info.exception){ + throw Exception(0, + 0, + info.exception); } - return StringBody(CORD_ec_to_cord(result)); + return String::Body(CORD_ec_to_cord(info.result)); }