--- parser3/src/main/untaint.C 2003/03/20 15:05:59 1.115.2.12.2.7 +++ parser3/src/main/untaint.C 2003/07/24 11:31:24 1.117 @@ -5,7 +5,7 @@ Author: Alexandr Petrosian (http://paf.design.ru) */ -static const char* IDENT_UNTAINT_C="$Date: 2003/03/20 15:05:59 $"; +static const char* IDENT_UNTAINT_C="$Date: 2003/07/24 11:31:24 $"; #include "pa_string.h" @@ -17,31 +17,61 @@ static const char* IDENT_UNTAINT_C="$Dat #include "pa_common.h" #include "pa_charset.h" #include "pa_request_charsets.h" +#include "pa_sapi.h" extern "C" { // author forgot to do that #include "ec.h" } -//#define PA_SQL +#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 +#define CORD_ec_append(x, c) \ + { \ + bool skip=false; \ + if(optimize) switch(c) { \ + case ' ': case '\r': case '\n': case '\t': \ + if(whitespace) \ + skip=true; /*skipping subsequent*/ \ + else \ + whitespace=true; \ + break; \ + default: \ + whitespace=false; \ + break; \ + } \ + if(!skip) { \ + if ((x)[0].ec_bufptr == (x)[0].ec_buf + CORD_BUFSZ) { \ + CORD_ec_flush_buf(x); \ + } \ + *((x)[0].ec_bufptr)++ = (c); \ + } \ + } + + #define escape(action) \ for(; fragment_length--; CORD_next(pos)) { \ char c=CORD_pos_fetch(pos); \ action \ } #define _default default: CORD_ec_append(result, c); break -#define encode(need_encode_func, prefix) \ +#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]); \ } else \ - CORD_ec_append(result, c); + CORD_ec_append(result, otherwise); #define to_char(c) CORD_ec_append(result, c) #define to_string(s) CORD_ec_append_cord(result, s) @@ -51,7 +81,7 @@ inline bool need_file_encode(unsigned ch return strchr( "*?'\"<>|" #ifndef WIN32 - ":\\~" + ":\\" #endif , c)!=0; } @@ -144,43 +174,42 @@ inline bool need_quote_http_header(const or marking ALL pieces of it with a @a lang when @a forced to, and propagating OPTIMIZE language bit. */ -String& String::append_to(String& dest, Untaint_lang lang, bool forced) const { - if(!*this) +String& String::append_to(String& dest, Language lang, bool forced) const { + if(is_empty()) return dest; // first: letters - dest.body=CORD_cat(dest.body, body); + dest.body< i(fragments); - while(i.has_next()) { - const Fragment& fragment=i.next(); + else if(lang&L_OPTIMIZE_BIT) { + for(Array_iterator 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==UL_TAINTED?lang:( - fragment.lang==UL_CLEAN?UL_CLEAN|UL_OPTIMIZE_BIT: // ORing with OPTIMIZED flag + 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 - Array_iterator i(fragments); - while(i.has_next()) { - const Fragment& fragment=i.next(); + for(Array_iterator i(fragments); i.has_next(); ) { + const Fragment fragment=i.next(); dest.fragments+=Fragment( - fragment.lang==UL_TAINTED?lang:fragment.lang, + fragment.lang==L_TAINTED?lang:fragment.lang, fragment.length); } } + ASSERT_STRING_INVARIANT(dest); return dest; } @@ -217,33 +246,62 @@ inline bool mail_header_char_valid_withi || c>='0' && c<='9' || strchr("!*+-/", c); } -inline void ec_append(CORD_ec result, CORD_pos& pos, size_t size) { +/** + RFC + Upper case should be used for hexadecimal digits "A" through "F" + The 8-bit hexadecimal value 20 (e.g., ISO-8859-1 SPACE) + may be represented as "_" +*/ +inline bool mail_header_nonspace_char(char c) { + return c != 0x20; +} +inline void ec_append(CORD_ec& result, bool& optimize, bool& whitespace, CORD_pos pos, size_t size) { while(size--) { CORD_ec_append(result, CORD_pos_fetch(pos)); CORD_next(pos); } } -CORD String::cstr_to_cord(Untaint_lang lang, - SQL_Connection *connection, +inline void pa_CORD_pos_advance(CORD_pos pos, size_t n) { + while(true) { + size_t avail=CORD_pos_chars_left(pos); + if(avail==0) { + CORD_next(pos); + if(!--n) + break; + } else if(avail=n + CORD_pos_advance(pos, n); + break; + } + } +} +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; CORD_set_pos(pos, body, 0); + 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(); + 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); - Untaint_lang to_lang=lang==UL_UNSPECIFIED?fragment.lang:lang; - - switch(to_lang & ~UL_OPTIMIZE_BIT) { - case UL_CLEAN: - case UL_TAINTED: - case UL_AS_IS: + 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 @@ -251,68 +309,64 @@ CORD String::cstr_to_cord(Untaint_lang l // for ^process{body} evaluation // tainted, untaint language: as-is - ec_append(result, pos, fragment_length); + ec_append(result, optimize, whitespace, pos, fragment_length); break; - case UL_FILE_SPEC: + case L_FILE_SPEC: // tainted, untaint language: file [name] escape( // Macintosh has problems with small Russian letter 'r' - if( c=='\xF0' && charsets && charsets->source().name()=="windows-1251" ) { + if( c=='\xF0' && charsets && charsets->source().NAME()=="WINDOWS-1251" ) { // fixing that letter for most common charset to_char('p'); } else // fallback to default - encode(need_file_encode, '_'); + encode(need_file_encode, '_', c); ); break; -#ifdef LATER - case UL_URI: + case L_URI: // tainted, untaint language: uri { - const char *substr= - CORD_to_const_char_star( - CORD_substr(body, fragment_begin, fragment_length)); - const void *client_ptr; - size_t client_size; + 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) - Charset::transcode( - charsets->source(), substr, fragment_length, - charsets->client(), client_ptr, client_size); - else { - client_ptr=substr; - client_size=fragment_length; - } + output=Charset::transcode(output, + charsets->source(), + charsets->client()); char c; - for(const char* src=(const char* )client_ptr; c=*src++; ) - encode(need_uri_encode, '%'); + for(const char* src=output.str; c=*src++; ) + encode(need_uri_encode, '%', c); } break; -#endif - case UL_HTTP_HEADER: + case L_HTTP_HEADER: // tainted, untaint language: http-field-content-text escape( - encode(need_uri_encode, '%'); + encode(need_uri_encode, '%', c); ); break; -#ifdef LATER - case UL_MAIL_HEADER: + case L_MAIL_HEADER: // tainted, untaint language: mail-header // http://www.ietf.org/rfc/rfc2047.txt if(charsets) { - const void *mail_ptr=fragment.ptr; - size_t mail_size=fragment_length; - const char* charset_name=charsets->mail().name()->cstr(); + 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?= - const char* src=(const char* )mail_ptr; bool to_quoted_printable=false; bool email=false; - for(const char* end=src+mail_size; src') email=false; } if(to_quoted_printable) // close - dest+=sprintf(dest, "?="); + to_string("?="); - } else { - memcpy(dest, fragment.ptr, fragment_length); - dest+=fragment_length; - } + } else + ec_append(result, optimize, whitespace, pos, fragment_length); break; -#endif - case UL_TABLE: + case L_TABLE: // tainted, untaint language: table escape(switch(c) { case '\t': to_char(' '); break; @@ -351,22 +400,21 @@ CORD String::cstr_to_cord(Untaint_lang l }); break; #ifdef PA_SQL - case UL_SQL: + case L_SQL: // tainted, untaint language: sql if(connection) { - dest+=connection->quote(0, - CORD_substr(body, fragment_begin, fragment_length), - fragment_length); -...^^^ + const char *fragment_str=body.mid(fragment_begin, fragment_length).cstr(); + // skip source [we use recoded version] + pa_CORD_pos_advance(pos, fragment_length); - dest+=connection->quote(dest, fragment.ptr, fragment_length); - else + to_string(connection->quote(fragment_str, fragment_length)); + } else throw Exception(0, 0, "untaint in SQL language failed - no connection specified"); break; #endif - case UL_JS: + case L_JS: escape(switch(c) { case '"': to_string("\\\""); break; case '\'': to_string("\\'"); break; @@ -376,7 +424,7 @@ CORD String::cstr_to_cord(Untaint_lang l _default; }); break; - case UL_XML: + case L_XML: escape(switch(c) { case '&': to_string("&"); break; case '>': to_string(">"); break; @@ -386,7 +434,7 @@ CORD String::cstr_to_cord(Untaint_lang l _default; }); break; - case UL_HTML: + case L_HTML: escape(switch(c) { case '&': to_string("&"); break; case '>': to_string(">"); break; @@ -396,34 +444,11 @@ CORD String::cstr_to_cord(Untaint_lang l }); break; default: - throw Exception(0, - 0, - "unknown untaint language #%d", - static_cast(to_lang)); // should never + SAPI::die("unknown untaint language #%d", + static_cast(to_lang)); // should never break; // never } -#ifdef LATER - if(to_lang & UL_OPTIMIZE_BIT) { - // optimizing whitespace - char *stop=dest; dest=start; - for(char c=start; src