--- parser3/src/main/untaint.C 2001/08/28 09:27:42 1.59 +++ parser3/src/main/untaint.C 2001/12/10 10:20:24 1.83 @@ -2,10 +2,10 @@ Parser: String class part: untaint mechanizm. Copyright(c) 2001 ArtLebedev Group(http://www.artlebedev.com) + Author: Alexander Petrosyan (http://paf.design.ru) - Author: Alexander Petrosyan (http://design.ru/paf) + $Id: untaint.C,v 1.83 2001/12/10 10:20:24 paf Exp $ */ -static const char *RCSId="$Id: untaint.C,v 1.59 2001/08/28 09:27:42 parser Exp $"; #include "pa_pool.h" #include "pa_string.h" @@ -15,6 +15,7 @@ static const char *RCSId="$Id: untaint.C #include "pa_globals.h" #include "pa_sql_connection.h" #include "pa_dictionary.h" +#include "pa_common.h" #define escape(action) \ { \ @@ -24,30 +25,35 @@ static const char *RCSId="$Id: untaint.C } #define _default default: *dest++=*src; break #define encode(need_encode_func, prefix) \ - default: \ if(need_encode_func(*src)) { \ static const char *hex="0123456789ABCDEF"; \ char chunk[3]={prefix}; \ chunk[1]=hex[((unsigned char)*src)/0x10]; \ chunk[2]=hex[((unsigned char)*src)%0x10]; \ - strncpy(dest, chunk, 3); dest+=3; \ + memcpy(dest, chunk, 3); dest+=3; \ } else \ *dest++=*src; \ break #define to_char(c) *dest++=c #define to_string(b, bsize) \ - strncpy(dest, b, bsize); \ + memcpy(dest, b, bsize); \ dest+=bsize; \ inline bool need_file_encode(unsigned char c){ + // theoretical problem with, for instance, "_2B" and "." fragments, + // they would yield the same + // because need_file_encode('_')=false + // but we need to delete such files somehow, getting names from ^index + if((c>='0') &&(c<='9') ||(c>='A') &&(c<='Z') ||(c>='a') &&(c<='z')) return false; return !strchr( + "_./()-" #ifdef WIN32 ":\\~" #endif - "./()_-", c); + , c); } inline bool need_uri_encode(unsigned char c){ if((c>='0') &&(c<='9') ||(c>='A') &&(c<='Z') ||(c>='a') &&(c<='z')) @@ -75,31 +81,20 @@ static const char * String_Untaint_lang_ just a flag, that value not stored */ "A", ///< leave all characters intact - "F", ///< filename - "H", ///< text in HTTP response header + "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 - "HTML", ///< HTML code (for editing) - "UHTML", ///< HTML code with USER chars + "XML", ///< ^dom:set xml + "HTML" ///< HTML code (for editing) }; // String -static bool typo_present(Array::Item *value, const void *info) { - Array *row=static_cast(value); - const char *src=static_cast(info); - - int partial; - row->get_string(0)->cmp(partial, src); - return - partial==0 || // full match - partial==1; // typo left column starts 'src' -} - /* HTTP-header = field-name ":" [ field-value ] CRLF @@ -142,260 +137,325 @@ inline bool need_quote_http_header(const return false; } -/** @todo maybe additional check "are all pieces are clean?" would be profitable? - @todo fix potential forigins_mode buf overrun -*/ -size_t String::cstr_bufsize(Untaint_lang lang) const { - return (lang==UL_AS_IS?size():size()*UNTAINT_TIMES_BIGGER*(forigins_mode?10:1)) +1; +String& String::append(const String& src, uchar lang, bool forced) { + // manually unrolled code to avoid do{if(const)} constructs + if(forced) + STRING_SRC_FOREACH_ROW( + APPEND(row->item.ptr, row->item.size, + lang, //forcing passed lang + row->item.origin.file, row->item.origin.line); + ) + else if(lang==UL_PASS_APPENDED) + STRING_SRC_FOREACH_ROW( + APPEND(row->item.ptr, row->item.size, + row->item.lang, // passing item's lang + row->item.origin.file, row->item.origin.line); + ) + else if(lang&UL_OPTIMIZE_BIT) // 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 + STRING_SRC_FOREACH_ROW( + APPEND(row->item.ptr, row->item.size, + row->item.lang==UL_TAINTED?lang:( + row->item.lang==UL_CLEAN?UL_CLEAN|UL_OPTIMIZE_BIT: // ORing with OPTIMIZED flag + row->item.lang + ), + row->item.origin.file, row->item.origin.line); + ) + else + STRING_SRC_FOREACH_ROW( + APPEND(row->item.ptr, row->item.size, + row->item.lang==UL_TAINTED?lang:row->item.lang, + row->item.origin.file, row->item.origin.line); + ); +break2: + return *this; +} + +size_t String::cstr_bufsize(Untaint_lang lang, + SQL_Connection *connection, + const char *charset) const { + size_t dest=1; // for terminating 0 + STRING_FOREACH_ROW( + uchar to_lang=lang==UL_UNSPECIFIED?row->item.lang:lang; + + switch(to_lang & ~UL_OPTIMIZE_BIT) { + case UL_CLEAN: + case UL_TAINTED: + case UL_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 + dest+=row->item.size; + break; + case UL_FILE_SPEC: + // tainted, untaint language: file [name] + dest+=row->item.size*3/* worst: Z->%XX */; + break; + case UL_URI: + // tainted, untaint language: uri + dest+=row->item.size*3/* worst: Z->%XX */; + break; + case UL_HTTP_HEADER: + // tainted, untaint language: http-field-content-text + dest+=row->item.size*3/* worst: Z->%XX */; + break; + case UL_MAIL_HEADER: + // tainted, untaint language: mail-header + if(charset) { + // Subject: Re: parser3: =?koi8-r?Q?=D3=C5=CD=C9=CE=C1=D2?= + dest+=row->item.size*3+MAX_STRING/* worst: =?charset?Q?=%XX?= */; + } else { + dest+=row->item.size; + } + break; + case UL_TABLE: + // tainted, untaint language: table + dest+=row->item.size; + break; + case UL_SQL: + // tainted, untaint language: sql + if(connection) + dest+=connection->quote(0, row->item.ptr, row->item.size); + break; + case UL_JS: + escape(switch(*src) { + case '"': case '\'': case '\n': case '\\': case '\xFF': + dest+=2; break; + default: + dest++; break; + }); + break; + case UL_XML: + escape(switch(*src) { + case '&': case '>': case '<': case '"': case '\'': + dest+= 6; break; + default: + dest++; break; + }); + break; + case UL_HTML: + escape(switch(*src) { + case '&': + case '>': + case '<': + case '"': + dest+=6; break; + default: + dest++; break; + }); + break; + } + ); +break2: + return dest; } -/** @todo fix theoretical \n mem overrun in TYPO replacements -*/ char *String::store_to(char *dest, Untaint_lang lang, SQL_Connection *connection, const char *charset) const { - // $MAIN:html-typo table - Dictionary *user_typo_dict=static_cast(pool().tag()); - Dictionary *typo_dict=user_typo_dict?user_typo_dict:default_typo_dict; - + // WARNING: + // before any changes check cstr_bufsize first!!! bool whitespace=true; + // expanded STRING_FOREACH_ROW here for debugging purposes + const Chunk *chunk=&head; \ + do { \ + const Chunk::Row *row=chunk->rows; \ + for(uint i=0; icount; i++, row++) { \ + if(row==append_here) \ + goto break2; \ + \ + uchar to_lang=lang==UL_UNSPECIFIED?row->item.lang:lang; + + char *start=dest; + + switch(to_lang & ~UL_OPTIMIZE_BIT) { + case UL_CLEAN: + case UL_TAINTED: + case UL_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 + memcpy(dest, row->item.ptr, row->item.size); + dest+=row->item.size; + break; + case UL_FILE_SPEC: + // tainted, untaint language: file [name] + escape( + encode(need_file_encode, '_'); + ); + break; + case UL_URI: + // tainted, untaint language: uri + escape(switch(*src) { + case ' ': to_char('+'); break; + default: encode(need_uri_encode, '%'); + }); + break; + case UL_HTTP_HEADER: + // tainted, untaint language: http-field-content-text + escape(switch(*src) { + case ' ': to_char('+'); break; + default: encode(need_uri_encode, '%'); + }); + break; + case UL_MAIL_HEADER: + // tainted, untaint language: mail-header + if(charset) { + // Subject: Re: parser3: =?koi8-r?Q?=D3=C5=CD=C9=CE=C1=D2?= + const char *src=row->item.ptr; + bool to_quoted_printable=false; + for(int size=row->item.size; size--; src++) { + if(*src & 0x80) { + if(!to_quoted_printable) { + dest+=sprintf(dest, "=?%.15s?Q?", charset); + to_quoted_printable=true; + } + dest+=sprintf(dest, "=%02X", *src & 0xFF); + } else { + *dest++=*src; + } + } + if(to_quoted_printable) // close + dest+=sprintf(dest, "?="); + } else { + memcpy(dest, row->item.ptr, row->item.size); + dest+=row->item.size; + } + break; + case UL_TABLE: + // tainted, untaint language: table + escape(switch(*src) { + case '\t': to_char(' '); break; + case '\n': to_char(' '); break; + _default; + }); + break; + case UL_SQL: + // tainted, untaint language: sql + if(connection) + dest+=connection->quote(dest, row->item.ptr, row->item.size); + else + throw Exception(0, 0, + this, + "untaint in SQL language failed - no connection specified"); + break; + case UL_JS: + escape(switch(*src) { + case '"': to_string("\\\"", 2); break; + case '\'': to_string("\\'", 2); break; + case '\n': to_string("\\n", 2); break; + case '\\': to_string("\\\\", 2); break; + case '\xFF': to_string("\\\xFF", 2); break; + _default; + }); + break; + case UL_XML: + escape(switch(*src) { + case '&': to_string("&", 5); break; + case '>': to_string(">", 4); break; + case '<': to_string("<", 4); break; + case '"': to_string(""", 6); break; + case '\'': to_string("'", 6); break; + _default; + }); + break; + case UL_HTML: + escape(switch(*src) { + case '&': to_string("&", 5); break; + case '>': to_string(">", 4); break; + case '<': to_string("<", 4); break; + case '"': to_string(""", 6); break; + _default; + }); + break; + default: + throw Exception(0, 0, + this, + "unknown untaint language #%d", + static_cast(row->item.lang)); // sould never + break; // never + } + + if(to_lang & UL_OPTIMIZE_BIT) { + // optimizing whitespace + char *stop=dest; dest=start; + for(char *src=start; srclink; \ + } while(chunk); \ + +break2: + return dest; +} + +char *String::cstr_debug_origins() const { + //_asm int 3; + char *result=(char *)malloc(size()+used_rows()*MAX_STRING*2); + char *dest=result; + const Chunk *chunk=&head; do { const Chunk::Row *row=chunk->rows; - for(size_t i=0; icount; i++, row++) { + for(uint i=0; icount; i++, row++) { if(row==append_here) goto break2; - Untaint_lang to_lang=lang==UL_UNSPECIFIED?row->item.lang:lang; - - char *dest_before_origins=dest; - - if(forigins_mode) { #ifndef NO_STRING_ORIGIN - if(row->item.origin.file) - dest+=sprintf(dest, "%s(%d)", - row->item.origin.file, - 1+row->item.origin.line); - else - dest+=sprintf(dest, "unknown"); + if(row->item.origin.file) + dest+=sprintf(dest, ORIGIN_FILE_LINE_FORMAT, + row->item.origin.file, + 1+row->item.origin.line); + else + dest+=sprintf(dest, ""); #endif - dest+=sprintf(dest, "#%s: ", - String_Untaint_lang_name[to_lang]); - } - char *dest_after_origins=dest; - - // WARNING: - // string can grow only UNTAINT_TIMES_BIGGER - switch(to_lang) { - case UL_CLEAN: - // clean piece - { // optimizing whitespace - const char *src=row->item.ptr; - for(int size=row->item.size; size--; src++) - switch(*src) { - case ' ': case '\n': case '\r': case '\t': - if(!whitespace) { - *dest++=*src; - whitespace=true; - } - break; - default: - whitespace=false; - *dest++=*src; - break; - } - } - break; - case UL_TAINTED: - // tainted piece, but undefined untaint language - // for VString.as_double of tainted values - // for ^process{body} evaluation - case UL_AS_IS: - // tainted, untaint language: as-is - memcpy(dest, row->item.ptr, row->item.size); - dest+=row->item.size; - break; - case UL_FILE_NAME: - // tainted, untaint language: file [name] - escape(switch(*src) { - case ' ': to_char('_'); break; - encode(need_file_encode, '+'); - }); - break; - case UL_URI: - // tainted, untaint language: uri - escape(switch(*src) { - case ' ': to_char('+'); break; - encode(need_uri_encode, '%'); - }); - break; - case UL_HTTP_HEADER: - // tainted, untaint language: http-header - if(need_quote_http_header(row->item.ptr, row->item.size)) { - *dest++='\"'; - escape(switch(*src) { - case '\"': to_string("\\\"", 2); break; - _default; - }); - *dest++='\"'; - } else { - memcpy(dest, row->item.ptr, row->item.size); - dest+=row->item.size; - } - break; - case UL_MAIL_HEADER: - // tainted, untaint language: mail-header - if(charset) { - // Subject: Re: parser3: =?koi8-r?Q?=D3=C5=CD=C9=CE=C1=D2?= - const char *src=row->item.ptr; - bool to_quoted_printable=false; - for(int size=row->item.size; size--; src++) { - if(*src & 0x80) { - if(!to_quoted_printable) { - dest+=sprintf(dest, "=?%.15s?Q?", charset); - to_quoted_printable=true; - } - dest+=sprintf(dest, "=%02X", *src & 0xFF); - } else { - *dest++=*src; - } - } - if(to_quoted_printable) // close - dest+=sprintf(dest, "?="); - } else { - memcpy(dest, row->item.ptr, row->item.size); - dest+=row->item.size; - } - break; - case UL_TABLE: - // tainted, untaint language: table - escape(switch(*src) { - case '\t': to_char(' '); break; - case '\n': to_char(' '); break; - _default; - }); - break; - case UL_SQL: - // tainted, untaint language: sql - if(connection) - dest+=connection->quote(dest, row->item.ptr, row->item.size); - else - THROW(0, 0, - this, - "untaint in SQL language failed - no connection specified"); - break; - case UL_JS: - escape(switch(*src) { - case '"': to_string("\\\"", 2); break; - case '\'': to_string("\\'", 2); break; - case '\n': to_string("\\n", 2); break; - case '\\': to_string("\\\\", 2); break; - case '\xFF': to_string("\\\xFF", 2); break; - _default; - }); - break; - case UL_HTML: - escape(switch(*src) { - case '&': to_string("&", 5); break; - case '>': to_string(">", 4); break; - case '<': to_string("<", 4); break; - case '"': to_string(""", 6); break; - //TODO: XSLT case '\'': to_string("'", 6); break; - _default; - }); - break; - case UL_USER_HTML: { - // tainted, untaint language: html-typo - if(!typo_dict) // never, always has default - THROW(0, 0, - this, - "untaint to user-html lang failed, no typo table"); - - char *html_for_typo= - (char *)malloc(row->item.size*2/* '\n' -> '\' 'n' */+1); - // note: - // there still is a possibility that user - // would not replace \n as she supposed to - // and rather replace \ and n into huge strings - // thus causing memory overrun - // this can be dealed by allocating *2 memory, but that's too expensive - size_t html_for_typo_size; - { // local dest - char *dest=html_for_typo; - escape(switch(*src) { - // convinient name for typo match "\n" - case '\r': - to_string("\\n", 2); // \r -> "\n" - if(size && src[1]=='\n') { // \r\n -> remove \n - size--; src++; - } - break; - case '\n': - to_string("\\n", 2); - break; - //TODO: XSLT case '\'': to_string("'", 6); break; - _default; - }); - *dest=0; - html_for_typo_size=dest-html_for_typo; - } - // typo table replacements - const char *src=html_for_typo; - do { - // there is a row where first column starts 'src' - if(Table::Item *item=typo_dict->first_that_starts(src)) { - // get a=>b values - const String& a=*static_cast(item)->get_string(0); - const String& b=*static_cast(item)->get_string(1); - // overflow check: - // b allowed to be max UNTAINT_TIMES_BIGGER then a - if(b.size()>UNTAINT_TIMES_BIGGER*a.size()) { - pool().set_tag(0); // avoid recursion - THROW(0, 0, - &b, - "is %g times longer then '%s', " - "while maximum, handled by Parser, is %d", - ((double)b.size())/a.size(), - a.cstr(), - UNTAINT_TIMES_BIGGER); - } - - // skip 'a' in 'src' - src+=a.size(); - // write 'b' to 'dest' - b.store_to(dest); - // skip 'b' in 'dest' - dest+=b.size(); - } else - *dest++=*src++; - } while(*src); - break; - } - default: - THROW(0, 0, + uchar show_lang=row->item.lang & ~UL_OPTIMIZE_BIT; + if(show_lang>=sizeof(String_Untaint_lang_name)/sizeof(String_Untaint_lang_name[0])) + throw Exception(0, 0, this, - "unknown untaint language #%d of %d piece", - static_cast(row->item.lang), - i); // never - break; // never - } + "unknown untaint language #%d", + static_cast(show_lang)); // sould never - if((lang==UL_UNSPECIFIED?row->item.lang:lang)!=UL_CLEAN) - whitespace=false; + dest+=sprintf(dest, "#%s%s: ", + String_Untaint_lang_name[show_lang], + row->item.lang & UL_OPTIMIZE_BIT?".O":""); + char *dest_after_origins=dest; - if(forigins_mode) - if(dest==dest_after_origins) // never moved==optimized space - dest=dest_before_origins; - else { - for(char *p=dest_after_origins; pitem.ptr, row->item.size); + dest+=row->item.size; - to_char('\n'); - } + remove_crlf(dest_after_origins, dest); + to_char('\n'); } chunk=row->link; } while(chunk); + break2: - return dest; + *dest=0; + return result; }