--- parser3/src/main/untaint.C 2001/08/01 12:08:40 1.58 +++ parser3/src/main/untaint.C 2002/02/28 15:40:40 1.98 @@ -1,11 +1,11 @@ /** @file Parser: String class part: untaint mechanizm. - Copyright(c) 2001 ArtLebedev Group(http://www.artlebedev.com) + Copyright(c) 2001, 2002 ArtLebedev Group (http://www.artlebedev.com) + Author: Alexandr Petrosian (http://paf.design.ru) - Author: Alexander Petrosyan (http://design.ru/paf) + $Id: untaint.C,v 1.98 2002/02/28 15:40:40 paf Exp $ */ -static const char *RCSId="$Id: untaint.C,v 1.58 2001/08/01 12:08:40 parser Exp $"; #include "pa_pool.h" #include "pa_string.h" @@ -15,6 +15,14 @@ static const char *RCSId="$Id: untaint.C #include "pa_globals.h" #include "pa_sql_connection.h" #include "pa_dictionary.h" +#include "pa_common.h" +#include "pa_charset.h" + +//#define DEBUG_STRING_APPENDS_VS_EXPANDS + +#ifdef DEBUG_STRING_APPENDS_VS_EXPANDS +ulong string_string_shortcut_economy=0; +#endif #define escape(action) \ { \ @@ -24,30 +32,35 @@ 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 +88,20 @@ 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 @@ -142,267 +144,396 @@ 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 +//#include "pa_sapi.h" +/** + appends other String, + marking all tainted pieces of it with @a lang. + or marking ALL pieces of it with a @a lang when @a forced to, + and propagating OPTIMIZE language bit. + + using architecture advantage: after string-to-string-append string never modified. + algorithm: + if no language-change specified and src not yet appended to some other string[last_chunk!=0] + shrinking dest last_chunk[preparing it for linking], + ///shrinking src last_chunk[preparing it to be linked, consequent dest.appends would go there], + linking[dest.last_chunk = src.head] + if some language-change specified or src already appended to some other string[last_chunk==0] + cloning pieces. */ -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) { + if(!last_chunk) // growth stopped [we're appended as string to somebody] + throw Exception(0, 0, + this, + "string growth stopped (append string)"); + + if(src.is_empty()) + return *this; + + // without language-chage, not-appended-before, big[not fitting our tail] string? + if(lang==UL_PASS_APPENDED + && src.last_chunk + && (uint(&last_chunk->rows[last_chunk->count]-append_here) < src.used_rows())) { +#ifdef DEBUG_STRING_APPENDS_VS_EXPANDS + string_string_shortcut_economy+=src.used_rows()*sizeof(String::Chunk::Row); +#endif + + // using fact: + // src.head.count initally equeals this.head.count and shrinks-only, + // so can't be more than this.head.count, + // which means that we know that + // src.head would fit into this.head + if(is_empty()) { // our head is empty + // they have more than head? we need all head : we need only filled-part of head + Chunk *src_head_link=src.head.rows[src.head.count].link; + size_t head_count=src_head_link?src.head.count:(src.append_here-src.head.rows); + // "your head is my head" + memcpy(head.rows, src.head.rows, sizeof(Chunk::Row)*(head_count)); + if(src_head_link) { + // "your body is my body" + head.rows[head.count=head_count].link=src_head_link; + // "your last_chunk is mine now" + last_chunk=src.last_chunk; + // "your append_here is mine now" + append_here=src.append_here; + } else { + // "your last_chunk is mine now" + last_chunk=&head; + // "your append_here is recalc-mine now" + append_here=head.rows+head_count; + } + } else { // our head contains something + // "chopping off my tail-reserve" + last_chunk->count=append_here-last_chunk->rows; + // "you is my tail" + append_here->link=&src.head; + // "your last_chunk is mine now" + last_chunk=src.last_chunk; + // "your append_here is mine now" + append_here=src.append_here; + } + + // stop-growing mark + src.last_chunk=0; + return *this; + } + + // 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); + ); +/* + for(Chunk::Row *row=last_chunk->rows; rowlink==(void*)0xcdcdcdcd) + _asm int 3;*/ + return *this; +} + +size_t String::cstr_bufsize(Untaint_lang lang, + SQL_Connection *connection, + Charset *buf_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*6*3/* worst utf8 x 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(buf_charset) { + // Subject: Re: parser3: =?koi8-r?Q?=D3=C5=CD=C9=CE=C1=D2?= + dest+= + row->item.size*3+ + buf_charset->name().size()+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; + } + ); + return dest; } -/** @todo fix theoretical \n mem overrun in TYPO replacements -@todo rename base_64 to quoted_printable [invalid name now] -*/ 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; - + Charset *store_to_charset) const { + // WARNING: + // before any changes check cstr_bufsize first!!! bool whitespace=true; - const Chunk *chunk=&head; - do { - const Chunk::Row *row=chunk->rows; - for(size_t 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"); -#endif - dest+=sprintf(dest, "#%s: ", - String_Untaint_lang_name[to_lang]); - } - char *dest_after_origins=dest; + STRING_FOREACH_ROW( + uchar to_lang=lang==UL_UNSPECIFIED?row->item.lang:lang; + + char *start=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; + 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 + const void *client_ptr; + size_t client_size; + Charset::transcode(pool(), + pool().get_source_charset(), row->item.ptr, row->item.size, + pool().get_client_charset(), client_ptr, client_size); + { + const char *src=(const char *)client_ptr; + for(int size=client_size; size--; src++) + 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(store_to_charset) { + const void *mail_ptr; + size_t mail_size; + Charset::transcode(pool(), + pool().get_source_charset(), row->item.ptr, row->item.size, + *store_to_charset, mail_ptr, mail_size); + + // 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; + for(int size=mail_size; size--; src++) { + if(*src & 0x80) { + if(!to_quoted_printable) { + dest+=sprintf(dest, "=?%s?Q?", store_to_charset->name().cstr()); + to_quoted_printable=true; } + dest+=sprintf(dest, "=%02X", *src & 0xFF); + } else { + *dest++=*src; + } } - 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 + if(to_quoted_printable) // close + dest+=sprintf(dest, "?="); + + } else { 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_base_64=false; - for(int size=row->item.size; size--; src++) { - if(*src & 0x80) { - if(!to_base_64) { - dest+=sprintf(dest, "=?%.15s?Q?", charset); - to_base_64=true; - } - dest+=sprintf(dest, "=%02X", *src & 0xFF); - } else { - *dest++=*src; - } - } - if(to_base_64) // 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); - // empty 'a' | 'b' checks - if(a.size()==0 || b.size()==0) { - pool().set_tag(0); // avoid recursion - THROW(0, 0, - typo_dict->origin_string(), - "typo table column elements must not be empty"); - } - // 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); - dest+=b.size(); - } else - *dest++=*src++; - } while(*src); - break; - } - default: - THROW(0, 0, - this, - "unknown untaint language #%d of %d piece", - static_cast(row->item.lang), - i); // never - break; // never } + 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((lang==UL_UNSPECIFIED?row->item.lang:lang)!=UL_CLEAN) - whitespace=false; - - if(forigins_mode) - if(dest==dest_after_origins) // never moved==optimized space - dest=dest_before_origins; - else { - for(char *p=dest_after_origins; plink; - } 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; + + STRING_FOREACH_ROW( +IFNDEF_NO_STRING_ORIGIN( + 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, ""); +); + 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", + static_cast(show_lang)); // sould never + + dest+=sprintf(dest, "#%s%s: ", + String_Untaint_lang_name[show_lang], + row->item.lang & UL_OPTIMIZE_BIT?".O":""); + char *dest_after_origins=dest; + + memcpy(dest, row->item.ptr, row->item.size); + dest+=row->item.size; + + remove_crlf(dest_after_origins, dest); + to_char('\n'); + ); + + *dest=0; + return result; +}