--- parser3/src/main/untaint.C 2003/02/14 12:59:10 1.115.2.9 +++ parser3/src/main/untaint.C 2003/04/18 10:12:05 1.116 @@ -1,11 +1,11 @@ /** @file Parser: String class part: untaint mechanizm. - Copyright(c) 2001-2003 ArtLebedev Group (http://www.artlebedev.com) + Copyright(c) 2001, 2003 ArtLebedev Group (http://www.artlebedev.com) Author: Alexandr Petrosian (http://paf.design.ru) */ -static const char* IDENT_UNTAINT_C="$Date: 2003/02/14 12:59:10 $"; +static const char* IDENT_UNTAINT_C="$Date: 2003/04/18 10:12:05 $"; #include "pa_pool.h" #include "pa_string.h" @@ -17,18 +17,23 @@ static const char* IDENT_UNTAINT_C="$Dat #include "pa_dictionary.h" #include "pa_common.h" #include "pa_charset.h" -#include "pa_request_charsets.h" + +//#define DEBUG_STRING_APPENDS_VS_EXPANDS + +#ifdef DEBUG_STRING_APPENDS_VS_EXPANDS +ulong string_string_shortcut_economy=0; +#endif #define escape(action) \ { \ - const char* src=fragment->ptr; \ - for(int size=fragment->size; size--; src++) \ + const char *src=row->item.ptr; \ + for(int size=row->item.size; size--; src++) \ action \ } #define _default default: *dest++=*src; break #define encode(need_encode_func, prefix) \ if(need_encode_func(*src)) { \ - static const char* hex="0123456789ABCDEF"; \ + static const char *hex="0123456789ABCDEF"; \ char chunk[3]={prefix}; \ chunk[1]=hex[((unsigned char)*src)/0x10]; \ chunk[2]=hex[((unsigned char)*src)%0x10]; \ @@ -47,7 +52,7 @@ inline bool need_file_encode(unsigned ch return strchr( "*?'\"<>|" #ifndef WIN32 - ":\\~" + ":\\" #endif , c)!=0; } @@ -66,7 +71,7 @@ inline bool need_http_header_encode(unsi // -static const char* String_Untaint_lang_name[]={ +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 @@ -126,7 +131,7 @@ quoted-pair = "\" CHAR if(strchr("()<>@,;:\\\"/[]?={} \t", *ptr)) */ -inline bool need_quote_http_header(const char* ptr, size_t size) { +inline bool need_quote_http_header(const char *ptr, size_t size) { for(; size--; ptr++) if(strchr(";\\\"= \t" /* excluded ()<>@, :/ ? []{} */, *ptr)) return true; @@ -135,69 +140,121 @@ inline bool need_quote_http_header(const //#include "pa_sapi.h" /** - appends to other String, + 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.chunk] + if some language-change specified or src already appended to some other string[last_chunk==0] + cloning pieces. */ -String& String::append_to(String& dest, uchar lang, bool forced) const { - if(is_empty()) - return dest; - - // without language-chage string? - if(lang==UL_PASS_APPENDED) { - // can simply append elements - ssize_t delta=fused-(dest.fallocated-dest.fused); - if(delta>0) // they don't fit? - dest.expand(delta); - - // felements[fused...]=src.felements[0..fused], fused+=src.fused - memcpy(&dest.felements[dest.fused], felements, sizeof(felements[0])*fused); - dest.fused+=fused; - return dest; +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, + 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.chunk.count initally equeals this.head.chunk.count and shrinks-only, + // so can't be more than this.head.chunk.count, + // which means that we know that + // src.head.chunk would fit into this.head.chunk + if(is_empty()) { // our head.chunk is empty + // they have more than head.chunk? we need all head.chunk : we need only filled-part of head.chunk + Chunk *src_head_link=src.head.chunk.rows[src.head.chunk.count].link; + size_t head_count=src_head_link?src.head.chunk.count:(src.append_here-src.head.chunk.rows); + // "your head.chunk is my head.chunk" + memcpy(head.chunk.rows, src.head.chunk.rows, sizeof(Chunk::Row)*(head_count)); + if(src_head_link) { + // "your body is my body" + head.chunk.rows[head.chunk.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.chunk; + // "your append_here is recalc-mine now" + append_here=head.chunk.rows+head_count; + } + } else { // our head.chunk contains something + // "chopping off my tail-reserve" + last_chunk->count=append_here-last_chunk->rows; + // "you is my tail" + append_here->link=&src.head.chunk; + // "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_FOREACH_FRAGMENT( - dest.APPEND(fragment->ptr, fragment->size, + STRING_SRC_FOREACH_ROW( + APPEND(row->item.ptr, row->item.size, lang, //forcing passed lang - fragment->origin.file, fragment->origin.line); + row->item.origin.file, row->item.origin.line); ) else if(lang==UL_PASS_APPENDED) - STRING_FOREACH_FRAGMENT( - dest.APPEND(fragment->ptr, fragment->size, - fragment->lang, // passing item's lang - fragment->origin.file, fragment->origin.line); + 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_FOREACH_FRAGMENT( - dest.APPEND(fragment->ptr, fragment->size, - fragment->lang==UL_TAINTED?lang:( - fragment->lang==UL_CLEAN?UL_CLEAN|UL_OPTIMIZE_BIT: // ORing with OPTIMIZED flag - fragment->lang + 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 ), - fragment->origin.file, fragment->origin.line); + row->item.origin.file, row->item.origin.line); ) else - STRING_FOREACH_FRAGMENT( - dest.APPEND(fragment->ptr, fragment->size, - fragment->lang==UL_TAINTED?lang:fragment->lang, - fragment->origin.file, fragment->origin.line); + 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); ); - - return dest; +/* + 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, - const Request_charsets *charsets) const { + Charset *buf_charset) const { size_t dest=1; // for terminating 0 - STRING_FOREACH_FRAGMENT( - uchar to_lang=lang==UL_UNSPECIFIED?fragment->lang:lang; + STRING_FOREACH_ROW( + uchar to_lang=lang==UL_UNSPECIFIED?row->item.lang:lang; switch(to_lang & ~UL_OPTIMIZE_BIT) { case UL_CLEAN: @@ -210,46 +267,46 @@ size_t String::cstr_bufsize(Untaint_lang // for ^process{body} evaluation // tainted, untaint language: as-is - dest+=fragment->size; + dest+=row->item.size; break; case UL_FILE_SPEC: // tainted, untaint language: file [name] - dest+=fragment->size*3/* worst: Z->%XX */; + dest+=row->item.size*3/* worst: Z->%XX */; break; case UL_URI: // tainted, untaint language: uri - dest+=fragment->size*6*3/* worst utf8 x worst Z->%XX */; + 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+=fragment->size*3/* worst: Z->%XX */; + dest+=row->item.size*3/* worst: Z->%XX */; break; case UL_MAIL_HEADER: // tainted, untaint language: mail-header - if(charsets) { + if(buf_charset) { int parts_count=1; - const char* prev=fragment->ptr; - size_t size=fragment->size; - while(const char* comma_at=(const char*)memchr(prev, ',', size)) { + const char *prev=row->item.ptr; + size_t size=row->item.size; + while(const char *comma_at=(const char*)memchr(prev, ',', size)) { parts_count++; size-=comma_at-prev; prev=comma_at+1; } // Subject: Re: parser3: =?koi8-r?Q?=D3=C5=CD=C9=CE=C1=D2?= dest+= - fragment->size*3+ - parts_count*(charsets->mail().name()->size()+MAX_STRING/* worst: =?charset?Q?=%XX?= */); + row->item.size*3+ + parts_count*(buf_charset->name().size()+MAX_STRING/* worst: =?charset?Q?=%XX?= */); } else - dest+=fragment->size; + dest+=row->item.size; break; case UL_TABLE: // tainted, untaint language: table - dest+=fragment->size; + dest+=row->item.size; break; case UL_SQL: // tainted, untaint language: sql if(connection) - dest+=connection->quote(0, fragment->ptr, fragment->size); + dest+=connection->quote(0, row->item.ptr, row->item.size); break; case UL_JS: escape(switch(*src) { @@ -318,14 +375,25 @@ static bool mail_header_char_valid_withi } char *String::store_to(char *dest, Untaint_lang lang, SQL_Connection *connection, - const Request_charsets *charsets) const { + Charset *store_to_charset, + const char *store_to_charset_name) const { // WARNING: // before any changes check cstr_bufsize first!!! bool whitespace=true; -// STRING_FOREACH_FRAGMENT( - element_type *last = felements+fused; element_type *fragment = felements; for(; fragmentrows; \ + uint countdown=chunk->count; \ + while(row!=(*this).append_here) { \ + if(countdown==0) { \ + chunk=row->link; \ + row=chunk->rows; \ + countdown=chunk->count; \ + }; +//*/ +// STRING_FOREACH_ROW( - uchar to_lang=lang==UL_UNSPECIFIED?fragment->lang:lang; + uchar to_lang=lang==UL_UNSPECIFIED?row->item.lang:lang; char *start=dest; @@ -340,16 +408,16 @@ char *String::store_to(char *dest, Untai // for ^process{body} evaluation // tainted, untaint language: as-is - memcpy(dest, fragment->ptr, fragment->size); - dest+=fragment->size; + memcpy(dest, row->item.ptr, row->item.size); + dest+=row->item.size; break; case UL_FILE_SPEC: // tainted, untaint language: file [name] escape( // Macintosh has problems with small Russian letter 'r' - if( *src=='\xF0' && charsets && *charsets->source().name()=="windows-1251" ) { + if( *src=='\xF0' && pool().get_source_charset().name()=="windows-1251" ) { // fixing that letter for most common charset - to_char('p'); + to_char('p'); } else // fallback to default encode(need_file_encode, '_'); ); @@ -358,17 +426,11 @@ char *String::store_to(char *dest, Untai // tainted, untaint language: uri const void *client_ptr; size_t client_size; - if(charsets) - Charset::transcode(charsets->pool(), - charsets->source(), fragment->ptr, fragment->size, - charsets->client(), client_ptr, client_size); - else { - client_ptr=fragment->ptr; - client_size=fragment->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; + const char *src=(const char *)client_ptr; for(int size=client_size; size--; src++) encode(need_uri_encode, '%'); } @@ -382,20 +444,19 @@ char *String::store_to(char *dest, Untai case UL_MAIL_HEADER: // tainted, untaint language: mail-header // http://www.ietf.org/rfc/rfc2047.txt - if(charsets) { + if(store_to_charset && store_to_charset_name) { const void *mail_ptr; size_t mail_size; - Charset::transcode(charsets->pool(), - charsets->source(), fragment->ptr, fragment->size, - charsets->mail(), mail_ptr, mail_size); - CharPtr charset_name=charsets->mail().name()->cstr(); + 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; + const char *src=(const char *)mail_ptr; bool to_quoted_printable=false; bool email=false; - for(const char* end=src+mail_size; src(to_lang)); // should never break; // never @@ -500,41 +561,44 @@ char *String::store_to(char *dest, Untai break; }; } else // piece without optimization +//debug whitespace=false; -// ); - } - + row++; countdown--; \ + } +/* + ); +*/ return dest; } -char *String::cstr_debug_origins(Pool& pool) const { +char *String::cstr_debug_origins() const { //_asm int 3; - char *result=new(pool) char[size()+fused*MAX_STRING*2]; + char *result=(char *)malloc(size()+used_rows()*MAX_STRING*2); char *dest=result; - STRING_FOREACH_FRAGMENT( + STRING_FOREACH_ROW( IFNDEF_NO_STRING_ORIGIN( - if(fragment->origin.file) + if(row->item.origin.file) dest+=sprintf(dest, ORIGIN_FILE_LINE_FORMAT, - fragment->origin.file, - 1+fragment->origin.line); + row->item.origin.file, + 1+row->item.origin.line); else dest+=sprintf(dest, ""); ); - uchar show_lang=fragment->lang & ~UL_OPTIMIZE_BIT; + 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(Exception::undefined_type, - Exception::undefined_source, + throw Exception(0, + this, "unknown untaint language #%d", static_cast(show_lang)); // sould never dest+=sprintf(dest, "#%s%s: ", String_Untaint_lang_name[show_lang], - fragment->lang & UL_OPTIMIZE_BIT?".O":""); + row->item.lang & UL_OPTIMIZE_BIT?".O":""); char *dest_after_origins=dest; - memcpy(dest, fragment->ptr, fragment->size); - dest+=fragment->size; + memcpy(dest, row->item.ptr, row->item.size); + dest+=row->item.size; remove_crlf(dest_after_origins, dest); to_char('\n');