--- parser3/src/main/untaint.C 2002/12/15 15:58:34 1.113 +++ parser3/src/main/untaint.C 2003/03/20 14:19:10 1.115.2.12.2.6 @@ -1,50 +1,49 @@ /** @file Parser: String class part: untaint mechanizm. - Copyright(c) 2001, 2002 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: 2002/12/15 15:58:34 $"; +static const char* IDENT_UNTAINT_C="$Date: 2003/03/20 14:19:10 $"; + -#include "pa_pool.h" #include "pa_string.h" #include "pa_hash.h" #include "pa_exception.h" #include "pa_table.h" #include "pa_globals.h" -#include "pa_sql_connection.h" #include "pa_dictionary.h" #include "pa_common.h" #include "pa_charset.h" +#include "pa_request_charsets.h" + +extern "C" { // author forgot to do that +#include "ec.h" +} -//#define DEBUG_STRING_APPENDS_VS_EXPANDS +//#define PA_SQL -#ifdef DEBUG_STRING_APPENDS_VS_EXPANDS -ulong string_string_shortcut_economy=0; +#ifdef PA_SQL +#include "pa_sql_connection.h" #endif #define escape(action) \ - { \ - const char *src=row->item.ptr; \ - for(int size=row->item.size; size--; src++) \ - action \ + for(; fragment_length--; CORD_next(pos)) { \ + char c=CORD_pos_fetch(pos); \ + action \ } -#define _default default: *dest++=*src; break +#define _default default: CORD_ec_append(result, c); break #define encode(need_encode_func, prefix) \ - 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]; \ - memcpy(dest, chunk, 3); dest+=3; \ - } else \ - *dest++=*src; \ - break -#define to_char(c) *dest++=c -#define to_string(b, bsize) \ - memcpy(dest, b, bsize); \ - dest+=bsize; \ + 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); +#define to_char(c) CORD_ec_append(result, c) +#define to_string(s) CORD_ec_append_cord(result, s) inline bool need_file_encode(unsigned char c){ // russian letters and space ENABLED @@ -71,7 +70,7 @@ 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 @@ -131,212 +130,57 @@ 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; return false; } -//#include "pa_sapi.h" + /** - appends other String, + appends to 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(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 +String& String::append_to(String& dest, Untaint_lang lang, bool forced) const { + if(!*this) + return dest; - // 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; - } + // first: letters + dest.body=CORD_cat(dest.body, body); - // stop-growing mark - src.last_chunk=0; - return *this; + // next: fragment infos + + if(lang==UL_PASS_APPENDED) // without language-chage string? + dest.fragments.append(fragments, 0, fragments.count()); + else if(forced) //forcing passed lang? + dest.fragments+=Fragment(lang, length()); + else if(lang&UL_OPTIMIZE_BIT) { + Array_iterator i(fragments); + while(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 + 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(); + dest.fragments+=Fragment( + fragment.lang==UL_TAINTED?lang:fragment.lang, + fragment.length); + } } - // 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) { - int parts_count=1; - 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+= - row->item.size*3+ - parts_count*(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; } @@ -367,35 +211,34 @@ RFC without "=", or one could not differ "=E0" and "russian letter a" and without "_", or in would mean 0x20 */ -static bool mail_header_char_valid_within_Qencoded(char c) { +inline bool mail_header_char_valid_within_Qencoded(char c) { return c>='A' && c<='Z' || c>='a' && c<='Z' || c>='0' && c<='9' || strchr("!*+-/", c); } -char *String::store_to(char *dest, Untaint_lang lang, - SQL_Connection *connection, - Charset *store_to_charset, - const char *store_to_charset_name) const { - // WARNING: - // before any changes check cstr_bufsize first!!! +inline void ec_append(CORD_ec result, 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, + const Request_charsets *charsets) const { + CORD_ec result; CORD_ec_init(result); + CORD_pos pos; CORD_set_pos(pos, body, 0); + bool whitespace=true; -//debug - const Chunk *chunk=&(*this).head.chunk; \ - const Chunk::Row *row=chunk->rows; \ - 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?row->item.lang:lang; + size_t fragment_start=0; + size_t fragment_end; + for(Array_iterator i(fragments); i.has_next(); fragment_start=fragment_end) { + const Fragment& fragment=i.next(); + size_t fragment_length=fragment.length; + fragment_end=fragment_start+fragment_length; - char *start=dest; + Untaint_lang to_lang=lang==UL_UNSPECIFIED?fragment.lang:lang; switch(to_lang & ~UL_OPTIMIZE_BIT) { case UL_CLEAN: @@ -408,204 +251,179 @@ char *String::store_to(char *dest, Untai // for ^process{body} evaluation // tainted, untaint language: as-is - memcpy(dest, row->item.ptr, row->item.size); - dest+=row->item.size; + ec_append(result, pos, fragment_length); break; case UL_FILE_SPEC: // tainted, untaint language: file [name] escape( // Macintosh has problems with small Russian letter 'r' - if( *src=='\xF0' && pool().get_source_charset().name()=="windows-1251" ) { + if( c=='\xF0' && charsets && charsets->source().name()=="windows-1251" ) { // fixing that letter for most common charset - to_char('p'); - continue; - } - // fallback to default - encode(need_file_encode, '_'); + to_char('p'); + } else // fallback to default + encode(need_file_encode, '_'); ); break; +#ifdef LATER 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++) + const char *substr= + CORD_to_const_char_star( + CORD_substr(body, fragment_start, fragment_length)); + const void *client_ptr; + size_t client_size; + if(charsets) + Charset::transcode( + charsets->source(), substr, fragment_length, + charsets->client(), client_ptr, client_size); + else { + client_ptr=substr; + client_size=fragment_length; + } + + char c; + for(const char* src=(const char* )client_ptr; c=*src++; ) encode(need_uri_encode, '%'); } break; +#endif case UL_HTTP_HEADER: // tainted, untaint language: http-field-content-text escape( encode(need_uri_encode, '%'); ); break; +#ifdef LATER case UL_MAIL_HEADER: // tainted, untaint language: mail-header // http://www.ietf.org/rfc/rfc2047.txt - if(store_to_charset && store_to_charset_name) { - 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); + if(charsets) { + const void *mail_ptr=fragment.ptr; + size_t mail_size=fragment_length; + 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; + const char* src=(const char* )mail_ptr; bool to_quoted_printable=false; bool email=false; - for(const char *end=src+mail_size; src') + *dest++=c; + if(c=='>') email=false; } if(to_quoted_printable) // close dest+=sprintf(dest, "?="); } else { - memcpy(dest, row->item.ptr, row->item.size); - dest+=row->item.size; + memcpy(dest, fragment.ptr, fragment_length); + dest+=fragment_length; } break; +#endif case UL_TABLE: // tainted, untaint language: table - escape(switch(*src) { + escape(switch(c) { case '\t': to_char(' '); break; case '\n': to_char(' '); break; _default; }); break; +#ifdef PA_SQL case UL_SQL: // tainted, untaint language: sql - if(connection) - dest+=connection->quote(dest, row->item.ptr, row->item.size); + if(connection) { + dest+=connection->quote(0, + CORD_substr(body, fragment_start, fragment_length), + fragment_length); +...^^^ + + dest+=connection->quote(dest, fragment.ptr, fragment_length); else throw Exception(0, - this, + 0, "untaint in SQL language failed - no connection specified"); break; +#endif 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; + 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 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; + 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 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; + escape(switch(c) { + case '&': to_string("&"); break; + case '>': to_string(">"); break; + case '<': to_string("<"); break; + case '"': to_string("""); break; _default; }); break; default: throw Exception(0, - this, + 0, "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 *src=start; srcitem.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, - 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'); - ); +#endif + } - *dest=0; - return result; + return CORD_ec_to_cord(result); }