--- parser3/src/main/untaint.C 2003/01/29 11:06:44 1.115.2.3 +++ parser3/src/main/untaint.C 2003/07/24 11:31:24 1.117 @@ -1,44 +1,79 @@ /** @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: 2003/01/29 11:06:44 $"; +static const char* IDENT_UNTAINT_C="$Date: 2003/07/24 11:31:24 $"; + -#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" +#include "pa_sapi.h" + +extern "C" { // author forgot to do that +#include "ec.h" +} + +#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) \ - { \ - const char *src=fragment->ptr; \ - for(int size=fragment->size; size--; src++) \ - action \ + for(; fragment_length--; CORD_next(pos)) { \ + char c=CORD_pos_fetch(pos); \ + action \ } -#define _default default: *dest++=*src; 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; \ +#define _default default: CORD_ec_append(result, c); break +#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, otherwise); +#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 @@ -46,7 +81,7 @@ inline bool need_file_encode(unsigned ch return strchr( "*?'\"<>|" #ifndef WIN32 - ":\\~" + ":\\" #endif , c)!=0; } @@ -65,7 +100,7 @@ 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 @@ -125,155 +160,56 @@ 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 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. */ -String& String::append_to(String& dest, uchar lang, bool forced) const { +String& String::append_to(String& dest, Language lang, bool forced) const { if(is_empty()) return dest; - // without language-chage, small[fitting dest tail] string? - if(lang==UL_PASS_APPENDED && fused<=(dest.fallocated-dest.fused)) { - // felements[fused...]=src.felements[0..fused], fused+=src.fused - memcpy(&dest.felements[dest.fused], felements, sizeof(felements[0])*fused); - dest.fused+=fused; - return dest; - } - - // manually unrolled code to avoid do{if(const)} constructs - if(forced) - STRING_FOREACH_FRAGMENT( - dest.APPEND(fragment->ptr, fragment->size, - lang, //forcing passed lang - fragment->origin.file, fragment->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); - ) - 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 - ), - fragment->origin.file, fragment->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); - ); - - return dest; -} - -size_t String::cstr_bufsize(Untaint_lang lang, - SQL_Connection *connection, - const Request_charsets *charsets) const { - size_t dest=1; // for terminating 0 - STRING_FOREACH_FRAGMENT( - uchar to_lang=lang==UL_UNSPECIFIED?fragment->lang:lang; - - switch(to_lang & ~UL_OPTIMIZE_BIT) { - case UL_CLEAN: - case UL_TAINTED: - case UL_AS_IS: - // clean piece + // first: letters + dest.body<size; - break; - case UL_FILE_SPEC: - // tainted, untaint language: file [name] - dest+=fragment->size*3/* worst: Z->%XX */; - break; - case UL_URI: - // tainted, untaint language: uri - dest+=fragment->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 */; - break; - case UL_MAIL_HEADER: - // tainted, untaint language: mail-header - if(charsets && charsets->mail) { - int parts_count=1; - const char *prev=fragment->ptr; - size_t size=fragment->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?= */); - } else - dest+=fragment->size; - break; - case UL_TABLE: - // tainted, untaint language: table - dest+=fragment->size; - break; - case UL_SQL: - // tainted, untaint language: sql - if(connection) - dest+=connection->quote(0, fragment->ptr, fragment->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; + // next: fragment infos + + if(lang==L_PASS_APPENDED) // without language-change? + dest.fragments.append(fragments); + else if(forced) //forcing passed lang? + dest.fragments+=Fragment(lang, length()); + 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==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 + for(Array_iterator i(fragments); i.has_next(); ) { + const Fragment fragment=i.next(); + dest.fragments+=Fragment( + fragment.lang==L_TAINTED?lang:fragment.lang, + fragment.length); + } + } + + ASSERT_STRING_INVARIANT(dest); return dest; } @@ -304,29 +240,68 @@ 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, - const Request_charsets *charsets) 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(; fragmentlang:lang; +/** + 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); + } +} +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; body.set_pos(pos, 0); - char *start=dest; + bool whitespace=true; - switch(to_lang & ~UL_OPTIMIZE_BIT) { - case UL_CLEAN: - case UL_TAINTED: - case UL_AS_IS: + 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(); + size_t fragment_length=fragment.length; + fragment_end=fragment_begin+fragment_length; + //fprintf(stderr, "%d, %d\n", fragment.lang, fragment.length); + + 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 @@ -334,205 +309,146 @@ 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; + 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( *src=='\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; - case UL_URI: + case L_URI: // 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; - } - { - const char *src=(const char *)client_ptr; - for(int size=client_size; size--; src++) - encode(need_uri_encode, '%'); + 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) + output=Charset::transcode(output, + charsets->source(), + charsets->client()); + + char c; + for(const char* src=output.str; c=*src++; ) + encode(need_uri_encode, '%', c); } break; - 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; - case UL_MAIL_HEADER: + case L_MAIL_HEADER: // tainted, untaint language: mail-header // http://www.ietf.org/rfc/rfc2047.txt - if(charsets && charsets->mail) { - const void *mail_ptr; + if(charsets) { size_t mail_size; - Charset::transcode(*charsets->pool, - *charsets->source, fragment->ptr, fragment->size, - *charsets->mail, mail_ptr, 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') + to_char(c); + if(c=='>') email=false; } if(to_quoted_printable) // close - dest+=sprintf(dest, "?="); + to_string("?="); - } else { - memcpy(dest, fragment->ptr, fragment->size); - dest+=fragment->size; - } + } else + ec_append(result, optimize, whitespace, pos, fragment_length); break; - case UL_TABLE: + case L_TABLE: // tainted, untaint language: table - escape(switch(*src) { + escape(switch(c) { case '\t': to_char(' '); break; case '\n': to_char(' '); break; _default; }); break; - case UL_SQL: +#ifdef PA_SQL + case L_SQL: // tainted, untaint language: sql - if(connection) - dest+=connection->quote(dest, fragment->ptr, fragment->size); - else + if(connection) { + const char *fragment_str=body.mid(fragment_begin, fragment_length).cstr(); + // skip source [we use recoded version] + pa_CORD_pos_advance(pos, fragment_length); + + to_string(connection->quote(fragment_str, fragment_length)); + } else throw Exception(0, - ConstStringPtr(this), + 0, "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; +#endif + case L_JS: + 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; + case L_XML: + 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; + case L_HTML: + 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, - ConstStringPtr(this), - "unknown untaint language #%d", - static_cast(to_lang)); // should never + SAPI::die("unknown untaint language #%d", + static_cast(to_lang)); // should never break; // never } - - if(to_lang & UL_OPTIMIZE_BIT) { - // optimizing whitespace - char *stop=dest; dest=start; - for(char *src=start; srcorigin.file) - dest+=sprintf(dest, ORIGIN_FILE_LINE_FORMAT, - fragment->origin.file, - 1+fragment->origin.line); - else - dest+=sprintf(dest, ""); -); - uchar show_lang=fragment->lang & ~UL_OPTIMIZE_BIT; - if(show_lang>=sizeof(String_Untaint_lang_name)/sizeof(String_Untaint_lang_name[0])) - throw Exception(0, - ConstStringPtr(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":""); - char *dest_after_origins=dest; - - memcpy(dest, fragment->ptr, fragment->size); - dest+=fragment->size; - - remove_crlf(dest_after_origins, dest); - to_char('\n'); - ); - - *dest=0; - return result; + return StringBody(CORD_ec_to_cord(result)); }