From 1f15fa4796c5fa09f72c0c2c534dac4fcfe2a168 Mon Sep 17 00:00:00 2001 From: Tatsuhiro Tsujikawa Date: Wed, 31 Oct 2012 00:42:07 +0900 Subject: [PATCH] Parse Content-Disposition header field value conforming to RFC 6266 This change enables curl to parse filename*= parameter in Content-Disposition header, which is described in RFC 6266 and RFC 5987. It also handles quoted characters in filename= parameter properly. The filename*= parameter can contain percent-encoded string and its charset. This implementation does not use charset and does not unescape percent-encoded string. --- src/tool_cb_hdr.c | 366 ++++++++++++++++++++++++++++++++++++++++++++++----- tests/data/test1311 | 2 +- tests/data/test1313 | 2 +- tests/data/test1340 | 4 +- tests/data/test1341 | 4 +- 5 files changed, 339 insertions(+), 39 deletions(-) diff --git a/src/tool_cb_hdr.c b/src/tool_cb_hdr.c index ef340f7..8b619c2 100644 --- a/src/tool_cb_hdr.c +++ b/src/tool_cb_hdr.c @@ -35,6 +35,9 @@ static char *parse_filename(const char *ptr, size_t len); +static ssize_t parse_content_disposition(char *dest, size_t destlen, + const char *in, size_t len); + /* ** callback for CURLOPT_HEADERFUNCTION */ @@ -46,7 +49,6 @@ size_t tool_header_cb(void *ptr, size_t size, size_t nmemb, void *userdata) struct OutStruct *heads = hdrcbdata->heads; const char *str = ptr; const size_t cb = size * nmemb; - const char *end = (char*)ptr + cb; /* * Once that libcurl has called back tool_header_cb() the returned value @@ -86,31 +88,12 @@ size_t tool_header_cb(void *ptr, size_t size, size_t nmemb, void *userdata) if(hdrcbdata->honor_cd_filename && (cb > 20) && checkprefix("Content-disposition:", str)) { const char *p = str + 20; - - /* look for the 'filename=' parameter - (encoded filenames (*=) are not supported) */ - for(;;) { + ssize_t rv; + char rawval[1024]; + rv = parse_content_disposition(rawval, sizeof(rawval), p, cb - 20); + if(rv > 0) { char *filename; - size_t len; - - while(*p && (p < end) && !ISALPHA(*p)) - p++; - if(p > end - 9) - break; - - if(memcmp(p, "filename=", 9)) { - /* no match, find next parameter */ - while((p < end) && (*p != ';')) - p++; - continue; - } - p += 9; - - /* this expression below typecasts 'cb' only to avoid - warning: signed and unsigned type in conditional expression - */ - len = (ssize_t)cb - (p - str); - filename = parse_filename(p, len); + filename = parse_filename(rawval, rv); if(filename) { outs->filename = filename; outs->alloc_filename = TRUE; @@ -119,7 +102,6 @@ size_t tool_header_cb(void *ptr, size_t size, size_t nmemb, void *userdata) outs->fopened = FALSE; outs->stream = NULL; hdrcbdata->honor_cd_filename = FALSE; - break; } else return failure; @@ -129,6 +111,331 @@ size_t tool_header_cb(void *ptr, size_t size, size_t nmemb, void *userdata) return cb; } +static int is_lws(const char c) +{ + /* In strict definition, \r and \n are not included in linear white + space, but they are included for convenience because the header + field value ends with \r\n. */ + return c == ' ' || c == '\t' || c == '\r' || c == '\n'; +} + +static int in_iso8859_1(unsigned char c) +{ + return (0x20u <= c && c <= 0x7eu) || 0xa0u <= c; +} + +static int is_alpha(const char c) +{ + return ('A' <= c && c <= 'Z') || ('a' <= c && c <= 'z'); +} + +static int is_hex(const char c) +{ + return ('A' <= c && c <= 'F') || ('a' <= c && c <= 'f') || ISDIGIT(c); +} + +static int in_rfc2616_http_token(const char c) +{ + return is_alpha(c) || ISDIGIT(c) || + c == '!' || c == '#' || c == '$' || c == '%' || + c == '&' || c == '\'' || c == '*' || c == '+' || + c == '-' || c == '.' || c == '^' || c == '_' || + c == '`' || c == '|' || c == '~'; +} + +static int in_rfc2978_mime_charset(const char c) +{ + return is_alpha(c) || ISDIGIT(c) || + c == '!' || c == '#' || c == '$' || c == '%' || + c == '&' || c == '\'' || c == '+' || + c == '-' || c == '^' || c == '_' || + c == '`' || c == '{' || c == '}' || c == '~'; +} + +static int in_rfc5987_attr_char(const char c) +{ + return in_rfc2616_http_token(c) && c != '*' && c != '\'' && c != '%'; +} + +/* Returns nonzero if [first, last) == s in case-insensitive + manner. */ +static int strieq(const char *first, const char *last, const char *s) +{ + for(; first != last && *s && + Curl_raw_toupper(*first) == Curl_raw_toupper(*s); ++first, ++s); + return first == last && !*s; +} + +typedef enum { + CD_BEFORE_DISPOSITION_TYPE, + CD_AFTER_DISPOSITION_TYPE, + CD_DISPOSITION_TYPE, + CD_BEFORE_DISPOSITION_PARM_NAME, + CD_AFTER_DISPOSITION_PARM_NAME, + CD_DISPOSITION_PARM_NAME, + CD_BEFORE_VALUE, + CD_AFTER_VALUE, + CD_QUOTED_STRING, + CD_TOKEN, + CD_BEFORE_EXT_VALUE, + CD_CHARSET, + CD_LANGUAGE, + CD_VALUE_CHARS, + CD_VALUE_CHARS_PCT_ENCODED1, + CD_VALUE_CHARS_PCT_ENCODED2 +} content_disposition_parse_state; + +typedef enum { + CD_FILENAME_FOUND = 1, + CD_EXT_FILENAME_FOUND = 1 << 1 +} content_disposition_parse_flag; + +static int fnstore(int in_file_parm, char **dpp, size_t *dlenp, char c) +{ + if(in_file_parm) { + if(*dlenp == 0) + return -1; + else { + *(*dpp)++ = c; + --*dlenp; + } + } + return 0; +} + +/* +** Parses Content-Disposition header field value and stores the +** filename in dest, whose maximum capacity is destlen. This parser +** largely conforms to RFC 6266, except that: +** +** * it does not check unescaped pct-encoded byte sequence is valid +** according to the specified charset when RFC 5987 Encoded is used. +** +** * it does not unescape pct-encoded string when RFC 5987 Encoding is +** used. +** +** This function returns the length of extracted filename in dest if +** it succeeds, or -1. If dest is not big enough to store filename to +** be extracted, this function fails. +** +*/ +static ssize_t parse_content_disposition(char *dest, size_t destlen, + const char *in, size_t len) +{ + const char *p = in, *eop = in + len, *mark_first = NULL, *mark_last = NULL; + int state = CD_BEFORE_DISPOSITION_TYPE; + /* becomes nonzero if p is in filename parameter. */ + int in_file_parm = 0; + /* bitwise-OR of content_disposition_parse_flag to indicate which + parameter has already been parsed. */ + int flags = 0; + /* becomes nonzero if current character is escaped by '\' in + quoted-string. */ + int quoted_seen = 0; + /* To suppress warnings */ + char *dp = dest; + size_t dlen = destlen; + + for(; p != eop; ++p) { + switch(state) { + case CD_BEFORE_DISPOSITION_TYPE: + if(in_rfc2616_http_token(*p)) + state = CD_DISPOSITION_TYPE; + else if(!is_lws(*p)) + return -1; + break; + case CD_AFTER_DISPOSITION_TYPE: + case CD_DISPOSITION_TYPE: + if(*p == ';') + state = CD_BEFORE_DISPOSITION_PARM_NAME; + else if(is_lws(*p)) + state = CD_AFTER_DISPOSITION_TYPE; + else if(state == CD_AFTER_DISPOSITION_TYPE || + !in_rfc2616_http_token(*p)) + return -1; + break; + case CD_BEFORE_DISPOSITION_PARM_NAME: + if(in_rfc2616_http_token(*p)) { + mark_first = p; + state = CD_DISPOSITION_PARM_NAME; + } + else if(!is_lws(*p)) + return -1; + break; + case CD_AFTER_DISPOSITION_PARM_NAME: + case CD_DISPOSITION_PARM_NAME: + if(*p == '=') { + if(state == CD_DISPOSITION_PARM_NAME) + mark_last = p; + in_file_parm = 0; + if(strieq(mark_first, mark_last, "filename*")) { + if((flags & CD_EXT_FILENAME_FOUND) == 0) + in_file_parm = 1; + else + return -1; + state = CD_BEFORE_EXT_VALUE; + } + else if(strieq(mark_first, mark_last, "filename")) { + if(flags & CD_FILENAME_FOUND) + return -1; + if((flags & CD_EXT_FILENAME_FOUND) == 0) + in_file_parm = 1; + state = CD_BEFORE_VALUE; + } + else { + /* ext-token must be characters in token, followed by "*" */ + if(mark_first != mark_last - 1 && *(mark_last - 1) == '*') + state = CD_BEFORE_EXT_VALUE; + else + state = CD_BEFORE_VALUE; + } + if(in_file_parm) { + dp = dest; + dlen = destlen; + } + } + else if(is_lws(*p)) { + mark_last = p; + state = CD_AFTER_DISPOSITION_PARM_NAME; + } + else if(state == CD_AFTER_DISPOSITION_PARM_NAME || + !in_rfc2616_http_token(*p)) { + return -1; + } + break; + case CD_BEFORE_VALUE: + if(*p == '"') { + quoted_seen = 0; + state = CD_QUOTED_STRING; + } + else if(in_rfc2616_http_token(*p)) { + if(fnstore(in_file_parm, &dp, &dlen, *p) == -1) + return -1; + state = CD_TOKEN; + } + else if(!is_lws(*p)) + return -1; + break; + case CD_AFTER_VALUE: + if(*p == ';') + state = CD_BEFORE_DISPOSITION_PARM_NAME; + else if(!is_lws(*p)) + return -1; + break; + case CD_QUOTED_STRING: + if(*p == '\\' && quoted_seen == 0) + quoted_seen = 1; + else if(*p == '"' && quoted_seen == 0) { + if(in_file_parm) + flags |= CD_FILENAME_FOUND; + state = CD_AFTER_VALUE; + } + else { + /* TEXT which is OCTET except CTLs, but including LWS. We only + accept ISO-8859-1 chars. */ + quoted_seen = 0; + if(!in_iso8859_1(*p)) + return -1; + if(fnstore(in_file_parm, &dp, &dlen, *p) == -1) + return -1; + } + break; + case CD_TOKEN: + if(in_rfc2616_http_token(*p)) { + if(fnstore(in_file_parm, &dp, &dlen, *p) == -1) + return -1; + } + else if(*p == ';') { + if(in_file_parm) + flags |= CD_FILENAME_FOUND; + state = CD_BEFORE_DISPOSITION_PARM_NAME; + } + else if(is_lws(*p)) { + if(in_file_parm) + flags |= CD_FILENAME_FOUND; + state = CD_AFTER_VALUE; + } + else + return -1; + break; + case CD_BEFORE_EXT_VALUE: + if(*p == '\'') + /* Empty charset is not allowed */ + return -1; + else if(in_rfc2978_mime_charset(*p)) + state = CD_CHARSET; + else if(!is_lws(*p)) + return -1; + break; + case CD_CHARSET: + if(*p == '\'') + /* We ignore charset for now. */ + state = CD_LANGUAGE; + else if(!in_rfc2978_mime_charset(*p)) + return -1; + break; + case CD_LANGUAGE: + if(*p == '\'') { + if(in_file_parm) { + dp = dest; + dlen = destlen; + } + state = CD_VALUE_CHARS; + } + else if(*p != '-' && !is_alpha(*p) && !ISDIGIT(*p)) + return -1; + break; + case CD_VALUE_CHARS: + if(in_rfc5987_attr_char(*p)) { + if(fnstore(in_file_parm, &dp, &dlen, *p) == -1) + return -1; + } + else if(*p == '%') { + if(fnstore(in_file_parm, &dp, &dlen, *p) == -1) + return -1; + state = CD_VALUE_CHARS_PCT_ENCODED1; + } + else if(*p == ';' || is_lws(*p)) { + if(in_file_parm) + flags |= CD_EXT_FILENAME_FOUND; + if(*p == ';') + state = CD_BEFORE_DISPOSITION_PARM_NAME; + else + state = CD_AFTER_VALUE; + } + else if(!in_rfc5987_attr_char(*p)) + return -1; + break; + case CD_VALUE_CHARS_PCT_ENCODED1: + case CD_VALUE_CHARS_PCT_ENCODED2: + /* We just check that "%" is followed by 2 hex bytes so that + they forms pct-encoded string properly. */ + if(is_hex(*p)) { + if(fnstore(in_file_parm, &dp, &dlen, *p) == -1) + return -1; + if(state == CD_VALUE_CHARS_PCT_ENCODED1) + state = CD_VALUE_CHARS_PCT_ENCODED2; + else + state = CD_VALUE_CHARS; + } + else + return -1; + break; + } + } + switch(state) { + case CD_BEFORE_DISPOSITION_TYPE: + case CD_AFTER_DISPOSITION_TYPE: + case CD_DISPOSITION_TYPE: + case CD_AFTER_VALUE: + case CD_TOKEN: + case CD_VALUE_CHARS: + return destlen - dlen; + default: + return -1; + } +} + /* * Copies a file name part and returns an ALLOCATED data buffer. */ @@ -147,13 +454,6 @@ static char *parse_filename(const char *ptr, size_t len) copy[len] = '\0'; p = copy; - if(*p == '\'' || *p == '"') { - /* store the starting quote */ - stop = *p; - p++; - } - else - stop = ';'; /* if the filename contains a path, only use filename portion */ q = strrchr(copy, '/'); diff --git a/tests/data/test1311 b/tests/data/test1311 index e47647c..9bd7e99 100644 --- a/tests/data/test1311 +++ b/tests/data/test1311 @@ -16,7 +16,7 @@ Server: test-server/fake Content-Length: 6 Connection: close Content-Type: text/html -Content-Disposition: filename=name1311; charset=funny; option=strange +Content-Disposition: inline; filename=name1311; charset=funny; option=strange 12345 diff --git a/tests/data/test1313 b/tests/data/test1313 index 2331ae9..55dcc3f 100644 --- a/tests/data/test1313 +++ b/tests/data/test1313 @@ -56,7 +56,7 @@ Host: %HOSTIP:%HTTPPORT Accept: */* - + 12345 diff --git a/tests/data/test1340 b/tests/data/test1340 index 5153889..1a4a3de 100644 --- a/tests/data/test1340 +++ b/tests/data/test1340 @@ -16,7 +16,7 @@ Server: test-server/fake Content-Length: 6 Connection: close Content-Type: text/html -Content-Disposition: filename=name1340; charset=funny; option=strange +Content-Disposition: inline; filename=name1340; charset=funny; option=strange 12345 @@ -68,7 +68,7 @@ Server: test-server/fake Content-Length: 6 Connection: close Content-Type: text/html -Content-Disposition: filename=name1340; charset=funny; option=strange +Content-Disposition: inline; filename=name1340; charset=funny; option=strange diff --git a/tests/data/test1341 b/tests/data/test1341 index e12ee31..94b8de4 100644 --- a/tests/data/test1341 +++ b/tests/data/test1341 @@ -16,7 +16,7 @@ Server: test-server/fake Content-Length: 6 Connection: close Content-Type: text/html -Content-Disposition: filename=name1341; charset=funny; option=strange +Content-Disposition: inline; filename=name1341; charset=funny; option=strange 12345 @@ -68,7 +68,7 @@ Server: test-server/fake Content-Length: 6 Connection: close Content-Type: text/html -Content-Disposition: filename=name1341; charset=funny; option=strange +Content-Disposition: inline; filename=name1341; charset=funny; option=strange curl: Saved to filename '%PWD/log/name1341' -- 1.7.10.4