Name

rfc2045 — RFC 2045 (MIME) parsing library

Synopsis

#include <rfc822.h>
#include <rfc2045.h>

cc ... -lrfc2045 -lrfc822

DESCRIPTION

The rfc2045 library parses MIME-formatted messages. The rfc2045 library is used to:

1) Parse the structure of a MIME formatted message

2) Examine the contents of each MIME section

3) Optionally rewrite and reformat the message.

Creating an rfc2045 structure

#include <rfc2045.h>

struct rfc2045 *ptr=rfc2045_alloc();
void rfc2045_parse(struct rfc2045 *ptr, const char *txt, size_t cnt);

struct rfc2045 *ptr=rfc2045_fromfd(int fd);
struct rfc2045 *ptr=rfc2045_fromfp(FILE *fp);

void rfc2045_free(struct rfc2045 *ptr);

void rfc2045_error(const char *errmsg)
{
        perror(errmsg);
        exit(0);
}

The rfc2045 structure is created from an existing message. The function rfc2045_alloc() allocates the structure, then rfc2045_parse() is called to initialize the structure based on the contents of a message. txt points to the contents of the message, and cnt contains the number of bytes in the message.

Large messages are parsed by calling rfc2045_parse() multiple number of times, each time passing a portion of the overall message. There is no need to call a separate function after the entire message is parsed -- the rfc2045 structure is created dynamically, on the fly.

rfc2045_alloc() returns NULL if there was insufficient memory to allocate the structure. The rfc2045_parse() also allocates memory, internally, however no error indication is return in the event of a memory allocation failure. Instead, the function rfc2045_error() is called, with errmsg set to "Out of memory". rfc2045_error() is also called by rfc2045_alloc() - it also calls rfc2045_error(), before returning a NULL pointer.

The rfc2045_error() function is not included in the rfc2045 library, it must be defined by the application to report the error in some appropriate way. All functions below will use rfc2045_error() to report an error condition (currently only insufficient memory is reported), in addition to returning any kind of an error indicator. Some functions do not return an error indicator, so rfc2045_error() is the only reliable way to detect a failure.

The rfc2045_fromfd() function initializes an rfc2045 structure from a file descriptor. It is equivalent to calling rfc2045_alloc(), then reading the contents of the given file descriptor, and calling rfc2045_parse(). The rfc2045_fromfp() function initializes an rfc2045 structure from a FILE.

After the rfc2045 structure is initialized, the functions described below may be used to access and work with the contents of the structure. When the rfc2045 structure is no longer needed, the function rfc2045_free() deallocates and destroys the structure.

Structure of a MIME message


struct rfc2045 {

        struct rfc2045 *parent;

        struct rfc2045 *firstpart;
        struct rfc2045 *next;
        int             isdummy;
        int             rfcviolation;
} ;

The rfc2045 structure has many fields, only some are publicly documented. A MIME message is represented by a recursive tree of linked rfc2045 structures. Each instance of the rfc2045 structure represents a single MIME section of a MIME-formatted message.

The top-level structure that represents the entire message is created by the rfc2045_alloc() function. The remaining structures are created dynamically by rfc2045_parse(). Any rfc2045 structure, except ones whose isdummy flag is set, may be used as an argument to any function described in the following chapters.

The rfcviolation field in the top-level rfc2045 indicates any errors found while parsing the MIME message. rfcviolation is a bitmask of the following flags:

RFC2045_ERR8BITHEADER

Illegal 8-bit characters in MIME headers.

RFC2045_ERR8BITCONTENT

Illegal 8-bit contents of a MIME section that declared a 7bit transfer encoding.

RFC2045_ERR2COMPLEX

The message has too many MIME sections, this is a potential denial-of-service attack.

RFC2045_ERRBADBOUNDARY

Ambiguous nested multipart MIME boundary strings. (Nested MIME boundary strings where one string is a prefix of another string).

In each rfc2045 structure that represents a multipart MIME section (or one that contains message/rfc822 content) the firstpart pointer points to the first MIME section in the multipart MIME section (or the included "message/rfc822" MIME section). If there are more than one MIME sections in a multipart MIME section firstpart->next gets you the second MIME section, firstpart->next->next gets you the third MIME section, and so on. parent points to the parent MIME section, which is NULL for the top-level MIME section.

Not all MIME sections are created equal. In a multipart MIME section, there is an initial, unused, "filler" section before the first MIME delimiter (see RFC 2045 for more information). This filler section typically contains a terse message saying that this is a MIME-formatted message. This is not considered to be a "real" MIME section, and all MIME-aware software must ignore those. These filler sections are designated by setting the isdummy field to a non-zero value. All rfc2045 structures that have isdummy set should be ignored, and skipped over, when traversing the rfc2045 tree.

Basic MIME information


const char *content_type, *content_transfer_encoding,
           *content_character_set;

void rfc2045_mimeinfo(const struct rfc2045 *ptr,
        &content_type, &content_transfer_encoding,
        &content_character_set);

off_t start_pos, end_pos, start_body, nlines, nbodylines;

void rfc2045_mimepos(const struct rfc2045 *ptr,
        &start_pos, &end_pos, &start_body, &nlines,
        &nbodylines);

The rfc2045_mimeinfo() function returns the MIME content type, encoding method, and the character set of the given MIME section. Where the MIME section does not specify any property, rfc2045_mimeinfo() automatically supplies a default value. The character set is only meaningful for MIME sections with a text content type, however it is still defaulted for other sections. It is not permissible to supply a NULL pointer for any argument to rfc2045_mimeinfo().

The rfc2045_mimepos() function locates the position of the given MIME section in the original message. It is not permissible to supply a NULL pointer for any argument to rfc2045_mimepos(). All arguments must be used.

start_pos and end_pos point to the starting and the ending offset, from the beginning of the message, of this MIME section. nlines is initialized to the number of lines of text in this MIME section. start_pos is the start of MIME headers for this MIME section. start_body is the start of the actual content of this MIME section (after all the MIME headers, and the delimiting blank line), and nbodylines is the number of lines of actual content in this MIME section.


const char *id=rfc2045_content_id(
                       const struct rfc2045 *ptr);

const char *desc=rfc2045_content_description(
                       const struct rfc2045 *ptr);

const char *lang=rfc2045_content_language(
                       const struct rfc2045 *ptr);

const char *md5=rfc2045_content_md5(
                       const struct rfc2045 *ptr);

These functions return the contents of the corresponding MIME headers. If these headers do not exist, these functions return an empty string, "", NOT a null pointer.


char *id=rfc2045_related_start(const struct rfc2045 *ptr);

This function returns the start attribute of the Content-Type: header, which is used by multipart/related MIME content. This function returns a dynamically-allocated buffer, which must be free(3)-ed after use (a null pointer is returned if there was insufficient memory for the buffer, and rfc2045_error() is called).


const struct rfc2045 *ptr;

const char *disposition=ptr->content_disposition;

char *charset;
char *language;
char *value;

int error;

error=rfc2231_decodeType(rfc, "name", &charset,
                         &language, &value);
error=rfc2231_decodeDisposition(rfc, "name", &charset,
                                &language, &value);

These functions and structures provide a mechanism for reading the MIME attributes in the Content-Type: and Content-Disposition: headers. The MIME content type is returned by rfc2045_mimeinfo(). The MIME content disposition can be accessed in the content_disposition directly (which may be NULL if the Content-Disposition: header was not specified).

rfc2231_decodeType() reads MIME attributes from the Content-Type: header, and rfc2231_decodeType() reads MIME attributes from the Content-Disposition: header. These functions understand MIME attributes that are encoded according to RFC 2231.

These functions initialize charset, language, and value parameters, allocating memory automatically. It is the caller's responsibility to use free() to return the allocated memory. A NULL may be provided in place of a parameter, indicating that the caller does not require the corresponding information.

charset and language will be set to an empty string (not NULL) if the MIME parameter does not exist, or is not encoded according to RFC 2231, or does not specify its character set and/or language. value will be set to an empty string if the MIME parameter does not exist.


char *url=rfc2045_content_base(struct rfc2045 *ptr);

char *url=rfc2045_append_url(const char *base, const char *url);

These functions are used to work with multipart/related MIME content. rfc2045_content_base() returns the contents of either the Content-Base: or the Content-Location: header. If both are present, they are logically combined. rfc2045_append_url() combines two URLs, base and url, and returns the absolute URL that results from the combination.

Both functions return a pointer to a dynamically-allocated buffer that must be free(3)-ed after it is no longer needed. Both functions return NULL if there was no sufficient memory to allocate the buffer. rfc2045_content_base() returns an empty string in the event that there are no Content-Base: or Content-Location: headers. Either argument to rfc2045_append_url() may be a NULL, or an empty string.

Decoding a MIME section


void rfc2045_cdecode_start(struct rfc2045 *ptr,
        int (*callback_func)(const char *, size_t, void *),
        void *callback_arg);

int rfc2045_cdecode(struct rfc2045 *ptr, const char *stuff,
        size_t nstuff);

int rfc2045_cdecode_end(struct rfc2045 *ptr);

These functions are used to return the raw contents of the given MIME section, transparently decoding quoted-printable or base64-encoded content. Because the rfc2045 library does not require the message to be read from a file (it can be stored in a memory buffer), the application is responsible for reading the contents of the message and calling rfc2045_cdecode().

The rfc2045_cdecode_start() function begins the process of decoding the given MIME section. After calling rfc2045_cdecode_start(), the application must the repeatedly call rfc2045_cdecode() with the contents of the MIME message between the offsets given by the start_body and end_pos return values from rfc2045_mimepos(). The rfc2045_cdecode() function can be called repeatedly, if necessary, for successive portions of the MIME section. After the last call to rfc2045_cdecode(), call rfc2045_cdecode_end() to finish up (rfc2045_cdecode() may have saved some undecoded content in an internal part, and rfc2045_cdecode_end() flushes it out).

rfc2045_cdecode() and rfc2045_cdecode_end() repeatedly call callback_func(), passing it the decoded contents of the MIME section. The first argument to callback_func() is a pointer to a portion of the decoded content, the second argument is the number of bytes in this portion. The third argument is callback_arg.

callback_func() is required to return zero, to continue decoding. If callback_func() returns non-zero, the decoding immediately stops and rfc2045_cdecode() or rfc2045_cdecode_end() terminates with callback_func's return code.

Rewriting MIME messages

This library contains functions that can be used to rewrite a MIME message in order to convert 8-bit-encoded data to 7-bit encoding, or to convert 7-bit encoded data to full 8-bit data, if possible.


struct rfc2045 *ptr=rfc2045_alloc_ac();
int necessary=rfc2045_ac_check(struct rfc2045 *ptr, int mode);

int error=rfc2045_rewrite(struct rfc2045 *ptr,
                int fdin,
                int fdout,
                const char *appname);

int rfc2045_rewrite_func(struct rfc2045 *p, int fdin,
        int (*funcout)(const char *, int, void *), void *funcout_arg,
        const char *appname);

When rewriting will be used, the rfc2045_alloc_ac() function must be used to create the initial rfc2045 structure. This function allocates some additional structures that are used in rewriting. Use rfc2045_parse() to parse the message, as usual. Use rfc2045_free() in a normal way to destroy the rfc2045 structure, when all is said and done.

The rfc2045_ac_check() function must be called to determine whether rewriting is necessary. mode must be set to one of the following values:

RFC2045_RW_7BIT

We want to generate 7-bit content. If the original message contains any 8-bit content it will be converted to 7-bit content using quoted-printable encoding.

RFC2045_RW_8BIT

We want to generate 8-bit content. If the original message contains any 7-bit quoted-printable content it should be rewritten as 8-bit content.

The rfc2045_ac_check() function returns non-zero if there's any content in the MIME message that should be converted, OR if there are any missing MIME headers. rfc2045_ac_check() returns zero if there's no need to rewrite the message. However it might still be worthwhile to rewrite the message anyway. There are some instances where it is desirable to provide defaults for some missing MIME headers, but they are too trivial to require the message to be rewritten. One such case would be a missing Content-Transfer-Encoding: header for a multipart section.

Either the rfc2045_rewrite() or the rfc2045_rewrite_func() function is used to rewrite the message. The only difference is that rfc2045_rewrite() writes the new message to a given file descriptor, fdout, while rfc2045_rewrite_func() repeatedly calls the funcout function. Both function read the original message from fdin. funcout receives to a portion of the MIME message, the number of bytes in the specified portion, and funcout_arg. When either function rewrites a MIME section, an informational header gets appended, noting that the message was converted by appname.