/**
* Create a `struct MHD_PostProcessor`.
*
* A `struct MHD_PostProcessor` can be used to (incrementally) parse
* the data portion of a POST request. Note that some buggy browsers
* fail to set the encoding type. If you want to support those, you
* may have to call #MHD_set_connection_value with the proper encoding
* type before creating a post processor (if no supported encoding
* type is set, this function will fail).
*
* @param connection the connection on which the POST is
* happening (used to determine the POST format)
* @param buffer_size maximum number of bytes to use for
* internal buffering (used only for the parsing,
* specifically the parsing of the keys). A
* tiny value (256-1024) should be sufficient.
* Do NOT use a value smaller than 256. For good
* performance, use 32 or 64k (i.e. 65536).
* @param iter iterator to be called with the parsed data,
* Must NOT be NULL.
* @param iter_cls first argument to @a iter
* @return NULL on error (out of memory, unsupported encoding),
* otherwise a PP handle
* @ingroup request
*/
struct MHD_PostProcessor *
MHD_create_post_processor (struct MHD_Connection *connection,
size_t buffer_size,
MHD_PostDataIterator iter, void *iter_cls)
{
struct MHD_PostProcessor *ret;
const char *encoding;
const char *boundary;
size_t blen;
if ((buffer_size < 256) || (connection == NULL) || (iter == NULL))
mhd_panic (mhd_panic_cls, __FILE__, __LINE__, NULL);
encoding = MHD_lookup_connection_value (connection,
MHD_HEADER_KIND,
MHD_HTTP_HEADER_CONTENT_TYPE);
if (encoding == NULL)
return NULL;
boundary = NULL;
if (!MHD_str_equal_caseless_n_ (MHD_HTTP_POST_ENCODING_FORM_URLENCODED, encoding,
strlen (MHD_HTTP_POST_ENCODING_FORM_URLENCODED)))
{
if (!MHD_str_equal_caseless_n_ (MHD_HTTP_POST_ENCODING_MULTIPART_FORMDATA, encoding,
strlen (MHD_HTTP_POST_ENCODING_MULTIPART_FORMDATA)))
return NULL;
boundary =
&encoding[strlen (MHD_HTTP_POST_ENCODING_MULTIPART_FORMDATA)];
/* Q: should this be "strcasestr"? */
boundary = strstr (boundary, "boundary=");
if (NULL == boundary)
return NULL; /* failed to determine boundary */
boundary += strlen ("boundary=");
blen = strlen (boundary);
if ((blen == 0) || (blen * 2 + 2 > buffer_size))
return NULL; /* (will be) out of memory or invalid boundary */
if ( (boundary[0] == '"') && (boundary[blen - 1] == '"') )
{
/* remove enclosing quotes */
++boundary;
blen -= 2;
}
}
else
blen = 0;
buffer_size += 4; /* round up to get nice block sizes despite boundary search */
/* add +1 to ensure we ALWAYS have a zero-termination at the end */
if (NULL == (ret = malloc (sizeof (struct MHD_PostProcessor) + buffer_size + 1)))
return NULL;
memset (ret, 0, sizeof (struct MHD_PostProcessor) + buffer_size + 1);
ret->connection = connection;
ret->ikvi = iter;
ret->cls = iter_cls;
ret->encoding = encoding;
ret->buffer_size = buffer_size;
ret->state = PP_Init;
ret->blen = blen;
ret->boundary = boundary;
ret->skip_rn = RN_Inactive;
return ret;
}
/**
* Decode multipart POST data.
*
* @param pp post processor context
*/
static int
post_process_multipart (struct MHD_PostProcessor *pp,
const char *post_data,
size_t post_data_len)
{
char *buf;
size_t max;
size_t ioff;
size_t poff;
int state_changed;
buf = (char *) &pp[1];
ioff = 0;
poff = 0;
state_changed = 1;
while ((poff < post_data_len) ||
((pp->buffer_pos > 0) && (state_changed != 0)))
{
/* first, move as much input data
as possible to our internal buffer */
max = pp->buffer_size - pp->buffer_pos;
if (max > post_data_len - poff)
max = post_data_len - poff;
memcpy (&buf[pp->buffer_pos], &post_data[poff], max);
poff += max;
pp->buffer_pos += max;
if ((max == 0) && (state_changed == 0) && (poff < post_data_len))
{
pp->state = PP_Error;
return MHD_NO; /* out of memory */
}
state_changed = 0;
/* first state machine for '\r'-'\n' and '--' handling */
switch (pp->skip_rn)
{
case RN_Inactive:
break;
case RN_OptN:
if (buf[0] == '\n')
{
ioff++;
pp->skip_rn = RN_Inactive;
goto AGAIN;
}
/* fall-through! */
case RN_Dash:
if (buf[0] == '-')
{
ioff++;
pp->skip_rn = RN_Dash2;
goto AGAIN;
}
pp->skip_rn = RN_Full;
/* fall-through! */
case RN_Full:
if (buf[0] == '\r')
{
if ((pp->buffer_pos > 1) && (buf[1] == '\n'))
{
pp->skip_rn = RN_Inactive;
ioff += 2;
}
else
{
pp->skip_rn = RN_OptN;
ioff++;
}
goto AGAIN;
}
if (buf[0] == '\n')
{
ioff++;
pp->skip_rn = RN_Inactive;
goto AGAIN;
}
pp->skip_rn = RN_Inactive;
pp->state = PP_Error;
return MHD_NO; /* no '\r\n' */
case RN_Dash2:
if (buf[0] == '-')
{
ioff++;
pp->skip_rn = RN_Full;
pp->state = pp->dash_state;
goto AGAIN;
}
pp->state = PP_Error;
break;
}
/* main state engine */
switch (pp->state)
{
case PP_Error:
return MHD_NO;
case PP_Done:
/* did not expect to receive more data */
pp->state = PP_Error;
return MHD_NO;
case PP_Init:
/**
* Per RFC2046 5.1.1 NOTE TO IMPLEMENTORS, consume anything
* prior to the first multipart boundary:
*
* > There appears to be room for additional information prior
* > to the first boundary delimiter line and following the
* > final boundary delimiter line. These areas should
* > generally be left blank, and implementations must ignore
* > anything that appears before the first boundary delimiter
* > line or after the last one.
*/
(void) find_boundary (pp,
pp->boundary,
pp->blen,
&ioff,
PP_ProcessEntryHeaders, PP_Done);
break;
case PP_NextBoundary:
if (MHD_NO == find_boundary (pp,
pp->boundary,
pp->blen,
&ioff,
PP_ProcessEntryHeaders, PP_Done))
{
if (pp->state == PP_Error)
return MHD_NO;
goto END;
}
break;
case PP_ProcessEntryHeaders:
pp->must_ikvi = MHD_YES;
if (MHD_NO ==
process_multipart_headers (pp, &ioff, PP_PerformCheckMultipart))
{
if (pp->state == PP_Error)
return MHD_NO;
else
goto END;
}
state_changed = 1;
break;
case PP_PerformCheckMultipart:
if ((pp->content_type != NULL) &&
(0 == strncasecmp (pp->content_type,
"multipart/mixed",
strlen ("multipart/mixed"))))
{
pp->nested_boundary = strstr (pp->content_type, "boundary=");
if (pp->nested_boundary == NULL)
{
pp->state = PP_Error;
return MHD_NO;
}
pp->nested_boundary =
strdup (&pp->nested_boundary[strlen ("boundary=")]);
if (pp->nested_boundary == NULL)
{
/* out of memory */
pp->state = PP_Error;
return MHD_NO;
}
/* free old content type, we will need that field
for the content type of the nested elements */
free (pp->content_type);
pp->content_type = NULL;
pp->nlen = strlen (pp->nested_boundary);
pp->state = PP_Nested_Init;
state_changed = 1;
break;
}
pp->state = PP_ProcessValueToBoundary;
pp->value_offset = 0;
state_changed = 1;
break;
case PP_ProcessValueToBoundary:
if (MHD_NO == process_value_to_boundary (pp,
&ioff,
pp->boundary,
pp->blen,
PP_PerformCleanup,
PP_Done))
{
if (pp->state == PP_Error)
return MHD_NO;
break;
}
break;
case PP_PerformCleanup:
/* clean up state of one multipart form-data element! */
pp->have = NE_none;
free_unmarked (pp);
if (pp->nested_boundary != NULL)
{
free (pp->nested_boundary);
pp->nested_boundary = NULL;
}
pp->state = PP_ProcessEntryHeaders;
state_changed = 1;
break;
case PP_Nested_Init:
if (pp->nested_boundary == NULL)
{
pp->state = PP_Error;
return MHD_NO;
}
if (MHD_NO == find_boundary (pp,
pp->nested_boundary,
pp->nlen,
&ioff,
PP_Nested_PerformMarking,
PP_NextBoundary /* or PP_Error? */ ))
{
if (pp->state == PP_Error)
return MHD_NO;
goto END;
}
break;
case PP_Nested_PerformMarking:
/* remember what headers were given
globally */
pp->have = NE_none;
if (pp->content_name != NULL)
pp->have |= NE_content_name;
if (pp->content_type != NULL)
pp->have |= NE_content_type;
if (pp->content_filename != NULL)
pp->have |= NE_content_filename;
if (pp->content_transfer_encoding != NULL)
pp->have |= NE_content_transfer_encoding;
pp->state = PP_Nested_ProcessEntryHeaders;
state_changed = 1;
break;
case PP_Nested_ProcessEntryHeaders:
pp->value_offset = 0;
if (MHD_NO ==
process_multipart_headers (pp, &ioff,
PP_Nested_ProcessValueToBoundary))
{
if (pp->state == PP_Error)
return MHD_NO;
else
goto END;
}
state_changed = 1;
break;
case PP_Nested_ProcessValueToBoundary:
if (MHD_NO == process_value_to_boundary (pp,
&ioff,
pp->nested_boundary,
pp->nlen,
PP_Nested_PerformCleanup,
PP_NextBoundary))
{
if (pp->state == PP_Error)
return MHD_NO;
break;
}
break;
case PP_Nested_PerformCleanup:
free_unmarked (pp);
pp->state = PP_Nested_ProcessEntryHeaders;
state_changed = 1;
break;
default:
mhd_panic (mhd_panic_cls, __FILE__, __LINE__, NULL); /* should never happen! */
}
AGAIN:
if (ioff > 0)
{
memmove (buf, &buf[ioff], pp->buffer_pos - ioff);
pp->buffer_pos -= ioff;
ioff = 0;
state_changed = 1;
}
}
END:
if (ioff != 0)
{
memmove (buf, &buf[ioff], pp->buffer_pos - ioff);
pp->buffer_pos -= ioff;
}
if (poff < post_data_len)
{
pp->state = PP_Error;
return MHD_NO; /* serious error */
}
return MHD_YES;
}
/**
* Process url-encoded POST data.
*
* @param pp post processor context
* @param post_data upload data
* @param post_data_len number of bytes in upload_data
* @return MHD_YES on success, MHD_NO if there was an error processing the data
*/
static int
post_process_urlencoded (struct MHD_PostProcessor *pp,
const char *post_data,
size_t post_data_len)
{
size_t equals;
size_t amper;
size_t poff;
size_t xoff;
size_t delta;
int end_of_value_found;
char *buf;
char xbuf[XBUF_SIZE + 1];
buf = (char *) &pp[1];
poff = 0;
while (poff < post_data_len)
{
switch (pp->state)
{
case PP_Error:
return MHD_NO;
case PP_Done:
/* did not expect to receive more data */
pp->state = PP_Error;
return MHD_NO;
case PP_Init:
equals = 0;
while ((equals + poff < post_data_len) &&
(post_data[equals + poff] != '='))
equals++;
if (equals + pp->buffer_pos > pp->buffer_size)
{
pp->state = PP_Error; /* out of memory */
return MHD_NO;
}
memcpy (&buf[pp->buffer_pos], &post_data[poff], equals);
pp->buffer_pos += equals;
if (equals + poff == post_data_len)
return MHD_YES; /* no '=' yet */
buf[pp->buffer_pos] = '\0'; /* 0-terminate key */
pp->buffer_pos = 0; /* reset for next key */
MHD_http_unescape (NULL, NULL, buf);
poff += equals + 1;
pp->state = PP_ProcessValue;
pp->value_offset = 0;
break;
case PP_ProcessValue:
/* obtain rest of value from previous iteration */
memcpy (xbuf, pp->xbuf, pp->xbuf_pos);
xoff = pp->xbuf_pos;
pp->xbuf_pos = 0;
/* find last position in input buffer that is part of the value */
amper = 0;
while ((amper + poff < post_data_len) &&
(amper < XBUF_SIZE) &&
(post_data[amper + poff] != '&') &&
(post_data[amper + poff] != '\n') &&
(post_data[amper + poff] != '\r'))
amper++;
end_of_value_found = ((amper + poff < post_data_len) &&
((post_data[amper + poff] == '&') ||
(post_data[amper + poff] == '\n') ||
(post_data[amper + poff] == '\r')));
/* compute delta, the maximum number of bytes that we will be able to
process right now (either amper-limited of xbuf-size limited) */
delta = amper;
if (delta > XBUF_SIZE - xoff)
delta = XBUF_SIZE - xoff;
/* move input into processing buffer */
memcpy (&xbuf[xoff], &post_data[poff], delta);
xoff += delta;
poff += delta;
/* find if escape sequence is at the end of the processing buffer;
if so, exclude those from processing (reduce delta to point at
end of processed region) */
delta = xoff;
if ((delta > 0) && (xbuf[delta - 1] == '%'))
delta--;
else if ((delta > 1) && (xbuf[delta - 2] == '%'))
delta -= 2;
/* if we have an incomplete escape sequence, save it to
pp->xbuf for later */
if (delta < xoff)
{
memcpy (pp->xbuf, &xbuf[delta], xoff - delta);
pp->xbuf_pos = xoff - delta;
xoff = delta;
}
/* If we have nothing to do (delta == 0) and
not just because the value is empty (are
waiting for more data), go for next iteration */
if ((xoff == 0) && (poff == post_data_len))
continue;
/* unescape */
xbuf[xoff] = '\0'; /* 0-terminate in preparation */
xoff = MHD_http_unescape (NULL, NULL, xbuf);
/* finally: call application! */
pp->must_ikvi = MHD_NO;
if (MHD_NO == pp->ikvi (pp->cls, MHD_POSTDATA_KIND, (const char *) &pp[1], /* key */
NULL, NULL, NULL, xbuf, pp->value_offset,
xoff))
{
pp->state = PP_Error;
return MHD_NO;
}
pp->value_offset += xoff;
/* are we done with the value? */
if (end_of_value_found)
{
/* we found the end of the value! */
if ((post_data[poff] == '\n') || (post_data[poff] == '\r'))
{
pp->state = PP_ExpectNewLine;
}
else if (post_data[poff] == '&')
{
poff++; /* skip '&' */
pp->state = PP_Init;
}
}
break;
case PP_ExpectNewLine:
if ((post_data[poff] == '\n') || (post_data[poff] == '\r'))
{
poff++;
/* we are done, report error if we receive any more... */
pp->state = PP_Done;
return MHD_YES;
}
return MHD_NO;
default:
mhd_panic (mhd_panic_cls, __FILE__, __LINE__, NULL); /* should never happen! */
}
}
return MHD_YES;
}
