[Fix] More fixes to end of headers detection

src/client/rspamc.c

@@ -1045,7 +1045,7 @@ rspamc_mime_output (FILE *out, ucl_object_t *result, GString *input,
 	gboolean is_spam = FALSE;
 	gchar *json_header, *json_header_encoded, *sc;
 
-	headers_pos = rspamd_string_find_eoh (input);
+	headers_pos = rspamd_string_find_eoh (input, NULL);
 
 	if (headers_pos == -1) {
 		rspamd_fprintf (stderr,"cannot find end of headers position");

src/libmime/message.c

@@ -218,8 +218,14 @@ process_raw_headers (struct rspamd_task *task, GHashTable *target,
 				tmp ++;
 			}
 
+			if (p + 1 == end) {
+				new->raw_len = end - new->raw_value;
+			}
+			else {
+				new->raw_len = p - new->raw_value;
+			}
+
 			new->value = tmp;
-			new->raw_len = p - new->raw_value;
 			new->decoded = g_mime_utils_header_decode_text (new->value);
 
 			if (new->decoded != NULL) {
@@ -1300,44 +1306,6 @@ rspamd_message_from_data (struct rspamd_task *task, GByteArray *data,
 	}
 }
 
-static inline const gchar *
-rspamd_message_find_body_start (const gchar *headers_end, const gchar *body_end)
-{
-	const gchar *p = headers_end;
-	enum {
-		st_start = 0,
-		st_cr,
-		st_lf,
-	} state = st_start;
-
-	if (headers_end + 1 >= body_end) {
-		return headers_end;
-	}
-
-	switch (state) {
-	case st_start:
-		if (*p == '\r') {
-			p ++;
-			state = st_cr;
-		}
-		else if (*p == '\n') {
-			p ++;
-			state = st_lf;
-		}
-		break;
-	case st_cr:
-		if (*p == '\n' && p < body_end) {
-			/* CRLF */
-			p ++;
-		}
-		break;
-	case st_lf:
-		break;
-	}
-
-	return p;
-}
-
 gboolean
 rspamd_message_parse (struct rspamd_task *task)
 {
@@ -1354,7 +1322,7 @@ rspamd_message_parse (struct rspamd_task *task)
 	struct received_header *recv, *trecv;
 	const gchar *p;
 	gsize len;
-	goffset hdr_pos;
+	goffset hdr_pos, body_pos;
 	gint i;
 	gdouble diff, *pdiff;
 	guint tw, *ptw, dw;
@@ -1450,15 +1418,12 @@ rspamd_message_parse (struct rspamd_task *task)
 			str.str = tmp->data;
 			str.len = tmp->len;
 
-			hdr_pos = rspamd_string_find_eoh (&str);
+			hdr_pos = rspamd_string_find_eoh (&str, &body_pos);
 
 			if (hdr_pos > 0 && hdr_pos < tmp->len) {
-				static const gchar *body_start;
-
-				body_start = rspamd_message_find_body_start (p + hdr_pos,
-						p + len);
 				task->raw_headers_content.begin = (gchar *) (p);
-				task->raw_headers_content.len = body_start - p;
+				task->raw_headers_content.len = hdr_pos;
+				task->raw_headers_content.body_start = p + body_pos;
 
 				if (task->raw_headers_content.len > 0) {
 					process_raw_headers (task, task->raw_headers,

src/libserver/dkim.c

@@ -1793,7 +1793,11 @@ rspamd_dkim_check (rspamd_dkim_context_t *ctx,
 	/* First of all find place of body */
 	p = task->msg.begin;
 	body_end = task->msg.begin + task->msg.len;
-	body_start = task->msg.begin + task->raw_headers_content.len;
+	body_start = task->raw_headers_content.body_start;
+
+	if (!body_start) {
+		return DKIM_RECORD_ERROR;
+	}
 
 	/* Start canonization of body part */
 	if (!rspamd_dkim_canonize_body (&ctx->common, body_start, body_end)) {
@@ -2049,12 +2053,16 @@ rspamd_dkim_sign (struct rspamd_task *task,
 	/* First of all find place of body */
 	p = task->msg.begin;
 	body_end = task->msg.begin + task->msg.len;
-	body_start = task->msg.begin + task->raw_headers_content.len;
+	body_start = task->raw_headers_content.body_start;
 
 	if (len > 0) {
 		ctx->common.len = len;
 	}
 
+	if (!body_start) {
+		return NULL;
+	}
+
 	/* Start canonization of body part */
 	if (!rspamd_dkim_canonize_body (&ctx->common, body_start, body_end)) {
 		return NULL;

src/libserver/task.h

@@ -142,7 +142,11 @@ struct rspamd_task {
 	GMimeMessage *message;							/**< message, parsed with GMime						*/
 	GPtrArray *parts;								/**< list of parsed parts							*/
 	GPtrArray *text_parts;							/**< list of text parts								*/
-	rspamd_ftok_t raw_headers_content;				/**< list of raw headers							*/
+	struct {
+		const gchar *begin;
+		gsize len;
+		const gchar *body_start;
+	} raw_headers_content;				/**< list of raw headers							*/
 	GPtrArray *received;							/**< list of received headers						*/
 	GHashTable *urls;								/**< list of parsed urls							*/
 	GHashTable *emails;								/**< list of parsed emails							*/

src/libutil/str_util.c

@@ -1413,7 +1413,7 @@ rspamd_substring_search_twoway (const gchar *in, gint inlen,
 
 
 goffset
-rspamd_string_find_eoh (GString *input)
+rspamd_string_find_eoh (GString *input, goffset *body_start)
 {
 	const gchar *p, *c = NULL, *end;
 	enum {
@@ -1459,6 +1459,10 @@ rspamd_string_find_eoh (GString *input)
 				}
 				else {
 					/* We have \r\r[^\n] */
+					if (body_start) {
+						*body_start = p - input->str + 1;
+					}
+
 					return p - input->str;
 				}
 			}
@@ -1474,6 +1478,9 @@ rspamd_string_find_eoh (GString *input)
 		case got_lf:
 			if (*p == '\n') {
 				/* We have \n\n, which is obviously end of headers */
+				if (body_start) {
+					*body_start = p - input->str + 1;
+				}
 				return p - input->str;
 			}
 			else if (*p == '\r') {
@@ -1517,11 +1524,21 @@ rspamd_string_find_eoh (GString *input)
 			break;
 		case got_linebreak_lf:
 			g_assert (c != NULL);
+			if (body_start) {
+				/* \r\n\r\n */
+				*body_start = p - input->str;
+			}
+
 			return c - input->str;
 		}
 	}
 
 	if (state == got_linebreak_lf) {
+		if (body_start) {
+			/* \r\n\r\n */
+			*body_start = p - input->str;
+		}
+
 		return c - input->str;
 	}
 

src/libutil/str_util.h

@@ -269,7 +269,7 @@ goffset rspamd_substring_search_twoway (const gchar *in, gint inlen,
  * Hence, to obtain the real EOH position, it is also required to skip
  * space characters
  */
-goffset rspamd_string_find_eoh (GString *input);
+goffset rspamd_string_find_eoh (GString *input, goffset *body_start);
 
 
 #define rspamd_ucl_emit_gstring(o, t, target) \