[CritFix] Switch from ragel to C for Content-Type parsing

src/CMakeLists.txt

@@ -106,7 +106,6 @@ SET(RAGEL_DEPENDS "${CMAKE_SOURCE_DIR}/src/ragel/smtp_address.rl"
 	"${CMAKE_SOURCE_DIR}/src/ragel/smtp_ip.rl"
 	"${CMAKE_SOURCE_DIR}/src/ragel/smtp_whitespace.rl"
 	"${CMAKE_SOURCE_DIR}/src/ragel/smtp_received.rl"
-	"${CMAKE_SOURCE_DIR}/src/ragel/content_type.rl"
 	"${CMAKE_SOURCE_DIR}/src/ragel/content_disposition.rl")
 RAGEL_TARGET(ragel_smtp_addr
 	INPUTS ${CMAKE_SOURCE_DIR}/src/ragel/smtp_addr_parser.rl
@@ -118,11 +117,6 @@ RAGEL_TARGET(ragel_smtp_received
 	DEPENDS ${RAGEL_DEPENDS}
 	COMPILE_FLAGS -T1
 	OUTPUT ${CMAKE_CURRENT_BINARY_DIR}/smtp_received_parser.rl.c)
-RAGEL_TARGET(ragel_content_type
-	INPUTS ${CMAKE_SOURCE_DIR}/src/ragel/content_type_parser.rl
-	DEPENDS ${RAGEL_DEPENDS}
-	COMPILE_FLAGS -G2
-	OUTPUT ${CMAKE_CURRENT_BINARY_DIR}/content_type.rl.c)
 RAGEL_TARGET(ragel_content_disposition
 	INPUTS ${CMAKE_SOURCE_DIR}/src/ragel/content_disposition_parser.rl
 	DEPENDS ${RAGEL_DEPENDS}

src/libmime/content_type.c

@@ -71,27 +71,286 @@ rspamd_content_type_add_param (rspamd_mempool_t *pool,
 	}
 }
 
-struct rspamd_content_type *
-rspamd_content_type_parse (const gchar *in,
-		gsize len, rspamd_mempool_t *pool)
+static struct rspamd_content_type *
+rspamd_content_type_parser (const gchar *in, gsize len, rspamd_mempool_t *pool)
 {
+	guint obraces = 0, ebraces = 0, qlen = 0;
+	const gchar *p, *c, *end, *pname_start = NULL, *pname_end = NULL;
 	struct rspamd_content_type *res = NULL, val;
-	rspamd_ftok_t srch;
+	gboolean eqsign_seen = FALSE;
+	enum {
+		parse_type,
+		parse_subtype,
+		parse_after_subtype,
+		parse_param_name,
+		parse_param_after_name,
+		parse_param_value,
+		parse_param_value_after_quote,
+		parse_space,
+		parse_quoted,
+		parse_comment,
+	} state = parse_space, next_state = parse_type;
+
+	p = in;
+	c = p;
+	end = p + len;
+	memset (&val, 0, sizeof (val));
+	val.lc_data = (gchar *)in;
+
+	while (p < end) {
+		switch (state) {
+		case parse_type:
+			if (g_ascii_isspace (*p) || *p == ';') {
+				/* We have type without subtype */
+				val.type.begin = c;
+				val.type.len = p - c;
+				state = parse_after_subtype;
+			} else if (*p == '/') {
+				val.type.begin = c;
+				val.type.len = p - c;
+				state = parse_space;
+				next_state = parse_subtype;
+				p++;
+			} else {
+				p++;
+			}
+			break;
+		case parse_subtype:
+			if (g_ascii_isspace (*p) || *p == ';') {
+				val.subtype.begin = c;
+				val.subtype.len = p - c;
+				state = parse_after_subtype;
+			} else {
+				p++;
+			}
+			break;
+		case parse_after_subtype:
+			if (*p == ';' || g_ascii_isspace (*p)) {
+				p++;
+			} else if (*p == '(') {
+				c = p;
+				state = parse_comment;
+				next_state = parse_param_name;
+				obraces = 1;
+				ebraces = 0;
+				pname_start = NULL;
+				pname_end = NULL;
+				eqsign_seen = FALSE;
+				p++;
+			} else {
+				c = p;
+				state = parse_param_name;
+				pname_start = NULL;
+				pname_end = NULL;
+				eqsign_seen = FALSE;
+			}
+			break;
+		case parse_param_name:
+			if (*p == '=') {
+				pname_start = c;
+				pname_end = p;
+				state = parse_param_after_name;
+				eqsign_seen = TRUE;
+				p++;
+			} else if (g_ascii_isspace (*p)) {
+				pname_start = c;
+				pname_end = p;
+				state = parse_param_after_name;
+			} else {
+				p++;
+			}
+			break;
+		case parse_param_after_name:
+			if (g_ascii_isspace (*p)) {
+				p++;
+			} else if (*p == '=') {
+				if (eqsign_seen) {
+					/* Treat as value start */
+					c = p;
+					eqsign_seen = FALSE;
+					state = parse_space;
+					next_state = parse_param_value;
+					p++;
+				} else {
+					eqsign_seen = TRUE;
+					p++;
+				}
+			} else {
+				if (eqsign_seen) {
+					state = parse_param_value;
+					c = p;
+				} else {
+					/* Invalid parameter without value */
+					c = p;
+					state = parse_param_name;
+					pname_start = NULL;
+					pname_end = NULL;
+				}
+			}
+			break;
+		case parse_param_value:
+			if (*p == '"') {
+				p++;
+				c = p;
+				state = parse_quoted;
+				next_state = parse_param_value_after_quote;
+			} else if (g_ascii_isspace (*p)) {
+				if (pname_start && pname_end && pname_end > pname_start) {
+					rspamd_content_type_add_param (pool, &val, pname_start,
+							pname_end, c, p);
 
-	val.lc_data = rspamd_mempool_alloc (pool, len);
-	memcpy (val.lc_data, in, len);
-	rspamd_str_lc (val.lc_data, len);
+				}
+
+				state = parse_space;
+				next_state = parse_param_name;
+				pname_start = NULL;
+				pname_end = NULL;
+			} else if (*p == '(') {
+				if (pname_start && pname_end && pname_end > pname_start) {
+					rspamd_content_type_add_param (pool, &val, pname_start,
+							pname_end, c, p);
+				}
+
+				obraces = 1;
+				ebraces = 0;
+				p++;
+				state = parse_comment;
+				next_state = parse_param_name;
+				pname_start = NULL;
+				pname_end = NULL;
+			} else {
+				p++;
+			}
+			break;
+		case parse_param_value_after_quote:
+			if (pname_start && pname_end && pname_end > pname_start) {
+				rspamd_content_type_add_param (pool, &val, pname_start,
+						pname_end, c, c + qlen);
+			}
 
-	if (rspamd_content_type_parser (val.lc_data, len, &val, pool)) {
+			if (g_ascii_isspace (*p)) {
+				state = parse_space;
+				next_state = parse_param_name;
+				pname_start = NULL;
+				pname_end = NULL;
+			} else if (*p == '(') {
+				obraces = 1;
+				ebraces = 0;
+				p++;
+				state = parse_comment;
+				next_state = parse_param_name;
+				pname_start = NULL;
+				pname_end = NULL;
+			} else {
+				state = parse_param_name;
+				pname_start = NULL;
+				pname_end = NULL;
+				c = p;
+			}
+			break;
+		case parse_quoted:
+			if (*p == '\\') {
+				/* Quoted pair */
+				if (p + 1 < end) {
+					p += 2;
+				} else {
+					p++;
+				}
+			} else if (*p == '"') {
+				qlen = p - c;
+				state = next_state;
+			} else {
+				p++;
+			}
+			break;
+		case parse_comment:
+			if (*p == '(') {
+				obraces++;
+				p++;
+			} else if (*p == ')') {
+				ebraces++;
+				p++;
+
+				if (ebraces == obraces && p < end) {
+					if (g_ascii_isspace (*p)) {
+						state = parse_space;
+					} else {
+						c = p;
+						state = next_state;
+					}
+				}
+			} else {
+				p++;
+			}
+			break;
+		case parse_space:
+			if (g_ascii_isspace (*p)) {
+				p++;
+			} else if (*p == '(') {
+				obraces = 1;
+				ebraces = 0;
+				p++;
+				state = parse_comment;
+			} else {
+				c = p;
+				state = next_state;
+			}
+			break;
+		}
+	}
+
+	/* Process leftover */
+	switch (state) {
+	case parse_type:
+		val.type.begin = c;
+		val.type.len = p - c;
+		break;
+	case parse_subtype:
+		val.subtype.begin = c;
+		val.subtype.len = p - c;
+		break;
+	case parse_param_value:
+		if (pname_start && pname_end && pname_end > pname_start) {
+			rspamd_content_type_add_param (pool, &val, pname_start,
+					pname_end, c, p);
+
+		}
+	case parse_param_value_after_quote:
+		if (pname_start && pname_end && pname_end > pname_start) {
+			rspamd_content_type_add_param (pool, &val, pname_start,
+					pname_end, c, c + qlen);
+		}
+		break;
+	default:
+		break;
+	}
+
+	if (val.type.len > 0) {
 		res = rspamd_mempool_alloc (pool, sizeof (val));
 		memcpy (res, &val, sizeof (val));
+	}
 
+	return res;
+}
+
+struct rspamd_content_type *
+rspamd_content_type_parse (const gchar *in,
+		gsize len, rspamd_mempool_t *pool)
+{
+	struct rspamd_content_type *res = NULL;
+	rspamd_ftok_t srch;
+	gchar *lc_data;
+
+	lc_data = rspamd_mempool_alloc (pool, len);
+	memcpy (lc_data, in, len);
+	rspamd_str_lc (lc_data, len);
+
+	if ((res = rspamd_content_type_parser (lc_data, len, pool)) != NULL) {
 		if (res->attrs) {
 			rspamd_mempool_add_destructor (pool,
 					(rspamd_mempool_destruct_t)g_hash_table_unref, res->attrs);
 		}
 
-
 		/* Now do some hacks to work with broken content types */
 		if (res->subtype.len == 0) {
 			res->flags |= RSPAMD_CONTENT_TYPE_BROKEN;
@@ -157,7 +416,7 @@ rspamd_content_type_parse (const gchar *in,
 		}
 	}
 	else {
-		msg_warn_pool ("cannot parse content type: %*s", (gint)len, val.lc_data);
+		msg_warn_pool ("cannot parse content type: %*s", (gint)len, lc_data);
 	}
 
 	return res;

src/libmime/smtp_parsers.h

@@ -26,9 +26,6 @@ int rspamd_smtp_recieved_parse (struct rspamd_task *task,
 		const char *data, size_t len, struct received_header *rh);
 int rspamd_smtp_addr_parse (const char *data, size_t len,
 		struct rspamd_email_address *addr);
-
-gboolean rspamd_content_type_parser (const char *data, size_t len,
-		struct rspamd_content_type *ct, rspamd_mempool_t *pool);
 gboolean rspamd_content_disposition_parser (const char *data, size_t len,
 		struct rspamd_content_disposition *cd, rspamd_mempool_t *pool);
 

src/ragel/content_type.rl

@@ -9,12 +9,11 @@
   comment        =   "(" (FWS? ccontent)* FWS? ")";
   CFWS           =   ((FWS? comment)+ FWS?) | FWS;
   qcontent = qtextSMTP | quoted_pairSMTP;
-  quoted_string = CFWS?
-                  (DQUOTE
+  quoted_string = (DQUOTE
                     (((FWS? qcontent)* FWS?) >Quoted_Str_Start %Quoted_Str_End)
-                  DQUOTE) CFWS?;
+                  DQUOTE);
   token = 0x21..0x27 | 0x2a..0x2b | 0x2c..0x2e | 0x30..0x39 | 0x41..0x5a | 0x5e..0x7e;
-  value = (quoted_string | (token | 0x3d)+) >Param_Value_Start %Param_Value_End;
+  value = (quoted_string | (token)+) >Param_Value_Start %Param_Value_End;
   attribute = (token+) >Param_Name_Start %Param_Name_End;
   parameter = CFWS? attribute FWS? "=" FWS? value CFWS?;
 

src/ragel/content_type_parser.rl

@@ -42,6 +42,7 @@
   }
 
   action Param_Name_Start {
+          printf("name start: %s\n", p);
     qstart = NULL;
     qend = NULL;
     pname_start = p;
@@ -50,25 +51,32 @@
 
 
   action Param_Name_End {
+          printf("name end: %s\n", p);
     if (qstart) {
       pname_start = qstart;
     }
-    if (qend && qend >= qstart) {
+    if (qstart && qend && qend >= qstart) {
       pname_end = qend;
     }
     else if (p >= pname_start) {
       pname_end = p;
     }
-    qstart = NULL;
-    qend = NULL;
+
+    if (qstart && qend) {
+      qstart = NULL;
+      qend = NULL;
+    }
   }
 
 
   action Param_Value_Start {
-    qstart = NULL;
-    qend = NULL;
+          printf("value start: %s\n", p);
+    if (qend) {
+      qstart = NULL;
+      qend = NULL;
+    }
 
-    if (pname_end) {
+    if (pname_end && !pvalue_start) {
       pvalue_start = p;
       pvalue_end = NULL;
     }
@@ -76,30 +84,40 @@
 
 
   action Param_Value_End {
-    if (pname_end) {
+          printf("value end: %s\n", p);
+    if (pname_end && pname_start) {
       if (qstart) {
         pvalue_start = qstart;
+
+        if (!qend) {
+          pvalue_end = NULL;
+        }
       }
+
       if (qend && qend >= qstart) {
-        pvalue_end = qend;
+        if (qstart) {
+          pvalue_end = qend;
+        }
+        else {
+          pvalue_end = NULL;
+        }
       }
-      else if (p >= pvalue_start) {
+      else if (!qstart && p >= pvalue_start) {
         pvalue_end = p;
       }
-      qstart = NULL;
-      qend = NULL;
 
-      if (pvalue_end && pvalue_end > pvalue_start && pname_end > pname_start) {
-        rspamd_content_type_add_param (pool, ct, pname_start, pname_end, pvalue_start, pvalue_end);
+      if (pname_start && pvalue_start && pvalue_end && pvalue_end > pvalue_start
+              && pname_end > pname_start) {
+        rspamd_content_type_add_param (pool, ct, pname_start, pname_end,
+                pvalue_start, pvalue_end);
+        pname_start = NULL;
+        pname_end = NULL;
+        pvalue_start = NULL;
+        pvalue_end = NULL;
+        qend = NULL;
+        qstart = NULL;
       }
     }
-
-    pname_start = NULL;
-    pname_end = NULL;
-    pvalue_start = NULL;
-    pvalue_end = NULL;
-    qend = NULL;
-    qstart = NULL;
   }
 
   action Quoted_Str_Start {